The decline in the quality of service of electric distribution utilities in Latin America was a topic requiring investigation because

of its effect on the sustainability of privatization. The quality of service could be a hidden cost of privatization because the regulation policies do not include the quality-ofservice issue. The aim of this study was to determine the influence of regulation, monitoring, and ownership on the quality of service of Latin American electric distribution utilities from 2002 to 2007. One of the relevant findings is that in Latin America, utilities under the rateof-return regulation, rather than under the regulation-by-incentives system, produce a better quality of service. The second relevant finding is that the private utilities under a monitoring policy with sanction provide a better quality of service than do utilities under a monitoring policy without sanction. The third relevant finding is that the private utilities offer a better quality of service than do the state-owned utilities. The fourth relevant finding is that the private utility controlled under a regulation-by-incentives (price cap or model firm) system offers a similar quality of service to that supplied by the state-owned utilities under rate-of-return regulation. The result confirms that complementing the regulation by incentives applied to the private utilities with a monitoring of quality of service with sanction produces a quality of service comparable with the service of the electric utilities under rate-of-return regulation. The results of the research indicate that the schemes of regulation, types of monitoring, and forms of ownership affect the quality of service. The research conclusions may help regulators and policy makers to implement policies to improve the quality of service. ____ Miguel Rvolo holds his M. Phil degree from the Maastricht School of Management, Master degree in Economy and Regulation from Universitat de Barcelona, Master degree in Business Administration from Universidad del Pacfico, and an Electrical Engineering degree from Universidad Nacional del Centro del Per. Currently he is a guest professor at the Regional Energy Integration Committee of Latin America (CIER). Miguel Rvolo is now a manager of distribution regulation in OSINERGMIN, the energy regulator in Peru. He is an energy specialist with a background in power engineering. He has more than 25 years of experience in the field. He has held a managerial position in a distribution utility. His most recent experience has been in the reform of the energy sector in particular in Peru and Latin American.

Influence of Regulation, Monitoring, and Ownership on the Quality of Service of Latin American Electric Distribution Utilities from 2002 to 2007 | Miguel Juan Rvolo Acevedo

Influence of Regulation, Monitoring, and Ownership on the Quality of Service of Latin American Electric Distribution Utilities from 2002 to 2007

Miguel Juan Rvolo Acevedo

the globally networked management school

Influence of Regulation, Monitoring, and Ownership on the Quality of Service of Latin American Electric Distribution Utilities from 2002 to 2007

Dissertation

To obtain the degree of Doctor of Business Administration at the Maastricht School of Management, under authority of the Director Dean Prof. P.P. De Gijsel

by

Miguel Juan Revolo Acevedo

Master of Business Administration, born in Huancayo (Peru)

MAASTRICHT SCHOOL OF MANAGEMENT April 2010

 2010 by M. Rvolo, Maastricht School of Management. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the publisher. ii

This dissertation is approved of by the Doctoral Supervisors: Prof. Dr. Esteban Serra Maastricht School of Management, The Netherlands

Abstract The decline in the quality of service of electric distribution utilities in Latin America was a topic requiring investigation because of its effect on the sustainability of privatization. The quality of service could be a hidden cost of privatization because the regulation policies do not include the quality-of-service issue. The aim of this study was to determine the influence of regulation, monitoring, and ownership on the quality of service of Latin American electric distribution utilities from 2002 to 2007. A panel data technique was appropriate for the analysis because of the need to incorporate cross-sectional and time-series data. The results of the research indicate that the schemes of regulation, types of monitoring, and forms of ownership affect the quality of service. The research conclusions may help regulators and policy makers to implement policies to improve the quality of service. Keywords: Regulation, Monitoring, Ownership, Quality, Utility.

Acknowledgments I thank Jesus Christ for his eternal and almighty love, who gives me the strength and wisdom to success in this challenged research. I also send my gratitude to my parents Alberto and Mila. Special thanks and appreciation go to my wife Katia Starke, who support and understanding of my frequent absence from the family during my doctoral studies. I would like to dedicate this dissertation to my children Miguel and Arlet. Also, I would like to thank the professors of CENTRUM and MSM and particularly express my gratitude to Dr. Fernando DAlessio who, as Director General, motivated me to complete this thesis. I also wish to thank my assessors, Dr. Esteban Serra of MSM and Dr. Jos Tavara of CENTRUM, for their valuable comments and suggestions. Further, I appreciated the feedback and advice of Erik de Bruijn, Tooraj Jamasb, Edwin Quintanilla, Xavier Fagedas, Sandford Berg, Anton Costas, Germa Bel, and Bernard Tenenbaum. My colleagues of the regulatory bodies ANEEL, ASEP, ARESEP, CNEE, CONELEC, OSINERGMIN, SEC, and SIGET also deserve recognition for their support in the process of data collection. Special recognition goes to my colleagues of the electric distribution utilities of Latin America who represented the following corporations: COPEL, CPFL, ENDESA, IBERDROLA, NEONERGIA, and SAESA-FRONTEL. I would like to express my thanks to CIER and Juan Carlos Belza for providing valuable information. I also recognize the Latin American consultants who provided me with specific contacts that facilitated the data collection. Finally, I want to thank the authorities of OSINERGMIN for trusting in the research.

Summary The lack of empirical work on how regulation, monitoring, and ownership affect the quality of service supplied by the Latin American electric distribution utilities formed the motivation for the study. The existing literature illustrates that private electric distribution utilities would worsen the quality of service, which could explain why Latin Americans have been rejecting privatization. The quality of service could be a hidden cost of privatization because the regulation policies do not include the quality-of-service issue. The research involved examining whether the electric distribution utilities that operate under a rate-of-return regulation scheme offer a better quality of service than do those utilities operating according to regulation by incentives. Theoretically, the controlled utilities under a rate-of-return regulation would offer a better quality of service, and the utilities under a regulation-by-incentives scheme would tend to diminish the quality of service to become more profitable. In addition, the research included analyzing whether the application of a monitoring policy with sanction improved the levels of quality of service and whether the type of ownership which determines the managers behavior, affected the quality of service. One of the relevant findings is that in Latin America, utilities under the rate-ofreturn regulation, rather than under the regulation-by-incentives system, produce a better quality of service. The second relevant finding is that the private utilities under a monitoring policy with sanction provide a better quality of service than do utilities under a monitoring policy without sanction. The third relevant finding is that the private utilities offer a better quality of service than do the state-owned utilities. The fourth relevant finding is that the private utility controlled under a regulation-by-incentives (price cap or model firm) system offers a similar quality of service to that supplied by the state-owned utilities under rate-of-return regulation. The result confirms that complementing the regulation by incentives applied to the private utilities with a monitoring of quality of service with sanction produces a quality of service comparable with the service of the electric utilities under rate-ofreturn regulation. Finally, the research shows that the policies of price regulation, the quality regulation, and the type of ownership affect the level of quality of service supplied by the Latin American electric distribution utilities in Latin America.

vi

Table of Contents List of Tables ......................................................................................................... ix List of Figures..........................................................................................................x 1. Introduction .........................................................................................1 1.1 Background of the Problem ..................................................................1 1.2 Statement of the Problem ......................................................................2 1.3 Purpose of the Study .............................................................................2 1.4 Significance of the Problem ..................................................................3 1.5 Research Questions ...............................................................................4 1.6 Hypotheses ............................................................................................5 1.7 Theoretical Framework .........................................................................6 1.8 Definition of Terms ...............................................................................9 1.9 Assumptions ..........................................................................................9 1.10 Limitations ............................................................................................9 1.11 Delimitations .......................................................................................10 1.12 Summary .............................................................................................10 2. Literature Review .............................................................................11 2.1 Electric Distribution Industry ..............................................................11 2.1.1 Electric Distribution Activity ..............................................................11 2.1.2 Objective of the Electric Industry .......................................................12 2.1.3 Natural Monopoly and Regulation ......................................................13 2.1.4 Parameters of Quality of Service ........................................................13 2.2 Quality of Service in the Electric Distribution Industry ......................14 2.2.1 Definition of Quality of Service ..........................................................14 2.2.2 Dimensions of the Quality of Service .................................................15 2.3 Regulation Schemes and Quality of Service .......................................17 2.3.1 Rate-of-Return Regulation ..................................................................17 2.3.2 Regulation by Incentives .....................................................................19 2.4 Quality-of-Service Monitoring............................................................22 2.5 Ownership and Quality of Service ......................................................26 2.6 Summary .............................................................................................30 2.7 Conclusion ..........................................................................................31 3. Methodology ......................................................................................33 3.1 Research Design ..................................................................................33 3.2 Appropriateness of the Design ............................................................37 3.3 Research Questions .............................................................................38 3.4 Hypotheses ..........................................................................................38 3.5 Population ...........................................................................................39 3.6 Informed Consent ................................................................................40 3.7 Sampling Frame ..................................................................................40 3.8 Confidentiality ....................................................................................41 3.9 Geographic Location ...........................................................................42 3.10 Instrumentation ...................................................................................42 3.11 Data Collection ...................................................................................42 3.12 Data Analysis ......................................................................................43 vii

3.13 Validity and Reliability .......................................................................44 3.13.1 General Regressions ............................................................................44 3.13.2 Estimations, Analysis, and Tests of the Model ...................................46 3.13.2.1 Correlation Matrix...............................................................................46 3.13.2.2 Multicolinearity Test ...........................................................................50 3.13.2.3 Specification Test................................................................................52 3.13.2.4 Autocorrelation Test ...........................................................................55 3.13.2.5 Heteroscedasticity Test .......................................................................55 3.13.3 Conclusions of the Test .......................................................................55 3.14 Summary .............................................................................................56 4. Presentation and Analysis of the Data.............................................57 4.1 Findings ..............................................................................................57 4.2 Summary .............................................................................................66 5. Conclusions and Recommendations ................................................68 5.1 Conclusions .........................................................................................68 5.2 Implications.........................................................................................69 5.3 Recommendations ...............................................................................70 References .............................................................................................................72 Appendix A: Informed Consent Form ...................................................................81 Appendix B: Research Questionnaire ....................................................................82 Appendix C: Sources of Information .....................................................................87 Appendix D: Distribution Frequency and Q-Q Graphs of the Variables ...............94 Appendix E: Stata Reports...................................................................................103

viii

List of Tables Table 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Page Schemes of Regulation in Latin American Utilities ..................................22 Monitoring Mechanisms in Latin American Utilities................................25 Types of Ownership in Latin American Utilities ......................................30 General Description of Dependent and Independent Variables ................35 General Description of Control Variables (Zit) .........................................36 Expected Relation between Economic and Structural Factors and Quality of Service...............................................................................37 Expected Relation between Political and Institutional Variables and Quality of Service...............................................................................37 Typology of the Utilities: Number of Cases (and Number of Observations) ............................................................................................40 Sample Population from the 10 Latin American Countries ......................41 Population and Market Participation of Distribution Utilities ...................41 Descriptive Statistics of the Model Variables ...........................................43 General Regressions ..................................................................................45 Pearson Correlation Coefficient between the SAIDI and the Independent Variables...............................................................................47 Pearson Correlation Coefficient between the SAIFI and the Independent Variables...............................................................................49 Variance Inflation Index (VIF) of the General Model ..............................50 Relation between Regulation and Monitoring Variables ..........................51 Variance Inflation Index (VIF) .................................................................52 Results of the F Test ..................................................................................53 Results of the Breusch-Pagan Test ............................................................53 Results of the Hausman Test .....................................................................54 Regression: Regulation (RORR and PCAP) and Ownership ....................58 Regression: Monitoring and Ownership ....................................................59 Summary of the Wald Test........................................................................65

ix

List of Figures Figure Page 1. Multidimensional factors of reform sector. ..................................................6 2. Econometric model. ...................................................................................34 3. Scheme of hypotheses. ...............................................................................38

1.

Introduction

In the 1990s, because of a severe electricity crisis, many Latin American countries privatized state-owned electric companies. The aims of this decision included attracting foreign capital to invest in new-generation power plants, improving access to service, enhancing quality of service, and cutting fiscal losses (Paliza, 1999). The focus of most evaluations of privatization has been operation and financial performance. However, electric utilities may pursue increased efficiency and profitability at the expense of workers, customers, and other groups through higher prices, reduced levels of employment, and low quality of service (Kessides, 2004; Shirley, 2004). After a decade of satisfaction with the government, the private utilities and consumers raised concerns regarding the incompleteness of the goals of privatization. The concerns resulted in a discussion on the fairness of the model and restarted the ideological debate on whether the state or the private sector should provide the electric service (Kikeri & Nellis, 2004; Millan, 2006). Latin Americans disagree or strongly disagree that the privatization of state companies has benefited their country (Panizza & Yaez, 2006). The diminishing quality of service is one of the main reasons for rejecting privatization in Latin American countries (Shirley, 2004). Understanding why the quality of service is declining may add knowledge to the debate. 1.1 Background of the Problem In Latin America, researchers have been conducting cross-sectional studies to measure productivity and performance, which influence the profitability of the utilities. The results of the studies show that the productivity and profitability of the private electric utilities have improved since the reform (privatization) through reduction of costs (Alva & Bonifaz, 2004; Estache & Rossi, 2004; Farsi & Filippini, 2004; Fisher, Gutierrez, & Serra, 2003; Rodrguez Pardina, Rossi, & Ruzzier, 1998; Rossi & Ruzzier, 2002; Rudnick & Zolezzi, 2001). Other researchers argued that even though utilities have improved efficiency, regulators have been unable to pass on the cost efficiency in terms of low prices to customers (Estache, Guasch, & Trujillo, 2003; Fisher et al., 2003). Furthermore, empirical evidence does not indicate definite conclusions about the effects of the privatization of electric distribution utilities in Latin America because some important variables, such as the quality of service, are missing in the analysis (Estache & Rossi, 2004). According to Macedo (2004), researchers could improve cost-efficiency studies by incorporating a measure of quality of service in the analysis because a trade-off between the quality of service and the reduction of costs exists. Moreover, researchers have identified that the level of quality of service in the electric industry is diminishing progressively, negatively affecting consumers acceptance of the reform (Berg, 2006; Costas, 2006; Shirley, 2004). Therefore, a relevant issue requiring research involved understanding why electric 1

distribution utilities are lowering the quality of service. The privatization process seems to omit a policy of quality of service because the regulatory regime does not provide explicit signals that oblige the utilities to invest in and to expand on the quality of service (Costas, 2006). In developed countries, researchers have made some effort to explain empirically why the utilities lower quality of service. The studies occurred mainly in the telecommunications sector (Ai & Sappington, 1998; Ai & Sappington, 2002; Clements, 2001). Recently, in Latin America, Andres, Foster, and Guash (2006) elaborated on an analysis of the changes of performance indicators of the electric distribution utilities and found that privatization caused significant improvements in labor productivity, efficiency, and product or service quality in the short term, in contrast to the findings of other researchers. Nonetheless, Andres et al. recognized that ownership does matter and recommended further research on the differences among regulatory regimes, investors, and sector characteristics. Thus, addressing the quality-of-service problem appropriately required research to determine why the quality of service in Latin American electric utilities diminished after the reform. Latin America constitutes a natural laboratory where countries coexist with different regulation regimes, mechanisms of monitoring, and ownership. This particular structure presented an opportunity to test how these differences affect the level of quality of service. 1.2 Statement of the Problem In Latin America, 75% of the population rejects privatization (Panizza & Yaez, 2006; Shirley, 2004). The promised reform benefits, which included lower prices, improved quality of service, and increased level of access, have not materialized (Millan, 2006). The result included risking foreign investment of about $10 billion because of the threat of nationalization (Kikeri & Nellis, 2004). After the privatization in Latin America, researchers analyzed prices and costs extensively (Alva & Bonifaz, 2004; Estache & Rossi, 2004; Farsi & Filippini, 2004; Fisher et al., 2003; Rodrguez Pardina et al., 1998; Rudnick & Zolezzi, 2001). However, few studies included an analysis of the quality of service. No studies exist to indicate the factors that influence the quality of service provided by the electric distribution utilities (Andres et al., 2006; Estache & Rossi, 2004; Jamasb, Motta, Newbery, & Pollitt, 2005); thus, this study may have closed the gap. The results of this study may help regulators and policy makers to design future policies in Latin America and other developing countries. 1.3 Purpose of the Study The essence of this quantitative research was to examine the influence that regulation regimes, mechanisms of monitoring, and ownership have on the quality

of service provided by Latin American electric distribution utilities. The study involved the development of an econometric model based on the models of Ai and Sappington (1998), Clements (2001), and Ter-Martirosyan (2003). The research included independent variables related to the level of quality of service; independent variables related to electric distribution infrastructure, operations and maintenance, and market and performance indices; and dummy variables related to the regulation regime, monitoring quality of service, and ownership. The research results may close a gap in the literature because the purpose of the empirical study was to determine the effect of the schemes of regulation, types of monitoring, and forms of ownership on the quality of service, all key factors in the design of government policies. The findings may aid in designing policies of price, quality of service, and ownership that allow for the optimum development of the electric distribution utility. They may be applied in countries with similar characteristics to Latin America. 1.4 Significance of the Problem One important reason for this study was that many researchers who tested the results of the reform in Latin America neglected the issue of quality of service (Estache & Rossi, 2004). In the first decade after privatization, as a product of the lack of information, researchers focused mainly on performance evaluation, which helped to explain the improvement in cost efficiency and profitability of the privatized firms. Recently, however, researchers have recommended an analysis of the quality-of-service factor due to its noticeable reduction (Berg, 2006; Costas, 2006; Millan, 2006; Shirley, 2004). A second reason for the study was that in Latin America, a lack of studies that involve analyzing the factors of regulation regimes, ownership, and environmental characteristics and their influence on the decisions of the electric distribution companies regarding the provision of quality of service is evident (Andres et al., 2006). The results of this study illustrated whether a significant relationship exists between the regulatory regimes, monitoring mechanisms, and ownership forms and the quality of service of electric distribution utilities in Latin America. The regulation regime, constructed on the economic regulation theory, is a fundamental issue that influences the economical decision making of the utility (Kidokoro, 2002; Sheshinski, 1976; Spence, 1975). The point that regulation regimes are related to the economic signals, which influence the level of investments and expenses oriented to quality of service, was important and required analysis in the context of Latin American countries because applying the outcomes of studies conducted in developed countries would not be appropriate (Estache & Rossi, 2004) . Similarly, empirical proof of the influence that the mechanisms of regulation have on the provision of quality of service in Latin America was required (Kriehn, 2005; Lewis & Sappington, 1991).

In addition, analyzing the influence of the type of ownership on the level of quality of service of Latin American electric distribution utilities was important. The analysis resulted in insight into the theoretical approach of incomplete contracts, which includes the argument that state-owned utilities provide better quality of service than do private utilities when full specification of the standards of quality of service and costs recognition is not evident (Hart, Shleifer, & Vishny, 1997). Moreover, the study incorporated panel data, following the recommendation of many researchers, to capture the temporal variability of the factors affecting the level of quality of service (Jamasb et al., 2005). The independent and dependent variables, related to quality of service, were numerical because the data included indices and measurable values, which required a quantitative research method. Based on the empirical works of Neuberg (1977), Ai and Sappington (1998), Clements (2001), and Ter-Martirosyan (2003), an improved econometric model emerged to capture the relationship between the quality of service and the independent variables of capital, labor, and demand. Neuberg (1977) elaborated a heuristic argument supporting the use of network variables as structural variablesthe length of networks, capacity of transformer, and level of consumption per consumerto relate to the production efficiency of an electric distribution utility. Later, Ai and Sappington (1998) and Clements (2001) produced an econometric analysis that related telecommunications infrastructure and labor to the quality-of-service index. Ter-Martirosyan (2003) improved the econometric assessment for application to the electric industry in the United States of America testing the relationship between the variables of capital, labor, and regulatory regimes and the level of quality of service. The aim of this study was to determine how the political and institutional variables relate to the level of quality of service. Modification of the econometric model elaborated on by the above-mentioned researchers occurred to capture the interrelation between the regulatory regimes, the monitoring mechanisms, and the ownership forms and the quality of service in the Latin American setting. The study was longitudinal and incorporated panel data from a 6-year period (2002-2007). Use of STATA software supported the analyses. Data collection occurred through a field survey of primary and secondary sources. Data sources included regulatory bodies, nongovernmental organization databases, and the utilities of Latin American countries (Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela). 1.5 Research Questions The aim of the research was to determine how the schemes of regulation, monitoring, and ownership influence the quality of service supplied by the electric distribution utilities of Latin America. The following research questions guided the study: 4

R1: What is the influence of regulation schemes on the quality of service provided by the electric distribution utilities in Latin America? R2: What is the influence of monitoring mechanisms on the quality of service provided by the electric distribution utilities in Latin America? R3: What is the influence of ownership forms on the quality of service provided by the electric distribution utilities in Latin America? R4: Do electric distribution utilities in Latin America under the scheme of rate-ofreturn regulation offer a better quality of service than do private utilities regulated under the price-cap or model-firm scheme? 1.6 Hypotheses In order to answer the first question of the research, the following hypotheses will be contrasted: H1: The RORR regime provides a better quality of service than the price-cap regime. H2: The RORR regime provides a better quality of service than the model-firm regime. Answering the second research question required contrasting the following hypotheses: H3: In the case of state-owned utilities, adopting a monitoring regime with sanction positively affects the quality of service compared to the quality of service provided by the utilities not affected by sanctions. H4: Private utilities respond better than do state-owned utilities in terms of quality of service when the regime of monitoring involves sanction. Answering the third research question involved testing the following hypothesis: H5: The private utility does not provide a better quality of service than does the state-owned utility. Answering the fourth research question included contrasting the following hypotheses: H6: The utilities under rate-of-return regulation provide a better quality of service than do the private utilities under price-cap regulation.

H7: The utilities under rate-of-return regulation provide a better quality of service than do the private utilities under model-firm regulation. 1.7 Theoretical Framework The study occurred within the field of economic regulation, from which emerged theories supporting private firms operating natural monopoly markets under the supervision of a regulator. The electricity sector reform reflects multidimensional factors that interact. The exogenous factors considered include politics, economics, social factors, and institutional endowment. The endogenous factors include regulatory policies and market structure. Together, as illustrated in Figure 1, the factors influence the outcomes of the reform: (a) prices, (b) quality of service, (c) access to service, and (d) profits (Berg, 2001).

Politics, Economics, and Social Context

Regulatory Body Autonomy Price Regulation Quality Regulation

Government Policy

Ownership Private Sectorial Law Transfer Stability Laws

Outcome Tariff Quality of Service Access Profits

Electric Utility Infrastructure

Public Corporative Governance

Operation & Maintenance Market

Figure 1. Multidimensional factors of reform sector.

Therefore, the behavior of an electric distribution utility in terms of pricing, cost cutting, quality of service, and network expansion follows mainly the rules set in the regulation policy and the related market structure (Berg, 2005). Different schemes or regimes governed the setting of prices. The early rate-of-return regulation (cost-plus regulation) was the traditional scheme with the prices reflecting all the costs incurred by the firms under certain conditions in the rate base (Averch & Johnson, 1962; Gmez & Rothwell, 2003). This signal resulted in overinvestment and excess of the quality of service (Baldwin & Cave, 1999). In the 1980s, the theory of regulation evolved, presenting a form of competition to the natural monopoly of electricity, gas, and water. Introduction of price-cap

regulation and yardstick competition aided in setting prices based on cost efficiency, with the price related to the maximum price for a period of 4 to 5 years that a firm could charge the consumer. This scheme implicitly guided the firms to reduce costs and assure the consumer of no tariff increment (Gmez & Rothwell, 2003; Shleifer, 1985). The theories of principal-agent and right of property indicate that because private firms focus on profitability, they perform better than do public firms (Perotti, 2004; Pollitt, 1995; Pollitt, 1997; Shleifer & Vishny, 1994). Most of the empirical studies occurred in developed countries, and developing countries reflected the same results. Few researchers cast doubt on these findings. Pollitt (1995) found that private and public firms in the United Kingdom perform similarly, but in the case of Latin America, Fisher et al. (2003) discovered that public Chilean firms perform better than do private firms. Thus, the outcomes are not conclusive. None of the researchers addressed quality of service, which nowadays is an important factor that affects the sustainability of privatization. Only recently have researchers identified the decline in the quality of service as an important problem that requires a solution because such a problem can significantly affect the outcomes of efficiency of private firms (Millan, 2006; Nellis, 2004; Shirley, 2004). Spence (1975) and Sheshinski (1976) developed an economic model that illustrated that under the price-cap scheme, the firms that cut costs adversely affected the level of quality of service because the managers concentrated more on profits. Since the end of the 1990s, researchers have examined the issue. Some studies in the telecommunications industry showed that under price-cap regulation, the quality of service of firms had decreased or remained the same (Ai & Sappington, 1998; Ai & Sappington, 2002; Clements, 2001). In 2003, an empirical study on electrical distribution utilities in the United States indicated that the price-cap regime affected the system average interruption duration index (SAIDI) and the system average interruption frequency index (SAIFI) of quality of service (Ter-Martirosyan, 2003). Berg (2006) proposed that the factor that could explain the diminishing quality of service was the regulation scheme adopted during privatization. Lewis and Sappington (1991) and Noam (1990) expressed that the mechanism of monitoring could be a factor. However, Hart et al. (1997) emphasized that the incomplete contracts between the firms and government were the cause. In light of these findings, the research involved developing an econometric model. Ai and Sappington (1998) expanded an econometric model to analyze the effect that price-cap regulation had on quality for the telecommunications industry and found no systematic link between the two. Clements (2001) discovered a correlation between regulation-by-incentives schemes and the declining quality of service. TerMartirosyan (2003), based on the preceding works, analyzed the electric industry and found that the regulation-by-incentives scheme and some kind of monitoring mechanism affect the quality of service.

Ter-Martirosyan (2003) developed an econometric model and considered the following independent variables to explain the level of quality of service: service territory, average income per capita, length of aerial network, length of underground network, ratio of generation to total energy sold, and weather. Because the cited model does not include a variable to explain the avoidance of power outages and interruptions in the supply, this study included another variable to reflect investment in switches (Brown, 2002; Gnen, 1986). Also, the previous model did not indicate a variable for the costs of operations and maintenance. With respect to the variables of control, no previous models have reflected either the type of ownership or the mechanism of monitoring. This research included both these variables considering the corresponding categories to promote a more precise explanation of their influence on the quality of service. The study included the following factors as dependent variables: SAIFI, which measures the degree of continuity of the electric service, and SAIDI, which measures the time firms take to reconnect the electricity after an interruption. The independent variables included the following: total length of aerial network, total length of underground network, number of switches, costs of operations and maintenance, and annual sales of energy per firm. Neuberg (1977); Salvanes and Tjtta (1998); Ter-Martirosyan (2003); Estache and Rossi (2004); Andres et al. (2006); and Farsi, Fillipini used the first three independent variables extensively. However, according to Brown (2002) and Gnen (1986), following a theory for the calculation of the reliability of the electric distribution systems, the number of switches represents more accurately the outcomes of the quality of service due to its link with the distributors capacity of maneuver to avoid interruptions or reduce the time of recovery of the electric service. Andres et al. (2006) believed that the variables of quality of service, including those indicated previously, relate to variables of productivity (number of employees) and variables of efficiency (loss of energy). The variables reflected the administrative management of the firm and formed part of the model. The model did not include the variable related to the weather, proposed by Ter-Martirosyan (2003), because in Latin America, the temperature variations in the main cities are not extreme. The independent dummy variables included, first, the type of regulation scheme, evaluated according to the rate-of-return regulation, price-cap regulation, and model-firm regulation. The second dummy variable was the mechanism of monitoring, evaluated according to its grade of enforcement as high, medium, or low. The last dummy variable was the type of ownership, private or public. According to Centeno and Serra (2007), in Latin America, evaluation of the private dummy variable should occur according to the origin of the investor, subclassified as European, American, or regional/national. In addition, subclassifications of the state-owned dummy variable included controlled state-owned firm, state-owned influenced firm, and state-owned independent firm.

1.8 Definition of Terms Electric distribution utility: An electric distribution utility is a firm that commercializes electricity in concession (Jamasb & Pollitt, 2007). Model-firm regulation: Model-firm regulation is a type of regulation by incentives that introduced competence through a model firm based on the principle of the economic adapted system and recognition of capital using the principle of the new replacement value (Arnau, Mocarquer, Rudnick, & Voscoboinik, 2007). Price-cap regulation: Price-cap regulation is a type of regulation by incentives. The objective is to improve economical efficiency through a maximum price set for a period. The RPI-X mechanism allows firms to profit from the gap between the price cap and the realized cost (Gmez & Rothwell, 2003). Quality-of-service monitoring: Quality-of-service monitoring involves a regulator controlling the level of quality of service provided by the electric distribution utility (Holt, 2004). Rate-of-return regulation: Rate-of-return regulation is a scheme of regulation that includes recognition of the physical assets of the utility, also called cost-plus regulation (Gmez & Rothwell, 2003). State-owned enterprise: A state-owned enterprise is a firm owned and controlled by the state (Hinds, Sanchez, & Schap, 1991). 1.9 Assumptions The first assumption was that the SAIDI and SAIFI (indices of quality of service) applied to the electric distribution utilities are standardized international indices. The second assumption was that the variables of the studied framework have a degree of applicability and generalizability to the electrical sector in developing countries. The final assumption was that the distribution utilities under study with the same scheme of regulation would respond similarly in the Latin American context. 1.10 Limitations The research reflected the following limitations: 1. 2. The study was limited to the availability of information for the selected period (2002-2007). The validity of the study was limited to the homogeneity of the studied population.

3.

The study was limited to regions with characteristics similar to Latin American countries.

1.11 Delimitations The study involved the collection of information on the electric distribution utilities in Latin America. Mexicos regulation regime is not comparable with the regulatory regimes of the other Latin American countries, so Mexico did not form part of the study. The focus of the study was to examine the effect of regulation, monitoring, and ownership on the level of quality of service provided by electric distribution utilities with more than 100 000 consumers for the 2002 to 2007 period. 1.12 Summary The decline in the quality of service of electric distribution utilities in Latin America was a topic requiring investigation because of its effect on the sustainability of privatization (Millan, 2006; Nellis, 2004; Shirley, 2004). Quality of service could be a hidden cost of privatization because the regulation policies do not include the quality-of-service issue (Costas, 2006). A factor related to the regulatory regimes that provide economic incentives to the firms could explain the reduction in quality (Berg, 2006; Kidokoro, 2002; Sheshinski, 1976; Spence, 1975). In addition, the monitoring mechanism of the regulator to supervise the utilities could be a factor in the level of quality (Ajodhia & Hakvoort, 2005; Kriehn, 2005; Lewis & Sappington, 1991; Noam, 1990). Furthermore, the ownership of the utility could be a factor due to the existence of incomplete contracts where the government has not explicitly set the requirement for the quality of service, giving the private firms an incentive to reduce costs, which affects the level of quality of service (Hart et al., 1997). The study involved conducting an empirical analysis of the influence of the factors of regulation, monitoring, and ownership on quality of service in Latin American electric utilities. The results helped to close the gap in the literature, especially in terms of the Latin American context, making the study significant. Chapter 2 includes detailed discussion of the characteristics of the electric distribution system and the quality of service. The chapter further illustrates information on the regulatory regimes, monitoring mechanisms, and ownership forms with respect to their influence on the quality of service.

10

2.

Literature Review

The literature indicates that regulation schemes, quality-of-service monitoring, and ownership influence the level of quality of service provided by the electric distribution utilities (Baron, 1981; Berg, 2006; Clements, 2001; Costas, 2006; Hart et al., 1997; Kidokoro, 2002; Noam, 1990; Sheshinski, 1976; Spence, 1975; TerMartirosyan, 2003). Berg (2001) developed a conceptual framework to outline key links among public policy, basic industry conditions, and sector performance. The regulatory policies establish the rules (price setting and quality standards) that restrain the behavior of the firm, which influences the performance of the sector in terms of pricing, cost cutting, providing the service, and expanding the network. The regulation-by-incentives approach mostly adopted in sector reforms does not explicitly reflect consideration of allowances for quality of service. The electric distribution utilities operate under a natural monopoly where differentiation of quality of service does not occur. This particular feature causes the elasticity of demand not to affect the price, giving the utilities strong motivation to cut costs, which deteriorates the quality of service (Baron, 1981; Kidokoro, 2002; Sheshinski, 1976; Spence, 1975). Lewis and Sappington (1991) found that the quality of service is higher when the quality of service is verifiable. When an imperfect public monitor of delivery quality is available in the natural counterpart, the welfare of both the buyer and the supplier improves with the accuracy of the monitor. In the case of the electric industry, the public monitor is the regulator. Furthermore, the incomplete contracts of concession between the government and private firms drive the private distribution utilities, in their goal to make more profit, to reduce or eliminate the investments for improving or maintaining the electric distribution infrastructure, which affects the quality of service (Hart et al., 1997). According to Costas (2006) and Berg (2006), the level of quality of service has been diminishing progressively, negatively affecting consumers acceptance of the privatization, putting at risk the legitimacy and credibility of the reform. The purpose of the literature review is to understand the electric distribution utility industry, to identify the regulation schemes and their influence on the quality of service, to explain the types of quality-of-service monitoring, and to examine the influence of ownership. 2.1 Electric Distribution Industry This section reflects the main characteristics of the electric distribution industry and their relation to quality of service. 2.1.1 Electric Distribution Activity Electricity is essential for modern life, and electric utilities provide the service. The product has the following technical and economic features: (a) electricity cannot be 11

stored, (b) investment in electricity involves a long-run recovery, (c) the electricity network has strong externalities, (d) the industry has strong economies of scale and scope, and (e) the network takes a long time to build (Guash & Spiller, 1999; Hllriegl, 2007). The activity of generation involves the production of electricity by hydropower or thermal power plants. Transmission is the action related to the highlevel voltage transportation of energy produced by the power plant to the cities or industries. Distribution involves delivering low-voltage electricity by local networks consisting of overhead or underground lines, cables, switchgear, transformers, control systems, and meters to transfer electricity from the transmission system to customers. The supply or retailing function includes metering, billing, and selling electricity to end users (Edvardsen & Frsund, 2002; Jamasb & Pollitt, 2007). Despite the considerable technological progress of the industry, the role of distribution networks within the electricity supply industry has largely remained unchanged (Jamasb & Pollitt, 2007). Network industries include technologies through which provision of services occurs over a network of spatially distributed points with distinct demand characteristics. Demand characteristics vary according to customer groups (urban or rural), space, and quantities supplied (Neuberg, 1977; Salvanes & Tjtta, 1998). 2.1.2 Objective of the Electric Industry The goals of the electric distribution business are to satisfy the customers need for electricity, to ensure the quality of electricity supply, and to yield profits for the owners. Public policy creates incentives involving behavioral restraints. The incentives relate to price, quality-of-service requirements, and mandates for system expansion. Sector regulators use cost-plus regulation, or rate-of-return regulation, and price-cap mechanisms to constrain prices (Berg, 2005). In particular, policy makers and regulators have widely failed to implement a formal treatment of quality of service. The oversight is especially problematical because of the interactions and trade-offs between utility costs (capital as well as operating and maintenance expenditures) and quality of service (Growitsch, Jamasb, & Pollitt, 2005). The public acceptability of prices, quality of service, and access to service all affect the political and, thus, the regulatory climate. The efficiency of firms depends on the ability of investors to capture the benefits of good performance and on the firms ability to reward key decision makers (Berg, 2001). Sector performance affects another element, the credibility of the system to private and public investors (Berg, 2005). According to Hart et al. (1997), the ownership of the utility can influence the quality of service through its goals of cost efficiency and profitability. The sustainability of the regulatory process depends on public acceptance of the outcomes, which determine the legitimacy of the process in the eyes of citizens (Berg, 2001).

12

2.1.3 Natural Monopoly and Regulation According to the economic concept of subadditivity, one firm providing a local distribution of electricity is less costly than are multiple firms, so the electric distribution activity is a natural monopoly (Growitsch et al., 2005; Kessides, 2004; Salvanes & Tjtta, 1998). Three major interest groups exist in the electric distribution business: customers, utilities, and the government. The utilities, owned either privately or by the state, aim to reach the outcomes imposed by owners and the regulator. Because the captive consumer expects both a reasonable price (tariff) and quality of service, regulation is necessary. The regulator must be independent, accountable, and resistant to corruption by either the private provider or the state. The regulation includes objectives for efficiency improvements to reduce costs and, hence, tariffs. Regulation is essential to protect consumers against monopoly power abuse and assures investors of fair treatment (Honkapuro, Lasilla, Partanen, Tahvanainen. & Viljainen, 2004; Kikeri & Nellis, 2004). Regulators must ensure that tariff setting allows the utility to recover its investments with a reasonable return (Rudnick & Zolezzi, 2001). Traditional regulation required that regulators have detailed information on the technology, costs, and consumerdemand attributes facing the firms they regulate and can somehow impose costminimization obligations on regulated firms, but in reality, regulators have incomplete information (Joskow, 2006). Regulation schemes have evolved over time. The traditional rate-of-return regulation has changed to a regulation-byincentives approach, but its adoption in recent years has resulted in declining quality of service caused by profit maximization (Growitsch et al., 2005). 2.1.4 Parameters of Quality of Service In the activity of electric distribution, a common distinction is evident among three quality dimensions: reliability, voltage quality, and commercial quality. Of the three dimensions, reliability is the most important quality feature in electric distribution because the entire electric supply depends critically on the functioning of the distribution networks. The cause of over 90% of the interruptions experienced by customers is faults in medium and low voltage (Ajodhia & Hakvoort, 2005). In summary, the electric distribution industry is a complex system that involves special characteristics of electricity, which imply a natural monopoly. Regulation is required to promote efficiency. The regulator by means of different schemes sets the tariffs that allow for the utilities and investors to recover their investment and for the customers to pay a fair tariff. The regulation-by-incentives approach enables the utilities to maximize their profits, which results in a lowering of the quality of service. The problem may affect approval of the regulatory policy and could weaken the sustainability of the 13

electric industry reform. Therefore, the research study involved investigating the influence of the parameters of regulation policies (mechanisms of regulation and monitoring) and ownership on the outcomes of the quality of service. 2.2 Quality of Service in the Electric Distribution Industry This section includes a discussion of the paradigms and the dimensions of quality of service in the electric distribution industry. 2.2.1 Definition of Quality of Service Garvin (1984) maintained that five major approaches reflect definitions of quality: (a) the transcendent approach of philosophy; (b) the product-based approach of economics; (c) the user-based approach of economics, marketing, and operations management; (d) the manufacturing-based approach of operations management; and (e) the value-based approach of operations management. According to the transcendent approach, quality is synonymous with innate excellence; quality is both absolute and universally recognizable, a mark of compromising standards and high achievement. Nevertheless, proponents of the view claim that one cannot define quality precisely; rather, according to Garvin, quality is a simple, unanalyzable property that people learn to recognize only through experience. The product-based approach indicates quality as a precise and measurable variable. Differences in quality reflect differences in quantity of some ingredient or attribute of a product. The approach lends a vertical or hierarchical dimension to quality because of the possible ranking of goods according to the amount of the desired attribute they possess (Garvin, 1984). The user-based approach definition reflects the opposite premise in that quality lies in the eyes of the beholder (Garvin, 1984, p. 27). Individual consumers have different wants or needs, and the goods that best satisfy their preferences are those that they regard as having the highest quality. The approach is an idiosyncratic and personal view of quality, one that is highly subjective. The manufacturing-based approach involves a focus on the supply side of the equation and a concern with engineering and manufacturing practice. Virtually all manufacturing-based definitions illustrate quality as conformance to requirements. Excellence equates with meeting specifications and with making it right the first time (Garvin, 1984, p. 27). The final approach is the value-based approach. The approach includes a definition of quality in terms of costs and prices. According to this view, a quality product is one that provides performance at an acceptable cost.

14

Thus, no consensus exists on the definition of quality because each paradigm includes emphasis on different dimensions of quality. The implication is that the relations of (a) utility of distribution to utility of generation, (b) utility to regulator, and (c) utility to consumer may be complicated. The goal of the utility is to satisfy the consumers quality demands, regardless of whether the service meets the various technical standards. However, the technical standards will be paramount for the regulator (Clements, 2001). 2.2.2 Dimensions of the Quality of Service Distinguishing among the three quality dimensions for electric distribution is common. First, commercial quality concerns the quality of relationships between the electric distribution utility and its consumers. For example, utilities improve commercial quality through timely installations or connections, prompt responses to customer complaints, efficient billing practices, safeguarding of customer accounts, and accuracy of customer information (Holt, 2004; Pinter & ReKettye, 2005). Second, power quality, also known as voltage quality, relates to a variety of disturbances in the voltage waveform. The main parameters of voltage quality are frequency, voltage magnitude and its variation, voltage dips, temporary or transient overvoltages, and harmonic distortion. From the customers perspective, a power quality problem is any electric supply condition that causes appliances to malfunction (Brown, 2002; Lpez, 2007). Third, the quality dimension of reliability, which measures the ability of the continuity of the network, includes two main elements: adequacy and security. Adequacy relates to the availability of a sufficient network capacity to guarantee consumers a long-term supply of electricity. Security involves the avoidance of power outages and interruptions in the supply to customers (Giacchino & Lesser, 2007; Gnen, 1986). Key definitions relating to distribution reliability include the following (Brown, 2002; Gnen, 1986): 1. 2. 3. 4. A contingency is an unexpected event, such as a fault or an open circuit. A fault is an interruption caused by a short circuit. An outage occurs when a piece of network is de-energized. A momentary interruption happens when a customer is de-energized for less than a few minutes. Most momentary interruptions result from reclosing or automated switching. A momentary interruption event consists of one or more momentary interruptions within several minutes. A sustained interruption occurs when a customer is de-energized for more than a few minutes. Most sustained interruptions result from open circuits and faults. 15

5. 6.

The reliability indices reflect a statistical aggregation of reliability data for welldefined sets of loads, components, or customers. Most reliability indices are averaged values of particular reliability characteristics for an entire system, operating region, substation service territory, or feeder. The most widely used reliability indices are averages that weigh each customer equally. Customer-based indices are popular with regulating authorities because a small residential customer is as important as is a large industrial customer. The indices have limitations but are generally good aggregate measures of reliability often used as reliability benchmarks and improvement targets (Brown, 2002). Two methods exist to calculate reliability. The majority of utilities calculate indices based on the number of customers per outage and the duration of the outage. The minority calculates indices based on the load lost, but their databases are not as extensive as the indices based on the number of customers. According to Burke (1994), various groups, such as the Institute of Electrical and Electronic Engineers, have defined indices. The formulae for customer-based indices follow: System average interruption frequency index (SAIFI): SAIFI = Total number of customer interruptions Total number of customers served /year

System average interruption duration index (SAIDI): SAIDI = Customer interruption duration hr/year Total number of customers served Although establishment of numerous dimensions of quality of service is possible, outage-related indices are the only relatively widely accepted measures of quality across the electric utilities. The principal reliability indices are average duration (SAIDI) and average frequency (SAIFI) of electric outages (Meyrick & Associates, 2002; Ter-Martirosyan, 2003). The customer average interruption duration index (CAIDI) is a measure of the length of an average interruption and a measure of utility response time to contingencies. CAIDI did not form part of this research because distorting the CAIDI is possible by increasing the number of short interruptions (Brown, 2002). To summarize, the quality of service of electric utilities is a broad concept that includes different dimensions. However, the literature on mechanisms of effective and quantitative measurements exhibits a focus on indices of the average number of interruptions and the average duration of interruptions. For the purpose of this research, the major dimensions of quality of service of the electric distribution system were the SAIDI and SAIFI.

16

2.3 Regulation Schemes and Quality of Service This section includes an assessment of the main regulation schemes used for the regulation of distribution tariffs to address their economic signals and relation to the quality of service. 2.3.1 Rate-of-Return Regulation The rate-of-return regulation or cost-plus regulation is a traditional scheme of regulation. The setting of the tariff involves two steps during the rate case or regulatory review. First, the rate-level determination includes (a) identifying allowed costs and investments and (b) setting an allowed rate of return so that the utility will receive the appropriate level of earnings on its investment (Gmez & Rothwell, 2003; Parker & Kirkpatrick, 2005). Second, the rate-structure determination involves setting tariffs for different customer classes and products, which permits the utility to recover the revenues required to earn its allowed rate of return (Gmez & Rothwell, 2003). During the regulatory review, tariffs are set based on a test period (generally, the previous accountancy period) and remain in effect until the next review. In practice, the utility or the regulator can initiate regulatory reviews (Laffont & Tirole, 1993). The utility can argue that the current tariffs are too low because the costs allowance or the allowed rate of return is too low. The regulator can argue the opposite to request a rate case (Gmez & Rothwell, 2003). After the utility and the regulatory staff present detailed accounting information and negotiation has occurred among the regulator, the agency staff, and the utility, the regulator determines the appropriate level of expenses and sets the allowed rate of return. The regulator audits the firms costs carefully to reveal cost padding and unnecessary capital expenditures to avoid increase of the asset base (Hllriegl, 2007; Parker & Kirkpatrick, 2005). The regulator attempts to choose a fair and reasonable rate of return for capital. Finally, the utility adjusts the tariffs to yield the new rate of return allowed by the regulator (Laffont & Tirole, 1993). The required revenues remain fixed during the regulatory lag (period between two consecutive rate revisions), which provides an incentive for the utility to reduce costs. The utility earns higher rates of return by incurring costs lower than the costs anticipated in the rate base. If costs are higher than anticipated, the utility earns less than the allowed rate of return (Gmez & Rothwell, 2003). Another important instrument that provides incentives for regulated firms to be efficient is the determination of whether to include a particular investment in the rate base. According to Gmez and Rothwell, many regulators use original cost valuation (the amount that the company originally paid for its plant and equipment less depreciation) and focus on the selection of the allowed rate of return.

17

An asymmetric information problem is evident throughout regulatory reviews. The problem results from an asymmetry of information between a regulator who wants work done and the electric utility that must do the work (Gmez & Rothwell, 2003). Another problem occurs when the allowed rate of return exceeds the actual cost of capital; the firm has a strong incentive to overinvest and inflate the capital stock, known as the Averch-Johnson effect (Averch & Johnson, 1962). Overcapitalizing relates to an oversupply of quality because quality is typically a capital-using attribute (Spence, 1975). Thus, both prices and quality levels will be too high. Empirical studies show that under rate-of-return regulation, existing reliability levels in the electric industry are generally higher than optimal from a social point of view (Ajodhia & Hakvoort, 2005). This situation includes additional costs and, thus, a higher price: Consumers will be paying too high a price for too high a quality level, so-called gold plating (Lpez, 2007). In addition, the quality of output may rise if rate-of-return regulation encourages capital intensity, and if capital is normally required to increase service quality, the result may be excessive quality (Baldwin & Cave, 1999; Sappington, 2005). If quality is capital intensive, quality levels will automatically tend to be high, and less need for explicit quality regulation will exist. In this scheme, the main responsibility of providing a good quality of service remains with the utility instead of with the regulator (Kahn, 1988). According to Kahn, the government supervisor intervenes only where he or she can set objective standards or after an event when the monopolistic performance has been obviously bad. Under rate-of-return regulation, regulators can indirectly escape the quality-regulation problem, but the escape comes at a cost of lower efficiency. Spence (1975) emphasized that under regimes of monopoly, supply of product characteristics does not usually occur under the pressure of the market, so regulation exhibits difficulties when price and quality are decision variables. The difficulties are informational, and without the necessary information, rate-of-return regulation may be attractive as second best. Consequently, rate-of-return regulation may be a substitute for quality regulation. Following the discussions of Kahn (1988) and Spence (1975), one might assume that under rate-of-return regulation, regulators do not directly regulate the quality of service; the utility sets the quality of service. Thus, rate-of-return regulation leads to higher levels of quality of service because the rate base includes the investments and allowances for maintaining and operating the electric installations. In addition, the regulator has no explicit need to deal with quality because the responsibility for a good quality of service lies with the utilities that must comply or face government intervention. Finally, under rate-of-return regulation, the utilities may overcapitalize their rate base to obtain both higher rates of return and higher levels of quality of service.

18

2.3.2 Regulation by Incentives Regulation by incentives is a form of utility regulation that strengthens the financial incentives to lower rates, lower costs, or improve nonprice performance compared with traditional rate-of-return regulation. The design and application of a regulationby-incentives plan include a collection of interrelated tasks: (a) set a baseline revenue requirement, (b) set the adjustment factors, and (c) design the control mechanism to meet specific objectives (Gmez & Rothwell, 2003). The most typical forms of incentive regulation with a long regulatory lag applied for setting tariffs in a power distribution utility are (a) price cap, (b) revenue cap, (c) yardstick competition, and (d) a hybrid system, a combination of price cap, revenue cap, and yardstick competition (Agrell, Bogetoft, & Tind, 2005; Gmez & Rothwell, 2003; Jamasb & Pollitt, 2007). Under price-cap regulation (price cap or revenue cap), maximum (but not required) prices for utility services are set for several years regardless of the utilitys own costs. Adjustment in subsequent years of maximum prices allowed during the first year occurs according to a prespecified set of economic indices and factors. For example, economic indices and factors could include (a) maximum price or tariff, in the case of price cap; (b) authorized utility revenues in a year, in the case of revenue cap; (c) annual change in prices (the inflation index); (d) productivity offset; and (e) the adjustment factor for unseen events (Gmez & Rothwell, 2003). In addition, price-cap regulation provides the firm with an incentive for cost efficiency. The formula has the following structure:

(PRICE)YEAR1 = (PRICE)YEAR0 {1+ (i X )} Z ,

where i represents the inflation rate; X is the productivity factor, which is revised each certain period; and the unforeseen events factor Z can include, for example, increased taxes, changes in enviromental law, natural disasters, or resstructing costs. Regulators do not recognize the accountancy costs in the rate base but rather concentrate simply on setting the initial price cap and on the revision of the productivity factor. The formal regulatory review must take place after a reasonable period, every 4 or 5 years. This policy induces the utility to pursue savings in costs and to avoid public hostility toward the regulatory regime in view of the utility making large profits with prices well above costs of supply (Kriehn, 2005). After the period, not only will adjustment of prices occur, but also the consumers will benefit from the efficiency gain (Cowan, 2006; Hllriegl, 2007). Moreover, under price-cap regulation, the regulator provides the utility with only high-level incentives that promote some regulatory objective without interfering in the details of running the firm (Ajodhia & Hakvoort, 2005). The central idea of price-cap regulation is to control the prices charged by the regulated utility rather than to control its earnings (Uri, 2001). As a result, price-cap 19

regulation gives firms an incentive to cut costs, which raises the concern that firms may achieve part of the cost reductions by decreasing quality (Lpez, 2007). Germinal papers by Spence (1975) and Sheshinski (1976) theoretically show that price caps led firms to reduce quality to cut costs and increase profits. The difficulty of effective quality regulation becomes apparent if one moves from rate-of-return regulation toward stricter price-cap regulation. Under rate-of-return regulation, no explicit need for quality regulation exists because quality is more or less safeguarded. However, under price-cap regulation, such safeguarding is not evident (Ajodhia & Hakvoort, 2005). According to Fraser (1994), two situations exist in the relationship between pricecap regulation and the reliability of service provided by a private monopoly. Excluding reliability, the firm tends to protect profits by lowering the quality of service; however, including quality-of-service costs eliminates the tendency. The exclusion of reliability of service from the price-cap formula means that although consumers experience protection from the cost increase, the protection occurs at the expense of quality of service. Empirical studies occurred mostly in the telecommunications industry. Ai and Sappington (1998) and Roycroft and Garcia-Murrilo (2000) found that although regulation by incentives affects the quality of service, the results are not conclusive. Nonetheless, Clements (2001) claimed that a lower quality of service relates to a monopoly environment under price-cap regulation. Weisman (2002) reported that Oregon and Idaho in the United States abandoned the regulation-by-incentives approach due to a reduction in the quality of service (as cited in Lpez, 2007). Ai and Sappington (2005) stated the following: The implications for service quality are not clear. Because higher levels of quality of service can increase revenues, incentive quality regulation can enhance incentives to provide high-quality service to customers. On the other hand, because lower levels of quality can reduce operating costs, incentive regulation can diminish incentives to provide high-quality service. Thus, the overall effect of incentive regulation on service quality is ambiguous as a theoretical matter. (p. 202) Kidokoro (2002) expressed that the use of price-cap regulation as a regulatory method poses a difficulty because even though a regulator gives the economic signals to make the utility lower its prices by imposing an upper limit on the price, the regulator cannot provide the utilities with incentives to improve quality of service. According to Mikkers and Shestalova (2003), the cost-reducing incentives are especially strong in the short term but may have an adverse effect on investment in long-term objectives. In particular, a firm can delay an upgrade or the installation of new capacity, which may not affect todays performance but may result in the deterioration of performance in the future, influencing the quality of service. Furthermore, Burns (2003) argued that the direction for electric distribution utilities 20

under price caps is to cut costs, especially operations and maintenance costs. The reductions could result in poorer reliability. An empirical study within the electric sector by Ter-Martirosyan (2003) illustrated that under regulation by incentives and in the absence of explicit regulation for quality of service, quality of service tends to decline. Ter-Martirosyan found that price-cap regulation led to worse quality performance in terms of an increase in the SAIFI. Ter-Martirosyan argued that price-cap regulation affects the cost structure of the firm; the impact on equipment (capital) is a long-term effect although the related changes in reliability may not be evident in the short term. Furthermore, TerMartirosyan emphasized that price-cap regulation affects the utilitys expenditure on operations and maintenance. Such cost reductions negatively affect the reliability of the service. Under yardstick competition, the regulator sets a price cap for a firm based on the average cost of the other companies in the sector and allows the firm to keep the difference between the cap and the realized cost (Shleifer, 1985). Tangers (2002) claimed that when regulating quantity, yardstick regulation results in lower quality than does individual regulation although, under the latter, the quality would be too high for some. In principle, the argument also holds for both price-cap and revenuecap regulation models. Mikkers and Shestalova (2003) indicated that yardstick regulation involves unlinked prices from companies costs, providing firms with strong incentives to reduce cost and improve efficiency but resulting in the deterioration of quality of service. Thus, price-cap, revenue-cap, and yardstick regulation schemes within the regulation-by-incentives approach together result in firms reducing the quality of service. The main hybrid mechanism for regulation in Latin American countries is the denominated model firm. Establishing the revenue of the utilities occurs based on the optimization of a model firm, against which all distribution concessionaire firms compete (Pollitt, 2005). Generally, the value-added distribution cost includes recognition of the remuneration of the electric assets and the expenses associated with (a) the network operations and maintenance, (b) capital of efficient electric installations, and (c) customer attention; the latter two expenses depend on the concession contract (Arnau et al., 2007; Giachino & Lesser, 2007). The regulator sets prices for distribution (the value-added distribution) relative to an inflation rate for the next 4 years (Pollitt, 2005). Rainieri and Rudnick (1997) emphasized the following: The model firm mechanism has proven effective in reducing costs and in conveying this greater efficiency to consumers of what is a simple mechanism where total distribution costs are reimbursed. If the tariffs set by the regulator fail to recognize properly the quality offered, we may have that the firm, when facing an incentive to reduce costs, has in hand a perverse incentive to reduce the quality offered, because it will try to protect its profits. (p. 287) 21

The hybrid mechanism mimics the later suggestions of price-cap regulation (Littlechild, 2003) and yardstick competition based on average costs in other similar firms (Shleifer, 1985). Tangers (2002) argued that no matter the type of regulationby-incentives approach used for price regulation, the utility always lowers the quality of service. Utilities only respond to explicit quality-of-service incentives. Based on the technical report of the Regional Energy Integration Commission (Comisin de Integracin Energtica Regional [CIER], 2007), Table 1 illustrates the schemes of regulation adopted by Latin American countries.
Table 1 Schemes of Regulation in Latin American Utilities Type of regulation Rate of return Price cap Revenue cap yardstick (model firm) Country Costa Rica, Ecuador, Honduras, Paraguay, & Venezuela Argentina, Bolivia, Brazil, Colombia, El Salvador, Nicaragua, & Uruguay Chile, Guatemala, Panama, & Peru No. electric distribution utilities 43 175 68 Period of review Anytime 4-5 years 4-5 years

In conclusion, regulation by incentives provides strong signals to the electric distribution utilities to improve their economic performance through cost reductions, which result in a decline in the quality of service. The specialists have not conducted any research in developing countries to test empirically the effect that different schemes of regulation have on the quality of service. In contrast, this study involved assessing the influence that the schemes of regulation have on the level of quality of service in Latin American countries. The outcomes of the study may promote clarity in the debate on the privatization of natural monopoly public services and may help policy makers and regulators to deal with the quality-ofservice problem appropriately. Solving the problem may help to consolidate the reform (privatization) in Latin America and encourage credibility and sustainability. 2.4 Quality-of-Service Monitoring Spence (1975) and Sheshinski (1976) believed that the monopoly firm under pricecap regulation has the incentive to lower costs, which adversely affects the quality of service, and Lewis and Sappington (1991) found that the quality of service is higher when the quality of service is verifiable. Kriehn (2005) stated that under price-cap regulation, regulators in developing countries would need to introduce service-quality measures and monitor standards. Urbiztondo (2000) explained that historically a difference in quality of service exists between developed and developing countries because the former use rate-of-return regulation (or cost-plus regulation), which induces the utility to provide a high quality of service because 22

the tariff permanently includes the investment and maintenance costs. Furthermore, Waddams Price, Brigham, and Fitzgerald (2002) indicated that one characteristic of network industries is the provision of a common level of quality across at least some groups of consumers (e.g., those supplied from the same section of a distribution system). In such cases, regulators have to determine for which services they should set standards and identify appropriate target levels and penalties. Lewis and Sappington (1991) and Noam (1990) emphasized that in an environment of price-cap or incentive regulation, linking quality performance to financial rewards is necessary. Otherwise, pressure for quality shortcuts exists. According to Baker and Trmolet (2000), in developing countries, the quality-of-service standards tend to be high for the following reasons: 1. 2. 3. Providers have often inherited operating structures and tariffs from largescale operations and are not used to considering low-cost options. Providers base investment designs on the standards of developed countries. Large private utility providers tend to focus on high-margin customers and frequently have no financial incentive to develop low-cost provision.

Joskow (2006) pointed out that quality-related incentives to cost control must include (a) two indices for distribution service interruption targeting the number of outages and the number of minutes per outage, (b) interruption payment obligations, (c) quality of telephonic attention to consumers, and (d) discretionary award based on surveys of customer satisfaction. Ajodhia and Hakvoort (2005) described the various regulatory instruments: Regulatory attention to network reliability has increased during the last few years. Different types of regulatory instruments can generally be classified as: a) indirect instruments, b) standards, and c) incentive schemes. Indirect instruments promote good quality of service by strengthening the information and negotiation position of customers. There are different methods to achieve this. The regulator may require the firm to publish information about its performance or it may publish comparative overviews itself. Exposure to public criticism can motivate a utility to consider consumers preferences for quality. Standards dictate a minimum performance level for a utility. Violation of the standard leads to a fine or tariff rebate. In general, two types of standards exist, overall and individual. Overall standards relate to network quality at the system level. Individual standards prescribe a minimum level of performance to be delivered to individual customers. Quality incentive schemes can be considered as an extension of a standard. Here, price and quality are closely related: The firms performance is compared to some quality target, and deviations result in a price adjustment that can be either a penalty or a reward. (pp. 214-215) Waddams Price et al. (2002) explained the following: 23

The economic regulator could incorporate quality concerns directly in the price cap, making the price or revenue allowed directly dependent on quality delivered. If quality increased, a company would be allowed to raise its price, recouping some of the costs directly from the consumers who had benefited, and a degradation of quality would similarly be reflected in lower allowed prices. (p. 6) Alternatively, the regulator could impose a particular level of quality (standard). Ideally, any movement away from the optimum level should result in the utility confronting costs that are equal to the value of the total consumer losses. However, the standards have proved a powerful incentive for the utilities concerned, which have responded by improving their performance in almost every measure (Waddams Price et al., 2002). The multidimensional nature of quality attributes complicates the measurement of quality of service, which means that, depending on their specific needs and expectations, customers perceive certain attributes to be of greater value than other attributes (Holt, 2004). Because quality is more important than quantity, Holt recommended starting with fewer indicators and adding to them as reliable data become available. Holt defined technical and commercial service standards: Minimum service standards of quality-of-service apply to technical, commercial and commodity standards. Utility regulators generally have more direct oversight with respect to technical and commercial standards. Technical standards apply to reliability issues, such as the number and duration of service interruptions. Commercial standards apply to the direct transactions between the utility and the end user. Such standards are expressed in terms of measures and represent the minimum performance level that regulators expect from the utilities. (p. 194) Regulators may apply rewards and penalties to each measure used to assess a utilitys service quality or to measures aggregated into an index of overall service quality (Holt, 2004). Whether the incentives are in the form of a lower allowed reward or a penalty makes no difference. Nevertheless, an attractive political and distributive argument exists that consumers who have suffered poor service should receive some compensation (Waddams Price et al., 2002). Rainieri and Rudnick (1997) pointed out the following: The distribution firms exposure to fines for offering an unsatisfactory service quality does not solve the issue of separating the incentive scheme with respect to quantity and quality. When the set tariffs underestimate the cost of quality, we will have that the substantial fines will only make the distribution firm offer a quality that is the minimum possible demanded. For instance, and thinking about the setting of a permitted voltage level

24

fluctuation band, we will have that to protect its profits, the firm will tend to stay close to the band floor. (p. 287) Lewis and Sappington (1991) noted that if both the buyer and supplier gain when an accurate public monitor of quality is available, they are likely to agree on institutional structures that facilitate third-party verification, such as testing and onsite inspections by independent parties. According to Laffont and Tirole (1993) and Sappington (2005), quality is verifiable when a third party (such as an enforcement agency) observes and, if necessary, documents the realized level of service quality. When quality is verifiable, basing financial rewards and penalties explicitly on realized service quality is possible. In contrast, enforcing such policies when realized levels of service quality are not verifiable will be difficult, if not impossible. Lewis and Sappington (1992) emphasized that the optimal regulatory policy depends critically on the regulators ability to monitor the firms activities. The regulator must measure the level of service accurately to enhance the quality of service provided by the utility. Moreover, the optimal regulatory policy will be sensitive to the information available to the regulator. Sappington (2005) argued that no simple policy solutions for effective monitoring of quality of service exist, but the solutions depend on the information available to the regulator, institutional settings, and consumer preferences (as cited in Jamasb & Pollitt, 2007). Taking into account the technical report of the CIER (2007), Table 2 illustrates the schemes for monitoring quality of service (with and without sanction) adopted by Latin American countries. With-sanction monitoring concerns the setting of a quality-of-service standard and its relation to penalties. Without-sanction monitoring involves the setting of a minimum quality-of-service standard with no penalties but includes measurement and publication of the results.
Table 2 Monitoring Mechanisms in Latin American Utilities Type of monitoring Monitoring (with sanction) Monitoring (without sanction) Country Argentina, Bolivia, Brazil, Colombia, Chile, Guatemala, Panama, & Peru Costa Rica, Ecuador, Honduras, Nicaragua, Paraguay, Uruguay, & Venezuela No. electric distribution utilities 237 49

To summarize, the electric distribution utilities in developed countries traditionally provide a good quality of service because of the extensive use of the rate-of-return regulation (or cost-plus regulation). However, in Latin America, as a product of the reform in the electric sector, the countries have mainly adopted the regulation-byincentives scheme for setting prices, which calls for the intervention of a regulator to monitor the quality of service. Regulators have had to adopt different schemes of 25

monitoring: with sanction and without sanction. Understanding how the mechanisms of monitoring influence the level of quality of service provided by the electric distribution utilities in Latin America is important because such information will help the regulator and policy makers to improve the regulation policy. The regulatory experience and the endowments of developing countries are different from those of developed countries. In this sense, the empirical results of the study may enlighten the economic regulation field because to date no empirical studies exist on this issue in Latin America. 2.5 Ownership and Quality of Service Growitsch et al. (2005) pointed out that since the 1990s, liberal models based on competition, economic incentives, and private ownership have been popular to achieve internal and external efficiency improvements in the public service sectors (telecommunications, electricity, gas, and water) to benefit the consumers in the form of lower costs and higher quality of service. Nevertheless, the reformed industry has widely neglected the quality-of-service factor. Yarrow (1986) indicated the following: Privatization is likely (but not inevitably) to lead managers to place greater emphasis on profit goals. Whether or not this in turn leads to an increase in economic efficiency depends upon a trade-off between market failures and deficiencies in governmental monitoring and control of state-owned firms. In particular, it depends heavily upon both the degree of competition in product markets and the firms regulatory environment. (p. 334) In addition, Atkinson and Halvorsen (1986) observed the following: The theoretical model is based on the hypothesis of utility-maximizing behavior by the managers of firms. Although utility-maximizing models of firm behavior have been available for many years, and the concept of utility-maximizing behavior underlies the property-rights approach to analyzing relative efficiency, a formal analysis of the joint implications of ownership type and regulation does not appear to have been undertaken previously. (p. 281) Pollitt (1995), in a postreform study for the U.K. electric distribution industry, concluded the following: On the balance of theoretical evidence would seem to be that there is probably little difference in the incentives created by different regulatory regimes, designed to reduce the exploitation of monopoly power, and that the magnitude of the distortion which they introduce with respect to the incentive to minimize costs is likely to be small and declining over time, as regulation improves. Even if the general theoretical conclusion seems to be that private regulated monopolies tend to overinvest relative to the 26

optimum, while state-owned firms tend to have too much labour relative to the optimum, this still leaves it to empirical testing to determine the relative efficiency of both private and municipal firms. (p. 21) In addition, Bagdadioglu, Waddams, and Weyman-Jones (1996) found that private electric distribution utilities show better technical- and scale-efficiency scores. However, the result does not necessarily imply the success of private ownership in electric distribution. Technical and scale-efficient publicly operated distribution organizations are evident. According to Purdy (1997), the link between ownership and performance remains largely unproven. Pollitt (2000) and Renzetti and Dupont (2003) emphasized that different theoretical arguments exist on why private ownership and market-oriented reforms might lead to greater efficiency. The main theories that sustain this declaration are (a) principal-agent theories, (b) property-rights theories, and (c) regulation theory. On balance, the theories indicate that reform will lead to improved economic efficiency. Shirley and Guash (2000) summed up the ownership or reform debate based on a review of some 50 empirical studies covering a variety of countries and sectors (as cited in Kikeri & Nellis, 2004). The summary reflects greater ambiguity about ownership in the theoretical literature than in the empirical literature. The clear majority of empirical studies illustrated that privatized, and private, firms perform better than do state enterprises, a finding that is robust across sectors and market structures and across developed and developing countries. Megginson and Netter (2000) suggested, in their assessment of empirical studies on the privatization experience worldwide, that strong evidence exists that privatization improves operating performance. Dewenter and Malatesta (2001) also claimed that government-owned firms were less efficient or, at least, less profitable than were privately owned firms. Concerning the effect of ownership, Macedo (2004) claimed that no evidence exists that private firms are more efficient than are state-owned enterprises. A positive, significant relationship between private firms and costs is evident. The result is in accordance with Pollitts 1995 empirical study, which does not reflect evidence of a higher efficiency of private firms. Finally, the assessment by Estache and Rossi (2004) is worthy of mention. The authors argued that, overall, the empirical evidence does not provide definitive conclusions about the effects of the change of ownership and the regulation-by-incentives scheme on the efficiency of electric distribution firms in Latin America because some important variables, such as the quality of service, are missing in the analysis. In another vein of economic theory, Holmstrom and Milgrom (1991), through a comprehensive theoretical contract framework, showed that when an agent receives strong incentives to pursue one objective, such as profits, he or she might neglect other objectives, such as quality of service. In addition, Hart et al. (1997) 27

constructed an argument, based on the contract theory, that state-owned enterprise in a narrow range of circumstances may be superior, especially when full specification of the quality of service is not available. Critics of privatization often argue that private contractors would cut quality in the process of cutting costs because contracts do not adequately guard against this possibility. Shleifer (1998) observed that the narrow set of circumstances in which government ownership is likely to be superior includes the following: (a) opportunities for cost reductions that lead to noncontractible deterioration of quality are significant; (b) innovation is relatively unimportant; (c) competition is weak, and consumer choice is ineffective; and (d) reputational mechanisms are weak. Perotti (2004) insisted that ownership makes a difference to incentives and, thus, actions because contracts are incomplete. In this context, ownership completes contracting because it assigns to the owner the set of residual control rights over noncontracted or unregulated contingencies. Therefore, state ownership is justified when explicit regulation is difficult to implement because of no verifiable contingencies. Kwoka (2005) asserted that studies of the performance effects of public versus private ownership, realized for the electric distribution utilities in North America, showed mixed evidence. Kwoka suggested that public enterprise (municipalities) might have an advantage in producing goods and services with quality attributes that are difficult to specify a priori. Kwoka elaborated on the debate: Evidence concerning quality of service is imperfect, but there are some relevant data on the most common measure of quality, namely, reliability of distribution service. The values of System Average Interruption Duration Index (SAIDI) showed that state owned utilities perform better than private utilities when the size of the state owned utility is small. (p. 636) Kwoka continued as follows: These results provide support for newer theories of public ownership, which identify possible advantages over private ownership in the provision of certain services. From a policy perspective, it cautions that the quest for superior performance is not simply a matter of prescribing privatization. There are identifiable circumstances in which public enterprise is an appropriate, if not perfect, policy prescription. Research and policy require a more sophisticated view of the effect of ownership on enterprise performance. (p. 639) Agrell et al. (2005) contended that irrespective of ownership (investor-owned or publicly owned utilities), any natural monopoly poses a risk to society by accruing excess profits and costs at the expense of the consumers. The problem is the principal-agent theory under asymmetric information, with society (the customers represented by a regulator) as the principal and the utility (and its manager) as the 28

agent. According to Baldwin and Cave (1999), in the case of state-owned utilities, public managers had substantial discretion over expenditure and had little personal interest in good financial results. Managers were, thus, tempted to gratify their own preferences, which often ran to substantial expenditure on gold plating the engineering and design aspects of capital. Hinds et al. (1991) emphasized that private firms belong to private shareholders who have an incentive to incur the costs of monitoring the actions of management, according to the property theory. Concentrated ownership makes monitoring of management economical, and close monitoring induces managers to maximize the wealth of the private owners. In the case of private firms, Parker and Kirkpatrick (2005) explained that in low-income countries, an international company provides commonly privatized services. Accordingly, one could expect the nature of the ownership of assets in developing countries to complicate the operation of a price cap, because under price-cap regulation, the utility with a given price cap will be able to make extra profit by degrading the quality of service. Hinds et al. (1991) claimed that a public enterprise consists of organizations owned and controlled by the government. Centeno and Serra (2007) categorized the public enterprises of Latin America as (a) controlled, (b) influenced, and (c) independent. Controlled involves public firms making decisions based on the wishes of politicians who have an interest in sharing the benefits of the utility with their party and consumers to gain votes. Influenced includes public firms basing their decisions on the technical criteria of board members who are restricted by public rules and regulations. Independent entails firms belonging to the government, but public rules and regulations do not influence managerial decisions. Centeno and Serra (2007) further stated that the main electric distribution utilities in Latin America belong to European and American investors and a small percentage of national or regional investors, whose culture influences decision making. The basis of European investors behavior is preserving a good reputation while American investors preference is recuperating their investments in the short term. National or regional investors, who generally do not have the background in and knowledge of the distribution business, are more aggressive in cutting costs. Finally, based on the technical report of the CIER (2007), Table 3 indicates the type of ownership (private and state owned or municipal) of the electric distribution utilities in Latin American countries.

29

Table 3 Types of Ownership in Latin American Utilities Type of ownership State Private State and private Country Ecuador, Paraguay, Uruguay, & Venezuela Chile, El Salvador, Honduras, Guatemala, Nicaragua, & Panama Argentina, Brazil, Colombia, Costa Rica, & Peru, No. electric distribution utilities 35 72 179

The decision-making behavior of the electric distribution utility to promote the quality of service will develop according to the type of ownership (private or public). Hart et al. (1997) argued that public enterprises would be superior to private firms especially when full specification of the quality of service was not available. However, according to the principal-agent theory, a state-owned utility will respond to the official representatives of the government and so could expend more in providing a better quality of service, while the private utility controlled by a private manager will maximize its profits. In this sense, one could argue that private firms under the regulation-by-incentives scheme have high incentives to cut costs, which may result in a reduction in the quality of service. Thus, the ownership of the utility constitutes a relevant factor in the economic behavior of the electric distribution utilities in Latin America. To test the real implications of ownership, the study involved categorizing the public and private utilities depending on the influence of participation and control of the government in decision making and depending on the origin of the investor. Because the reform in many Latin American countries remains incomplete, Latin America was a natural laboratory in which to test the influence of ownership on the quality of service, which may aid regulators and policy makers in improving regulation policies. 2.6 Summary The quality of service of electric distribution is a wide concept that includes different dimensions. Nonetheless, in terms of the mechanisms of effective and quantitative measures, the literature reflects a focus on indices related to the number and the average duration of interruptions. Thus, SAIDI and SAIFI formed the main dimension of the quality of service of the electric distribution system in this research. The regulation-by-incentives approach provides strong signals to the electric distribution utilities to improve the economic performance through reducing costs that result in a decline in the quality of service. Spence (1975), Sheshinski (1976), and Kidokoro (2002) proposed that the price-cap regulation enabled the utility to reduce costs in areas that adversely affect the quality of service. In the United States of America, few empirical studies have emerged in the field of telecommunications 30

and electricity; the outcomes showed that price-cap regulation affects the quality of service (Ai & Sappington, 1998; Clements, 2001; Ter-Martirosyan, 2003). The researchers have not conducted studies in developing countries to test empirically the effect that different regulation schemes have on the quality of service. Two price-regulation schemes exist: (a) rate-of-return regulation, known as the traditional regulation scheme, and (b) regulation by incentives, known as the new regulation scheme or modern regulation. Under the first scheme, the utility is responsible for the quality of service, so the utility has a strong interest in providing a good quality of service because the tariff includes recognition of the investments and costs. The second regulation scheme does not indicate explicit recognition of the investments and cost of the provision of quality of service because setting the tariffs occurs under a prospective model of efficiency reflected in the price cap. The utility is not responsible for the quality of service but rather interested in increasing its benefits by reducing costs, which affects the level of quality of service. Thus, the regulation scheme is a key factor that emerges from the economic fundamentals. The electric distribution utilities in developed countries traditionally provide a good quality of service because of the extensive use of the rate-of-return regulation (or cost-plus regulation). However, in Latin America, as a product of the reform in the electric sector, the countries have mainly adopted the regulation-by-incentives scheme for setting prices, which calls for the intervention of a regulator to monitor the quality of service. Regulators have had to use different schemes of monitoring: with sanction and without sanction. Determining whether the adopted monitoring mechanism has an influence on the quality of service provided by the electric distribution utilities in Latin America was important. Because the reform in many Latin American countries remains incomplete, Latin America was a natural laboratory in which to test the influence of ownership on the quality of service. The ownership of the utility constitutes a relevant factor in the economic behavior of the electric distribution utilities in Latin America. To test the real implications of ownership, the study involved categorizing the public and private utilities. 2.7 Conclusion One of the major outputs of the electric industry is the quality of service. Because the electric distribution activity occurs under natural monopoly, regulation is required to guarantee the investors a reasonable payment for their investments and to guarantee the consumers fair prices, quality of service, and access. The quality of service is a hidden cost of privatization due to the lack of a quality-of-service policy. An important question concerns why the quality of service has deteriorated. The answer may be evident in the theoretical arguments of Spence (1975), Sheshinski (1976), and Kidokoro (2002), who claimed that the schemes of regulation influence 31

the level of quality of service of the firms. A few researchers have conducted empirical studies to assess the quality of service in the U.S. telecommunications sector, and only one study involves the electric industry (Ai & Sappington, 1998; Clements, 2001; Ter-Martirosyan, 2003). Because the institutional endowment of developed countries and their regulatory bodies reflect more experience in regulation, the behavior of the utilities in developed and developing countries could be different. Before the 1990s, the utilities in developed countries provided a good quality of service due to the use of rate-of-return regulation while the utilities in developing countries did not exhibit any defined regulation scheme because the government had been using the state-owned enterprises as subsidy mechanisms. Regulation by incentives promotes cost cutting, which impacts adversely on the quality of service (Kidokoro, 2002; Sheshinski, 1976; Spence, 1975). Lewis and Sappington (1991) and Noam (1990) pointed out that monitoring is an important factor because a regulation scheme of price causes the utilities, in general, to reduce costs, which affects quality. Lack of monitoring could result in consumers receiving a poor quality of service. The type of ownership is another important factor that influences decision making on investments and expenditure for improving the quality of service. According to the theory of contracts, a public firm could be more efficient than could a private firm in the case of an incomplete contract. The private utility, through its goal to be more profitable, would have the incentive to cut costs, so affecting the quality of service not specified in the contract with the government. The study of how the factors of regulation, monitoring, and ownership influence the quality of service of Latin American electric distribution utilities may provide regulators and policy makers with knowledge to improve the regulation policies to support the credibility and sustainability of the reform (privatization).

32

3.

Methodology

The purpose of the study was to determine the influence of regulation, monitoring, and ownership on the level of quality of service provided by the electric distribution utilities in Latin America. Chapter 3 includes an explanation of the development of the research design that was required to fulfill the objective of the study. Descriptions of the population and the sample appear in the chapter. Finally, the chapter indicates details of the instrument, data analysis, and validity and reliability of the econometric model. 3.1 Research Design A quantitative paradigm was appropriate for the study because the aim of the research was to evaluate empirically the effect that regulation, monitoring, and ownership have on the level of quality of service. The main reason for selecting a quantitative paradigm was that the measurement of the independent variable, which denominated the quality of service, was numerical through the standardized system average interruption duration index (SAIDI) and system average interruption frequency index (SAIFI) (Burke, 1994). The indices aided in avoiding the use of qualitative (soft) factors of quality of service based on perceptions. The study included an econometric model with quality of service functioning as an explicative variable. According to Ter-Martirosyan (2003), the independent variables are both the SAIDI and the SAIFI. Ai and Sappington (1998, 2002) and Ter-Martirosyan (2003) emphasized that researchers should consider the explicative variables of capital, labor, and demand of the electric distribution utility. Thus, this study involved constructing an improved econometric model to consider the relation between the quality of service and the economic factors (capital, labor, demand and management) and the factors of design of the reform (scheme of regulation, mechanism of monitoring, and type of ownership). Figure 2 illustrates the econometric model.

33

Infrastructure (capital)

Operation & Maintenance (labor)

Level of Quality of Service SAIFI, SAIDI

Demand (consumption)

Productivity (management)

Regulation, Monitoring, and Ownership

Figure 2. Econometric model.

The dependent variables, SAIDI and SAIFI, represented the degree of continuity of the electric distribution service. The independent variables related to electric infrastructure, represented by the length of the distribution networks and the quantity of the installed equipment (switches and transformers). These components reflected the level of investment executed by the utilities. Furthermore, the quantity of equipment included specifically equipment for protection and sectionalizing the distribution networks due to their high incidence in the quality indices of the electric distribution system (Brown, 2002; Gnen, 1986). The independent variables related to labor were the number of workers and the cost of operations and maintenance. These factors indicated the level of expenses of the utilities that influenced the operation of the distribution installations. The independent variables of market (number of users and consumption) and management (number of employees and losses of energy) illustrated the grade of enterprise efficiency. The variables formed part of the econometric model through ratios (see Table 4).

34

Table 4 General Description of Dependent and Independent Variables Variable SAIDI SAIFI LA Definition Measures the time to recover the electric service after an interruption Measures the degree of continuity of the electric service Measures the kilometers of the aerial networks of medium voltage over the kilometers of aerial networks of low voltage Measures the quantity of switches per kilometer of medium-voltage networks (aerial & underground) Measures the quantity of transformers per kilometer of low-voltage networks (aerial & underground) Measures the annual expenses for the electric distributions operations & maintenance (US$ per year) Measures the annual sales of electric energy in MWh per consumer Measures the quantity of consumers per workers Measures the percentage of electric energy losses at the level of the distribution system Indicates the existence of underground networks Index System average interruption duration index (hrs/year) System average interruption frequency index (times/year) Length of aerial medium-voltage network/length of aerial low-voltage network No. switches/length of medium-voltage network No. distribution transformers/length of low-voltage network Expenses of operations and maintenance/electric distribution installations Annual sales of energy/consumer No. consumer (thousands)/workers Electric energy distribution losses (%)

SKM

SED

OM

MERCADO PROD (productivity) PERD (losses) LS1

0: Without underground network 1: With underground network

Representing the independent variables with ratios standardized the components to improve comparability among the studied utilities and to solve possible problems of colinearity. For example, the market variables, the number of users, and the consumption of energy correlate strongly. Thus, the following ratio was appropriate: consumption/number of consumers. The variables, used to contrast the corresponding hypotheses of this study, were the scheme of regulation (rate of return, price cap, and model firm), monitoring (with or without sanction), and ownership (public or private), as is evident in Table 5.

35

Table 5 General Description of Control Variables (Zit) Variable RORR PCAP EMDL TIPOFIS TIPOWN Description Rate-of-return regulation Price-cap regulation Model-firm regulation Type of monitoring Type of ownership Scale 1: Rate of return 0: Others 1: Price cap 0: Others 1: Model firm 0: Others 0: Without sanction 1: With sanction 0: Public 1: Private

The institutional variables related to policy of regulation, monitoring of the quality of service, and type of ownership had not been registered variations in the studied countries under the considered period (2002-2007) in this research. An aspect that was important to consider was that the dummy variables of regulation and monitoring correlate strongly because the sample utilities that use regulation by incentives are subject to a monitoring regime with sanction. In addition, the utilities that operate under rate-of-return regulation reflect a monitoring regime without sanction. The dummy variable of ownership exhibits a low correlation with the scheme of regulation and type of monitoring because, in some countries, the process of privatization is incomplete. This situation has resulted in the existence of private and public utilities operating under a regulation-by-incentives scheme. Thus, the structure of the research questions and the hypotheses reflected the purpose of estimating the influence that each institutional variable has on the quality of service. The electric distribution industry is a capital-intensive business, which involves long-lived assets, strong economies of scale, and low market-risk characteristics of long-term business. Consequently, the study involved a panel data technique to permit the evaluation of the variables and factors that determine the level of quality of service for a number of firms over i and t periods. The study included 58 electric distribution utilities from 10 countries in Latin America over 6 years. Following the heuristic model of the electric distribution systems and the economic theory, the coefficients of the independent variables should reflect the signs presented in Tables 6 and 7. The tables illustrate the expected relations between independent and dependent variables. The positive (+) sign represents an increment in the SAIDI and SAIFI, which influences the quality of service negatively. The negative () sign represents a decrease in the SAIDI and SAIFI, indicating an improvement in the quality of service.

36

Table 6 Expected Relation between Economic and Structural Factors and Quality of Service Variable Aerial network Underground network No. switches No. transformers Expenses of O&M No. consumers/employees Energy losses Note. O&M = operations and maintenance. SAIDI + + + + SAIFI + + + + Quality Worse Worse/Better Better Better/Worse Better Worse Worse

Table 7 Expected Relation between Political and Institutional Variables and Quality of Service Variable RORR PCAP EMDL TIPOFIS (without sanction) TIPOFIS (with sanction) TIPOWN (public) TIPOWN (private) SAIDI + + + + SAIFI + + + + Quality Better Worse Worse Worse Better Better Worse

3.2 Appropriateness of the Design The dependent variables, measured by a quantitative index, included the following: (a) technical and economic variables (investment, costs, market, and management), (b) institutional variables (regulation and monitoring), and (c) structure variable (ownership). A quantitative methodology was appropriate to explain the relation of these variables to the quality of service. Quantitative studies aid in clarifying causal relations. Other researchers conducting studies in the telecommunications and power sectors have used quantitative methods (Ai & Sappington, 1998; Clements, 2001; Ter-Martirosyan, 2003). Because the sample data included the characteristics of both cross-sectional and time-series information, a panel data technique was appropriate. The panel data allowed for control of the heterogeneity of the electric distribution utilities and enabled capture of the dynamics of the data (Greene, 2000; Verbeek, 2000; Wooldridge, 2001). The data set facilitated quantification of the effect of the variables included in the hypotheses, such as the independent variables, which influence the quality of service.

37

3.3 Research Questions The aim of the research was to determine how regulation, monitoring, and ownership influence the quality of service supplied by the electric distribution utilities of Latin America. The following research questions guided the study: R1: What is the influence of regulation schemes on the quality of service provided by the electric distribution utilities in Latin America? R2: What is the influence of monitoring mechanisms on the quality of service provided by the electric distribution utilities in Latin America? R3: What is the influence of ownership forms on the quality of service provided by the electric distribution utilities in Latin America? R4: Do electric distribution utilities in Latin America under the scheme of rate-ofreturn regulation offer a better quality of service than do private utilities regulated under the price-cap or model-firm scheme? 3.4 Hypotheses To answer the four research questions, creation of seven hypotheses occurred. Figure 3 shows the hypotheses considered in this research schematically. No cases of private utilities under rate-of-return regulation and without-sanction monitoring were apparent.
Property Private State-owned H6 H7 H5 H4

Scheme/type RORR PCAP EMDL Total With sanction Without sanction Total

Total

x x x x x

x x x x x x x

x x x

H1

H2

H3

Figure 3. Scheme of hypotheses.

38

Answering the first research question involved contrasting the following hypotheses: H1: The rate-of-return regulation regime provides a better quality of service than does the price-cap regime. H2: The rate-of-return regulation regime provides a better quality of service than does the model-firm regime. Answering the second research question required contrasting the following hypotheses: H3: In the case of state-owned utilities, adopting a monitoring regime with sanction positively affects the quality of service compared to the quality of service provided by the utilities not affected by sanctions. H4: Private utilities respond better than do state-owned utilities in terms of quality of service when the regime of monitoring involves sanction. Answering the third research question involved testing the following hypothesis: H5: The private utility does not provide a better quality of service than does the state-owned utility. Answering the fourth research question included contrasting the following hypotheses: H6: The utilities under rate-of-return regulation provide a better quality of service than do the private utilities under price-cap regulation. H7: The utilities under rate-of-return regulation provide a better quality of service than do the private utilities under model-firm regulation. 3.5 Population The study included the electric distribution utilities of 10 countries located in Central and South America. Table 8 illustrates the composition of the number of utilities and the total number of observations; the studied sample included 58 cases and 309 observations. The sample did not reflect private electric distribution utilities under a rate-of-return regulation scheme or a regime of monitoring without sanction because, in Latin America, the countries that implemented infrastructure sector reform (privatization) have adopted a regulation-by-incentives scheme and a monitoring mechanism with sanction.

39

Table 8 Typology of the Utilities: Number of Cases (and Number of Observations) Institutional variable Regulation Scheme/structural variable RORR PCAP EMDL Total With sanction Without sanction Total Ownership Private 16 (134) 23 (83) 39 (217) 39 (217) 39 (217) State-owned 8 (32) 2 (12) 9 (48) 19 (92) 11 (60) 8 (32) 19 (92) Total 8 (32) 18 (146) 32 (131) 58 (309) 50 (277) 8 (32) 58 (309)

Monitoring

Reduction of the initial 16 countries to 10 was required. Removal of Honduras, Nicaragua, and Venezuela from the sample occurred due to lack of technical, economic, and market information for the period 2002 to 2007. Furthermore, Colombia did not form part of the sample because the country uses quality indices (DES and FES) other than SAIDI and SAIFI to measure the quality of service. Similarly, removal of Costa Rica from the sample occurred because the country reflects different quality indices (DEPI and FPI). Finally, Panama did not form part of the sample due to lack of technical information with relation to the number of installed switches in the medium-voltage network. Thus, the study included electric distribution utilities from the following 10 Latin American countries: Argentina, Bolivia, Brazil, Chile, Ecuador, El Salvador, Guatemala, Paraguay, Peru, and Uruguay. 3.6 Informed Consent In the first stage of the research, representatives of the regulators, managers of the utilities, and representatives of certain nongovernmental institutions of the initial 16 countries received the informed consent form (see Appendix A) and the questionnaire. The informed consent form indicated that the participants had decided to take part in the research of their own free will. The consent form included a description of the research and its nature and aim to give the utilities the confidence to participate in the research. 3.7 Sampling Frame The electric distribution utilities presented diverse characteristics due to their own sectors and market structures, which required operating with different schemes of regulations, types of monitoring, and types of ownership. Table 9 indicates the number of utilities per country. Table 10 includes details related to the utilities in the study sample. The utilities per country reflected market shares varying between 35% and 100%, with an average of 67%.

40

Table 9 Sample Population from the 10 Latin American Countries Ownership Country Argentina Bolivia Brazil Chile Ecuador El Salvador Guatemala Paraguay Peru Uruguay Total No. utilities 3 1 21 8 6 4 1 1 12 1 58 Private 3 1 19 8 0 4 1 0 3 0 39 State-owned 0 0 2 0 6 0 0 1 9 1 19

Table 10 Population and Market Participation of Distribution Utilities Market participation Country Argentina Bolivia Brazil Chile Ecuador El Salvador Guatemala Paraguay Peru Uruguay Total No. utilities 3 1 21 8 6 4 1 1 12 1 58 Total of energy selling (MWh) 56% 30% 64% 94% 56% 100% 37% 100% 89% 100% 67% No. customers 45% 36% 63% 83% 50% 100% 43% 100% 67% 100% 62%

3.8 Confidentiality Coding of the data submitted by the regulators, utilities, and nongovernmental organizations of Latin America occurred. Identification of the electric distribution utilities was only possible during the processing of data. Similarly, coding of the collected information from secondary sources to safeguard the confidentiality of all sources. The final report shows only coded and aggregated information.

41

3.9 Geographic Location The study included the electric distribution utilities of 10 countries located in Central and South America. The countries were Argentina, Bolivia, Brazil, Chile, Ecuador, El Salvador, Guatemala, Paraguay, Peru, and Uruguay. 3.10 Instrumentation Data collection required use of a questionnaire (see Appendix B). The questionnaire reflected simplicity and clarity in the definition of the variables and their quantitative measurement. The purpose of the questionnaire was to gather the information needed for the econometric model. The collected data related directly to indices and statistics of (a) quantity of installations, (b) operations and maintenance costs, (c) market, (d) quality of service, and (e) performance. The information was registered data, so a validation test of the instrument (questionnaire) was not required because the collected data did not refer to qualitative or opinion data. Secondary data were required to supplement or validate the primary information. Trochim (2001) pointed out that a content analysis, evident in this research, makes the use of published data possible. In Latin America, data and statistics published by regulators and nongovernmental organizations were available to reduce the asymmetry of information. However, the most important utilities in Latin America published technical, market, financial, and performance information through their websites, which was useful in gathering and validating the data. The websites were an efficient source of information and allowed for a more detailed extension of the collected data. However, some of the published data was not at all consistent, which resulted in a return to cross validation. 3.11 Data Collection Data sources were both primary and secondary. The required data for the research were technical, economic, and commercial information. The electric distribution utilities were the primary sources of information. The utilities located in Argentina, Brazil, Chile, Guatemala, Paraguay, and Peru participated by completing the designed questionnaire. Secondary sources of information included completed questionnaires from the Electrical Energy Agency of Brazil (ANEEL), the National Electricity Council of Ecuador (CONELEC), the General Superintendence of Telecommunication and Electricity of El Salvador (SIGET), and the Supervisory Organism of Energy and Mining Investments of Peru (OSINERGMIN). In addition, data collection involved accessing information published on the websites of regulators and organizations, such as the Latin American Energy Organization (OLADE), the Regional Energy Integration Commission (CIER), the Electric Energy Distribution Utilities Association (ABRADEE), and the World Banks database of the electric 42

distribution utilities of Latin American Countries (LAC). Data collection through various sources allowed for the revision of the reliability of the reported information. The electronic addresses of the regulators, utilities, and nongovernmental organizations appear in Appendix C. 3.12 Data Analysis The definition of each variable of the econometric model is evident in Table 4. The purpose of data analysis was to identify unusual values, outliers, discontinuities, and other peculiarities that allow the generation of errors that do not permit identification of the existent relations among the variables of the model. The study involved evaluating 58 electric distribution utilities of 10 Latin American countries over a period of 6 years (2002-2007). The total of processed and observed cases was 309. The descriptive statistics of the model variables appear in Table 11. Worthy of mention is that analysis involved reviewing collected data over the base of the initial information using data from different sources. Telephonic conversations with and e-mails from the key representatives of the utilities, regulators, and nongovernmental institutions aided in verifying the data gathered. The process simplified the decisions on maintaining or retiring the collected data when the information registered did not correspond to a normal characteristic (outliers).
Table 11 Descriptive Statistics of the Model Variables Variable SAIDI SAIFI LA SKM SED OYM MERCADO PROD PERD M 17.54489 12.27796 1.457087 1.204401 2.420194 0.742492 4.00123 17.54489 12.27796 SD 14.38136 8.101269 1.270385 1.049598 2.01642 0.5395982 1.797268 14.38136 8.101269 Bias 1.732712 1.136072 2.125137 1.634239 1.629569 1.689237 0.561542 1.732712 1.136072 Kurtosis 5.939676 3.818405 7.04154 5.779523 5.189165 6.753061 3.245647 5.939676 3.818405 Variability coefficient 0.819 0.659 0.871 0.871 0.833 0.726 0.449 0.819 0.659 No. cases 309 309 309 309 309 309 309 309 309 N 58 58 58 58 58 58 58 58 58

The indices of capital (LA, SKM, and SED) show a moderate variability with standard deviation under its mean reflected in the variation coefficient where the values for all the cases are higher than 0.8. Similarly, the indices of quality of service, SAIDI and SAIFI, show a moderate variability due to the standard deviation being under its mean. Nonetheless, the variation coefficient in the SAIDI (0.819) is substantially higher than the value obtained in the SAIFI (0.659). Related to the normal curve, these variables have different asymmetries with different 43

grades to the right. The variable related to productivity is almost symmetric, as is shown through the values of median, mean, and kurtosis. No data transformation was required because the panel data models (nonlinear) are not based on criteria related to variable distribution. Appendix D illustrates the frequency of distribution and the Q-Q graphs of each variable of the model. 3.13 Validity and Reliability Taking into account the validated data of 58 electric distribution utilities of Latin America for the 2002 to 2007 period, regressions were required to explain the dependent variables, SAIDI and SAIFI. The estimations involved using unbalanced panel data due to the lack of data for the period considered in the study. The general specification of the regression model was the following:

Y = + X + Z + it IID (0, 2 )
it it it

(1)
it

In Equation 1, Y represents the dependent variables, SAIDI and SAIFI, and is the autonomous coefficient. The independent variables reflect two groups: X represents the independent variables (indices), and Z is the control variables (dummy variables). The general regression can be presented as follows:
SAIDIit = + 1LAit + 2SKMit + 3SEDit + 4 OMit + 5 MERCADO it + 6 PRODit + 7 PERDit + 8 LS1it + 1TIPOWN + 2TIPOFIS + 3 REGU + it

(2)

SAIFIit = + 1LAit + 2SKMit + 3SEDit + 4OMit + 5MERCADO it + 6 PRODit + 7 PERDit + 8LS1it + 1TIPOWN + 2TIPOFIS + 3 REGU + it

(3)

Table 4 includes descriptions of each variable of the econometric model shown in Equations 2 and 3. 3.13.1 General Regressions A regression of the econometric model was necessary to verify the relations between independent and dependent variables. Table 12 shows the relations between the dependent variables, SAIDI and SAIFI, and the independent variables. The signs of the capital, labor market, and performance variables were the ones expected, and the coefficients are significant. In addition, the signs of the LA, SKM, SED, OM, and PERD variables show the theoretical expected relation from Table 6. The presented model reflects the influence of the independent variables with levels

44

of reliability (p value) of 1% and 5%, while the model is significant with a p value of 1%, according to the Wald test.
Table 12 General Regressions SAIDI Constant LA SKM SED OM MERCADO PROD PERD LS1 TIPOWN TIPOFIS REGU Wald chi2(11) Prob. chi^2 Observations *p < 0.05. ** p < 0.01. 7.356476 (4.764111) 2.854153** (0.8090492) -2.989169** (1.008276) -1.016365* (0.4969278) 2.223509* (1.000083) -0.5427978 (0.4423807) 1.862785 (3.470459) 43.27146* (17.85262) -0.5915788 (1.483238) -14.16416** (5.523327) 9.606753 (6.864056) 5.748643* (2.58658) 74.03 0.0000 309 SAIFI 10.17481* (4.001961) 1.343546* (0.5268945) -1.160788 (0.6299439) -0.4203361 (0.3176643) 1.032096* (0.4994154) -0.1331915 (0.2643485) -1.506952 (1.624087) 28.86231* (11.75306) -0.7470988 (0.9149828) -9.014742** (2.718519) -9.8286437 (4.618635) -0.1711495 (1.429165) 66.12 0.0000 309

Furthermore, to explain the SAIDI variable, the dummy variables, REGU (regulation regime) and TIPOWN (type of ownership), are significant with a p value of 5% and 1% respectively, while the TIPOFIS (type of monitoring) variable is nonsignificant. To explain the SAIFI variable, TIPOWN is significant with a p value of 1%. The obtained results for the variables TIPOFIS, TIPOWN, and REGU with respect to the independent variable SAIFI could be explained by the high correlation with the regulation and monitoring variables. These variables relate to the development of the power activity adopted by the countries considered in this study. For example, in the case of countries like Argentina, Bolivia, Brazil, Chile, El Salvador, Guatemala, and Peru, the utilities operate under a regulation-byincentives scheme and a mechanism of monitoring by sanction. In addition, 45

Ecuador, Paraguay, and Uruguay have state-owned utilities operating under rate-ofreturn regulation and a mechanism of monitoring without sanction. To verify the correlation supposition between the dummy variables, analysis of the variance inflation index (VIF) and its correlation matrix occurred. 3.13.2 Estimations, Analysis, and Tests of the Model The following sections illustrate tests of the model relating to the correlation variables, the multicolinearity, the autocorrelation, and the specification of the model. Furthermore, a proposed specification to solve the problem of heteroscedasticity is formulated. 3.13.2.1 Correlation Matrix The correlation matrix shows the relationships between variables and tests whether the correlation index of the variables in absolute value is higher. The existence of higher indices indicates the possible presence of colinearity. Table 13 shows that the LA and PERD variables have positive and significant relationships with the SAIDI. In addition, the SKM, SED, OM, and MERCADO variables have negative and significant relationships with the SAIDI. The results indicate the expected signs between the independent and dependent variables. The positive indices cause an increment in the quality index, which means a decrease in the quality of service. The negative indices cause a reduction in the quality index, which means an improvement in the quality of service.

46

Table 13 Pearson Correlation Coefficient between the SAIDI and the Independent Variables
SAIDI SAIDI LA SKM SED OM MERCADO PROD PERD LS1 TIPOFIS TIPOWN REGU RORR PCAP 1,000 0.134* -0.421* -0.226* -0.116* -0.543* -0.098 0.117* -0.074 0.105 -0.365* 0.256* -0.105 -0.206* LA 1,000 -0.262* 0.424* -0.298* 0.084 -0.277* 0.109 -0.342* 0.128* 0.097 -0.211* -0.128* 0.432* SKM SED OM MERCADO PROD PERD LS1 TIPOFIS TIPOWN REGU RORR PCAP

1,000 0.016 0.273* 0.552* 0.305* -0.206* 0.171* 0.110 0.306* 0.054 -0.110 0.064

1,000 -0.132* 0.386* -0.018 0.024 -0.039 -0.046 0.108 -0.317* 0.046 0.357*

1,000 0.227* 0.088 -0.333* 0.161* 0.173* 0.175* 0.234* -0.173* -0.095

1,000 0.044 -0.205* 0.153* 0.006 0.391* -0.247* -0.006 0.331*

1,000 -0.385* 0.295* 0.475* 0.305* 0.463* -0.475* -0.025

1,000 -0.234* -0.523* -0.179* -0.546* 0.523* 0.075

1,000 -0.038 -0.250* 0.137* 0.038 -0.225*

1,000 0.522* 0.689* -1.000 0.322*

1,000 0.146* -0.522* 0.446*

1,000 -0.689* -0.465*

1,000 -0.322*

1,000

*p < 0.05.

47

The institutional variables reflect a relation, on average stronger than the other variables. In particular, the TIPOWN and TIPOFIS variables have correlation coefficients of 0.52, and the REGU and TIPOFIS variables show a correlation of 0.69. Both correlations are significant. Separation of the REGU variable into the RORR (rate-of-return regulation) and PCAP (price-cap regulation) variables helped to determine the correlation grade between the different schemes of regulation and the TIPOFIS variable. A correlation of -1.0 is evident between RORR and TIPOFIS. Explanation of the obtained result includes that the utilities under rate-of-return regulation only exhibit monitoring without sanction. This fact reflects the presence of multicolinearity in the regressions. A possible solution could be not using the three institutional variables simultaneously. Table 14 shows the results of the correlations between the SAIFI and the independent variables. A different level of correlation is apparent compared to the correlations with the SAIDI. The fact that the signs and significance of each variable remain the same is relevant. The pattern of correlation between the institutional variables remains strong. Importantly, the perfect correlation between the RORR and TIPOFIS variables stays the same. The other variables, PCAP, TIPOFIS, and TIPOWN, show similar correlations to those presented in Table 13.

48

Table 14 Pearson Correlation Coefficient between the SAIFI and the Independent Variables
SAIFI SAIFI LA SKM SED OM MERCADO PROD PERD LS1 TIPOFIS TIPOWN REGU RORR PCAP 1,000 0.256* -0.420* -0.105 -0.235* -0.459* -0.303* 0.362* -0.146* -0.092 -0.434* -0.048 0.092 -0.049 LA 1,000 -0.262* 0.424* -0.298* 0.084 -0.277* 0.109 -0.342* 0.128* 0.097 -0.211* -0.128* 0.432* SKM SED OM MERCADO PROD PERD LS1 TIPOFIS TIPOWN REGU RORR PCAP

1,000 0.016 0.273* 0.552* 0.305* -0.206* 0.171* 0.110 0.306* 0.054 -0.110 0.064

1,000 -0.132* 0.386* -0.018 0.024 -0.039 -0.046 0.108 -0.317* 0.046 0.357*

1,000 0.227* 0.088 -0.333* 0.161* 0.173* 0.175* 0.234* -0.173* -0.095

1,000 0.044 -0.205* 0.153* 0.006 0.391* -0.247* -0.006 0.331*

1,000 -0.385* 0.295* 0.475* 0.305* 0.463* -0.475* -0.025

1,000 -0.234* -0.523* -0.179* -0.546* 0.523* 0.075

1,000 -0.038 -0.250* 0.137* 0.038 -0.225*

1,000 0.522* 0.689* -1.000 0.322*

1,000 0.146* -0.522* 0.446*

1,000 -0.689* -0.465*

1,000 -0.322*

1,000

*p < 0.05.

49

3.13.2.2 Multicolinearity Test The purpose of the test was to determine whether a strong level of multicolinearity was present between the explicative variables. A strong level of multicolinearity would require retiring some variables from or including new variables in the model. A regression incorporating the endogenous and exogenous variables of the model was necessary to determine the presence of multicolinearity. The VIF obtained is evident in Table 15.
Table 15 Variance Inflation Index (VIF) of the General Model Variable TIPOFIS REGU MERCADO TIPOWN PROD LA PERD SKM SED LS1 OM Mean VIF VIF 3.81 3.26 2.78 2.38 1.97 1.96 1.95 1.86 1.60 1.59 1.37 2.23 1/VIF 0.262246 0.306757 0.360267 0.420631 0.508265 0.511097 0.512556 0.536950 0.626496 0.629896 0.729956

The indices of the institutional variables are higher than 2, which indicates a higher correlation between the variables. This analysis is valid for the regressions of the two models (SAIDI and SAIFI) due to both equations being explained by the same independent variables. Furthermore, the TIPOFIS and TIPOWN variables can generate multicolinearity problems. The difference in percentage between the maximum value VIF (TIPOFIS) and the mean VIF is 71%, which may be an indication of multicolinearity. The following actions may address the problem: (a) estimate the model taking into account only one of the two dummies, and (b) separate the REGU variable into RORR and PCAP. The obtained VIF is explained through the data in the sample because the RORR variable in all the cases is only related to the monitoring variable with sanction. The same situation is evident with the PCAP and EMDL (model-firm regulation) variables; these variables only relate to the monitoring variable with sanction. All the relations appear in Table 16.

50

Table 16 Relation between Regulation and Monitoring Variables Type of monitoring Regulation scheme RORR PCAP EMDL With sanction 0 X X Without sanction X 0 0

Because considering only one dummy variable in the model estimation would not allow for contrasting the hypotheses of the research, the second alternative solution was more appropriate. Therefore, estimation of two equations occurred. The first equation included the regulation variables of RORR and PCAP but excluded the TIPOFIS variable because TIPOFIS correlates exactly with the mentioned regulation variables. Thus, the equations appeared as follows:
SAIDIit = + 1LAit + 2SKM it + 3SED it + 4OM it + 5MERCADOit + 6 PRODit + 7 PERDit + 8LS1it + 1TIPOWN + 2 PCAP + 3 RORR + it SAIFIit = + 1LAit + 2SKM it + 3SED it + 4OM it + 5 MERCADOit + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2 PCAP + 3 RORR + it

(4)

(5)

The second equation involved retaining the variables of monitoring (TIPOFIS) and ownership but excluding the regulation-scheme variables:
SAIDI it = + 1LA it + 2SKM it + 3SED it + 4 OM it + 5 MERCADO it + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2TIPOFIS + it

(6)

SAIFIit = + 1LAit + 2SKM it + 3SEDit + 4OM it + 5MERCADOit + 6PRODit + 7 PERDit + 8LS1it + 1TIPOWN + 2TIPOFIS + it

(7)

Determining the presence of multicolinearity in the new equations included estimating the VIF. Table 17 shows that the VIF mean values are 2.03 and 1.96. Additionally, the difference between the TIPOFIS variable, which has the higher index, and the VIF mean value is 34%. Because the differences are lower than are those obtained in the general regression, both specifications partially solved the multicolinearity problem.

51

Table 17 Variance Inflation Index (VIF) Equation 4 and 5 Variable RORR MERCADO TIPOWN PROD LA PERD PCAP SKM SED LS1 OM Mean VIF VIF 2.99 2.78 2.38 1.97 1.96 1.95 1.91 1.86 1.60 1.59 1.37 2.03 1/VIF 0.334396 0.360267 0.420631 0.508265 0.511097 0.512556 0.522897 0.536950 0.626496 0.629896 0.729956 Variable TIPOFIS MERCADO TIPOWN PROD LA SKM PERD LS1 SED OM Mean VIF Equation 6 and 7 VIF 2.63 2.61 2.32 1.94 1.91 1.86 1.79 1.59 1.55 1.36 1.96 1/VIF 0.380210 0.382708 0.431541 0.514525 0.524864 0.537810 0.558160 0.630432 0.645453 0.735996

3.13.2.3 Specification Test Three types of panel data models are relevant depending on the treatment of heterogeneity. The first type is the data pool model based on an estimation of minimum square, which reflects the assumption that individual heterogeneity does not exist, i = 0. This model is the simplest to analyze because of the omission of space and time dimensions of the grouped data and calculation of the usual regression of ordinary mean square (OMS). The second type is the fixed effects model, through which coefficients are obtained by the within estimator. The heterogeneity, j, is reflected in constant terms that vary between individuals. This model was not appropriate for the research because any invariable element of the time dimension is eliminated. The third type is the random effects model (generalized minimum squares), where the heterogeneity, i, is a random term with the following distribution: 2 . This model is more efficient but is only consistent when and the i IID(0, ) j explicative variables do not correlate. To determine which of the models (data pool, fixed effects, or random effects) would be more appropriate to observe the data characteristics, the following tests were necessary: the F test, the Breusch-Pagan test, and the Hausman test. The F test was required to verify whether the data pool model would be the most appropriate to make the econometric estimations. The null hypothesis contrasts the existence of intercepts, because individuals exist in the sample, with the presence of a single intercept. If the statistic F value was higher than the criticalvalue showed in the table 18, a model with individual effects (fixed or random) would be appropriate; if not, a data pool model would be suitable. The result of the F test, 52

apparent in Table 18, reflects that for all cases, the p values were lower than 0.05. Thus, the data pool model was not fitting.
Table 18 Results of the F Test SAIDI MODEL 1 MODEL 2 F(57, 243) = 10.75 Prob > F = 0.0000 F(57, 243) = 11.16 Prob > F = 0.0000 SAIFI F(57, 243) = 14.03 Prob > F = 0.0000 F(57, 243) = 14.03 Prob > F = 0.0000

The Breusch-Pagan test was required to determine whether the variance of the individual heterogeneity term was equal to zero. To accept this hypothesis, one would need to use the data pool model. Table 19, shows the outcomes of the test, with the null hypothesis being the use of the data pool model and the alternative hypothesis being the use of a nest model (fixed effects or random effects). The results for each case showed a p value lower than 0.05, which indicates significant evidence to reject the null hypothesis. Thus, determining which of the remaining models (fixed effects or random effects) was the most efficient was necessary.
Table 19 Results of the Breusch-Pagan Test SAIDI MODEL 1 MODEL 2 chi2(1) Prob > chi2 chi2(1) Prob > chi2 = = = = 274.07 0.0000 275.78 0.0000 chi2(1) Prob > chi2 chi2(1) Prob > chi2 SAIFI = = = = 319.34 0.0000 319.31 0.0000

The Hausman test was required to determine which of the two methods, fixed or random effects, would be the most appropriate. Table 20 shows the result, which indicated acceptance of the null hypothesis for all equations (4, 5, 6, and 7). The difference between the coefficients of random and fixed effects was not systematic, so the random effects method was appropriate. The dummy variables were not considered in the statistic tests because one of the comparative models (fixed effects) does not allow for the use of dichotomous variables.

53

Table 20 Results of the Hausman Test SAIDI MODEL 1 MODEL 2 chi2(8) Prob > chi2 chi2(1) Prob > chi2 = = = = 5.12 0.7442 5.82 0.6673 chi2(8) Prob > chi2 chi2(1) Prob > chi2 SAIFI = = = = 10.64 0.2231 10.67 0.2209

Because the Hausman test showed that the random effects model reflected consistent and efficient estimators, modification of the equation of the econometric model was necessary. The result is consistent with Greene (2000) who highlighted the convenience of using random effects panel data where the number of individuals is superior to the number of periods. In this case, the data corresponded to 58 electric distribution utilities for a period of 6 years. The specification test indicated use of the random effects model, so the general equation to be estimated appeared as follows:

Yit = + X it + Z it + u i + it
A model of random effects allowed for the supposition that each cross section includes a different intercept. Expression of this model was the following:

(8)

Yit = i + X it + Z it + it
Rewriting of the proposed regressions in Equations 4 and 5 occurred as follows:
SAIDIit = + 1LA it + 2SKM it + 3SED it + 4OM it + 5 MERCADOit + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2 PCAP + 3 RORR + i + it SAIFIit = + 1LA it + 2 SKM it + 3SED it + 4 OM it + 5 MERCADOit + 6 PRODit + 7 PERDit + 8 LS1it + 1TIPOWN + 2 PCAP + 3 RORR + i + it

(9)

(10)

(11)

In addition, rewriting of the proposed regressions in Equations 6 and 7 occurred as follows:

SAIDI it = + 1LA it + 2SKM it + 3SED it + 4OM it + 5 MERCADO it + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2TIPOFIS + i + it

(12)

SAIFI it = + 1 LA it + 2 SKM it + 3SED it + 4 OM it + 5 MERCADO it + 6 PROD it (13) + 7 PERD it + 8 LS1it + 1TIPOWN + 2TIPOFIS + i + it

54

3.13.2.4 Autocorrelation Test Analysis of the presence of autocorrelation in the formulated regressions was required through the LBI test of autocorrelation proposed by Baltagi and Wu (1999), which is a modification of the Durbin-Watson test. Furthermore, the test was appropriate to verify unbalanced panel data with random effects. Similar to the Hausman test, the Baltagi-Wu test eliminates the dichotomous variables when measuring the first differences. The statistic value verifies the presence or absence of autocorrelation. In this case, a value closer to 2 would indicate the absence of autocorrelation. Nonetheless, the test has an implicit limitation in the supposition that the temporal dimension of the unbalanced panel data must be wide (Baltagi & Wu, 1999). The Baltagi-Wu statistic showed values of 1.50 for the SAIDI and 1.98 for the SAIFI. Evidence of correlation is stronger in the SAIDI, while the evidence related to SAIFI illustrates no correlation of the errors. The relevant calculus appears in Appendix E. 3.13.2.5 Heteroscedasticity Test When the residual variance of a cross-section unit is not constant, violation of the supposition of homoscedasticity of the variance occurs. In this study, detection of the problem was not possible with the traditional methods due to the use of unbalanced panel data, which do not allow for reliable outcomes. Thus, assuming that the heteroscedasticity problem exists was necessary. Solving the problem required adoption of an estimation of generalized minimum squares equation with standard robust errors to correct some possible problems in the variance of the residuals (Greene, 2000). 3.13.3 Conclusions of the Test The model specification test indicated the presence of autocorrelation and heteroscedasticity. The available econometric tools of calculus do not allow for correction of both problems simultaneously. Thus, correcting only one of the problems was possible. One option was correcting the autocorrelation to analyze the sample composition, which reflects a lower number of observations in the temporal dimension and a higher number of observations in the transversal dimension. Because the sample includes only 6 years of observation, the dependency of error temporal problem could exist due to the highest persistent existence in the distribution segment, where the growth is produced fundamentally by vegetative demand. In contrast, the sample composition reflected a higher probability of heteroscedasticity because the heterogeneity of the utilities reflected a low 55

probability that the error variance would be constant. Considering that the number of utilities was higher than the number of panel periods, the potential consequences of the heteroscedasticity were more important compared to autocorrelation. Thus, correction of the heteroscedasticity problem occurred using a panel data model with a robust generalized estimation. 3.14 Summary The panel data econometric model adopted for the study allowed for the combination of cross-sectional and time-series information. The combination was necessary to assess the influence that the regulation scheme, monitoring mechanism, and ownership type had on the quality of service provided by the electric distribution utilities, with more than 100 000 consumers, between 2002 and 2007. Validation of the data used in the study occurred through different primary and secondary sources. The study included 58 electric distribution utilities and 309 observations, which represent 66% of the Latin American market. The multicolinearity (VIF) and autocorrelation (Baltagi-Wu) tests helped to validate the econometric model. In addition, the econometric model of unbalanced panel data with random effects passed the Hausman test successfully. Appendix E includes the test results of the model validation and specification. The STATA econometric software aided in performing the calculus. Chapter 4 reflects a presentation of the tests of the hypotheses, taking into account the econometric model developed in this chapter.

56

4.

Presentation and Analysis of the Data

The analysis included the electric distribution utilities of 10 countries located in Central and South America. From the validated data and the econometric model developed for the research, probing of the hypotheses transpired using 309 cases of 58 electric distribution utilities. 4.1 Findings Taking into account that the endogenous variable of the model is explained by the institutional variables of regulation and monitoring and that both variables are perfectly correlated (see Tables 13 and 14), using two equations was necessary. The first equations retained the RORR (rate-of-return regulation) and PCAP (price-cap regulation) variables and excluded the TIPOFIS (type of monitoring) variable, rewritten as follows (Equations 10 and 11 from chapter 3):
SAIDIit = + 1LA it + 2SKM it + 3SED it + 4 OM it + 5 MERCADOit + 6 PRODit + 7 PERDit + 8 LS1it + 1TIPOWN + 2 PCAP + 3 RORR + i + it SAIFIit = + 1LA it + 2SKM it + 3SED it + 4 OM it + 5 MERCADOit + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2 PCAP + 3 RORR + i + it

These equations facilitated probing of specific hypotheses: H1, H2, H6, and H7. The second equations retained the monitoring variables of monitoring (TIPOFIS) and ownership (TIPOWN), rewritten as follows (Equations 12 and 13 from chapter 3):
SAIDIit = + 1LA it + 2SKM it + 3SED it + 4OM it + 5 MERCADO it + 6 PROD it + 7 PERD it + 8LS1it + 1TIPOWN + 2TIPOFIS + i + it SAIFIit = + 1LA it + 2SKM it + 3SED it + 4 OM it + 5 MERCADOit + 6 PROD it + 7 PERD it + 8 LS1it + 1TIPOWN + 2TIPOFIS + i + it

These equations facilitated probing of H3, H4, and H5. The hypothesis tests included examining the value of the coefficients of the regression variables. The Wald test helped to verify the significance of the results of the hypotheses. The results of the first and second equations for the system average interruption duration index (SAIDI) and the system average interruption frequency index (SAIFI) appear in Tables 21 and 22. Table 21 shows the results of the first equations, including the institutional variables of regulation (RORR and PCAP) and the ownership type (TIPOWN). Table 22 reflects the results of the second

57

equations, including the institutional variable of monitoring (TIPOFIS) and the ownership type (TIPOWN).
Table 21 Regression: Regulation (RORR and PCAP) and Ownership SAIDI Constant LA SKM SED OM MERCADO PROD PERD LS1 TIPOWN PCAP RORR Wald chi2(11) Prob chi^2 Observations *p < 0.05. **p < 0.01. 28.46052** (5.979177) 2.854153** (0.8090492) -2.989169** (1.008276) -1.016365* (0.4969278) 2.223509* (1.243231) -0.05427978 (0.4423807) 1.862785 (3.470459) 43.27146* (17.85262) -0.5915788 (1.483238) -14.16416* (5.523327) -5.748643* (2.58658) -21.10404** (6.482052) 74.03 0.0000 309 SAIFI 18.11895** (3.043269) 1.343546* (0.5268945) -1.160788 (0.6299439) -0.4203361 (0.3385594) 1.032096* (0.4994154) -0.1331915 (0.2643485) -1.506952 (1.624087) 28.86231* (11.75306) -0.7470988 (0.9149828) -9.014742** (2.718519) 0.1711495 (1.429164) -7.944138* (4.126756) 66.12 0.0000 309

58

Table 22 Regression: Monitoring and Ownership SAIDI Constant LA SKM SED OM MERCADO PROD PERD LS1 TIPOFIS TIPOWN Wald chi2(11) Prob chi^2 Observations *p < 0.05. **p < 0.01 10.57828* (4.652487) 2.379052** (0.8602537) -3.065697** (1.055833) -1.229937* (0.5212016) 2.322827* (1.037354) -0.7211557 (0.4250488) 2.447953 (3.459857) 35.51765* (17.13544) -0.5572629 (1.47929) 18.46125** (6.226994) -15.75624** (5.55823) 74.03 0.0000 309 SAIFI 10.03413* (3.958654) 1.357288** (0.5201759) -1.153259 (0.6296633) -0.4126054 (0.3042976) 1.025657* (0.4991371) -0.1340548 (0.2612908) -1.52528 (1.607574) 29.38259* (11.69655) -0.7531871 (0.9165113) 8.054049* (4.024678) -8.964998** (2.672312) 66.12 0.0000 309

Presentation of the findings illustrates probing of the seven hypotheses following the order of the four research questions. The first research question read as follows: What is the influence of regulation schemes on the quality of service provided by the electric distribution utilities in Latin America? H1 and H2 related to the first research question. H1 indicated that electric distribution utilities under rate-of-return regulation would, on average, provide a better quality of service than would utilities under price-cap regulation. Table 21 shows that the coefficient of the RORR variable exhibited a negative sign in relation to the SAIDI and SAIFI, a significant result at 1% for the SAIDI and 5% for the SAIFI. The case of the dummy variable PCAP showed the coefficient sign as negative and significant at 5% with respect to the SAIDI and statistically zero in the case of the SAIFI. Nonetheless, verifying H1 correctly required the Wald test. Such types of tests are a generalization of the t test (Cameron & Trivedi, 2005) and reflect the significance of the linear and nonlinear parameter combinations. The results indicated that 59

rejection of the hypothesis was not possible due to the difference between the coefficients being zero. Thus, in both cases (the SAIDI and the SAIFI), the regulated utilities under the rate-of-return scheme supplied a better quality of service than did the utilities regulated under the price-cap scheme. H2 indicated that the electric distribution utilities regulated under a rate-of-return scheme would provide a better quality of service than would utilities regulated under the model-firm scheme (a kind of yardstick competition model). Verifying this hypothesis involved contrasting the coefficient of the RORR variable to determine whether its value was different to zero. Table 21 shows that the coefficient of the RORR variable has a negative relation to the SAIDI and SAIFI. Both cases indicated rejection of the null hypothesis where the coefficients of the RORR variable were equal to zero. The result was in line with the theoretical approach of Spence (1975) and Sheshinski (1976), who stated that the utilities under rate-of-return regulation valued privileged the investments and expenses to supply a good quality of service. In contrast, the private utilities under a high-powered price regulation (price cap or model firm), in their goal to be more efficient and profitable, reduce their costs, affecting the quality of service adversely. Thus, the utilities under rate-of-return regulation provided a better quality of service than did the utilities under a modelfirm scheme of regulation. The second research question read as follows: What is the influence of monitoring mechanisms with sanction on the quality of service provided by the electric distribution utilities in Latin America? H3 and H4 related to the second research question. H3 indicated that in the case of state-owned utilities, the adoption of a monitoring regime with sanction would positively affect the quality of service compared to the quality of service provided by utilities not affected by sanctions. To contrast H3, implementation of a dummy variable, TIPOFIS, was required. TIPOFIS exhibited the value of 1 when a utility would face sanctions due to not fulfilling the standard quality of service (higher degree of monitoring) and reflected the value of 0 when the utility would not face economic sanction (lower degree of monitoring). The contrast of H3 involved the coefficients of the TIPOFIS and TIPOWN variables. In particular, the hypothesis reflected that state-owned utilities with sanction would provide better levels of quality of service than would other stateowned utilities without sanction. Assessing the following was necessary:
tipofis [TIPOFIS = 1] + tipown [TIPOWN = 0] = tipofis [TIPOFIS = 0] + tipown [TIPOWN = 0]

This is equivalent to the following formula: tipofis [TIPOFIS = 1] = 0. Hence, the test was reduced to verify the statistical significance of the TIPOFIS variable and to indicate the sign of the coefficient of this variable.

60

Table 22 shows that the TIPOFIS variable reflected a positive relation with the SAIDI and the SAIFI, with significant p values at 1% and 5% respectively. The positive relation indicates that the state-owned utilities under a monitoring scheme with sanction provided a lower quality of service, which was unexpected. The sample may explain the result: The state-owned utilities under monitoring with sanction are located in Peru and Brazil. The results were consistent with the fact that the Peruvian and Brazilian state-owned utilities, predominantly, at first, exhibit a lower quality of service than do their Latin American peers. The stated-owned utilities under a higher degree of monitoring coexist in countries, such as Peru and Brazil, with both state-owned and private utilities under a regulation-by-incentives (price cap or model firm) scheme. In addition, the utilities under monitoring without sanction operate in countries that have adopted rate-ofreturn regulation. Thus, the result illustrated in a sense how public managers make decisions aligned with the formal and explicit requirements imposed for the rules and regulation of public expenditure or the accomplishment of explicit requirements imposed by the regulation policy for the development of the electric distribution activity. The regulation-by-incentives scheme can affect efficient public management because this type of regulation does not explicitly distinguish or enforce investments and expenses for the maintenance or improvement of the quality of service. Regulators set a cap and allow the firm to manage the costs with the aim of encouraging efficiency in the service (Joskow, 2006). Under this condition, the supposition that public managers exhibit management comparable to that of their peers in private utilities is false. The legal and managerial restrictions do not allow the state-owned utilities access to bank financing to make relevant investments, resulting in difficulty acquiring goods and services to guarantee the good performance of the utility. Furthermore, the fact that state-owned utilities submitted to the pressure of politicians could be impeded to develop independent management (Serra, 2008). Another explanation could be that the better behavior of the managers under monitoring without sanction, which allows for a better quality of service, could relate to the application of rate-of-return regulation, which explicitly includes the investments and operations and maintenance expenses for the quality of service. In this case, the explicit recognition in the tariff enables public managers to fulfill the obligations to provide a good quality of service. According to rate-of-return regulation, public managers behave as is expected, favoring the provision of a goldplated quality of service. Kahn (1988) pointed out that, under rate-of-return regulation, the quality of service is the responsibility of the firm. Therefore, the public manager under these circumstances cannot behave rationally to the economic signals because the regulation reflects the explicit responsibility of the utility to execute the investments and expenses related to quality of service to avoid reducing its tariffs. Thus, the result can be highly influenced for the type of regulation (rate of return or regulation by incentives) of the state-owned utilities. 61

In conclusion, the result of H3 was unexpected. The state-owned utilities with sanction did not respond to the quality policy signals because they did not favor the quality of service unlike the state-owned utilities without sanctions. The outcome is acceptable due to public managers in Latin America make decisions aligned with the formal and explicit requirements imposed for the rules and regulation of public expenditure or the accomplishment of explicit requirements imposed by the regulation policy for the development of the electric distribution activity. In this sense, rate-of-return regulation, which explicitly includes the investments and operations and maintenance expenses for the quality of service together with a monitoring regime without sanction works better than a regulation by incentives even though is accompanied by a monitoring regime with sanction. H4 indicated that private utilities would respond better than would state-owned utilities in terms of quality of service when the regime of monitoring involved sanctions. For this test, a dummy variable, TIPOFIS = 1, was required to separate only the electric distribution utilities with sanction monitoring so that, in this case, the dummy variable TIPOWN would provide the answer to H4. Table 22 shows that TIPOWN has a negative and significant relationship to both the SAIDI and SAIFI (p value of 1%). Furthermore, the TIPOFIS variable has a positive relationship with both dependent variables. In the case of the SAIDI, TIPOFIS has a significant relation of 1% while in the case of the SAIFI, TIPOFIS has a significant relation of 5%. To ensure a correct contrast of the hypothesis, verifying whether the sum of the coefficients of the TIPOFIS and TIPOWN variables was less than the coefficient of the TIPOFIS variable was necessary. This hypothesis could appear as follows: tipofis + tipown = tipofis . Nevertheless, in the analysis of the hypothesis, elimination of the coefficient of the TIPOFIS variable, which is of particular interest, would occur. Therefore, rewriting the hypothesis was required. Because the coefficient of the TIPOFIS variable was statistically different from zero, both sides of the expression could be divided by the TIPOFIS coefficient. The following hypothesis test resulted: H 4 : tipown / tipofis = 0 . The Wald contrast was appropriate to evaluate the hypothesis. The outcomes of the test resulted in rejecting the null hypothesis of equality of the coefficients at 1% of significance in the case of the SAIDI and 5% in the case of the SAIFI. Consequently, the combination of ownership (private) and monitoring with sanction causes a better quality of service than do state-owned utilities under the same monitoring regime. One explanation of this result is that Latin American private utilities are responding to a monitoring policy of quality of service through sanctions. The result is in line with Holt (2004), Lewis and Sappington (1991), Noam (1990), and Waddams Price et al. (2002) who recommended setting penalties for the electric distribution utilities regulated by incentives (price cap or model

62

firm) that do not reflect the minimum requirements of quality of service established in the regulations and standards. Another explanation is that the private electric distribution utilities regulated by an incentive policy accompanied by a high-powered quality monitoring policy are providing a better quality of service than are the state-owned utilities. Thus, monitoring with sanctions for supplying a poor quality of service could solve the problem. In conclusion, H4 corroborated the theory that private electric distribution utilities subjected to monitoring with sanctions provide a better quality of service. The third research question read as follows: What is the influence of ownership forms on the quality of service provided by the electric distribution utilities in Latin America? H5 related to the third research question under the premise of incomplete contracts theory formulated by Hart et al. (1997). The theory indicates that clear specification of quality of service is not evident in the concession contracts signed between the government and the private utilities. Therefore, the private utilities have not been a better alternative in offering and guaranteeing good quality of service. H5 indicated that the private utility would not provide a better quality of service than would the state-owned utility. In this test, analysis of the coefficient of the dummy variable TIPOWN reflected the result. Tables 21 and 22 show that the TIPOWN variable has a negative relationship with the SAIDI and SAIFI, both significant at 1% with the exception of the SAIDI equation in Table 21 where the significance is 5%. The outcomes were robust and illustrated that the private utilities provided a better quality of service than that supplied by the state-owned utilities. Thus, the result of the test was contrary to the theoretical formulation defined in H5. The result confirmed that measuring, monitoring, and sanctioning related to the quality of service supplied by the Latin American electric distribution utilities can occur, so governments should specify the quality of service within the contract. According to the Centraal Planbureau (2004), Netherlands Bureau for Economic Policy Analysis, quality of service should be contractible, and certain criteria should exist to debate privatization of a service sector. Hart et al. (1997) and the Centraal Planbureau (2004) indicated that privatization of a firm or service should occur only if the private firm can ensure a better quality of service than the service offered by the state-owned enterprise. Such an assurance may be evident within the contract. Because competition is not feasible in a network industry due to the characteristics of a natural monopoly, the contractibility of the quality of service is a crucial factor. Furthermore, the outcome illustrated that modification of the behavior of the private firm is possible even though the private manager focuses on the reduction of costs to be more profitable, as evident in the theories formulated by Spence (1975) and Sheshinski (1976). Researchers addressing the regulation-by-incentives schemes emphasized that the private manager, in his or her goal to obtain higher economic 63

benefits, reduces the quality of service (Ai & Sappington, 2005; Burns, 2003; Kidokoro, 2002; Mikkers & Shestalova, 2003; Tangers, 2002). The outcome of H4 showed that their view is not completely true because modification of the behavior of private managers is possible through a monitoring policy with sanctions. In summary, the private Latin American electric distribution utilities supplied a better quality of service than did the state-owned electric distribution utilities. The fourth research question read as follows: Do electric distribution utilities in Latin America under the scheme of rate-of-return regulation offer a better quality of service than do private utilities regulated under the price-cap or model-firm scheme? This question emerged through the research deductive process due to discovering that the private electric distribution utilities offer a better quality of service than do the state-owned utilities. In this sense, a worthwhile question was whether the utilities under rate-of-return regulation continue offering a better quality of service than do the private utilities. H6 and H7 related to this question. H6 indicated that the utilities under rate-of-return regulation would provide a better quality of service than would the private utilities under price-cap regulation. Table 21 shows that the dummy variable RORR has a negative relation to the SAIDI and the SAIFI with a significant p value of 1% and 5% respectively. Furthermore, in the case of the SAIDI, the PCAP variable reflected a significant p value of 5%. The TIPOWN variable related to the SAIDI and the SAIFI with a significant p value of 5% and 1% respectively. Verifying the hypothesis required analysis, through the Wald contrast, of whether the sum of the TIPOWN and PCAP parameters was equal to the RORR parameter. The results, for both cases, SAIDI and SAIFI, showed that the differential was not significant. In conclusion, the effect of the rate-of-return regulation on the quality of service would be equal to the combination of the pricecap regulation and private ownership. H7 indicated that the utilities under rate-of-return regulation would provide a better quality of service than would the private utilities under model-firm regulation. The dummy variables TIPOWN and RORR required analysis. Table 21 shows that, in the case of the SAIDI, both variables reflected a significant relationship (TIPOWN at 5%, RORR at 1%). In the case of the SAIFI, the TIPOWN and RORR were statistically different from zero (at 1% and 5% respectively). Verifying this result required examination of whether the TIPOWN variable coefficient was equal to the RORR variable coefficient. The Wald test illustrated that for both cases, SAIDI and SAIFI, the combination of marginal effects of private ownership and model firm was statistically equal to zero over the quality of service supplied under rate-ofreturn regulation. Thus, the electric distribution utilities under rate-of-return regulation offered a similar quality of service to the service of private electric distribution utilities under a regulation-by-incentives (price cap or model firm) scheme. This empirical result did not corroborate the theoretical assessment formulated by Spence (1975), Sheshinski (1976), and Kidokoro (2002) and was not consistent with the expected 64

relationship presented in Table 7. Nonetheless, the result was consistent with the outcomes obtained in testing H4 and H5. According to these results, the private utilities under a mechanism of monitoring with sanction would offer a better quality of service than would the electric distribution utilities under a rate-of-return regulation. According to economic literature, a system of incentive regulation combined with a high-powered mechanism of monitoring (with sanction) could be the best option to safeguard the quality of service (Centraal Planbureau, 2004). The results of H6 and H7 were important to show that no major differences existed in the level of quality of service supplied under both regulation schemes (rate of return and regulation by incentives) and type of ownership (private). Furthermore, the results confirm that applying regulation by incentives to private firms accompanied by a mechanism of monitoring with sanctions, in case of nonaccomplishment of standards related to quality of service, contributed to a better quality of service (Centraal Planbureau, 2004). In conclusion, the private firms under regulation by incentives (price cap and model firm) offered a quality of service similar to the quality offered by firms under rate-of-return regulation. Table 23 illustrates the contrast that allowed for testing of the hypotheses of this research and shows the Wald statistic accompanied by its p value. Generally, acceptance of the null hypothesis occurred where the p value was higher than 0.05.
Table 23 Summary of the Wald Test Hypothesis H1 SAIDI H0: pcap = rorr Chi2(1) = 6.23 Prob > chi2 = 0.0126 H0: rorr = 0 Chi2(1) = 10.60 Prob > chi2 = 0.0011 H0: tipofis = 0 Chi2(1) = 8.79 Prob > chi2 = 0.0030 H0: tipown / tipofis = 0 chi2(1) = 14.93 Prob > chi2 = 0.0001 H0: tipown = 0 Chi2(1) = 8.04 Prob > chi2 = 0.0046 H0: rorr = tipown + pcap Chi2(1) = 0.08 Prob > chi2 = 0.7840 H0: rorr = tipown Chi2(1) = 1.83 Prob > chi2 = 0.1764 SAIFI H0: pcap = rorr Chi2(1) = 3.84 Prob > chi2 = 0.0500 H0: rorr = 0 Chi2(1) = 3.71 Prob > chi2 = 0.0542 H0: tipofis = 0 Chi2(1) = 4.00 Prob > chi2 = 0.0454 H0: tipown / tipofis = 0 Chi2(1) = 5.27 Prob > chi2 = 0.0217 H0: tipown = 0 Chi2(1) = 11.25 Prob > chi2 = 0.0008 H0: rorr = tipown + pcap Chi2(1) = 0.06 Prob > chi2 = 0.8035 H0: rorr = tipown Chi2(1) = 0.08 Prob > chi2 = 0.7735

H2

H3

H4

H5

H6

H7

65

4.2 Summary Chapter 4 reflected the empirical analysis of the influence of schemes of regulation, types of monitoring, and forms of ownership on the quality of service. Answering the four research questions involved contrasting seven hypotheses. The purpose of contrasting H1 and H2 was to analyze whether rate-of-return regulation resulted in a better quality of service than did regulation by incentives. The results showed that the RORR variable was significant, so the electric distribution utilities under a rateof-return regulation scheme provided a better quality of service than did utilities under a regulation-by-incentives scheme. The purpose of testing H3 and H4 was to determine whether the mechanism of monitoring with sanction improved the quality of service. H3 involved contrasting the TIPOWN and TIPOFIS dummy variables to compare state-owned utilities with or without sanction. The result showed that state-owned utilities with sanction did not provide a better quality of service than did their peers without sanctions. The scheme of regulation affected the behavior of the Latin American electric distribution utilities. The result was contrary to the expected outcome of this test. The econometric analysis of H4 indicated that the private utilities under the mechanism of monitoring with sanction improved the quality of service compared to state-owned utilities. The result aligned with recent theories illustrating that electric distribution utilities, under a mechanism of monitoring with sanction, provided a better quality of service (Holt, 2004; Lewis & Sappington, 1991; Noam, 1990; Waddams Price et al., 2002). The purpose of testing H5 was to contrast if the private utility would not provide a better quality of service than would the state-owned utility. The result is contrary to the theoretical formulation of incomplete contracts (Hart et al., 1997). The result confirm that measuring, monitoring, and sanctioning related to the quality of service supplied by the Latin American electric distribution utilities can occur, so governments should specify the quality of service within the contract. Furthermore, the outcome illustrated that modification of the behavior of the private firm is possible. The purpose of testing H6 and H7 was to contrast the effect of a rate-of-return regulation scheme with the effect of a regulation-by-incentives scheme (price cap and model firm) applied to private electric distribution utilities. The results of H6 and H7 showed that the utilities under rate-of-return regulation offered a similar quality of service to the quality offered by the private electric distribution utilities under price-cap and model-firm regulation. Thus, the electric distribution utilities under rate-of-return regulation offered a similar quality of service to the service of private electric distribution utilities under a regulation-by-incentives (price cap or model firm) scheme.

66

The contrast that allowed for testing of the hypotheses of this research is summarized in Table 23. The calculus of the test and its respective outcomes for each of the hypotheses, using STATA software, appear in Appendix E. Chapter 5 includes the conclusions of the study and illustrates recommendations based on the findings.

67

5.

Conclusions and Recommendations

The motivation to conduct the study included the lack of empirical work on the influence of regulation, monitoring, and ownership on the quality of service supplied by Latin American electric distribution utilities. The existing literature illustrates that the private electric distribution utilities would reduce the quality of service, which could be why the Latin American population is opposed to privatization. This research study covered the period between 2002 and 2007, which included the monitoring policy of quality applied to utilities under regulation by incentives in the studied countries. One aspect of the research was to determine whether the electric distribution utilities operating under a rate-of-return regulation scheme offered a better quality of service than did the utilities under a regulation-by-incentives scheme. Theoretically, the controlled utilities under a rate-of-return regulation would offer a better quality of service. Another aspect of the research was to establish whether the application of a monitoring policy with sanction improved the levels of quality of service. The final aspect of the research involved examining the influence of the type of ownership, which determines the managers behavior, on the quality of service. Measuring quality of service through quantitative and qualitative variables is possible. Because service continuity is the most important characteristic of the quality of service and measurement of service continuity involves indices, a quantitative method was required for this research. Chapter 5 reflects the conclusions, implications, and recommendations that resulted from the study. 5.1 Conclusions The empirical evidence obtained in Latin America illustrates that rate-of-return regulation provides a better quality of service than does regulation by incentives. The result aligns with the expectations evident in H1 and H2 and with the formulations of Spence (1975) and Sheshisnki (1976). In theory a rate-of-return regulation scheme promotes a better quality of service than does a regulation-byincentives scheme. A policy of quality implemented in Latin America includes an administrative mechanism that authorizes the regulatory body to carry out the monitoring of quality and to implement sanctions against utilities not accomplishing the established quality of service. The result of H4 shows that the private utilities under a monitoring-with-sanction policy provide a better quality of service than do the utilities under a monitoring-without-sanction policy. This finding is important in designing price and quality policy regulation in countries with characteristics similar to Latin America. The outcome is in accordance with the theoretical results expected by Lewis and Sappington (1991), Noam (1990), and Kriehn (2005), who

68

pointed out that a regulation-by-incentives scheme accompanied by a policy of monitoring with sanction would improve the quality of service. The result of H5 shows that the private utilities offer a better quality of service than do the state-owned utilities. The finding is not consistent with the theory of incomplete contracts and indicates that specification of the quality of service is possible in the concession contracts. Therefore, the private utilities can develop activities of distribution with technical and economical efficiency without diminishing the quality of service. The result obtained in H3 illustrates that the state-owned utilities under a monitoring regime without sanction offer a better quality of service than do the state-owned utilities under a monitoring regime with sanction. The outcome is acceptable due to public managers in Latin America make decisions aligned with the formal and explicit requirements imposed for the rules and regulation of public expenditure or the accomplishment of explicit requirements imposed by the regulation policy for the development of the electric distribution activity. The reason for the outcome may be the supposition that public managers exhibit management comparable to that of their peers in private utilities is false. The results of H6 and H7 show that the private utilities controlled under a regulation-by-incentives (price cap or model firm) scheme offer a similar quality of service to that supplied by the state-owned utilities controlled under a rate-of-return regulation scheme. The result aligns with the formulations of Holt (2004), Lewis y Sappington (1991), Noam (1990), and Waddams Price et al., (2002). The obtained result is important because it indicates that complementing the regulation-byincentives scheme applied to the private utilities with a monitoring of quality of service with sanction results in a quality of service comparable with the service supplied by the electric utilities under a rate-of-return regulation scheme. Finally, the policies of regulation and monitoring in the electric sector are tightly related with the quality of service. This assessment could help in the implementation of economic signals to motivate the private or state-owned electric distribution utilities to attain acceptable levels of quality of service, an issue considered key in guaranteeing the sustainability of the service. 5.2 Implications In the decade of the 1990s, the lack of quality-of-service policies in the implementation of privatization affected the quality of service negatively. This research study involved analyzing quality of service between 2002 and 2007, a period that reflects implementation and consolidation of a policy of quality of service through the measurement of minimum standards of quality and the application of sanctions. Regulatory bodies could apply sanctions if the electric utilities did not accomplish the minimum standards of quality of service.

69

The results are consistent with the postulation that the regulation-by-incentives scheme accompanied by a policy of quality results in a quality of service similar to that supplied under the rate-of-return regulation scheme. The finding illustrates that a quality-monitoring approach may modify the signal evident under the regulationby-incentives scheme. In this sense, the argument that the utilities regulated by incentives, in their goals to improve profitability, avoid investments and costs, which reduces the quality of service, weakens in the presence of a quality-of-service monitoring mechanism that includes sanctions. In Latin America, the standards of quality of service have been established through specific regulations. Such regulations indicate the minimum levels of quality of service that electric distribution utilities must supply. To safeguard the attainment of quality of service, the governments delegate the task of monitoring to regulatory bodies. Some regulatory bodies have established procedures to measure and control quality of service, including implementing sanctions when an electric utility does not abide by the requirements. Thus, specifying the quality of service as part of the concession contract is possible. Consequently, the presence of private utilities could become beneficial for the development of the distribution activity under the theoretical supposition that the private utility does not diminish the quality of service when under the scrutiny of a regulatory body. The policy of quality of service is a key factor in the design of the reform (privatization) of a sector with the characteristics of a natural monopoly. Thus, complementing the mechanisms of measurement, control, and sanction with aspects related to the organizational structure and the availability of human resources is beneficial to enable the regulatory body to fulfill its monitoring task efficiently. This implies that is very important to have not only a policy of quality but also an efficient regulator. The fact that state-owned utilities under a regulation-by-incentives scheme do not react to the policies of quality of service implies that the economic signals that impose both price and quality policies were not considered. A solution to reach that public managers behave similar to their private peers may be to adopt policies of public management that let public managers overcome legal-administrative restrictions and operate with a higher autonomy concerning political power. An alternative could be the implementation of corporative governance or further publicprivate partnership. 5.3 Recommendations The first recommendation is to analyze the quality policy critically to revise qualitymonitoring procedures (measurement, control, and sanction) and the design of the distribution tariff to reflect that the allowed income will depend on the supplied quality. Improving the regulation-by-incentives scheme will involve introducing the factor Q (quality), which will indicate explicitly the allowed income related to supplying the target level of quality of service. If the utility attains the targeted 70

quality of service, the utility may maintain its income, but if the quality diminishes, subsequent reductions in the allowed tariff should occur. The second recommendation includes ensuring that the national agenda reflects the appropriate treatment of state-owned electric utilities. This emerges from the fact that state-owned utilities under a regulation-by-incentives scheme do not react to the policies of quality of service. The government could explore the possibility of establishing policies of corporative governance or private-public partnership to protect the utilities against political influence and ensure a higher degree of flexibility related to reviewing the expenses of public control regulations. The third recommendation is to extend the period and population studied to achieve better levels of significance. Calling upon regulators, electric utilities, and international organisms of the sector to construct a sector database containing standardized information on market, infrastructure, quality, income, tariffs, and management may be advantageous. The fourth recommendation involves improving the explained variables of the model used in this research by separating the variable of operations and maintenance into two components: (a) costs and (b) reparations. Addressing these aspects in this study was not possible due to unavailability of the information. The fifth recommendation is to research the influence that the institutional variables of regulation, monitoring, and ownership have on each of the variables of result expected in the model of the reform design: tariffs, quality of service, level of access, and profitability. The sixth recommendation includes developing a new econometric test for the analysis of unbalanced panel data to solve the problems of both autocorrelation and heteroscedasticity. The final recommendation is to replicate the study in other public service sectors and regions of the world in an attempt to confirm the findings of this research.

71

References Agrell, P., Bogetoft, P., & Tind, J. (2005). DEA and dynamic yardstick competition in Scandinavian electricity distribution. Journal of Productivity Analysis, 23(2), 173-201. Ai, C., & Sappington, D. E. (1998). The impact of state incentive regulation on the U.S. telecommunications industry. Gainesville: University of Florida, Department of Economics. Ai, C., & Sappington, D. E. (2002). The impact of state incentive regulation on the U.S. telecommunications industry. Journal of Regulatory Economics, 22(2), 133-159. Ai, C., & Sappington, D. E. (2005). Reviewing the impact of incentive regulation on U.S. telephone service quality. Utilities Policy, 13(3), 201-210. Ajodhia, V., & Hakvoort, R. (2005). Economic regulation of quality in electricity distribution networks. Utilities Policy, 13(3), 211-221. Alva, S., & Bonifaz, L. (2004). Eficiencia relativa en el servicio de distribucin elctrica en el Per durante el periodo 1997-2000: Un estudio de fronteras. En: Fernndez-Baca, Jorge (comp.) Experiencias de regulacin en el Per [Relative efficiency in the electrical distribution service in Peru for the period 1997-2000: A frontier study. In J. Fernndez-Baca (Ed.), Experiences of Regulation in Peru (pp. 143-189)]. Lima, Peru: Universidad del Pacifico. Andres, L., Foster, V., & Guash, L. (2006). The impact of privatization on the performance of the infrastructure sector: The case of electricity distribution in Latin America (WP No. 3936). Washington, D.C.: World Bank Policy Research. Arnau, A., Mocarquer, S., Rudnick, H., & Voscoboinik, E. (2007). Stimulating efficient distribution. IEEE Power & Energy Magazine, 5(4), 50-67. Atkinson, S., & Halvorsen, R. (1986). The relative efficiency of public and private firms in a regulated environment: The case of U.S. electric utilities. Journal of Public Economics, 29(3), 281-294. Averch, H., & Johnson, L. (1962). Behaviour of the firm under regulatory constraint. American Economic Review, 52(5),1053-1069. Bagdadioglu, N., Waddams, C., & Weyman-Jones, T. (1996). Efficiency and ownership in electricity distribution: A non-parametric model of the Turkish experience. Energy Economics, 18(1-2), 1-23. 72

Baker, B., & Trmolet, S. (2000). Utility reform: Regulating quality standards to improve access for the poor. Public Policy for the Private Sector, 219(10), 1-4. Baldwin, C., & Cave, M. (1999). Understanding regulation. Oxford, England: Oxford University Press. Baltagi, D., & Wu, P. (1999). Unequally spaced panel data regressions with AR (1) disturbances. Econometric Theory, 15(6), 814-823. Baron, D. (1981). Price regulation, product quality, and asymmetric information. The American Economic Review, 71(1), 212-220. Berg, S. (2001). Infrastructure regulation: Risk, return, and performance. Gainesville: University of Florida, Public Utility Research Center. Berg, S. (2005). Public policy. Gainesville: University of Florida, Public Utility Research Center. Berg, S. (2006, November). Institutional requirements for second-generation infrastructure reform: Processes and performance in developing countries. Paper presented at the seminar Regulation Reform and Regulators, Lima, Peru. Brown, R. (2002). Electric power distribution reliability. New York: Marcel Dekker. Burke, J. (1994). Power distribution engineering. New York: Marcel Dekker. Burns, R. E. (2003, December). The impact of price caps and rate freezes on service quality. Paper presented at the Thirty-Third Plenary Session of the Harvard Electricity Policy Group, Point Clear, Alabama. Cameron, C., & Trivedi, P. (2005). Microeconometrics: Methods and applications. Cambridge, England: Cambridge University Press. Centeno, C., & Serra, E. (2007). Cambios en la gestin de las distribuidoras elctricas en busca de la eficiencia [Changes in the electric distribution management for reaching effiency]. Lima, Peru: OSINERGMIN. Centraal Planbureau. (2004). Better safe than sorry? Reliability policy in network industries. La Haya, Netherlands: Author. Clements, M. (2001). Trends in local telephone quality service. Quarterly Bulletin, 21(2), 123-130. 73

Comisin de Integracin Energtica Regional. (2007). Informe sectorial energtico: Informaciones de los sectores electricidad y gas Natural en la regin [Report of the power sector: Electricity and natural gas data in the region]. Montevideo, Uruguay: Author. Costas, A. (2006). Regulacin y calidad de los servicios pblicos liberalizados [Regulation and quality of liberalized public service] (Paper No. 2). Barcelona, Spain: University of Barcelona, Department of Political Economy. Cowan, S. (2006). Network regulation. Oxford Review of Economic Policy, 22(2), 248-259. Dewenter, K., & Malatesta, P. (2001). State-owned and privately owned firms: An empirical analysis of profitability, leverage, and labor intensity. The American Economic Review, 91(1), 320-334. Edvardsen, D., & Frsund, F. (2002). International benchmarking of electricity distribution utilities. Resource and Energy Economics, 25(4), 353-371. Estache, A., Guasch, J. L., & Trujillo, L. (2003). Price caps, efficiency pay-offs and infrastructure contract renegotiation in Latin America. (WP No. 3129). Washington, D.C.: The World Bank. Estache, A., & Rossi, M. (2004). Have consumers benefited from the reforms in the electricity distribution sector in America? Washington, DC: World Bank. Farsi, M., & Filippini, M. (2004). Regulation and measuring cost efficiency with panel data models: Application to electricity distribution utilities. Journal of Review of Industrial Organization, 25(1), 1-19. Fisher, R., Gutierrez, R., & Serra, P. (2003). The effects of privatization on firms and on social welfare: The Chilean case (Occasional Paper No. R-456). Washington, DC: The GW Centre for the Study of Globalization. Fraser, R. (1994). Price, quality and regulation: An analysis of price capping and the reliability of electricity supply. Journal of Energy Economics, 16(3), 175-183. Garvin, D. (1984). What does product quality really mean? Journal of Sloan Management, 26(1), 25-43. Giacchino, J., & Lesser, L. (2007). Fundamentals of energy regulation. Vienna, VA: Public Utilities Reports.

74

Gmez, T., & Rothwell, G. (2003). Electricity economics: Regulation and deregulation. Piscataway, NJ: IEEE Press. Gnen, T. (1986). Electric power distribution system engineering. New York: McGraw-Hill. Greene, W. (2000). Econometric analysis. New York: Prentice Hall. Growitsch, C., Jamasb, T., & Pollitt, M. (2005). Quality of service, efficiency and scale in network industries: An analysis of European electricity distribution. Cambridge, England: University of Cambridge. Guash, L., & Spiller, P. (1999). Managing the regulatory process: Design, concepts, issues, and the Latin America and Caribbean story. Washington, DC: World Bank. Hart, O., Shleifer, A., & Vishny, R. W. (1997). The proper scope of government: Theory and an application to prisons. The Quarterly Journal of Economics, 112(4), 1127-1161. Hinds, S., Sanchez, N., & Schap, D. (1991). Public enterprise: Retrospective review and prospective theory. In Backhaus & Wagner. Handbook of Public Finance (pp. 277-300). NY: Kluwer Academic Publisher. Hllriegl, H. (2007, May). Incentive regulation and investment behavior in electricity industries (Discussion Paper, 1-10). Holmstrom, B., & Milgrom, P. (1991). Multitask principal-agent analyses: Incentive contracts, asset ownership, and job design. Journal of Law, Economics and Organization, 7, 24-52. Holt, L. (2004). Utility service quality: Telecommunications, electricity, water. Journal of Utilities Policy, 13(3), 189-200. Honkapuro, S., Lasilla, J., Partanen, J., Tahvanainen, K., &Viljainen, S. (2004). Effects of benchmarking of electricity distribution companies in Nordic countriesComparison between different benchmarking methods. Retrieved July, 31, 2007, from Helsinki University of Technology Web site: http://powersystems.tkk.fi/nordac2004/papers/nordac2004_honkapuro_et_ al_paper.pdf Jamasb, T., Motta, R., Newbery, D., & Pollitt, M. (2005). Electricity sector reform in developing countries: A survey of empirical evidence on determinants and performance. Washington, DC: World Bank.

75

Jamasb, T., & Pollitt, M. (2007). Incentive regulation and benchmarking of electricity distribution networks: From Britain to Switzerland. Cambridge, England: University of Cambridge. Joskow, P. (2006). Incentive regulation in theory and practice: Electricity distribution and transmission networks. Cambridge, MA: MIT, Center for Energy and Environmental Policy Research. Kahn, A. (1988). The economics of regulation: Principles and institutions. Cambridge, MA: MIT Press. Kessides, I. N. (2004). Reforming infrastructure: Privatization, regulation and competition. Washington, DC: World Bank. Kidokoro, Y. (2002). The effects of regulatory reform on quality. Journal of the Japanese and International Economies, 16(1), 135-146. Kikeri, S., & Nellis, J. (2004). An assessment of privatization. The World Bank Research Observer, 19(1), 87-118. Kriehn, C. (2005). Quality regulation without regulating quality. Berln, Germany: Ifo Institute for Economic Research. Kwoka, J. (2005). The comparative advantage of public ownership: Evidence from U.S. electric utilities. Canadian Journal of Economics, 38(2), 622-640. Laffont, J., & Tirole, J. (1993). A theory of incentives in procurement and regulation. Cambridge, MA: The MIT Press. Lewis, T., & Sappington, D. (1991). Incentives for monitoring quality. RAND Journal of Economics, 22(3), 370-384. Lewis, T., & Sappington, D. (1992). Incentives for conservation and qualityimprovement by public utilities. American Economic Review, 82(5), 13211340. Littlechild, S. (2003). Reflections on incentive regulation. Journal of Review of Network Economics, 2(4), 289-315. Lpez, R. (2007). Electricity distribution: Is quality really regulated or just given? Pars: University of Paris. Macedo, L. (2004). Regulation, competition and ownership in electricity distribution companies: The effects on efficiency. London: City University.

76

Megginson, W., & Netter, J. M. (2000). From the state to the market: A survey of empirical studies on privatization. Journal of Economic Literature, 39(2), 321-389. Meyrick & Associates. (2002). Electricity service quality incentives scoping paper. Canberra, Australia: Queensland Competition Authority. Mikkers, M., & Shestalova, V. (2003). Yardstick competition and reliability of supply in public utilities. London: University of Bath. Millan, J. (2006). Entre el Mercado y el Estado. Tres dcadas de reformas en el Sector Elctrico de Amrica Latina [Between market and state. Three decades of reform in the Latin American electrical sector]. Washington, DC: Banco Interamericano de Desarrollo. Nellis, J. (2004). Privatization in Latin America. Washington, DC: World Bank, Center for Global Development. Neuberg, L. (1977). Two issues in the municipal ownership of electric power distribution systems. The Bell Journal of Economics, 8(1), 303-323. Noam, E. (1990). The quality of regulation in regulating quality: A proposal for an integrated incentive approach to telephone service performance. In M. Einhorn (Ed.), Price caps and incentive regulation in telecommunications (pp. 167-189). Boston: Kluwer. Paliza, R. (1999). Impacto de las privatizaciones en el Per [Impact of privatization in Peru]. Revista de Estudios Econmicos, 4, 9-37. Panizza, H., & Yaez, M. (2006). Why Are Latin Americans so unhappy about reforms? Journal of Applied Economics, 8(5), 1-29. Parker, D., & Kirkpatrick, C. (2005). Regulating prices and profits in utility industries in low-income economies: Rate of return, price cap or slidingscale regulation? International Journal of Public Sector Management, 18(3), 241-255. Perotti, E. (2004). State ownership: A residual role? (WP No. 3407). Washington, DC: World Bank. Pinter, J., & ReKettye, G, (2005). Quantitative methods and models in measuring customer satisfaction in the electricity industry. Sixteenth Annual Conference of POMS, April, 2005.Chicago, MI. Pollitt, M. G. (1995). Ownership and performance in electric utilities. Oxford, England: Oxford University Press. 77

Pollitt, M. G. (1997). The impact of liberalization on the performance of the electricity supply industry: An international survey. The Journal of Energy Literature, 3(2), 3-31. Pollitt, M. G. (2000). The declining role of the state in infrastructure investments in the UK (WP No. 001). Cambridge, England: University of Cambridge, Judge Institute of Management Studies. Pollitt, M. G. (2005). Electricity reform in Chile: Lessons for developing countries (WP No. 51). Cambridge, England: University of Cambridge, Department of Applied Economics. Purdy, M. (1997). The performance of privatized industries. Journal of Consumer Policy Review, 7(3), 115-116. Rainieri, R., & Rudnick, H. (1997). Analysis of service quality standards for distribution firms. In F. G. Morand & R. Rainieri (Eds.), (De)regulation and competition: The electric industry in Chile (pp. 259-294). Washington, DC: Ilades/Georgetown University. Renzetti, S., & Dupont, D. (2003). Ownership and performance of water utilities. The Journal of Corporate Environmental Strategy and Practice (GMI), 42(3), 9-19. Rodrguez Pardina, M., Rossi, M., & Ruzzier, C. (1998). Fronteras de eficiencia en el sector de distribucin de energa elctrica: La experiencia Sudamericana [Efficiency frontiers in the power distribution sector: The South American experience]. Retrieved May, 24, 2007, from site: http://www.aaep.org.ar/espa/anales/resumen_98/rodriguezpardina_rossi_ruzzier.htm Rossi, M., & Ruzzier, C. (2002). On the regulatory application of efficiency measures. Journal of Utilities Policy, 9(2), 81-92. Roycroft, T., & Garcia-Murrilo, M. (2000). Trouble reports as an indicator of service quality: The influence of competition, technology, and regulation. Journal of Telecommunications Policy, 24(10), 947-967. Rudnick, H., & Zolezzi, J. (2001). Electric sector deregulation and restructuring in Latin America: Lessons to be learnt and possible ways forward. IEE Proceedings Generation, Transmission and Distribution, 148(2), 180-184. Salvanes, K., & Tjtta, S. (1998). A test for natural monopoly with application to Norwegian electricity distribution. Journal of Review of Industrial Organization, 13(6), 669-685. 78

Sappington, D. E. (2005). Regulating service quality: A survey. Journal of Regulatory Economics, 27(2), 123-154. Serra, E. (2008). Mejoramiento de la eficiencia en la gestin de las distribuidoras elctricas estatales [Improvement in the efficiency of state-owned electrical distribution utilities management]. Lima, Peru: OSINERGMIN. Sheshinski, E. (1976). Price, quality and quantity regulation in monopoly situations. Journal of Economics, 43(17), 127-137. Shirley, M. (2004). Why is reform so unpopular in Latin America? (WP No. 4). St. Louis, Missouri: Ronald Coase Institute. Shirley, M., & Walsh, P. (2000). Public versus private ownership: The current state of the debate. Washington, DC: World Bank. Shleifer, A. (1985). A theory of yardstick competition. Journal of Economics, 16(3), 319-327. Shleifer, A. (1998). State versus private ownership. Journal of Economic Perspectives, 12(4), 133-150. Shleifer, A., & Vishny, R. (1994). Politicians and firms. Quarterly Journal of Economics, 109(4), 995-1025. Spence, A. M. (1975). Monopoly, quality and regulation. The Bell Journal of Economics, 6(2), 417-429. Tangers, T. (2002). Regulation of cost and quality under yardstick competition (WP No. 573). Stockholm: The Research Institute of Industrial Economics. Tangers, T. (2008). Yardstick competition and quality. Journal of Economics & Management Strategy, 18(2), 589-613. Ter-Martirosyan, A. (2003). The effects of incentive regulation on quality service in electricity markets. (WP). Washington D.C.: George Washington University. Trochim, W. (2001). The research methods knowledge base. Cincinnati, OH: Atomic Dog. Urbiztondo, S. (2000, August). La regulacin de la calidad en el servicio elctrico: Una evaluacin en base a principios tericos y la experiencia internacional [The quality of service regulation: An analysis based on the theoretical principles and international experience]. Paper presented at the 79

annual meeting of the Argentinean Association of Political Economy, Crdoba, Argentina. Uri, N. (2001). Technical efficiency, allocative efficiency, and the implementation of a price cap plan in telecommunications in the United States. Journal of Applied Economics, 4(1), 163-186. Verbeek, M. (2000). A guide to modern econometrics. West Sussex, England: John Wiley & Sons. Waddams Price, C., Brigham, B., & Fitzgerald, L. (2002). Service quality in regulated monopolies (WP No. CCR 02-4). Norwich, England: Centre for Competition & Regulation. Wooldridge, J. (2001). Introduccin a la econometra: Un enfoque moderno [Introductory econometrics: A modern approach]. Mxico: Thomson Learning Ibero. Yarrow, G. (1986). Privatization in theory and practice. Journal of Economic Policy, 2(4), 324-377.

80

Appendix A: Informed Consent Form Lima, dd mm yy Mr. XXXX Representatives XXXX Lima Peru I agree to participate in the research entitled Influence of Regulation, Monitoring, and Ownership on the Quality of Service of Latin American Electric Distribution Utilities from 2002 to 2007, conducted by Miguel Revolo under the supervision of Dr. Esteban Serra. I understand that participation is entirely voluntary; I can withdraw my consent at any time without penalty and have the results of the participation, to the extent that they can be identified as mine, returned to me, removed from the research records, or destroyed. The following points have been explained to me: 1. 2. At this stage of the research, the aim is to determine which factors/attributes would explain the diminishing quality of service. The benefit that I may expect from this research is to have access to the results that illustrate the factors/attributes that influence the level of the quality of service. Participation will involve completing the questionnaire using official information related to the utility in terms of market, installations, costs, and index of performance. I understand that I may refuse to answer any question and may discontinue participating at any time. No discomfort or stress is foreseen. No risks are foreseen. The information I provide will be kept confidential and will not be released in any identifiable form without my prior consent.

3.

4. 5. 6.

The investigator will answer any further questions about the research, now or during the course of the project. I can contact the researcher at (00 511) 224 0487 or e-mail the researcher at a20029973@pucp.edu.pe and mrevolo@osinerg.gob.pe

Name of investigator

Name of participant

Signature of investigator, Date

Signature of participant, Date

PLEASE SIGN BOTH COPIES OF THIS FORM. KEEP ONE, AND RETURN THE OTHER TO THE INVESTIGATOR.

81

Appendix B: Research Questionnaire 1. Preliminary Aspects The purpose of this questionnaire is to collect information concerning the national electric distribution industry from the following Latin American countries: Argentina, Brazil, Bolivia, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, Panama, Paraguay, Peru, Uruguay, and Venezuela. The focus of the information is on the electric distribution activity but may include activities in generation and transmission. The information, required annually, will relate to a 6-year period: 2002-2007. 2. General Information Name of country: Name of utility: Name of informer: Date: 3. Market Information Market information is required for the period of 2002 to 2007, corresponding to the following variables: number of consumers, annual energy sales, and annual amount of sales (US$ M) for medium-tension (MT: > 1 KV y < 30KV) and low-voltage (LV: < 1 KV) networks. Please complete Table 1.
Table 1 Commercial Information MT: Energy sales (MWh) LV: Energy sales (MWh) MT: Amount of sales (US$ M) LV: Amount of sales (US$ M)

Year 2002 2003 2004 2005 2006 2007

MT: No. consumers

LV: No. consumers

82

4.

Technical Information The technical information of the installments of the distribution systems is required for the period of 2002 to 2007, corresponding to the following variables: length of aerial network for both MT and LV, length of underground network for MT and LV, number of switches and number of MT/LV transformers, and capacity of MT/LV transformers. Please complete Table 2.

Table 2 Technical Information of Distribution Installments Posts of cutting and protection network Transformers MV/LV

Medium voltage Length of underLength ground of aerial network network Year (km) (km) 2002 2003 2004 2005 2006 2007

Low voltage Length of Length underof aerial ground network network (km) (km)

1 Phase

2 Phases

3 Phases

No.

Capacity (MVA)

5.

Quality of Service Information The quality of service information of the installments of the distribution systems is required for the period of 2002 to 2007, corresponding to the following variables: system average interruption duration index (SAIDI) and system average interruption frequency index (SAIFI). Please complete Table 3.

Table 3 Indices of Quality of Service SAIFI (no. interruptions/year) SAIDI (hrs interruption/year)

Year 2002 2003 2004 2005 2006 2007

83

6.

Financial and Cost Information The financial and cost information of the distribution utility is required for the period of 2002 to 2007, corresponding to the following variables: fixed assets, depreciation, fixed net assets, distribution operation cost, distribution maintenance cost, commercialization cost, and general expenses. Please complete Table 4.

Table 4 Financial and Cost Information Capital cost (US$ M) Fixed net Fixed assets assets Depreciation Distribution exploitation cost (US$ M) Commercialization General expenses

Year 2002 2003 2004 2005 2006 2007

Operation Maintenance

7.

Regulatory and Monitoring Information The regime of regulation and mechanism of monitoring information applied to set prices and to monitor the activity of distribution is required for the period of 2002 to 2007, corresponding to the following variables: rate-of-return regulation (cost-plus regulation), price-cap regulation, and model-firm regulation (price cap and yardstick). Please report on the monitoring using the following categories: high powered, medium powered, and low powered. High-powered monitoring involves the regulator measuring and fining the utility according to the level of its quality of service. Medium-powered monitoring involves the regulator being able to measure but not fine. Lowpowered monitoring involves the regulator having an unimplemented standard of quality of service or having no standard. Please complete Table 5.

84

Table 5 Regulatory and Monitoring Information Regulatory scheme Model firm (price cap + yardstick) Monitoring mechanism

Year 2002 2003 2004 2005 2006 2007

Rate of return

Price cap

High powered

Medium powered

Low powered

8.

Ownership Information The ownership information of the distribution utility is required for the period of 2002 to 2007, corresponding to the following variables: Private firm European, American (north, central, or south), or national firm owned by private investorsand state-owned firmowned by the government. Please complete Table 6 by using an X to mark the appropriate column per year.

Table 6 Ownership Information Ownership Year 2002 2003 2004 2005 2006 2007 Private State-owned

9.

Additional Information Additional information related to the distribution utility is required for the period of 2002 to 2007, corresponding to the following variables: number of employees, percentage of energy losses, exchange rate of local currency, and the weather (use an X to mark the relevant weather condition). Please complete Table 7.

85

Table 7 Additional Information Other indices Rate of exchange (local currency per US$1) Weather

Year 2002 2003 2004 2005 2006 2007

No. employees

% energy losses

Rainy

Warm

Sunny

Please forward the completed research questionnaire to the following e-mail addresses: a20029973@pucp.edu.pe and mra1988@yahoo.com

Comments:

86

Appendix C: Sources of Information 1. Argentina Primary information resulted from the completed research questionnaires received from the key authorities of the main distribution utilities. Verification of the information occurred through checking the annual reports available on the utilities websites. Another source of information was the annual reports (2002-2007) from de Asociacin de Distribuidores de Energa Elctrica de la Repblica de Argentina (ADEERA) [Electricity Distribution Association of the ArgentineRepublic]. Further sources included the financial reports published on the website of the Comisin Nacional de Valores de Argentina (CNV) [Argentina National Commission of Values]. 2. Bolivia The completed research questionnaires from the electric distribution utilities yielded the data. Confirming the information involved contrasting the data with statistical information published by the Superintendecia de Electricidad (SIE) [Superintendence of Electricity]. 3. Brazil Data collection resulted in completed research questionnaires from seven distribution utilities. In addition, an Agencia Nacional de Energa Elctrica (ANEEL) [National Bureau of Electrical Energy] representative completed the research questionnaire and provided access to the website. The annual reports and statistical information published on the websites of 18 electric distribution utilities provided further information. Validation of all the information occurred through comparison with the information published by the Asociacin Brasilea de Distribuidores de Energa Elctrica (ABRADEE) [Brazilian Association of Electrical Distribution Utilities] and the information published on the ANEEL website. 4. Chile The completed research questionnaires from the electric distribution utilities yielded the data. Further information resulted from websites and printed annual reports. Validation of the information involved comparison to data and statistical information published by the Comisin Nacional de Energa (CNE) [National Commission of Energy].

87

5.

Ecuador Primary information resulted from the completed research questionnaire of the Consejo Nacional de Electricidad (CONELEC) [National Council of Electricity]. In addition, the statistics of the electrical sector published by CONELEC complemented the data and information received. The contribution of the specialists of CONELEC to validate the collected data was integral.

6.

El Salvador Primary information emerged through the completed research questionnaire of the Superintendecia General de Electricidad y Telecomunicaciones (SIGET) [General Superintendence of Telecommunication and Electricity]. Validating the information sent by SIGET involved using the data in the report Istmo Centroamericano: Estadsticas del Subsector Elctrico [Centro-American Isthmus: Statistics of the Electric Sector] published by the Comisin Econmica para Amrica Latina y el Caribe (CEPAL) [Economic Commission for the Caribbean and Latin America].

7.

Guatemala A private distribution utility reported information through the research questionnaire. Contrasting of the data with the statistics published by the Comisin Nacional de Energa Elctrica (CNEE) [National Commission of Electrical Energy] occurred. In addition, verifying the data and information involved consulting the data of the report Istmo Centroamericano: Estadsticas del Subsector Elctrico [Centro-American Isthmus: Statistics of the Electric Sector] published by CEPAL.

8.

Paraguay The information emerged from electronic and printed annual reports (2002-2007) published by Administracin Nacional de Electricidad (ANDE) [National Administration of Electricity]. Nonetheless, some unpublished data were specifically asked to ANDE to complete the information.

9.

Peru The information was evident in the technical and commercial database of the Tariff Regulation Office of Organismo Supervisor de la Inversin en Energa y Minera (OSINERGMIN) [Supervisory Agency of Investment in Mining and Energy]. The information corresponded to validated information published in the annual report and in the technical reports that 88

support the regulation of the new replacement value of the distribution electrical installations and its corresponding valuation of new and retired installations. 10. Uruguay La Administracin Nacional de Usinas Elctricas de Uruguay [National Administration of Uruguay Power Station] was the main source of technical, commercial, and economic information. Another source was the Unidad Reguladora de Servicios de Energa y Agua (URSEA) [Regulatory Body of Water and Energy Services]. Overall, the World Bank website that contains the benchmarking data of the electric distribution sector in the Latin America and Caribbean Region between 1995 and 2005 proved useful.

89

Table C1 Electrical Distribution Utilities Country Argentina Argentina Argentina Bolivia Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Acronym EDELAP EDENOR EDESUR ELECTROPAZ S.A. AMPLA BANDEIRANTE CAIUA CELB CELPA CELPE CELTINS CEMAT COELBA COELCE COPEL COSERN CPFL CPFLP EEB ELEKTRO Empresa Distribuidora La Plata Empresa Distribuidora y Comercializadora Norte S.A. Empresa Distribuidora Sur S.A. Electricidad de la Paz S.A. Ampla Energa S.A. Bandeirante Energa S.A. Caiu Distribuio de Energia S.A. Companhia Energetica da Borborema S.A. Centrais Eltricas do Para S.A. Companhia Energtica de Pernambuco Companhia de Energia Eltrica do Estado do Tocantins Centrais Eltricas Matogrossenses S.A. Companhia de Eletricidade do Estado da Bahia Companhia Energtica do Cer Companhia Paranaense de Energa Companhia Energtica do Rio Grande do Norte Companhia Paulista de Fora e Luz Companhia Piratininga de Fora e Luz Piratininga Empresa Eltrica Bragantina S.A. Elektro Eletricidade e Servio Utility Website http://www.edelap.com.ar/ http://www.edenor.com.ar/edenor/index.jsp http://www.edesur.com.ar/index2.htm http://www.electropaz.com.bo/ http://www.ampla.com/ http://www.bandeirante.com.br/energia/ http://www.gruporede.com.br/caiua/ http://www.borborema.energisa.com.br/ http://www.gruporede.com.br/celpa/ http://www.coelba.com.br http://www.gruporede.com.br/celtins/ http://www.gruporede.com.br/cemat/ http://www.coelba.com.br http://www.coelce.com.br/ http://www.copel.com/hpcopel/root/index.jsp http://www.cosern.com.br/ http://www.cpfl.com.br/paulista http://www.cpfl.com.br/piratininga http://www.gruporede.com.br/bragantina/ http://www.elektro.com.br/

90

Brazil Brazil Brazil Brazil Brazil Chile Chile Chile Chile Chile Chile Chile

ELETROPAULO ENERGIPE ENERSUL ESCELSA SAELPA CGE CHILECTRA CHILQUINTA CONAFE ELECDA EMELECTRIC FRONTEL

Metropolitana Eletricidade de Sao Paulo Empresa Energtica de Sergipe S.A. Empresa Energtica de Mato Grosso do Sul S.A. Esprito Santo Centrais Eltricas S/A Sociedade Annima de Eletrificao da Paraba Compaa General de Electricidad S.A. Chilectra S.A. Chilquinta Energa S.A. Compaa Nacional de Fuerza Elctrica S.A. Empresa Electrica de Antofagasta S.A. EMELECTRIC Empresa Elctrica de la Frontera S.A.

http://www.eletropaulo.com.br/ http://www.energipe.com.br/ http://www.enersul.com.br/ http://www.escelsa.com.br/energia/ http://www.paraiba.energisa.com.br/ http://www.cgedistribucion.cl/ http://www.chilectra.cl http://www.chilquinta.cl/ http://www.conafe.cl/ http://www.emel.cl/realinfo40lector/home.do?sitio=1 http://www.emel.cl/realinfo40lector/home.do?sitio=1 http://portal.saesa.cl:7778/portal/page?_ pageid=230,415173&_dad=portal&_schema = PORTAL&_requestedpageid=PAG_INICIO_FRO http://portal.saesa.cl:7778/portal/page?pageid= 213,414598&_dad=portal&_schema=PORTAL http://www.categ-sg.com/ http://www.centrosur.com.ec/ http://www.eeq.com.ec/ http://www.eersa.com.ec/eersa.php http://www.emelgur.com.ec/Site/home.php http://www.emeloro.gov.ec/ http://www.aeselsalvador.com/ http://www.aeselsalvador.com/

Chile Ecuador Ecuador Ecuador Ecuador Ecuador Ecuador

SAESA CATEG D CENTROSUR EEQ EERSA GUAYAS-LOS ROS ORO

Sociedad Austral de Electricidad S.A. Corporacin para la Administracin Temporal Elctrica de Guayaquil Empresa Elctrica Centro Sur Empresa Elctrica Quito Empresa Elctrica Riobamba S.A. Empresa Elctrica Guayas Los Ros S.A. Empresa Elctrica El Oro S.A. Compaa de Alumbrado elctrico de San Salvador S.A. Compaa de Alumbrado elctrico de Santa Ana S.A.

El salvador CAESS El salvador CLESA

91

El salvador DELSUR El salvador EEO Guatemala Paraguay Peru Peru Peru Peru Peru Peru Peru Peru Peru Peru Peru Peru Uruguay EEGSA ANDE EDLN ELC ELN ELNM ELOR ELPU ELS ELSE ELSM ENOSA LDS SEAL UTE

Distribuidora de Electricidad del Sur S.A. Empresa Elctrica de Oriente S.A. Empresa Elctricas de Guatemala S.A. Administracin Nacional de Electricidad Empresa de Distribucin Elctrica de Lima Norte Electro Centro S.A. Electro Norte Electro Norte Medio S.A. - Hidrandina S.A. Electro Oriente S.A. Electro Puno S.A.A. Electro Sur S.A. Electro Sur Este S.A. Electro Sur Medio S.A. Electro Nor Oeste S.A. Luz del Sur S.A. Sociedad Elctrica del Sur Oeste S.A. Administracin Nacional de Usinas elctricas de Uruguay

http://www.delsur.com.sv/ http://www.aeselsalvador.com/ http://www.eegsa.com/ http://www.ande.gov.py/ http://www.edelnor.com.pe/principal2.htm http://www.distriluz.com.pe/electrocentro http://www.distriluz.com.pe/ensa/ http://www.distriluz.com.pe/hidrandina/ http://www.elor.com.pe/?intro=ok http://www.electropuno.com.pe/ http://www.electrosur.com.pe/website/ http://www.else.com.pe/ http://www.surmedio.com.pe/ http://www.distriluz.com.pe/enosa http://www.luzdelsur.com.pe/ http://www.seal.com.pe/ http://www.ute.com.uy/

92

Table C2 List of Organizations and Regulatory Bodies Country Argentina Name Asociacin de Distribuidores de Energa Elctrica de la Repblica Argentina [Electricity Distribution Association of the Argentine Republic] Ente Nacional Regulador de la Electricidad [National Regulatory Body of Electricity] Superintendencia de Electricidad [Superintendence of Electricity] Instituto Nacional de Estadstica/ Servicios Pblicos [National Institute of Public Services Statistics] Agncia Nacional de Energia Eltrica [National Bureau of Electrical Energy] Associao Brasileira de Distribuidores de Energia Eltrica [Brazilian Association of Electrical Distribution Utilities] Comisin Nacional de Energa [National Commission of Energy] Consejo Nacional de Electricidad [National Council of Electricity] Centro Nacional de Control de Energa [National Center of Power Control] Superintendecia General de Electricidad y Telecomunicaciones [General Superintendence of Telecommunication and Electricity] Consejo Nacional de Energa Elctrica [National Commission of Electrical Energy] Administracin Nacional de Electricidad [National Administration of Electricity] Organismo Supervisor de la Inversin en Energa y Minera [Supervisory Agency of Investment in Mining and Energy] Unidad Reguladora de Servicios de Energa y Agua [Regulatory Body of Water and Energy Services] Website http://www.adeera.com.ar/

Argentina

http://www.enre.gov.ar/

Bolivia Bolivia

http://www.superele.gov.bo http://www.ine.gov.bo/indice/general .aspx?codigo=40109 http://www.aneel.gov.br/ http://www.abradee.org.br

Brazil Brazil

Chile Ecuador Ecuador El Salvador

http://www.cne.cl/ http://www.conelec.gov.ec/ http://www.cenace.org.ec/ http://www.siget.gob.sv

Guatemala

http://www.cnee.gob.gt/

Paraguay Peru

http://www.ande.gov.py http://www.osinerg.gob.pe/

Uruguay

http://www.ursea.gub.uy/

Table C3 List of International Organizations Description Comisin Econmica para Amrica Latina (CEPAL) [Economic Commission for the Caribbean and Latin America] Banco Mundial [World Bank] Organizacin Latinoamericana de Energa (OLADE) [Latin American Organization of Energy] Comisin de Integracin Energtica Regional (CIER) [Regional Energy Integration Commission] Website http://www.eclac.cl/ http://www.worldbank.org/ http://www.olade.org.ec http://www.cier.org.uy/

93

Appendix D: Distribution Frequency and Q-Q Graphs of the Variables

Figure D1. Distribution frequencies of the SAIDI variable.

Figure D2. Q-Q graph of the SAIDI variable.

94

Figure D3. Distribution frequencies of the SAIFI variable.

Figure D4. Q-Q graph of the SAIFI variable.

95

Figure D5. Distribution frequencies of the LA variable.

Figure D6. Q-Q graph of the LA variable.

96

Figure D7. Distribution frequencies of the SKM variable.

Figure D8. Q-Q graph of the SKM variable.

97

Figure D9. Distribution frequencies of the SED variable.

Normal Q-Q Plot of SED

Expected Normal

-1

-2

-3 -3 0 3 6 9 12

Observed Value

Figure D10. Q-Q graph of the SED variable.

98

Figure D11. Distribution frequencies of the OM variable.

Figure D12. Q-Q graph of the OM variable.

99

Figure D13. Distribution frequencies of the MERCADO variable.

Figure D14. Q-Q graph of the MERCADO variable.

100

Figure D15. Distribution frequencies of the PROD variable.

Figure D16. Q-Q graph of the PROD variable.

101

Figure D17. Distribution frequencies of the PERD variable.

Figure D18. Q-Q graph of the PERD variable.

102

Appendix E: Stata Reports

log: C:\Program Files\Stata9\apendice_e.log log type: text opened on: 28 Sep 2009, 10:18:28 . . . . . . . . *----------------------------------------------------------------------------* APENDICE E *----------------------------------------------------------------------------*----------------------------------------------------------------------------* Tabla N 11 Regresin general *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis regu, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042 om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524 prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 tipofis | 9.606753 6.864056 1.40 0.162 -3.846549 23.06006 regu | 5.748643 2.58658 2.22 0.026 .6790393 10.81825 _cons | 7.356476 4.764111 1.54 0.123 -1.98101 16.69396 -------------+---------------------------------------------------------------sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis regu, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544 tipofis | 8.286437 4.618635 1.79 0.073 -.7659212 17.33879

103

regu | -.1711495 1.429165 -0.12 0.905 -2.972262 2.629963 _cons | 10.17481 4.001961 2.54 0.011 2.331115 18.01851 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. *----------------------------------------------------------------------------. * Tablas N 12 y 13 Matrices de correlaciones . *----------------------------------------------------------------------------. corr saidi la skm sed om mercado prod perd ls1 tipofis tipown regu rorr pcap (obs=309)
| saidi la skm sed om mercado prod perd ls1 tipofis tipown regu rorr pcap -------------+---------------------------------------------------------------------------------------------------------------------------saidi | 1.0000 la | 0.1342 1.0000 skm | -0.4212 -0.2624 1.0000 sed | -0.2257 0.4236 0.0155 1.0000 om | -0.1157 -0.2977 0.2728 -0.1320 1.0000 mercado | -0.5433 0.0841 0.5519 0.3861 0.2271 1.0000 prod | -0.0983 -0.2774 0.3050 -0.0177 0.0879 0.0436 1.0000 perd | 0.1172 0.1093 -0.2057 0.0243 -0.3325 -0.2049 -0.3848 1.0000 ls1 | -0.0737 -0.3415 0.1713 -0.0397 0.1605 0.1526 0.2950 -0.2344 1.0000 tipofis | 0.1049 0.1282 0.1103 -0.0462 0.1725 0.0067 0.4753 -0.5227 -0.0376 1.0000 tipown | -0.3650 0.0971 0.3059 0.1083 0.1747 0.3912 0.3052 -0.1793 -0.2502 0.5220 1.0000 regu | 0.2557 -0.2113 0.0540 -0.3168 0.2339 -0.2472 0.4633 -0.5460 0.1369 0.6885 0.1463 1.0000 rorr | -0.1049 -0.1282 -0.1103 0.0462 -0.1725 -0.0067 -0.4753 0.5227 0.0376 -1.0000 -0.5220 -0.6885 1.0000 pcap | -0.2059 0.4324 0.0641 0.3573 -0.0949 0.3309 -0.0247 0.0749 -0.2246 0.3217 0.4461 -0.4652 -0.3217 1.0000

. corr saifi la skm sed om mercado prod perd ls1 tipofis tipown regu rorr pcap (obs=309) | saifi la skm sed om mercado prod perd ls1 tipofis tipown regu rorr pcap -------------+---------------------------------------------------------------------------------------------------------------------------saifi | 1.0000 la | 0.2526 1.0000 skm | -0.4200 -0.2624 1.0000 sed | -0.1046 0.4236 0.0155 1.0000 om | -0.2348 -0.2977 0.2728 -0.1320 1.0000 mercado | -0.4588 0.0841 0.5519 0.3861 0.2271 1.0000 prod | -0.3028 -0.2774 0.3050 -0.0177 0.0879 0.0436 1.0000 perd | 0.3621 0.1093 -0.2057 0.0243 -0.3325 -0.2049 -0.3848 1.0000 ls1 | -0.1455 -0.3415 0.1713 -0.0397 0.1605 0.1526 0.2950 -0.2344 1.0000 tipofis | -0.0922 0.1282 0.1103 -0.0462 0.1725 0.0067 0.4753 -0.5227 -0.0376 1.0000 tipown | -0.4338 0.0971 0.3059 0.1083 0.1747 0.3912 0.3052 -0.1793 -0.2502 0.5220 1.0000 regu | -0.0484 -0.2113 0.0540 -0.3168 0.2339 -0.2472 0.4633 -0.5460 0.1369 0.6885 0.1463 1.0000 rorr | 0.0922 -0.1282 -0.1103 0.0462 -0.1725 -0.0067 -0.4753 0.5227 0.0376 -1.0000 -0.5220 -0.6885 1.0000 pcap | -0.0494 0.4324 0.0641 0.3573 -0.0949 0.3309 -0.0247 0.0749 -0.2246 0.3217 0.4461 -0.4652 -0.3217 1.0000

. . . .

*----------------------------------------------------------------------------* Tablas N 14 y 15 VIF *----------------------------------------------------------------------------regress saidi la skm sed om mercado prod perd ls1 tipown tipofis regu Number of obs F( 11, 297) Prob > F R-squared Adj R-squared Root MSE = = = = = = 309 23.43 0.0000 0.4646 0.4447 10.716

Source | SS df MS -------------+-----------------------------Model | 29594.2131 11 2690.38301 Residual | 34107.4653 297 114.83995 -------------+-----------------------------Total | 63701.6784 308 206.823631

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.141886 .6723328 3.19 0.002 .8187458 3.465025 skm | -1.976359 .7939278 -2.49 0.013 -3.538795 -.4139219 sed | -.7273491 .3825876 -1.90 0.058 -1.480275 .025577 om | 1.72813 1.324501 1.30 0.193 -.8784668 4.334727 mercado | -1.586804 .5660384 -2.80 0.005 -2.700758 -.4728493 prod | -.2880767 2.113088 -0.14 0.892 -4.446598 3.870445 perd | 55.91815 14.98591 3.73 0.000 26.42612 85.41017 ls1 | -1.002432 1.629442 -0.62 0.539 -4.209148 2.204284 tipown | -12.35052 2.055659 -6.01 0.000 -16.39602 -8.305016 tipofis | 12.05227 3.907115 3.08 0.002 4.363133 19.74141 regu | 4.792115 1.68868 2.84 0.005 1.468821 8.11541 _cons | 9.092071 4.889617 1.86 0.064 -.5306143 18.71476 -----------------------------------------------------------------------------. vif Variable | VIF 1/VIF -------------+---------------------tipofis | 3.81 0.262246 regu | 3.26 0.306757 mercado | 2.78 0.360267 tipown | 2.38 0.420631 prod | 1.97 0.508265 la | 1.96 0.511097 perd | 1.95 0.512556

104

skm | 1.86 0.536950 sed | 1.60 0.626496 ls1 | 1.59 0.629896 om | 1.37 0.729956 -------------+---------------------Mean VIF | 2.23

. regress saidi la skm sed om mercado prod perd ls1 tipown rorr pcap Source | SS df MS -------------+-----------------------------Model | 29594.2131 11 2690.38301 Residual | 34107.4653 297 114.83995 -------------+-----------------------------Total | 63701.6784 308 206.823631 Number of obs F( 11, 297) Prob > F R-squared Adj R-squared Root MSE = = = = = = 309 23.43 0.0000 0.4646 0.4447 10.716

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.141886 .6723328 3.19 0.002 .8187458 3.465025 skm | -1.976359 .7939278 -2.49 0.013 -3.538795 -.4139219 sed | -.7273491 .3825876 -1.90 0.058 -1.480275 .025577 om | 1.72813 1.324501 1.30 0.193 -.8784668 4.334727 mercado | -1.586804 .5660384 -2.80 0.005 -2.700758 -.4728493 prod | -.2880767 2.113088 -0.14 0.892 -4.446598 3.870445 perd | 55.91815 14.98591 3.73 0.000 26.42612 85.41017 ls1 | -1.002432 1.629442 -0.62 0.539 -4.209148 2.204284 tipown | -12.35052 2.055659 -6.01 0.000 -16.39602 -8.305016 rorr | -21.6365 3.460032 -6.25 0.000 -28.44579 -14.82722 pcap | -4.792115 1.68868 -2.84 0.005 -8.11541 -1.468821 _cons | 30.72857 3.788303 8.11 0.000 23.27326 38.18389 -----------------------------------------------------------------------------. vif Variable | VIF 1/VIF -------------+---------------------rorr | 2.99 0.334396 mercado | 2.78 0.360267 tipown | 2.38 0.420631 prod | 1.97 0.508265 la | 1.96 0.511097 perd | 1.95 0.512556 pcap | 1.91 0.522897 skm | 1.86 0.536950 sed | 1.60 0.626496 ls1 | 1.59 0.629896 om | 1.37 0.729956 -------------+---------------------Mean VIF | 2.03 . regress saidi la skm sed om mercado prod perd ls1 tipown tipofis Source | SS df MS -------------+-----------------------------Model | 28669.4026 10 2866.94026 Residual | 35032.2758 298 117.557972 -------------+-----------------------------Total | 63701.6784 308 206.823631 Number of obs F( 10, 298) Prob > F R-squared Adj R-squared Root MSE = = = = = = 309 24.39 0.0000 0.4501 0.4316 10.842

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.832885 .6712621 2.73 0.007 .5118704 3.1539 skm | -1.886283 .8026259 -2.35 0.019 -3.465816 -.3067501 sed | -.9134143 .3813619 -2.40 0.017 -1.663918 -.1629108 om | 2.068624 1.334574 1.55 0.122 -.5577586 4.695007 mercado | -1.975768 .5556536 -3.56 0.000 -3.06927 -.8822658 prod | .3733843 2.124901 0.18 0.861 -3.808328 4.555096 perd | 43.76238 14.52962 3.01 0.003 15.16872 72.35603 ls1 | -.8676255 1.647912 -0.53 0.599 -4.110644 2.375393 tipown | -13.27804 2.053386 -6.47 0.000 -17.31901 -9.237065 tipofis | 18.22815 3.283059 5.55 0.000 11.76723 24.68906 _cons | 13.45532 4.696189 2.87 0.004 4.213428 22.69722 ------------------------------------------------------------------------------

105

. vif Variable | VIF 1/VIF -------------+---------------------tipofis | 2.63 0.380210 mercado | 2.61 0.382708 tipown | 2.32 0.431541 prod | 1.94 0.514525 la | 1.91 0.524864 skm | 1.86 0.537810 perd | 1.79 0.558160 ls1 | 1.59 0.630432 sed | 1.55 0.645453 om | 1.36 0.735996 -------------+---------------------Mean VIF | 1.96 . . . . . *----------------------------------------------------------------------------* Tabla N 16 Prueba F *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, fe Number of obs Number of groups = = 309 58 2 5.3 6 0.99 0.4454

Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0315 between = 0.0875 overall = 0.0870

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0984

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 3.826819 2.003076 1.91 0.057 -.1187881 7.772426 skm | -2.973421 4.191924 -0.71 0.479 -11.23057 5.283724 sed | -.4669423 1.354357 -0.34 0.731 -3.13472 2.200835 om | 2.001599 1.944835 1.03 0.304 -1.829287 5.832485 mercado | .3871478 .9269925 0.42 0.677 -1.438818 2.213114 prod | 2.92722 3.417124 0.86 0.392 -3.803743 9.658183 perd | 8.884762 34.79439 0.26 0.799 -59.65235 77.42187 ls1 | .0411012 2.376693 0.02 0.986 -4.640447 4.722649 tipown | (dropped) rorr | (dropped) pcap | (dropped) _cons | 9.804602 9.34041 1.05 0.295 -8.593899 28.2031 -------------+---------------------------------------------------------------sigma_u | 12.952493 sigma_e | 6.3130498 rho | .80804221 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 10.75 Prob > F = 0.0000 . xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0219 between = 0.0317 overall = 0.0426 Number of obs Number of groups = = 309 58 2 5.3 6 0.68 0.7095

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0185

-----------------------------------------------------------------------------saifi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.087557 .9965701 1.09 0.276 -.8754609 3.050576 skm | -.7591677 2.085566 -0.36 0.716 -4.867262 3.348927 sed | -.4144312 .6738195 -0.62 0.539 -1.741704 .9128412 om | 1.154967 .9675941 1.19 0.234 -.7509753 3.060909 mercado | .4266636 .4611973 0.93 0.356 -.481791 1.335118 prod | -1.282758 1.700087 -0.75 0.451 -4.631547 2.06603 perd | -3.14747 17.31091 -0.18 0.856 -37.24605 30.95111 ls1 | -.1118186 1.182452 -0.09 0.925 -2.440982 2.217345

106

tipown | (dropped) rorr | (dropped) pcap | (dropped) _cons | 11.69777 4.647041 2.52 0.012 2.544146 20.85139 -------------+---------------------------------------------------------------sigma_u | 7.4988421 sigma_e | 3.1408682 rho | .8507504 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 14.03 Prob > F = 0.0000 . xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0315 between = 0.0875 overall = 0.0870 Number of obs Number of groups = = 309 58 2 5.3 6 0.99 0.4454

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0984

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 3.826819 2.003076 1.91 0.057 -.1187881 7.772426 skm | -2.973421 4.191924 -0.71 0.479 -11.23057 5.283724 sed | -.4669423 1.354357 -0.34 0.731 -3.13472 2.200835 om | 2.001599 1.944835 1.03 0.304 -1.829287 5.832485 mercado | .3871478 .9269925 0.42 0.677 -1.438818 2.213114 prod | 2.92722 3.417124 0.86 0.392 -3.803743 9.658183 perd | 8.884762 34.79439 0.26 0.799 -59.65235 77.42187 ls1 | .0411012 2.376693 0.02 0.986 -4.640447 4.722649 tipown | (dropped) tipofis | (dropped) _cons | 9.804602 9.34041 1.05 0.295 -8.593899 28.2031 -------------+---------------------------------------------------------------sigma_u | 12.952493 sigma_e | 6.3130498 rho | .80804221 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 11.16 Prob > F = 0.0000 . xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0219 between = 0.0317 overall = 0.0426 Number of obs Number of groups = = 309 58 2 5.3 6 0.68 0.7095

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0185

-----------------------------------------------------------------------------saifi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.087557 .9965701 1.09 0.276 -.8754609 3.050576 skm | -.7591677 2.085566 -0.36 0.716 -4.867262 3.348927 sed | -.4144312 .6738195 -0.62 0.539 -1.741704 .9128412 om | 1.154967 .9675941 1.19 0.234 -.7509753 3.060909 mercado | .4266636 .4611973 0.93 0.356 -.481791 1.335118 prod | -1.282758 1.700087 -0.75 0.451 -4.631547 2.06603 perd | -3.14747 17.31091 -0.18 0.856 -37.24605 30.95111 ls1 | -.1118186 1.182452 -0.09 0.925 -2.440982 2.217345 tipown | (dropped) tipofis | (dropped) _cons | 11.69777 4.647041 2.52 0.012 2.544146 20.85139 -------------+---------------------------------------------------------------sigma_u | 7.4988421 sigma_e | 3.1408682 rho | .8507504 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 14.03 Prob > F = 0.0000

107

. . . . . .

*----------------------------------------------------------------------------* Tabla N 17 Test Breusch Pagan *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re Number of obs Number of groups = = 309 58 2 5.3 6 61.17 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0215 between = 0.5341 overall = 0.4466

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.850313 1.135184 2.51 0.012 .6253922 5.075234 skm | -2.985294 1.461793 -2.04 0.041 -5.850356 -.120231 sed | -1.015625 .6490139 -1.56 0.118 -2.287669 .2564192 om | 2.223066 1.601626 1.39 0.165 -.9160632 5.362196 mercado | -.5470919 .6906601 -0.79 0.428 -1.900761 .806577 prod | 1.85302 2.685033 0.69 0.490 -3.409549 7.115588 perd | 43.38239 22.4492 1.93 0.053 -.6172266 87.382 ls1 | -.5946484 1.929358 -0.31 0.758 -4.37612 3.186823 tipown | -14.15972 3.95957 -3.58 0.000 -21.92034 -6.399106 rorr | -21.11954 6.075586 -3.48 0.001 -33.02747 -9.211612 pcap | -5.74593 3.428813 -1.68 0.094 -12.46628 .9744191 _cons | 28.47059 5.339038 5.33 0.000 18.00627 38.93491 -------------+---------------------------------------------------------------sigma_u | 9.5620541 sigma_e | 6.3130498 rho | .6964326 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xttest0 Breusch and Pagan Lagrangian multiplier test for random effects: saidi[empresa,t] = Xb + u[empresa] + e[empresa,t] Estimated results: | Var sd = sqrt(Var) ---------+----------------------------saidi | 206.8236 14.38136 e | 39.8546 6.31305 u | 91.43288 9.562054 Test: Var(u) = 0 chi2(1) = Prob > chi2 = 274.07 0.0000

. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 51.19 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343539 .6176917 2.18 0.030 .1328853 2.554192 skm | -1.160847 .8150014 -1.42 0.154 -2.758221 .4365262 sed | -.4203555 .3568219 -1.18 0.239 -1.119713 .2790026 om | 1.032141 .832109 1.24 0.215 -.5987626 2.663045 mercado | -.1331316 .3611211 -0.37 0.712 -.8409159 .5746527 prod | -1.506916 1.405992 -1.07 0.284 -4.262611 1.248778 perd | 28.8587 12.06561 2.39 0.017 5.210541 52.50687 ls1 | -.747044 1.004269 -0.74 0.457 -2.715375 1.221288

108

tipown | -9.014782 2.250958 -4.00 0.000 -13.42658 -4.602984 rorr | -7.943699 3.391534 -2.34 0.019 -14.59098 -1.296415 pcap | .1712461 1.94929 0.09 0.930 -3.649292 3.991784 _cons | 18.11914 2.89037 6.27 0.000 12.45412 23.78416 -------------+---------------------------------------------------------------sigma_u | 5.4951497 sigma_e | 3.1408682 rho | .75375365 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xttest0 Breusch and Pagan Lagrangian multiplier test for random effects: saifi[empresa,t] = Xb + u[empresa] + e[empresa,t] Estimated results: | Var sd = sqrt(Var) ---------+----------------------------saifi | 65.63056 8.101269 e | 9.865053 3.140868 u | 30.19667 5.49515 Test: Var(u) = 0 chi2(1) = Prob > chi2 = 319.34 0.0000

. xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0201 between = 0.5099 overall = 0.4313 Number of obs Number of groups = = 309 58 2 5.3 6 57.05 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.378907 1.10729 2.15 0.032 .2086579 4.549156 skm | -3.065553 1.475928 -2.08 0.038 -5.95832 -.1727869 sed | -1.22991 .642178 -1.92 0.055 -2.488556 .0287353 om | 2.322838 1.607322 1.45 0.148 -.8274564 5.473132 mercado | -.7213355 .6850428 -1.05 0.292 -2.063995 .6213237 prod | 2.447671 2.676039 0.91 0.360 -2.79727 7.692611 perd | 35.5204 22.13852 1.60 0.109 -7.87031 78.91111 ls1 | -.5573861 1.937765 -0.29 0.774 -4.355335 3.240563 tipown | -15.75599 3.889571 -4.05 0.000 -23.37941 -8.132568 tipofis | 18.46161 5.929548 3.11 0.002 6.839908 30.08331 _cons | 10.57843 7.170239 1.48 0.140 -3.474984 24.63183 -------------+---------------------------------------------------------------sigma_u | 9.6797412 sigma_e | 6.3130498 rho | .70157989 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xttest0 Breusch and Pagan Lagrangian multiplier test for random effects: saidi[empresa,t] = Xb + u[empresa] + e[empresa,t] Estimated results: | Var sd = sqrt(Var) ---------+----------------------------saidi | 206.8236 14.38136 e | 39.8546 6.31305 u | 93.69739 9.679741 Test: Var(u) = 0 chi2(1) = Prob > chi2 = 275.78 0.0000

. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re

109

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 between = 0.5088 overall = 0.4592

Number of obs Number of groups

= =

309 58 2 5.3 6 52.21 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357297 .5940478 2.28 0.022 .1929847 2.521609 skm | -1.154105 .8083905 -1.43 0.153 -2.738521 .4303117 sed | -.412841 .3473363 -1.19 0.235 -1.093608 .2679257 om | 1.026303 .8297051 1.24 0.216 -.5998887 2.652495 mercado | -.1331495 .3559875 -0.37 0.708 -.8308722 .5645733 prod | -1.524866 1.390445 -1.10 0.273 -4.250089 1.200357 perd | 29.33135 11.78658 2.49 0.013 6.230087 52.43261 ls1 | -.7524121 1.001705 -0.75 0.453 -2.715718 1.210894 tipown | -8.965161 2.16176 -4.15 0.000 -13.20213 -4.72819 tipofis | 8.048197 3.260244 2.47 0.014 1.658237 14.43816 _cons | 10.04268 3.859198 2.60 0.009 2.478789 17.60657 -------------+---------------------------------------------------------------sigma_u | 5.4373595 sigma_e | 3.1408682 rho | .749808 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xttest0 Breusch and Pagan Lagrangian multiplier test for random effects: saifi[empresa,t] = Xb + u[empresa] + e[empresa,t] Estimated results: | Var sd = sqrt(Var) ---------+----------------------------saifi | 65.63056 8.101269 e | 9.865053 3.140868 u | 29.56488 5.437359 Test: Var(u) = 0 chi2(1) = Prob > chi2 = . . . . . . 319.31 0.0000

*----------------------------------------------------------------------------* Tabla N 18 Test de Hausman *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, fe Number of obs Number of groups = = 309 58 2 5.3 6 0.99 0.4454

Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0315 between = 0.0875 overall = 0.0870

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0984

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 3.826819 2.003076 1.91 0.057 -.1187881 7.772426 skm | -2.973421 4.191924 -0.71 0.479 -11.23057 5.283724 sed | -.4669423 1.354357 -0.34 0.731 -3.13472 2.200835 om | 2.001599 1.944835 1.03 0.304 -1.829287 5.832485 mercado | .3871478 .9269925 0.42 0.677 -1.438818 2.213114 prod | 2.92722 3.417124 0.86 0.392 -3.803743 9.658183 perd | 8.884762 34.79439 0.26 0.799 -59.65235 77.42187 ls1 | .0411012 2.376693 0.02 0.986 -4.640447 4.722649 tipown | (dropped) rorr | (dropped)

110

pcap | (dropped) _cons | 9.804602 9.34041 1.05 0.295 -8.593899 28.2031 -------------+---------------------------------------------------------------sigma_u | 12.952493 sigma_e | 6.3130498 rho | .80804221 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 10.75 Prob > F = 0.0000 . estimates store FIXED . xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0215 between = 0.5341 overall = 0.4466 Number of obs Number of groups = = 309 58 2 5.3 6 61.17 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.850313 1.135184 2.51 0.012 .6253922 5.075234 skm | -2.985294 1.461793 -2.04 0.041 -5.850356 -.120231 sed | -1.015625 .6490139 -1.56 0.118 -2.287669 .2564192 om | 2.223066 1.601626 1.39 0.165 -.9160632 5.362196 mercado | -.5470919 .6906601 -0.79 0.428 -1.900761 .806577 prod | 1.85302 2.685033 0.69 0.490 -3.409549 7.115588 perd | 43.38239 22.4492 1.93 0.053 -.6172266 87.382 ls1 | -.5946484 1.929358 -0.31 0.758 -4.37612 3.186823 tipown | -14.15972 3.95957 -3.58 0.000 -21.92034 -6.399106 rorr | -21.11954 6.075586 -3.48 0.001 -33.02747 -9.211612 pcap | -5.74593 3.428813 -1.68 0.094 -12.46628 .9744191 _cons | 28.47059 5.339038 5.33 0.000 18.00627 38.93491 -------------+---------------------------------------------------------------sigma_u | 9.5620541 sigma_e | 6.3130498 rho | .6964326 (fraction of variance due to u_i) -----------------------------------------------------------------------------. estimates store RANDOM . hausman FIXED RANDOM ---- Coefficients ---| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | FIXED RANDOM Difference S.E. -------------+---------------------------------------------------------------la | 3.826819 2.850313 .9765062 1.650354 skm | -2.973421 -2.985294 .0118723 3.92879 sed | -.4669423 -1.015625 .5486824 1.188723 om | 2.001599 2.223066 -.2214669 1.103257 mercado | .3871478 -.5470919 .9342397 .6183072 prod | 2.92722 1.85302 1.0742 2.113606 perd | 8.884762 43.38239 -34.49763 26.58352 ls1 | .0411012 -.5946484 .6357496 1.387893 -----------------------------------------------------------------------------b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 5.12 Prob>chi2 = 0.7442 . xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0219 between = 0.0317 overall = 0.0426 Number of obs Number of groups = = 309 58 2 5.3 6

Obs per group: min = avg = max =

111

corr(u_i, Xb)

= -0.0185

F(8,243) Prob > F

= =

0.68 0.7095

-----------------------------------------------------------------------------saifi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.087557 .9965701 1.09 0.276 -.8754609 3.050576 skm | -.7591677 2.085566 -0.36 0.716 -4.867262 3.348927 sed | -.4144312 .6738195 -0.62 0.539 -1.741704 .9128412 om | 1.154967 .9675941 1.19 0.234 -.7509753 3.060909 mercado | .4266636 .4611973 0.93 0.356 -.481791 1.335118 prod | -1.282758 1.700087 -0.75 0.451 -4.631547 2.06603 perd | -3.14747 17.31091 -0.18 0.856 -37.24605 30.95111 ls1 | -.1118186 1.182452 -0.09 0.925 -2.440982 2.217345 tipown | (dropped) rorr | (dropped) pcap | (dropped) _cons | 11.69777 4.647041 2.52 0.012 2.544146 20.85139 -------------+---------------------------------------------------------------sigma_u | 7.4988421 sigma_e | 3.1408682 rho | .8507504 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 14.03 Prob > F = 0.0000 . estimates store FIXED . xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 51.19 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343539 .6176917 2.18 0.030 .1328853 2.554192 skm | -1.160847 .8150014 -1.42 0.154 -2.758221 .4365262 sed | -.4203555 .3568219 -1.18 0.239 -1.119713 .2790026 om | 1.032141 .832109 1.24 0.215 -.5987626 2.663045 mercado | -.1331316 .3611211 -0.37 0.712 -.8409159 .5746527 prod | -1.506916 1.405992 -1.07 0.284 -4.262611 1.248778 perd | 28.8587 12.06561 2.39 0.017 5.210541 52.50687 ls1 | -.747044 1.004269 -0.74 0.457 -2.715375 1.221288 tipown | -9.014782 2.250958 -4.00 0.000 -13.42658 -4.602984 rorr | -7.943699 3.391534 -2.34 0.019 -14.59098 -1.296415 pcap | .1712461 1.94929 0.09 0.930 -3.649292 3.991784 _cons | 18.11914 2.89037 6.27 0.000 12.45412 23.78416 -------------+---------------------------------------------------------------sigma_u | 5.4951497 sigma_e | 3.1408682 rho | .75375365 (fraction of variance due to u_i) -----------------------------------------------------------------------------. estimates store RANDOM . hausman FIXED RANDOM

112

---- Coefficients ---| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | FIXED RANDOM Difference S.E. -------------+---------------------------------------------------------------la | 1.087557 1.343539 -.2559814 .7820543 skm | -.7591677 -1.160847 .4016795 1.919729 sed | -.4144312 -.4203555 .0059242 .5715863 om | 1.154967 1.032141 .1228258 .4937946 mercado | .4266636 -.1331316 .5597952 .2868702 prod | -1.282758 -1.506916 .2241577 .9557625 perd | -3.14747 28.8587 -32.00617 12.41324 ls1 | -.1118186 -.747044 .6352254 .6242084 -----------------------------------------------------------------------------b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 10.64 Prob>chi2 = 0.2231 . xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0315 between = 0.0875 overall = 0.0870 Number of obs Number of groups = = 309 58 2 5.3 6 0.99 0.4454

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0984

-----------------------------------------------------------------------------saidi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 3.826819 2.003076 1.91 0.057 -.1187881 7.772426 skm | -2.973421 4.191924 -0.71 0.479 -11.23057 5.283724 sed | -.4669423 1.354357 -0.34 0.731 -3.13472 2.200835 om | 2.001599 1.944835 1.03 0.304 -1.829287 5.832485 mercado | .3871478 .9269925 0.42 0.677 -1.438818 2.213114 prod | 2.92722 3.417124 0.86 0.392 -3.803743 9.658183 perd | 8.884762 34.79439 0.26 0.799 -59.65235 77.42187 ls1 | .0411012 2.376693 0.02 0.986 -4.640447 4.722649 tipown | (dropped) tipofis | (dropped) _cons | 9.804602 9.34041 1.05 0.295 -8.593899 28.2031 -------------+---------------------------------------------------------------sigma_u | 12.952493 sigma_e | 6.3130498 rho | .80804221 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 11.16 Prob > F = 0.0000 . estimates store FIXED . xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0201 between = 0.5099 overall = 0.4313 Number of obs Number of groups = = 309 58 2 5.3 6 57.05 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.378907 1.10729 2.15 0.032 .2086579 4.549156 skm | -3.065553 1.475928 -2.08 0.038 -5.95832 -.1727869 sed | -1.22991 .642178 -1.92 0.055 -2.488556 .0287353 om | 2.322838 1.607322 1.45 0.148 -.8274564 5.473132 mercado | -.7213355 .6850428 -1.05 0.292 -2.063995 .6213237 prod | 2.447671 2.676039 0.91 0.360 -2.79727 7.692611

113

perd | 35.5204 22.13852 1.60 0.109 -7.87031 78.91111 ls1 | -.5573861 1.937765 -0.29 0.774 -4.355335 3.240563 tipown | -15.75599 3.889571 -4.05 0.000 -23.37941 -8.132568 tipofis | 18.46161 5.929548 3.11 0.002 6.839908 30.08331 _cons | 10.57843 7.170239 1.48 0.140 -3.474984 24.63183 -------------+---------------------------------------------------------------sigma_u | 9.6797412 sigma_e | 6.3130498 rho | .70157989 (fraction of variance due to u_i) -----------------------------------------------------------------------------. estimates store RANDOM . hausman FIXED RANDOM ---- Coefficients ---| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | FIXED RANDOM Difference S.E. -------------+---------------------------------------------------------------la | 3.826819 2.378907 1.447912 1.669197 skm | -2.973421 -3.065553 .0921321 3.923502 sed | -.4669423 -1.22991 .7629682 1.19243 om | 2.001599 2.322838 -.3212382 1.094942 mercado | .3871478 -.7213355 1.108483 .624525 prod | 2.92722 2.447671 .4795493 2.124982 perd | 8.884762 35.5204 -26.63564 26.8428 ls1 | .0411012 -.5573861 .5984873 1.376131 -----------------------------------------------------------------------------b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 5.82 Prob>chi2 = 0.6673 . xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, fe Fixed-effects (within) regression Group variable (i): empresa R-sq: within = 0.0219 between = 0.0317 overall = 0.0426 Number of obs Number of groups = = 309 58 2 5.3 6 0.68 0.7095

Obs per group: min = avg = max = F(8,243) Prob > F = =

corr(u_i, Xb)

= -0.0185

-----------------------------------------------------------------------------saifi | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.087557 .9965701 1.09 0.276 -.8754609 3.050576 skm | -.7591677 2.085566 -0.36 0.716 -4.867262 3.348927 sed | -.4144312 .6738195 -0.62 0.539 -1.741704 .9128412 om | 1.154967 .9675941 1.19 0.234 -.7509753 3.060909 mercado | .4266636 .4611973 0.93 0.356 -.481791 1.335118 prod | -1.282758 1.700087 -0.75 0.451 -4.631547 2.06603 perd | -3.14747 17.31091 -0.18 0.856 -37.24605 30.95111 ls1 | -.1118186 1.182452 -0.09 0.925 -2.440982 2.217345 tipown | (dropped) tipofis | (dropped) _cons | 11.69777 4.647041 2.52 0.012 2.544146 20.85139 -------------+---------------------------------------------------------------sigma_u | 7.4988421 sigma_e | 3.1408682 rho | .8507504 (fraction of variance due to u_i) -----------------------------------------------------------------------------F test that all u_i=0: F(57, 243) = 14.03 Prob > F = 0.0000 . estimates store FIXED . xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 Number of obs Number of groups = = 309 58 2

Obs per group: min =

114

between = 0.5088 overall = 0.4592 Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed) Wald chi2(10) Prob > chi2

avg = max = = =

5.3 6 52.21 0.0000

-----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357297 .5940478 2.28 0.022 .1929847 2.521609 skm | -1.154105 .8083905 -1.43 0.153 -2.738521 .4303117 sed | -.412841 .3473363 -1.19 0.235 -1.093608 .2679257 om | 1.026303 .8297051 1.24 0.216 -.5998887 2.652495 mercado | -.1331495 .3559875 -0.37 0.708 -.8308722 .5645733 prod | -1.524866 1.390445 -1.10 0.273 -4.250089 1.200357 perd | 29.33135 11.78658 2.49 0.013 6.230087 52.43261 ls1 | -.7524121 1.001705 -0.75 0.453 -2.715718 1.210894 tipown | -8.965161 2.16176 -4.15 0.000 -13.20213 -4.72819 tipofis | 8.048197 3.260244 2.47 0.014 1.658237 14.43816 _cons | 10.04268 3.859198 2.60 0.009 2.478789 17.60657 -------------+---------------------------------------------------------------sigma_u | 5.4373595 sigma_e | 3.1408682 rho | .749808 (fraction of variance due to u_i) -----------------------------------------------------------------------------. estimates store RANDOM . hausman FIXED RANDOM ---- Coefficients ---| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | FIXED RANDOM Difference S.E. -------------+---------------------------------------------------------------la | 1.087557 1.357297 -.2697395 .800162 skm | -.7591677 -1.154105 .3949369 1.922522 sed | -.4144312 -.412841 -.0015903 .5773996 om | 1.154967 1.026303 .1286635 .4978232 mercado | .4266636 -.1331495 .5598131 .2932163 prod | -1.282758 -1.524866 .2421077 .9782426 perd | -3.14747 29.33135 -32.47882 12.67849 ls1 | -.1118186 -.7524121 .6405935 .6283146 -----------------------------------------------------------------------------b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic chi2(8) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 10.67 Prob>chi2 = 0.2209 . . . . *----------------------------------------------------------------------------* Prueba de Baltagi Wu *----------------------------------------------------------------------------xtregar saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re lbi Number of obs Number of groups = = 309 58 2 5.3 6 75.86 0.0000

RE GLS regression with AR(1) disturbances Group variable (i): empresa R-sq: within = 0.0164 between = 0.5434 overall = 0.4544

Obs per group: min = avg = max = Wald chi2(12) Prob > chi2 = =

corr(u_i, Xb)

= 0 (assumed)

------------------- theta -------------------min 5% median 95% max 0.3812 0.4494 0.4991 0.4991 0.4991 -----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.638007 1.116974 2.36 0.018 .4487777 4.827236 skm | -2.651438 1.37271 -1.93 0.053 -5.341899 .0390234 sed | -.9808999 .6293452 -1.56 0.119 -2.214394 .252594 om | 2.090006 1.633697 1.28 0.201 -1.111981 5.291992

115

mercado | -.7942308 .6554414 -1.21 0.226 -2.078872 .4904107 prod | 1.427104 2.575469 0.55 0.580 -3.620724 6.474931 perd | 53.75318 21.64402 2.48 0.013 11.33168 96.17467 ls1 | -.5624336 1.822857 -0.31 0.758 -4.135167 3.0103 tipown | -15.07301 3.67547 -4.10 0.000 -22.2768 -7.869225 rorr | -23.53636 5.714961 -4.12 0.000 -34.73748 -12.33524 pcap | -5.650007 3.207834 -1.76 0.078 -11.93725 .6372327 _cons | 29.53837 5.268649 5.61 0.000 19.212 39.86473 -------------+---------------------------------------------------------------rho_ar | .57358998 (estimated autocorrelation coefficient) sigma_u | 7.7301783 sigma_e | 5.6233033 rho_fov | .65394495 (fraction of variance due to u_i) -----------------------------------------------------------------------------modified Bhargava et al. Durbin-Watson = .92835547 Baltagi-Wu LBI = 1.5011637 . xtregar saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re lbi RE GLS regression with AR(1) disturbances Group variable (i): empresa R-sq: within = 0.0038 between = 0.5239 overall = 0.4726 Number of obs Number of groups = = 309 58 2 5.3 6 63.96 0.0000

Obs per group: min = avg = max = Wald chi2(12) Prob > chi2 = =

corr(u_i, Xb)

= 0 (assumed)

------------------- theta -------------------min 5% median 95% max 0.4986 0.5933 0.6489 0.6489 0.6489 -----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.337274 .6152158 2.17 0.030 .1314734 2.543075 skm | -.9693174 .7706549 -1.26 0.208 -2.479773 .5411385 sed | -.3917324 .3481061 -1.13 0.260 -1.074008 .2905431 om | .8648375 .8893911 0.97 0.331 -.878337 2.608012 mercado | -.3555429 .3714532 -0.96 0.338 -1.083578 .3724921 prod | -1.540736 1.443597 -1.07 0.286 -4.370134 1.288662 perd | 33.01436 12.07164 2.73 0.006 9.354373 56.67434 ls1 | -.7400503 1.030964 -0.72 0.473 -2.760703 1.280603 tipown | -9.074376 2.064832 -4.39 0.000 -13.12137 -5.02738 rorr | -8.41674 3.199208 -2.63 0.009 -14.68707 -2.146408 pcap | .2953455 1.794305 0.16 0.869 -3.221428 3.812119 _cons | 18.46722 2.897304 6.37 0.000 12.7886 24.14583 -------------+---------------------------------------------------------------rho_ar | .30548936 (estimated autocorrelation coefficient) sigma_u | 4.7215689 sigma_e | 3.2248736 rho_fov | .68189512 (fraction of variance due to u_i) -----------------------------------------------------------------------------modified Bhargava et al. Durbin-Watson = 1.4339979 Baltagi-Wu LBI = 1.9895325 . xtregar saidi la skm sed om mercado prod perd ls1 tipown tipofis, re lbi RE GLS regression with AR(1) disturbances Group variable (i): empresa R-sq: within = 0.0150 between = 0.5196 overall = 0.4387 Number of obs Number of groups = = 309 58 2 5.3 6 70.47 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

corr(u_i, Xb)

= 0 (assumed)

------------------- theta -------------------min 5% median 95% max 0.3899 0.4582 0.5077 0.5077 0.5077 -----------------------------------------------------------------------------saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.14534 1.092757 1.96 0.050 .0035762 4.287104 skm | -2.71148 1.391956 -1.95 0.051 -5.439665 .0167037 sed | -1.190812 .6263968 -1.90 0.057 -2.418527 .0369033

116

om | 2.216266 1.641702 1.35 0.177 -1.001411 5.433944 mercado | -.9745634 .6500558 -1.50 0.134 -2.248649 .2995225 prod | 1.96184 2.574089 0.76 0.446 -3.083282 7.006962 perd | 45.60566 21.36208 2.13 0.033 3.736753 87.47457 ls1 | -.5324315 1.831989 -0.29 0.771 -4.123063 3.058201 tipown | -16.63153 3.626708 -4.59 0.000 -23.73975 -9.52331 tipofis | 20.91376 5.595741 3.74 0.000 9.94631 31.88121 _cons | 9.306552 6.883392 1.35 0.176 -4.184649 22.79775 -------------+---------------------------------------------------------------rho_ar | .57358998 (estimated autocorrelation coefficient) sigma_u | 7.9233786 sigma_e | 5.6335632 rho_fov | .66421864 (fraction of variance due to u_i) -----------------------------------------------------------------------------modified Bhargava et al. Durbin-Watson = .92835547 Baltagi-Wu LBI = 1.5011637 . xtregar saifi la skm sed om mercado prod perd ls1 tipown tipofis, re lbi RE GLS regression with AR(1) disturbances Group variable (i): empresa R-sq: within = 0.0039 between = 0.5248 overall = 0.4725 Number of obs Number of groups = = 309 58 2 5.3 6 64.16 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

corr(u_i, Xb)

= 0 (assumed)

------------------- theta -------------------min 5% median 95% max 0.4985 0.5932 0.6489 0.6489 0.6489 -----------------------------------------------------------------------------saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.363029 .5939489 2.29 0.022 .198911 2.527148 skm | -.966153 .7690946 -1.26 0.209 -2.473551 .5412448 sed | -.3809265 .3413034 -1.12 0.264 -1.049869 .2880158 om | .8588339 .8871888 0.97 0.333 -.8800242 2.597692 mercado | -.3454268 .365662 -0.94 0.345 -1.062111 .3712575 prod | -1.569373 1.430742 -1.10 0.273 -4.373575 1.23483 perd | 33.43831 11.77934 2.84 0.005 10.35122 56.5254 ls1 | -.7412954 1.029239 -0.72 0.471 -2.758567 1.275976 tipown | -8.994335 2.00325 -4.49 0.000 -12.92063 -5.068037 tipofis | 8.555059 3.082368 2.78 0.006 2.513728 14.59639 _cons | 9.875502 3.791288 2.60 0.009 2.444715 17.30629 -------------+---------------------------------------------------------------rho_ar | .30548936 (estimated autocorrelation coefficient) sigma_u | 4.7212442 sigma_e | 3.2252325 rho_fov | .68181701 (fraction of variance due to u_i) -----------------------------------------------------------------------------modified Bhargava et al. Durbin-Watson = 1.4339979 Baltagi-Wu LBI = 1.9895325 . . . . *----------------------------------------------------------------------------* Tabla N 19 Regresiones con Dummies de Regulacin (RORR, PCAP y Propiedad) *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042

117

om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524 prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 rorr | -21.10404 6.482052 -3.26 0.001 -33.80863 -8.399451 pcap | -5.748643 2.58658 -2.22 0.026 -10.81825 -.6790393 _cons | 28.46052 5.979177 4.76 0.000 16.74154 40.17949 -------------+---------------------------------------------------------------sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544 rorr | -7.944138 4.126756 -1.93 0.054 -16.03243 .144155 pcap | .1711495 1.429165 0.12 0.905 -2.629963 2.972262 _cons | 18.11895 3.043269 5.95 0.000 12.15425 24.08365 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. . . . *----------------------------------------------------------------------------* Tabla N 20 Regresiones con Dummies de Supervisin y Propiedad *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 68.22 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0201 between = 0.5099 overall = 0.4313

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.379052 .8602537 2.77 0.006 .6929854 4.065118 skm | -3.065697 1.055833 -2.90 0.004 -5.135091 -.9963029 sed | -1.229937 .5212016 -2.36 0.018 -2.251474 -.2084011 om | 2.322827 1.037354 2.24 0.025 .2896511 4.356003 mercado | -.7211557 .4250488 -1.70 0.090 -1.554236 .1119246 prod | 2.447953 3.459857 0.71 0.479 -4.333241 9.229147 perd | 35.51765 17.13544 2.07 0.038 1.932798 69.1025 ls1 | -.5572629 1.47929 -0.38 0.706 -3.456619 2.342093 tipown | -15.75624 5.55823 -2.83 0.005 -26.65017 -4.862307

118

tipofis | 18.46125 6.226994 2.96 0.003 6.256567 30.66594 _cons | 10.57828 4.652487 2.27 0.023 1.459571 19.69699 -------------+---------------------------------------------------------------sigma_u | 9.6815861 sigma_e | 6.3130498 rho | .70165969 (fraction of variance due to u_i) -----------------------------------------------------------------------------. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 between = 0.5090 overall = 0.4593 Number of obs Number of groups = = 309 58 2 5.3 6 65.06 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357288 .5201759 2.61 0.009 .3377622 2.376814 skm | -1.153259 .6296633 -1.83 0.067 -2.387376 .0808588 sed | -.4126054 .3042976 -1.36 0.175 -1.009018 .183807 om | 1.025657 .4991371 2.05 0.040 .0473663 2.003948 mercado | -.1340548 .2612908 -0.51 0.608 -.6461754 .3780659 prod | -1.52528 1.607574 -0.95 0.343 -4.676067 1.625507 perd | 29.38259 11.69655 2.51 0.012 6.457774 52.30741 ls1 | -.7531871 .9165113 -0.82 0.411 -2.549516 1.043142 tipown | -8.964998 2.672312 -3.35 0.001 -14.20263 -3.727362 tipofis | 8.054049 4.024678 2.00 0.045 .165825 15.94227 _cons | 10.03413 3.958654 2.53 0.011 2.275306 17.79295 -------------+---------------------------------------------------------------sigma_u | 5.4264767 sigma_e | 3.1408682 rho | .74905556 (fraction of variance due to u_i) -----------------------------------------------------------------------------. . *----------------------------------------------------------------------------. * Tabla N 21 Resumen de Pruebas de Wald . *----------------------------------------------------------------------------. *----------------------------------------------------------------------------. * Hiptesis H1 . *----------------------------------------------------------------------------. . xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464 Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042 om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524 prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 rorr | -21.10404 6.482052 -3.26 0.001 -33.80863 -8.399451 pcap | -5.748643 2.58658 -2.22 0.026 -10.81825 -.6790393 _cons | 28.46052 5.979177 4.76 0.000 16.74154 40.17949 -------------+----------------------------------------------------------------

119

sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = pcap ( 1) rorr - pcap = 0 chi2( 1) = Prob > chi2 = 6.23 0.0126

. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544 rorr | -7.944138 4.126756 -1.93 0.054 -16.03243 .144155 pcap | .1711495 1.429165 0.12 0.905 -2.629963 2.972262 _cons | 18.11895 3.043269 5.95 0.000 12.15425 24.08365 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = pcap ( 1) rorr - pcap = 0 chi2( 1) = Prob > chi2 = . . . . . 3.84 0.0500

. *----------------------------------------------------------------------------* Hiptesis H2 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042 om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524

120

prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 rorr | -21.10404 6.482052 -3.26 0.001 -33.80863 -8.399451 pcap | -5.748643 2.58658 -2.22 0.026 -10.81825 -.6790393 _cons | 28.46052 5.979177 4.76 0.000 16.74154 40.17949 -------------+---------------------------------------------------------------sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = 0 ( 1) rorr = 0 chi2( 1) = Prob > chi2 = 10.60 0.0011

. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544 rorr | -7.944138 4.126756 -1.93 0.054 -16.03243 .144155 pcap | .1711495 1.429165 0.12 0.905 -2.629963 2.972262 _cons | 18.11895 3.043269 5.95 0.000 12.15425 24.08365 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = 0 ( 1) rorr = 0 chi2( 1) = Prob > chi2 = . . . . . . 3.71 0.0542

*----------------------------------------------------------------------------* Hiptesis H3 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 68.22 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0201 between = 0.5099 overall = 0.4313

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

121

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.379052 .8602537 2.77 0.006 .6929854 4.065118 skm | -3.065697 1.055833 -2.90 0.004 -5.135091 -.9963029 sed | -1.229937 .5212016 -2.36 0.018 -2.251474 -.2084011 om | 2.322827 1.037354 2.24 0.025 .2896511 4.356003 mercado | -.7211557 .4250488 -1.70 0.090 -1.554236 .1119246 prod | 2.447953 3.459857 0.71 0.479 -4.333241 9.229147 perd | 35.51765 17.13544 2.07 0.038 1.932798 69.1025 ls1 | -.5572629 1.47929 -0.38 0.706 -3.456619 2.342093 tipown | -15.75624 5.55823 -2.83 0.005 -26.65017 -4.862307 tipofis | 18.46125 6.226994 2.96 0.003 6.256567 30.66594 _cons | 10.57828 4.652487 2.27 0.023 1.459571 19.69699 -------------+---------------------------------------------------------------sigma_u | 9.6815861 sigma_e | 6.3130498 rho | .70165969 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test tipofis = 0 ( 1) tipofis = 0 chi2( 1) = Prob > chi2 = 8.79 0.0030

. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 between = 0.5090 overall = 0.4593 Number of obs Number of groups = = 309 58 2 5.3 6 65.06 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357288 .5201759 2.61 0.009 .3377622 2.376814 skm | -1.153259 .6296633 -1.83 0.067 -2.387376 .0808588 sed | -.4126054 .3042976 -1.36 0.175 -1.009018 .183807 om | 1.025657 .4991371 2.05 0.040 .0473663 2.003948 mercado | -.1340548 .2612908 -0.51 0.608 -.6461754 .3780659 prod | -1.52528 1.607574 -0.95 0.343 -4.676067 1.625507 perd | 29.38259 11.69655 2.51 0.012 6.457774 52.30741 ls1 | -.7531871 .9165113 -0.82 0.411 -2.549516 1.043142 tipown | -8.964998 2.672312 -3.35 0.001 -14.20263 -3.727362 tipofis | 8.054049 4.024678 2.00 0.045 .165825 15.94227 _cons | 10.03413 3.958654 2.53 0.011 2.275306 17.79295 -------------+---------------------------------------------------------------sigma_u | 5.4264767 sigma_e | 3.1408682 rho | .74905556 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test tipofis = 0 ( 1) tipofis = 0 chi2( 1) = Prob > chi2 = . . . . . . 4.00 0.0454

*----------------------------------------------------------------------------* Hiptesis H4 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Number of obs Number of groups = = 309 58

Random-effects GLS regression Group variable (i): empresa

122

R-sq:

within = 0.0201 between = 0.5099 overall = 0.4313

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

2 5.3 6 68.22 0.0000

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.379052 .8602537 2.77 0.006 .6929854 4.065118 skm | -3.065697 1.055833 -2.90 0.004 -5.135091 -.9963029 sed | -1.229937 .5212016 -2.36 0.018 -2.251474 -.2084011 om | 2.322827 1.037354 2.24 0.025 .2896511 4.356003 mercado | -.7211557 .4250488 -1.70 0.090 -1.554236 .1119246 prod | 2.447953 3.459857 0.71 0.479 -4.333241 9.229147 perd | 35.51765 17.13544 2.07 0.038 1.932798 69.1025 ls1 | -.5572629 1.47929 -0.38 0.706 -3.456619 2.342093 tipown | -15.75624 5.55823 -2.83 0.005 -26.65017 -4.862307 tipofis | 18.46125 6.226994 2.96 0.003 6.256567 30.66594 _cons | 10.57828 4.652487 2.27 0.023 1.459571 19.69699 -------------+---------------------------------------------------------------sigma_u | 9.6815861 sigma_e | 6.3130498 rho | .70165969 (fraction of variance due to u_i) -----------------------------------------------------------------------------. testnl _b[tipown]/_b[tipofis] = 0 (1) _b[tipown]/_b[tipofis] = 0 chi2(1) = Prob > chi2 = 14.93 0.0001

. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 between = 0.5090 overall = 0.4593 Number of obs Number of groups = = 309 58 2 5.3 6 65.06 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357288 .5201759 2.61 0.009 .3377622 2.376814 skm | -1.153259 .6296633 -1.83 0.067 -2.387376 .0808588 sed | -.4126054 .3042976 -1.36 0.175 -1.009018 .183807 om | 1.025657 .4991371 2.05 0.040 .0473663 2.003948 mercado | -.1340548 .2612908 -0.51 0.608 -.6461754 .3780659 prod | -1.52528 1.607574 -0.95 0.343 -4.676067 1.625507 perd | 29.38259 11.69655 2.51 0.012 6.457774 52.30741 ls1 | -.7531871 .9165113 -0.82 0.411 -2.549516 1.043142 tipown | -8.964998 2.672312 -3.35 0.001 -14.20263 -3.727362 tipofis | 8.054049 4.024678 2.00 0.045 .165825 15.94227 _cons | 10.03413 3.958654 2.53 0.011 2.275306 17.79295 -------------+---------------------------------------------------------------sigma_u | 5.4264767 sigma_e | 3.1408682 rho | .74905556 (fraction of variance due to u_i) -----------------------------------------------------------------------------. testnl _b[tipown]/_b[tipofis] = 0 (1) _b[tipown]/_b[tipofis] = 0 chi2(1) = Prob > chi2 = 5.27 0.0217

123

. . . . . .

*----------------------------------------------------------------------------* Hiptesis H5 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 68.22 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0201 between = 0.5099 overall = 0.4313

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.379052 .8602537 2.77 0.006 .6929854 4.065118 skm | -3.065697 1.055833 -2.90 0.004 -5.135091 -.9963029 sed | -1.229937 .5212016 -2.36 0.018 -2.251474 -.2084011 om | 2.322827 1.037354 2.24 0.025 .2896511 4.356003 mercado | -.7211557 .4250488 -1.70 0.090 -1.554236 .1119246 prod | 2.447953 3.459857 0.71 0.479 -4.333241 9.229147 perd | 35.51765 17.13544 2.07 0.038 1.932798 69.1025 ls1 | -.5572629 1.47929 -0.38 0.706 -3.456619 2.342093 tipown | -15.75624 5.55823 -2.83 0.005 -26.65017 -4.862307 tipofis | 18.46125 6.226994 2.96 0.003 6.256567 30.66594 _cons | 10.57828 4.652487 2.27 0.023 1.459571 19.69699 -------------+---------------------------------------------------------------sigma_u | 9.6815861 sigma_e | 6.3130498 rho | .70165969 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test tipown = 0 ( 1) tipown = 0 chi2( 1) = Prob > chi2 = 8.04 0.0046

. xtreg saifi la skm sed om mercado prod perd ls1 tipown tipofis, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0073 between = 0.5090 overall = 0.4593 Number of obs Number of groups = = 309 58 2 5.3 6 65.06 0.0000

Obs per group: min = avg = max = Wald chi2(10) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.357288 .5201759 2.61 0.009 .3377622 2.376814 skm | -1.153259 .6296633 -1.83 0.067 -2.387376 .0808588 sed | -.4126054 .3042976 -1.36 0.175 -1.009018 .183807 om | 1.025657 .4991371 2.05 0.040 .0473663 2.003948 mercado | -.1340548 .2612908 -0.51 0.608 -.6461754 .3780659 prod | -1.52528 1.607574 -0.95 0.343 -4.676067 1.625507 perd | 29.38259 11.69655 2.51 0.012 6.457774 52.30741 ls1 | -.7531871 .9165113 -0.82 0.411 -2.549516 1.043142 tipown | -8.964998 2.672312 -3.35 0.001 -14.20263 -3.727362 tipofis | 8.054049 4.024678 2.00 0.045 .165825 15.94227 _cons | 10.03413 3.958654 2.53 0.011 2.275306 17.79295 -------------+---------------------------------------------------------------sigma_u | 5.4264767 sigma_e | 3.1408682 rho | .74905556 (fraction of variance due to u_i) ------------------------------------------------------------------------------

124

. test tipown = 0 ( 1) tipown = 0 chi2( 1) = Prob > chi2 = . . . . . . 11.25 0.0008

*----------------------------------------------------------------------------* Hiptesis H6 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042 om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524 prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 rorr | -21.10404 6.482052 -3.26 0.001 -33.80863 -8.399451 pcap | -5.748643 2.58658 -2.22 0.026 -10.81825 -.6790393 _cons | 28.46052 5.979177 4.76 0.000 16.74154 40.17949 -------------+---------------------------------------------------------------sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = pcap + tipown ( 1) - tipown + rorr - pcap = 0 chi2( 1) = Prob > chi2 = 0.08 0.7840

. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544

125

rorr | -7.944138 4.126756 -1.93 0.054 -16.03243 .144155 pcap | .1711495 1.429165 0.12 0.905 -2.629963 2.972262 _cons | 18.11895 3.043269 5.95 0.000 12.15425 24.08365 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = pcap + tipown ( 1) - tipown + rorr - pcap = 0 chi2( 1) = Prob > chi2 = . . . . . . 0.06 0.8035

*----------------------------------------------------------------------------* Hiptesis H7 *----------------------------------------------------------------------------xtreg saidi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Number of obs Number of groups = = 309 58 2 5.3 6 74.03 0.0000

Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0216 between = 0.5339 overall = 0.4464

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

-----------------------------------------------------------------------------| Robust saidi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 2.854153 .8090492 3.53 0.000 1.268446 4.43986 skm | -2.989169 1.008276 -2.96 0.003 -4.965354 -1.012984 sed | -1.016365 .4969278 -2.05 0.041 -1.990325 -.0424042 om | 2.223509 1.000083 2.22 0.026 .2633832 4.183636 mercado | -.5427978 .4423807 -1.23 0.220 -1.409848 .3242524 prod | 1.862785 3.470459 0.54 0.591 -4.939189 8.664759 perd | 43.27146 17.85262 2.42 0.015 8.280964 78.26196 ls1 | -.5915788 1.483238 -0.40 0.690 -3.498672 2.315514 tipown | -14.16416 5.523327 -2.56 0.010 -24.98968 -3.338635 rorr | -21.10404 6.482052 -3.26 0.001 -33.80863 -8.399451 pcap | -5.748643 2.58658 -2.22 0.026 -10.81825 -.6790393 _cons | 28.46052 5.979177 4.76 0.000 16.74154 40.17949 -------------+---------------------------------------------------------------sigma_u | 9.615314 sigma_e | 6.3130498 rho | .69877605 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = tipown ( 1) - tipown + rorr = 0 chi2( 1) = Prob > chi2 = 1.83 0.1764

. xtreg saifi la skm sed om mercado prod perd ls1 tipown rorr pcap, re sa robust Random-effects GLS regression Group variable (i): empresa R-sq: within = 0.0075 between = 0.5075 overall = 0.4586 Number of obs Number of groups = = 309 58 2 5.3 6 66.12 0.0000

Obs per group: min = avg = max = Wald chi2(11) Prob > chi2 = =

Random effects u_i ~ Gaussian corr(u_i, X) = 0 (assumed)

126

-----------------------------------------------------------------------------| Robust saifi | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------la | 1.343546 .5268945 2.55 0.011 .3108516 2.37624 skm | -1.160788 .6299439 -1.84 0.065 -2.395456 .073879 sed | -.4203361 .3176643 -1.32 0.186 -1.042947 .2022745 om | 1.032096 .4994154 2.07 0.039 .0532604 2.010933 mercado | -.1331915 .2643485 -0.50 0.614 -.651305 .3849219 prod | -1.506952 1.624087 -0.93 0.353 -4.690103 1.6762 perd | 28.86231 11.75306 2.46 0.014 5.826745 51.89788 ls1 | -.7470988 .9149828 -0.82 0.414 -2.540432 1.046235 tipown | -9.014742 2.718519 -3.32 0.001 -14.34294 -3.686544 rorr | -7.944138 4.126756 -1.93 0.054 -16.03243 .144155 pcap | .1711495 1.429165 0.12 0.905 -2.629963 2.972262 _cons | 18.11895 3.043269 5.95 0.000 12.15425 24.08365 -------------+---------------------------------------------------------------sigma_u | 5.4943971 sigma_e | 3.1408682 rho | .7537028 (fraction of variance due to u_i) -----------------------------------------------------------------------------. test rorr = tipown ( 1) - tipown + rorr = 0 chi2( 1) = Prob > chi2 = 0.08 0.7735

. . log close log: C:\Program Files\Stata9\apendice_e.log log type: text closed on: 28 Sep 2009, 10:18:46 -----------------------------------------------------------------------------------------

127