Report No. 36624-BR Brazil How to Revitalize Infrastructure Investments in Brazil Public Policies for Better Private Participation (In Two Volumes) Volume II: Background Report January 10, 2007 Finance, Private Sector and Infrastructure Management Unit Latin America and the Caribbean Region Document of the World Bank ACKNOWLEDGEMENTS This report was prepared by Paulo Correa, Luis Albert0 Andres, Mariam Dayoub, Katherine S. Coleman(World Bank) and Rodrigo Moita (University of Illinois at Urbana Champaign), with inputs from Marianne Fay and Tito Yepes (Chapter 3); Maria Elena Pinglo (Chapter 4); Aymeric-Albin Meyer, Carlos Velez, Jayme Port0 Carreiro, Luiz Maurer,Gar0Batmanian, andRaffaella Mota(Chapter5); underthe supervisionof Susan Goldmark andJos6Luis Guasch. Background papers were commissioned to Caio Mario (FGV-SP); Claudio Considera (UFF); Cleveland Prates (FGV-SP); Marcel0 Neri (FGV-RJ); and Ronald0 Seroa da Motta (IPEA). Comments provided by Aymeric-Albin Meyer, Carlos Velez, Luiz Maurer,Jose GuilhermeReis,Angela Furtado,andMarianne Fayto anearlier draft of the documentwere greatly appreciated. The analytical underpinnings of this report are based on the work done at regional level by M. Fay and M. Morrison (Chapter 1); L. Serven and C. Calder6n (Chapter 3); L.A. Andres, S. Sirtaine, M.E.Pinglo, J.L. Guasch, and V. Foster (Chapter4); andJ.L. Guasch (Chapter 6). PPIAF support to the background paper (Chapter 7) "Regulatory Governance in Infrastructure Industries-Assessment and Measurement of BrazilianRegulators," PPIAF OccasionalPaper No. 3, preparedby P. Correa, C. Pereira (MSU), B. Muller (UNB) and M.Mello (UFPE), is alsoacknowledged. Armando Castelar Pinheiro(IPEA); Clive G. Harris (SASEI, World Bank); Jost Cltiudio Linhares (IADB); and Mark Roland Thomas (World Bank, PRMED) were the peer reviewersfor the study. The report consolidatesthree years of non-lendingtechnical assistance work providedby the World Bank to the Brazilian Government.Parts of the report or preliminary versions of the chapters were discussedin different occasions with country authorities, including the Ministry of Finance, the Ministry of Planning, the Civil Cabinet, and the Tribunal de Contasde UniCo. 3 HOW TO REVITALIZE INFRASTRUCTUREINVESTMENTSINBRAZIL: PUBLICPOLICIESFORBETTERPRIVATEPARTICIPATION Table of Contents Acronyms and Abbreviations .............................................................................. 8 PART I:TAKING STOCK OFTHE SITUATION ........................................................... .......................................... 11 1. INFRASTRUCTUREINBRAZIL 11 1.1 InternationalBenchmarking ...................................................................... 11 1.2 Access ByHouseholds .............................................................................. 1.3 Access by Firms........................................................................................ 13 23 2. 2.1. InfrastructureInvestmentsinthe Last Two Decades................................ INFRASTRUCTUREINVESTMENTS INBRAZIL.............................. 28 2.2. Why did Public InvestmentsDecline? ....................................................... 28 2.3. Where did all the privateinvestmentgo?................................................... 30 3.1. I s Infrastructurea Binding Constraint to GrowthinBrazil?...................... 3. INFRASTRUCTUREAND GROWTH: HOW MUCHIS NEEDED? 38 ...35 3.2. Infrastructure.Growthand SocialPerformance......................................... 38 41 PART 11: HOW TO REVITALIZE INFRASTRUCTUREINVESTMENTS .47 3.3. How Much i s Neededinthe LongRun...................................................... 4 4.1 Were PrivateInvestmentsAdequately Profitable?.................................... . PRIVATEPARTICIPATIONS ININFRASTRUCTUREINBRAZIL..51 51 51 4.2 Was Privatizationa Failure?..................................................................... 5 REGULATORY AND POLICY ISSUES .................................................... . 59 69 5.1 The TelecommunicationsSector.. ............................................................. 69 5.2 The Power Sector ...................................................................................... 5.3. The TransportSector................................................................................. 70 79 5.4 Water and Sanitation ................................................................................. 5.4 EnvironmentalRequirements.................................................................... 87 5.5. Other RegulatoryIssues ............................................................................ 95 97 6..CONTRACT RENEGOTIATION.............................................................. 100 7.1 Autonomy................................................................................................ 7 REGULATORY GOVERNANCEAND EFFECTIVENESS................ 107 7.2. Decision-Making..................................................................................... 112 111 ......................................................................................... 7.4. Accountability......................................................................................... 7.3. DecisionTools 115 7.5. Summaryof the Results.......................................................................... 116 117 8. CONCLUSIONSAND POLICY IMPLICATIONS .............................. 118 References ....................................................................................................... 128 4 Listof Annexes Annex 1: The Impact of Access to Infrastructure on Education Performance in Brazil.............133 Annex 2: Methodology for the Calculation of Concession Returns.................................. Annex 4: Recent Evolutions of Water and Sanitation Sector inBrazil.............................. Annex 3: The Impact of Privatization on Infrastructure Industries: Methodological Appendix..193 196 208 ListofTables Table 1.2. 1Brazil and selected states: infrastructure access share (ranking). 1991 and 2000......16 Table 1.2. 2 Brazil: level of access to infrastructure services according to income decile. 1991 and 20 Table 1.3. 2 Infrastructureservices as obstaclesto growth by industry......................................... Table 1.3. 1Infrastructureservices as obstaclesto growth by state.................................... 2000....................................................................................................................................... 24 Table 1.3. 3 Annual loss inproduction by state (inpercent).......................................................... 24 Table 1.3. 4 Annual loss inproduction by industry (inpercent) .................................................... 25 25 Table 1.3. 5 Average number of days and hours of interruption of infrastructure services by state ............................................................................................................................................... 26 Table 1.3. 6 Brazil. China. India and Malaysia: port transit times (in days).................................. Table 2.2. 1Brazil: investments in infrastructure by sector. 1981-90 (percent GDP)...............27 31 Table 2.2. 2 Brazil: summary of the effects of privatization on the performance of utility Table 2.3. 1. Private investments in infrastructure (in US$ billion). 1990-2004., ...................68 companies.............................................................................................................................. 36 Table 3.1. 1 Brazil: gross fixed capital formation by economic sector (in percent of GDP). 1971- 2000...................................................................................................... 40 Table 3.2. 1 Siio Paulo state: change in educational performance associated with access to infrastructure (inpercent)............................................................................. 46 Table 3.2. 2 Si30 Paulo state: results for difference in percentage for scenarios with good conditions in relation to those with bad conditions inthe presence of infrastructure variables (in percent)........................................................................................................................................... 46 Table 4.2. 1 Brazil and L A C without Brazil: pre- and post-privatization comparisons per sector Table 5.2. 1Disputes pertaining to property rights on energy in2001.............................. (inlevels) ..................................................................................................... 60 Table 5.2. 2 Firstnew energy auction results. 2008-2010 ............................................................. 72 74 Table 5.2. 3 Brazilian railway operators: gross revenues and operational profits (RS$ million). Table 5.3. 1Comparative analysis on the Brazilian sewerage policy................................. 2001 and 2004 ....................................................................................................................... 86 92 Table 6. 1L A C and selectedcountries: fraction of contracts renegotiated ......................... Table 5.3. 2 Summary of the main sector issues............................................................................ 94 102 Table 7.2. 1Decision-making: descriptive results of selectedquestions............................. 114 ListofFigures Figure 1.1. 1Electricity coverage. 2000.................................................................. -11 Figure 1.1. 2 Electricity: distributional losses. 2002 (% of output) ............................................... Figure 1.1.3 Paved road coverage. 1999....................................................................................... 11 12 12 Figure 1.1. 6 Improved water source coverage. 2002 (% of population with water access)..........13 Figure 1.1.5 Fixed telecommunications coverage. 2002............................................................... Figure 1.1.4 Paved road coverage. 1999....................................................................................... 13 14 Figure 1.2. 3 Brazil: population absolute growth. 1991/2000........................................................ Figure 1.2. 1Brazil: population growth rate (96). 1991/2000........................................... 14 5 Figure 1.2.4 Brazil: share of population with access to infrastructure services (96). 1991 and 2000 ............................................................................................................................................... 14 Figure 1.2. 5 Brazil: population growth per state, 1991/2000........................................................ Figure 1.2. 6 Brazil: population growth rate per state (%), 1991/2000.......................................... 15 15 Figure 1.2. 8 Brazil: population with water access growth rate per state (%), 1991/2000.............15 Figure 1.2. 7 Brazil: change inpopulation with water access per state, 1991/2000 ...................... Figure 1.2. 9 Brazil: population with telephone access growth rate per state (%), 1991/2000......15 Figure 1.2. 8 Brazil: change inpopulation with telephone access per state, 1991/2000................15 Figure 1.2. 10Brazil: change in population with sewerage access per state, 1991/2000 ..............15 Figure 1.2. 11Brazil: population with sewerage access growth rate per state (%), 1991/2000.....16 Figure 1.2. 12Brazil: change in population with electricity access per state, 1991/2000 ............. 16 16 Figure 1.2. 13 Brazil: population with electricity access growth rate per state (%), 1991/2000....16 Figure 1.2. 14 Brazil: probability of having access to infrastructure by municipality category ...17 Figure 1.2. 15 Brazil: share of population with sewerage access per household situation (%), 1991 and 2000 ................................................................................................................................ 18 Figure 1.2. 16 Brazil: share of population with water access per household situation (%), 1991 and 2000................................................................................................................................ 18 Figure 1.2. 17 Brazil: share of population with telephone access per household situation (%), 1991 and 2000 ....................................................................................................................... 19 Figure 1.2. 18 Brazil: share of population with infrastructure services in subnormal human agglomerates (%), 1991and 2000......................................................................................... 19 Figure 1.2. 19 Brazil: share of population with electricity access per household situation (%), 1991 and 2000 ....................................................................................................................... 19 Figure 1.2. 20 Brazil: share of population with infrastructure services per income deciles (%), 1991 and 2000 ....................................................................................................................... 21 Figure 1.2. 21 Brazil: probability of having access to infrastructure by gender and years of Figure 1.3. 2 Share of firms' electricity from own generator by size and state (inpercent)..........26 Figure 1.3. 1Firmsowing power generators by size and state.,........................................ Figure 1.2. 22 Brazil: probability of having access to infrastructure by age and race...................22 education ............................................................................................................................... 22 Figure 2.1. 1. Infrastructure investments in Brazil: total, public, and private (percent of GDP), 26 1980-2001 ............................................................................................. 29 Figure 2.1. 2 Investments in infrastructure: Chile and Colombia in 1980-2001(percent of GDP)30 Figure 2.2. 1Public administrationallocation of expenditures (percent of GDP), 1995-2002.....33 Figure 2.2. 2 Public sector investment in infrastructure (percent of GDP),................................... 35 Figure 2.3. 1Private investment ininfrastructure inBrazil, by sector (1994-2002) .................36 Figure 2.2. 3 Brazil: primary surplus and public investment in infrastructure (percent of GDP),-35 Figure 2.3. 2 Private investment ininfrastructure in Brazil, by modality (1994-2002) .................36 Figure 2.3. 3 Composition of private flows to infrastructure by investment modality for selected countries (in percent), 2003................................................................................................... 37 Figure 2.3. 4 Composition of private flows to infrastructure by sector for selected countries (in Figure 3.1. 1. Perceptions of obstacles to growth in Brazil (inpercent) .............................. percent), 2003 ........................................................................................................................ 37 41 Figure 3.2. 1 Brazil: average productivity (TFP) impact (gains and losses) of ICA variables (in Figure 3.2.2 Brazil: impact of ICA variables on firms' probability of exporting (inpercent) ......43 percent) ................................................................................................. 42 Figure 4.1. 2 Returns on equity for investments inenergy, telecommunications, and water.........52 Figure 4.1. 1Returns on infrastructure concessions inBrazil.......................................... 54 55 Figure 4.1. 4 Brazil: WACC per year, 1998-2003 ......................................................................... Figure 4.1. 3 Average concession return by country..................................................................... 58 6 Figure 4.1. 5 Brazil: Cost of equity for selected infrastructure sectors and country risk premium. 1996-2003 (inpercentage)..................................................................................................... 58 Figure 4.2. 1Brazil and L A C without Brazil: cross-period comparisons for labor productivity per sector.................................................................................................... 62 Figure 4.2. 2 Brazil and L A C without Brazil: cross period comparisons for distributional losses per sector (inlevels) .............................................................................................................. 63 Figure4.2. 3 BrazilandLAC without Brazil: cross periodcomparisons for coverage per sector (in Figure 4.2. 4. Brazil and L A C without Brazil: cross period comparisons for real price in local levels) .................................................................................................................................... 64 currency per sector (in levels) ............................................................................................... 66 Figure4.2. 5 Brazil and LAC without Brazil: cross periodcomparisons for price inUS dollars per Figure 7. 1Regulatory governance, effectiveness and industry performance..................... sector (inlevels) .................................................................................................................... 67 109 List of Boxes Box 1The Case of the Water Sector: I s There a Paradox?............................................................ Box 2 The New Legal Framework for the Natural Gas Industry (PLS No. 226/2005) ..............77 58 Box 3 RegulatingEssential Facilities inPractice: The Case of Port0 de Santos (Resolution ANTT Box 4 Environmental Bottlenecks inPractice................................................................................ No.9492005) ........................................................................................................................ 86 Box 5 Meetingthe Regulators: Where They Come from and Where They Go...................... 97 Box 6 "Buying Credibility" throughPartial RiskGuarantees (PRG) against Regulatory Risk...123 112 7 ACRONYMSAND ABBREVIATIONS ALL America LatinaLogist,ica S.A. Latin America LogisticsS.A. ANA AgCnciaNacionalde Aguas NationalWater Agency ANATEL AgCnciaNacionalde TelecomunicaCBes NationalTelecommunications Agency ANCINE AgCnciaNacionaldo Cinema NationalMovie Agency ANEEL AgCnciaNacionalde EnergiaElttrica NationalElectricity Agency A N S AgEnciaNacionalde Sadde NationalHealthAgency ANP AgCnciaNacionaldo Petr6leo NationalPetroleumAgency ANTAQ AgCnciaNacionalde TrasportesAquaviCios NationalWater Transport Agency ANTT AgCnciaNacionalde TransportesTerrestres NationalLand Transport Agency ANVISA AgEncia Nacionalde Vigilhcia Sanitiria NationalFoodand DrugAgency AP AutoridadePortuiria PortAuthority BNDES BancoNacionalde DesenvolvimentoEconamicoe NationalBank for EconomicandSocial Social Development BNDESpar EmpresasParticipantesdo Banco Nacionalde ParticipatingCompaniesof the National DesenvolvimentoEconBmicoe Social Bankfor Economicand Social Development CADE ConselhoAdministrativo de DefesaEconBmica EconomicDefenseCouncil CAP Conselhoda AutoridadePortuiria Port Authority Council CAPM Modelo de ValoraGBode Ativos CapitalAsset Pricing Model CCEE Cimara de ComercializaCbde EnergiaElktrica EnergyTradingChamber CE Custo Patrimonial Cost of Equity CETESB Companhiade Tecnologiade Saneamento Ambiental EnvironmentalTechnologyCompany CFN CompanhiaFerroviCia do NordesteSA. Railway Companyof NortheastS.A. CONAMA ConselhoNacionalde Meio Ambiente NationalEnvironmentalCouncil CONPORTOS Comissb Nacionalde SeguranCaWblica no Portos, NationalCommissionfor Public Safety in Terminaise Vias Navegiveis Ports COPASA Companhiade Saneament de MinasGerais Sewerage Company of Minas Gerais State CNPE ConselhoNacionalde PolfticaEnergetica NationalEnergy Policy Council CPFL CopmanhiaPaulistade ForGae Luz Electricity Companyof Si30 Paul0 State CREMA Contratode RestauraGbe ManutenGbda Malha Federal RoadNetwork Maintenance RodoviCiaFederal Contract CVRD CompanhiaVale de Rio Doce Vale do Rio Doce Company DEA Anilise de Envolt6ria de Dados DataEnvelopmentAnalysis DILIQ Diretoria de Licenciamentoe QualidadeAmbiental Office of Licensingand Environmental Quality DNEF DepartamentoNacionaldas Estradasde Ferro DNER DepartamentoNacionalde Estradas de Rodagem NationalHighway Agency DNIT DepartamentoNacionalde Infra-estruturade Department of Transport Infrastructure Transporte E&P ExploraCiio e ProduCgo Exploration andProduction EAP Asii Oriepl e Pacific0 East Asia and the Pacific ECA Europae AsiaCentral Europe & CentralAsia EDP Energiasde Portugal Portugal'sEnergy EFC Estradade FerroCarajis Carajis Railway EFVM Estradade FerroVit6ria aMinas Vit6ria to MinasRailway EIA Estudode Impact0Ambiental EnvironmentalImpactAssessment EMBASA EmpresaBaianade Aguae Saneamento Water and SewerageCompanyof Bahia State EPE Empresade PesquisaEnergetica EnergyResearchCompany FCA FerroviaCentro-AtlhticaS.A. FEEMA Fundal50Estadualde Engenhariado Meio Ambiente State EnvironmentalEngineeringFund FEPASA FerroviasPaulistasS.A. SBo Paul0State RailwayS.A. FERROBAN FerroviasBandeirantesS.A. BandeirantesRailwayS.A. FERRONORTE FerroviasNorte do Brad S.A. Railway of the North of Brazil S.A. FERROESTE Estradade FerroParan6Oeste S.A. ParaniWest RailwayS.A. FGTS Fundode Garantiapor Tempo de ServiCo PrivateSector SeveranceFund FGV FundaqBoGetdlioVargas Get6lio Vargas Foundation 8 FTC FerroviaTerezaCristin?S.A. TerezaCristinaRailwayS.A. FUNAI Fundaqgo Nacionaldo Indio National Foundationfor Indigenous Peoples FUNASA Fundaqb Nacionalde Sdde National HealthFoundation FUST Fundode Universalizagb dos Serviqos de UniversalizationFundfor Telecomunicaqdes TelecommunicationServices GDP Produto InternoBruto GrossDomesticProduct GWh Gigawatthora Gigawatthour IADB BancoInteramericanode Desenvolvimento Inter-AmericanDevelopment Bank IBAMA InstitutoBrasileirodo Meio Ambientee dos Recursos Brazilian Institutefro the Environmentand Naturais Renoviveis Renewable NaturalResources IBGE Instituto Brasileirode Geografiae Estatistica Brazilian Institutefor Geography and Statistics IBRD Banco InternacionalparaReconstruqioe InternationalBank for Reconstructionand Desenvolvimento Development ICA Andise do Clima de Investimentos InvestmentClimate Assessment EMS Impostosobre Circulagb de Mercadorias e Serviqos State Sales Tax ICs Pesquisasobre o Climade Investimentos InvestmentClimateSurvey IFC Corporaqb FinanceiraInternacional InternationalFinanceCorporation IGP-DI IndiceGeral de Preqos- DisponibilidadeInterna GeneralPriceIndex- Domestic Availability IGP-M hdice Geralde Preqosdo Mercado GeneralMarketPriceIndex IMO OrganizaqLoMaritima Intemacional InternationalMaritime Organization IPA-OG Indicede Pregopor Atacado - OfertaGlobal/ Wholesale PriceIndex-Global Miquinas e EquipamentosIndustriais Supply/IndustrialMachineryand Equipment IPCA hdice de Preqosao Consumidor Amplo ConsumerPriceIndex IPEA Institutode PesquisaEconBmicaAplicada Institutefor Applied EconomicResearch IPHAN Institutodo PatrimBnio Hist6rico e Artistic0 Nacional NationalInstitutefor Historic and Artistic Assets IPP Auto-produtorEnergCtico IndependentPowerProducer IRR Taxa Internade Retorno Internal Rateof Return ISPS Terminal Portuirio Intemacional InternationalShip andPortFacility IST hdide de Servigosde Telecomuniq6es TelecommunicationServices Index ITU Uniiio 1nt:rnacional paraas Telecomuniqdes InternationalTelecommunications Union IUCLIG ImpostoUnico sobre Combustiveis e Lubrificantes Tax on Liquid andGas Fuels and Liquidos e Gasosos Lubricants IUEE ImpostoUnico sobre EnergiaElttrica Tax on Electricity LAC AmCricaLatinae Caribe Latin Americaand the Caribbean MDGs Metasde Desenvolvimentodo Milenio Millennium Development Goals MENA Oriente MCdioe Africa do Norte Middle East and North Africa MME MinistCrioda Minas e Energia Ministry of MinesandEnergy MRS MRS LogisticaS.A. MRS LogisticsS.A. MT MinistCriodos Transportes Ministry of Transportation MWh Megawatt hora Megawatthour NOVOESTE FerroviaNovoesteS.A NovoesteRailwayS.A. OECD Organizagb paraCooperaqtioEconBmicae Organizationfor EconomicCooperation Desenvolvimento and Development OEMA Orgb Estadualde Meio Ambiente State EnvironmentInstitutes OGMO Orgb Gestor de M2o-de-Obra Labor ManagementOrganization PAEG Planode AqSo EconBmicado Govern0 Governmental EconomicAction Plan PAS00 Plano Alternativo de Serviqo de OfertaObrigat6ria Alternative Planof MandatorySupplied Services PCG GarantiaParcialde Crtdito PartialCredit Guarantee PLANASA Plano Nacionalde Saneamento NationalSeweragePlan PMGU Plano Geral de Metas de Universalizaqlo General Planon UniversalizationGoals PNAD PesquisaNacionalpor Amostrade Domicflios Brazilian HouseholdSurvey PNR NationalRoadPolicy PPI ParticipaqioPrivadaem Infra-Estrutura PrivateParticipationin Infrastructure PPP ParceriaWblico-Privada Public-PrivatePartnerships PRG GarantiaParcialde Risco PartialRisk Guarantee PSR Pesquisaem SistemasEnergCticos PowerSystems Research 9 RFFSA RedeFerroviLiaFederal S.A. FederalRailway Network S.A. RIMA Relat6riode Impact0Ambiental Environmental ImpactAssessment Report RoCE Retorno sobre o CapitalEmpregado Returnon CapitalEmployed ROE Retorno sobre PatrimBnio Returnon Equity S&P Standard& Poor's Standard& Poor's SABESP Companhiade SaneamentoBisico do Estadode SZo Sewerage Company of SZo Paulo State Paulo SANEPAR Companhiade SaneamentodoEstadodo Parani Water and Sewerage Companyof Parani State SISNASA SistemaNacionalde SaneamentoAmbiental National Sewerage System SNIS SistemaNacionalde Informa@essobre Saneamento National Sewerage Information System SNV SistemaNacionalde ViagBo NationalTransportInfrastructureSystem SOE Empresas Wblicas State-OwnedEnterprises TCU Tribunal de Contasda UniB? General AccountingOffice TKU Toneladaspor QuilBmetroUti1 Tons per UsedKilometer TV Valor Final Terminal Value VALEC FerroviasNorte Sul S.A. Norte SUI Railway S.A. VAT Impost0sobre o Valor Adicionado Value-AddedTax YPFB Dep6sitosPetroliferosFiscaisBolovianos Fiscal Oil Depositof Bolivia WACC Taxa de Custo MCdioPonderadodo Capital WeightedAverage Cost of Capital WDI IndicadoresGlobaisde Desenvolvimento World DevelopmentIndicators 10 How to Revitalize InfrastructureInvestments inBrazil: Public Policies for Better Private Participation PARTI:TAKING STOCK OFTHE SITUATION 1. INFRASTRUCTUREINBRAZIL This chapter compares the stock of infrastructure in Brazil with its international counterparts based on data available in the World Development Indicators (WDI) database. It also presents household access to infrastructure services inBrazil interms of regions, income levels and other socio-demographic characteristics, using data from the Brazilian Census and the Brazilian Household Survey (Pesquisu Nucional por Amostru de Domicilios, PNAD).Finally, it examines the needs of the private sector according to a 2003 World Bank Investment Climate Assessment (ICA) for Brazil. 1.1INTERNATIONAL BENCHMARKING Electricity 1.1. Electricity indicators show that Brazil ranks better than or near its peers in the LAC region (in aggregate). Per capita electricity consumption was 18 percent higher in Brazil than in L A C in 2002. The fact that Brazil's consumption i s only 63 percent the value of that of Europe & Central Asia (ECA) is likely due to the harsh winters inECA. For electricity coverage in2000, measured as a share of total population, Brazil was a leader in the L A C region with a coverage value above the other three regions (Figure 1.1. 1).When distributional losses were considered, the Brazilian level i s about the same as the L A C region - and not as low as East Asia & Pacific (EAP), ECA, and Middle East and NorthAfrica (MENA) (Figure 1.1.2). Figure 1.1.1Electricity coverage, 2000 Figure 1.1.2 Electricity: distributional losses, 2002 (% of DoDulationi (% Of output) I io01 ac, ^^ 1 -- I 26 90 80 32 70 c 60 50 ,"40 0 30 20 i o 0 Brazil Latin Europe8 Middle East EastAsia &SouthAsk Brazil LatinAmerica MiddleEast & EastAsia 8 SouthAsia America8 CentralAsia 8 North Pacfic &Caribbean NorthAfrica Pacific Caribbean Africa Note: For electricity coverage in2000, no data was Source: World Bank/ World Development available for Europe and Central Asia. Indicators. Source: World Bank / World Development Indicators. 11 Road Transportation 1.2. Brazil's paved roads network lags behind the average for L A C and other regions. For coverage, two indicators were computed. The first considers the length of paved roads per 100km2of territorial area. It shows that Brazil's coverage i s less than a third of the LAC average and less than one-tenth the coverage of South Asia (the region with the highest value) (Figure 1.1. 3). Second, the length of paved road per 1,000 inhabitants was calculated. Results show that Brazil's coverage i s less than one-third that of LAC, but at least the Brazilian figure is not as bad as those for E C A and EAP-highly populated regions, encompassing countries such as Bangladesh, China, and Lndia (Figure 1.1.4). It is important to keep inmindthat Brazilhas a vast area covered with forest (40 percent of its territory i s covered by the Amazon)' and a relatively low population density compared to the rest of the L A C regioa2 Figure 1.1.3 Pavedroad coverage, 1999 Figure 1.1.4 Pavedroad coverage, 1999 (kmof paved roadper 100km2of land) (kmof pavedroadper 1,000 inhabitants) 3 2.0 .-m E I 1.59 n LU .p E 1.5 r 0 .- 15 0 W 9 :1.0 U g 10 W n U W m 9 5 E 0.5 -a .- c (Y 0 >m 5 0 n 0.0 Brazil Latin Europe8 Middle East EastAsia 8 South Asia Brazil Latin Europe8 Middle East EastAsia & South Asia .x E h r i c a 8 CentralAsia 8 Mrth Pacific h r i c a 8 CentralAsia 8 Mrlh Pacific Caribbean Africa Caribbean Africa Source: World Bank / World Development Source: World Bank / World Development Indicators. Indicators. Telecommunications 1.3. In telecommunications, indicators show that figures for Brazil are much better than those for L A C and, in general, surpass those for other regions. Three indicators were assessed for fixed telecommunications: labor productivity, coverage and price of a three-minute call. First, for labor productivity, the number o f mainlines per employee in Brazil i s more than twice that o f L A C and at least four times the figures for other regions. Second, for coverage, Brazil has 53 additional mainlines per 1,000 inhabitants than LAC, and its total value i s close to that of ECA and much better than the other three regions (from 85 to 183 additional mainlines per 1,000 inhabitants) (Figure 1.1.5). Third, for the US$ price of a three-minute call, the cost inBrazil is one-third the average cost in LAC - a fact related to the Brazilian currency devaluation in 1999. This third indicator is similar to (inthe case of ECA) or higher than other regions. 1See Fonseca, V.I. and F.D.Por (2005). 2For example, in 2002, the population density in Brazil was about 20 inhabitants per km2,while it was around 30 inhabitants per km2inthe rest of LAC. 12 1.4. In terms of mobile telecommunications, the coverage indicator presents similar characteristics as those of fixed telecommunications: higher than L A C (11 additional mobiles per 1,000 inhabitants), close to ECA, and much better than values in the remaining regions. It i s worth noting that the expansion of fixed telecommunications requires higher investment levels than expansion of mobiles, and that these two goods are substitutes for consumers. Also, the difference in fixed coverage between Brazil and L A C (in absolute value) i s higher than the difference in mobile coverage, since telecommunications sector reforms were completed by 2002 in Brazil, but not in all of the region. As a result, restriction of fixed telecommunications in LAC led to a more pronounced expansion of the mobile segment inthe region overall than was observed inBrazil. Water 1.5. Water coverage in Brazil is at levels either comparable to or better than regional levels. The share of population with water access in Brazil i s the same as that for LAC and just one percentage point below the value for MENA - a region where droughts and deserts are common (Figure 1.1. 6). Water coverage in Brazil i s two percentage points less than coverage inECA. To reach ECA levels, Brazil would need to expand access to at least 3.6 million additional p e ~ p l e . ~ Figure 1.1.5 Fixed telecommunications coverage, Figure 1.1.6 Improved water source coverage, 2002 2002 (% of population with water access) (telephone mainlines per 1,000 inhabitants) -2507 237 I -" I 91 0 c - Bradl Latin Europe & East &!a & SouthAsia D- MiddleEast Brazil Latin Europe8 Middle East EastAsia &SouthAs al &rea & CenbalAsla & NsrihAfrica Pactic America8 CentralAsia 8 North Pacific al b Caribbean Caribbean Africa Source: International Telecommunications Union/ Source: World Bank / World Development ITU. Indicators 1.2 ACCESSBYHOUSEHOLDS4 1.6. Access to infrastructure services has increased significantly over the 1990s. Using census data on infrastructure services, the service that had the highest increase in access between 1991 and 2000 was fixed telephony. In 1991, around 25 million people This calculation considers the Brazilianpopulation to be 180millionpeople. M.Neri(2005). "Acesso aServiGosde Infra-Estrutura" served as backgroundpaper for this section. 13 had access to fixed lines, and in2000 the number increased to 63 million--an increase of 145 percent in the number of users. The population growth during this period was only 15.7 percent, as the following graphs show. 1.7. Despite its rapid growth, fixed telephony service is the infrastructure service that presents the lowest level of access (37.1 percent) when compared to other services. On the other extreme i s electricity, with an access level of 92.7 percent. Access to potable water has an intermediate rate of 75 percent, but significantly more than sewage services, at 50 percent (Figures 1.2.1, 1.2.2 and 1.2.3). Figure 1.2.1 Brazil: population growth rate (%), Figure 1.2.2 Brazil: population absolute growth, 1991/2000 1991/2000 I 35 M)o,ooo Total wim access With access wim access With access Total With a c c w With acces With access With access 10 telephone to Bewerage to water to eiectricity to telephone to Sewerage to water to electncitf ColleCbOn ColleCtion Figure 1.2.3 Brazil: share of population with access to infrastructure services (%), 1991and 2000 90 80 70 60 50 40 30 20 10 0 Telephone Electnaty /*1991.20001 Water Sewerage Source: IBGEICensoDemogrcifico 1991and 2000. HouseholdAccess by State 1.8. The inequality of access to infrastructure among regions becomes clear when states are ranked according to access. Figures 1.2.4 to 1.2.13 show information about population growth and access to infrastructure. The states with the highest rate of access are in the Southeast: S5o Paulo, Rio de Janeiro, Minas Gerais and the Distrito Federal. There i s more variability among the lower levels of the rankings, but states from the North and Northeast regions tend to be at this end of the spectrum. With respect to electricity, 99 percent of the populations of Distrito Federal and Rio de Janeiro and around 75 percent o f Acre and Piauihave access. Interms of telephone and water, access varies considerably across the different states. Inthe Distrito Federal, access to telephone services i s around 75 percent, while in Maranhgo, access does not reach 15 percent (Table 1.2.1). With respect to sewerage, 83 percent of the Distrito Federal has access and 14 only 3 percent has access in Tocantins. The situation for water services i s similar to telephony (Figure 1.2.7). Figure 1.2.4 Brazil: population growth per Figure 1.2.5 Brazil: population growth rate per state, 1991/2000 state (%), 1991/2000 Figure 1.2.6 Brazil: change inpopulation with Figure 1.2.7 Brazil: population with water water access per state, 1991/2000 access growth rate per state (%), 1991/2000 Figure 1.2.8 Brazil: change inpopulation with Figure 1.2.9 Brazil: population with telephone telephone access per state, 1991/2000 access growth rate per state (%), 1991/2000 1 Source: IBGEICenso Demogrdfico 1991 and 2000. 15 Figure 1.2. 10Brazil: change inpopulation with Figure 1.2. 11Brazil: population with sewerage sewerage access per state, 1991/2000 access growth rate per state (%), 1991/2000 Figure 1.2.12 Brazil: change inpopulation with Figure 1.2. 13Brazil: population with electricity electricity access per state, 1991/2000 access growth rateper state (%), 1991/2000 Source: IBGEICenso Demogrrifico 1991and 2000. 1.9. S5o Paulo state had the highest growth in number of people with access to infrastructure services, followed by Minas Gerais. This variation cannot be explained by the total change in the population of Siio Paulo, since it i s among the states with the smaller variation in population. However, population growth can explain the increase in access in the Northern states shown in the previous graphs. The growth in access i s not necessarily due to any real improvement inaccess to these types of services as compared to different regions inthe country. Table 1.2. 1Brazil and selected states: infrastructure access share (ranking), 1991and 2000 Water Sewerage Telephone Electricity 1991 2000 1991 2000 1991 2000 1991 2000 Brasil 67.51 75.18 Brad 31.79 44.39 Brasii 17.54 37.11 Brasil 82.42 92.69 SBo Paulo 89.51 92.29 DistritoFederal 73.19 82.57 DistritoFederal 44.54 75.33 Distrito Federal 91.80 98.92 DistritoFederal 87.78 87.68 SHo Pauio 73.34 80.11 SHo Paulo 26.91 55.80 Rio de Janeiro 93.36 98.90 Rio de Janeiro 813 3 81.75 MinasGerais 53.13 66.04 MatoGrossodo SUI 18.03 44.06 SBo Paulo 95.93 98.79 Pare 38.92 41.37 Piaui 1.09 3.98 Tocantins 5.79 18.94 Pare 54.13 75.86 Acre 41.82 33.83 Rondbnia 0.24 3.60 Alagoas 7.70 17.91 Acre 43.95 75.08 Rondania 31.28 29.20 Tocantins 0.00 2.85 MaranhHo 6.32 15.23 Piaui 69.21 74.32 Source: IBGEICensoDemogrcifico 1991and 2000. 16 SQaPauXo has the n o r ~ hstate of Amapri. Thi c ~ y he i r i ~ ~by the factcthat ~the ~ ~ ~ e to calculate these results, PN ,does not take into acco~nthe ~ of the northern states. 5%oPaufo conies in second place and T o c a n ~ ~ini sLast, with 95 ~ percent Iess of a chance than a similar person living in SBo Paulo. Access to both mobile and fixed t e l e ~ h o nservices are better d ~ s t r ~ ~a~~t oendgthe different states. The states ~ cccss to these services arc Distrito Federal and Rio Graride do lowest ~ r o b a ~ ~ l i t y 1,11. People living in urban areas e a higher chance of h j ~ f r a ~ t services,~With~ the exce ~ ~ u ~ of cell phones, the prob areas having access to any ~nfrast~uctur~ service is 95 percent Iess than for vidual living in a ~ ~ t e ~ r o p oarca.t ~The same result is obta~~ied l ~ n wheri we eas in three ~ateg~~ries with an increasing degree of ~ r ~ ~ ~ iThe~ a ~ ~ o ~ ~ i access to ~ n r a s t ~ ~ ~increases as we move from the rural to urban t u r e ofha to in Water Sewerapa Tebphone Elenncrty Cell phone 1.12. In alt these wed in ~ # ~ - s t ~ ~~d a rgd ~ l ~ ~ ~ ~ mdf#vez@s, wi services r e a ~ hai66 ~ ~ Looked at in a positit. h ~ ~ o ~ h that ~the~ pri e s e cant ~ e ~ ~toe hf i~s p~ os ~ ~ l asirtce ~sher ~situat~onwas far worse in 1991. For ~ ~ , example, in 1991 the chance of s o ~ ~ e ~fivirig in this type of ~ ~ ~ ~ o m ehavingt ai ~ n n e r ~ telephone line was 68 percent Iess than in ~ ~ The~ lowest .increase is in sewerage U service access, tvhich i n c r e a ~ eonly 15 percent duringthis period (Figure I,2.15). ~ 17 Figure 1.2. 15 Bra e of' 1.15. The costs of utitity services like el~trjc~tyare higher in isolated areas and in ~ # ~ - a ~~ ~e l ~ n~~like rslumsr jor#~ ~~ ~v ~Inl thescase of electricity, there: is ~ ~~ t ~ . evidence that the number of ~ ~ a n d e s tco~nect~ons the d i ~ ~ r i b u ~network in these i ~ c to i o n xcas i s four times higher than in other areas (Neri, 2003). The is the low ~~~~~~~~ revenue ~enera~ed these areas, which adversely affect the s by ~ and~the price ~ the~ p in r e ~ ~ i areas.~Dcspite this fact, thc ratc of access to all i n ~ r ~ ~ ~ r services r e ~ i ~ g ~ c ~ u i ~ c ~ e during dthc ~ period. ~ ~ e ~ Theaccessto~elleph~servicesincreasedhy ~ ~ ~-e ~ ~ ~ eleven times, while access to ~ e w e r a ~increased by 2.5 times. e I S c 1.2. 17 e uf'pop per house , 1991 and Wrbdn NOsi-urban l~olaied Rural area Rural Rural Rural Ruralarea area area urban wth urban viILa6)e nucleus others but Uriiege ~UCIBLS,oI?m HouseholdAccess by Income Level 1.16. In 2000, all income deciles had a rate of access superior to 90 percent. The privatization o f telecommunications benefited all classes o f income, but had a greater effect on the higher income deciles than the lower deciles. It i s possible to measure the relative gains of each income group by the distance between the lines in the graphs below; the distance increases as it moves along the income deciles. Access to both water and sewerage also increased duringthe 1990s. For example, in 1991less than 50 percent of the population in all income deciles had access to sewage services. The situation has greatly improved, as shown by Figure 1.2.20 and Table 1.2.2. Table 1.2.2 Brazil: level of access to infrastructure services according to income decile, 1991 and 2000 Electricity Telephone Water Sewerage 1991 2000 1991 2000 1991 2000 1991 2000 Total l408,173 157 ,753,999 63,0369828 46,676,064 15,408, 1 3,345,872 12,731,951 1,208,758 1,553,742 8,839,558 8,492,851 3,442,849 3,345,879 2 3,169,123 13,896,102 1,356,926 1,114,064 9,309,323 9,170,175 4,017,267 3,169,146 3 4,673,804 15,247,565 1,703,479 2,052,796 9,522,923 11,032,219 4,176,150 4,673,776 4 6,822,873 17,916,168 2,017,855 3,672,043 9,685,390 13,819,429 4,310,675 6,822,772 5 6,230,142 14,456,978 2,288,860 3,938,728 9,832,207 11,598,249 4,418,070 6,230,083 6 7,708,581 16,100,447 2,585,401 5,386,450 9,993,754 13,266,400 4,613,892 7,708,481 7 9,216,857 16,808,790 2,884,201 7,765,170 10,134,816 14,481,824 4,788,460 9,216,722 8 10,078,589 16,587,125 3,232,561 9,850,787 10,312,270 14,723,346 5,085,408 10,078,448 9 11,354,291 16,835,821 3,690,432 12,607,409 10,537,816 15,407,147 5,478,085 11,354,138 10 12,808,784 16,880,536 4,785,525 15,095,639 10,941,671 15,721,721 6,345,208 12,808,638 Source: IBGEICenso Demogrdfico 1991 and 2000. 20 \ II Figure 1.2.22 Brazil:p r ~ ~ of~ha~ i l i ~ ~ By race 3 2 5 2 1 5 1 a 5 0 .................................... 0.15 yrs 15 to 25 yrs 25 to 45 yrs 45 to 60yrs _ _ _ _ _ _ 60+yrs .......................................... ~.!!.E!?.~ ~ , , _ _ x _ x _ _ x , .*,.,'!!Ya3R! ..... %S"BYB.. .....!!!!!.%.E? .E!.EI'!Uhl............?!??!.? .............. .....: ~ . ~ .... ~ ~ ~ , W ~ ~ ! ~ 22 1.3 ACCESS FIRM$ BY 1.20. Firms located in the states of Amazonas, Goihs, Minas Gerais and Rio de Janeiro show the worst perception of the impact of infrastructure on their operations and growth prospects. These are the states for which both the share of firms grading the service as "a major or severe obstacle" and the average punctuation are significantly higher than national figures. For example, the share of firms grading telecommunications services as "a major or severe obstacle" in Amazonas (17 percent) and Goiis (14 percent) i s at least two times larger than the national figure (7 percent). The average grading results for Goiis and Amazonas are 1.11 and 1.13, respectively-- higher than the national mean (0.62). These two states present the worst results for the three infrastructure sectors considered. Among the three infrastructure services considered, transportation appears to be the most severe obstacle in Mato Grosso, with the caveat that the sample i s small (Table 1.3.1). 1.21. In general, firms from labor-intensive industries seem to be relatively more affected by infrastructure services, especially electricity and transportation. This is particularly true of the food and garment sectors. Table 1.3.2 indicates that firms from the electronics and machinery sectors report to be less severely affected by infrastructure services. While firms from the textile sector report to be damaged the most by electricity outages, chemicals and food are the two sectors that are damaged the most by the quality o f transportation services. Moreover, firms in the chemicals sector report relatively high damages caused by electrical services. As an exception within the labor-intensive industries, footwear does not appear to be among the industries that are the most affected by infrastructureservices. Based on World Bank (2005a). "Chapter 8: Infrastructure Services and Firm Performance in Brazil: The Need for Creative and Efficient Partnerships between the Public and Private Sectors." In: World Bank, "Brazil Investment Climate Assessment - Volume 11: Background Documents." Mimeo. Investment Climate Assessments (ICAs) are country reports that draw upon the results of firm-level surveys and other available data. They systematically analyze the conditions for private investment and enterprise growth ina country, drawing on the experience of local firms to pinpoint the areas where reform i s most needed to improve the private sector's productivity and competitiveness. B y providing a practical foundation for policy recommendations and involving local partners throughout the process, the assessments are designed to give greater impetus to policy reforms that can speed the private sector's growth. Produced by the World Bank Group in close partnership with a public or private institution in each country, ICAs are based on a survey of private enterprises designed to capture firms' experience in a range of areas - financing, governance, regulation, tax policy, labor relations, conflict resolution, infrastructure services, technology, and training, among others. Using a standard methodology, the assessment compares the survey findings with those in similar countries to evaluate how the country's private sector is competing. In Brazil, an Investment Climate Survey (ICs) was performed between July and October of 2003, covering 1,642 manufacturing establishments in 13 states. 23 Table 1.3. 1Infrastructureservices as obstacles to growth by state' State/Region No.of Telecommunications Electricity Transportation obs. No/ Large/ Mean No/ Large/ Mean No/ Large/ Mean Small Med. severe grade Small Med. severe grade Small Med. severe grade SBo Paul0 359 81 11 8 0.61 58 19 22 1.27 61 20 19 1.23 Rio Janeiro 121 71 20 9 0.85 85 17 25 1.4 54 20 25 1.36 MinasGerais 232 79 16 5 0.63 63 20 24 1.31 47 30 26 1.53 RioGrandeSul 190 87 8 5 0.55 55 19 15 1.10 63 30 11 1.09 PUan6 182 87 10 3 0.40 70 16 13 0.92 57 30 15 1.11 Bahia 76 85 11 4 0.49 66 14 20 1.06 66 20 17 1.06 SantaCatalina 174 82 13 5 0.52 52 17 14 0.95 60 30 14 1.17 Goiiis 82 62 24 14 1.11 36 28 36 2.00 35 30 33 1.85 Amazonas 24 58 25 17 1.13 37 42 21 1.70 25 30 46 2.21 Mato Grosso 38 84 11 5 0.53 53 16 18 1.10 42 30 32 1.82 Northeast' 238 84 11 5 0.52 52 12 20 1.00 67 20 13 1.oo Brazil 1,640 81 13 7 0.62 62 18 21 1.20 56 20 19 1.28 Notes: 'Percentage of firms responding to the question: "Grade the severity (on a four-point scale) of the following issues as a problem for the operation and growth of your business, if an issue poses a problem." 2Northeast includes the states of CearB, Maranhao and Parafba. Source: 2003 Brazil Investment Climate Survey. Table 1.3.2 Infrastructure services as obstacles to growth by industry Industry No. of Telecommunication Electricity Transportation Ob`' Largdsevere Meangrade Largdsevere Meangrade Largdsevere Meangrade Food 127 8 0.69 32 1.60 28 1.60 Textile 106 7 0.65 34 1.73 17 1.25 Garment 442 5 0.59 20 1.25 19 1.30 Footwear 172 6 0.50 19 1.10 18 1.17 Chemicals 84 6 0.58 24 1.25 32 1.81 Machinery 183 7 0.61 14 0.92 14 1.07 Electronics 79 6 0.55 12 0.80 17 1.23 Auto-parts 130 4 0.59 20 0.88 16 1.23 Furniture 315 7 0.63 17 1.07 19 1.23 Brazil 1,638 7 0.62 21 1.20 19 1.28 Note: Percentage of firms declaring "severe or large" barrier to growth by infrastructure service and industry. Source: 2003 Brazil Investment Climate Survey. 1.22. Losses in production due to outages in infrastructure services are higher in the states of Amazonas and Goihs. Electricity i s the infrastructure service reported to cause more losses among all states with the exception of Paraiba, in which transportation causes more losses than electricity. Amazonas ' firms experience the highest losses among states in all infrastructure services - however, the degree o f severity differs among them -with 5.26percent and0.98percentof firms' annualproductionlostdueto inadequacy o f electrical and telecommunication services, respectively. For all industrial sectors, the least harmful infrastructure service is water - only in Amazonas were its production losses over one percent (Table 1.3.3). 24 1.23. Footwear appears to be the sector with the highest losses in production (over ten percent) due to infrastructure outages--mostly caused by electricity, telecommunications and transportation services. Losses in the auto-parts sector are ranked second, with more than 8 percent of its annual production. In both cases, transportation is the infrastructure sector that causes the largest share of damages; thus, f i r m s in this sector tend to heavily rely on outsourcing. Food i s the third most affected sector (around 7 percent) while the electronics suffered the lowest losses in production (Table 1.3.4). Table 1.3. 3 Annual loss inproduction by state(inpercent) State Telecommunications Electricity Water Transportation Sgo Paul0 0.29 1.20 0.02 0.31 Minas Gerais 0.14 1.30 0.05 0.60 Rio Janeiro 0.59 3.25 0.12 0.85 Rio Grande Sul 0.19 0.87 0.01 0.08 Paran6 0.36 1.32 0.05 0.58 SantaCatarina 0.34 2.00 0.06 0.27 Goihs 0.92 2.43 0.07 1.14 Mato Grosso 0.42 0.67 0.02 0.53 Bahia 0.15 0.89 0.08 0.48 Ceara 0.19 1.49 0.02 0.26 Maranhb 0.01 1.24 0.00 0.23 Parafba 0.14 0.23 0.01 0.45 Amazonas 0.98 5.26 1.27 1.71 Note: The table reports mean values. Source: 2003 Brazil Investment Climate Survey. Table 1.3.4 Annual loss inproduction by industry (inpercent) Industry Telecommunication Electricity Water Transportation Food 1.83 2.96 0.61 2.16 Textile 1.11 3.38 0.00 2.30 Garment 1.02 2.50 1.23 3.95 Footwear 3.31 3.50 0.07 3.34 Chemicals 0.68 1.82 0.00 2.07 Machinery 1.29 1.70 0.03 1.57 Electronics 0.97 0.76 0.05 1.17 Auto-parts 1.16 2.46 0.08 4.71 Furniture 0.70 2.70 0.57 3.53 Note: The table reports mean values. Source: 2003 Brazil Investment Climate Survey. 1.24. Infrastructure losses consist of both the frequency and the duration of interruptions, indicating reliability problems. Electricity i s the infrastructure service that presented the greatest number of interruptions, on average, followed by telecommunications and water. This pattern holds for all the states in the sample. Transportation is the service with the smallest number of interruptions, but the longest ones induration(Table 1.3.5). 1.25. To mitigate the problem of reliability, a large share of Brazilian firms own and use power generators. More than 15 percent of the surveyed firms reportedto own power generators, and 50 percent of large firms. The share of micro-enterprisesowning power generators is muchlower. When disaggregatedby state, the highest shares of firms 25 owning these equipments are found in Mato Grosso (MT) and Amazonas (AM) (Figure 1.3.1). These results are similar to the findings for China and the Philippines (around 17 percent and 32 percent, respectively). For those firms that own power generators, "in- house" electricity production represents, on average, 10 percent of their electric needs; however, this share can reach more than 30 percent and 20 percent in Maranhtio (MA) and Amazonas (AM), respectively (Figure 1.3.2). Finally, a growing consensus i s that new investments in electricity generation are needed in the short-term to reduce the risk of future energy rationing. Table 1.3.5 Average number of days and hours of interruption of infrastructure services by state State Telecommunications Electricity Water Transport # of days # of hours # of days # of hours # of days # of hours # of days # of hours SHo Paulo 2 15.7 3 5.9 2 11.9 1 30.8 Rio Janeiro 4 18.4 5 4.8 3 14.0 1 13.1 Minas Gerais 1 9.8 3 20.3 1 10.2 2 45.2 Rio Grande SUI 2 14.0 6 13.8 2 36.8 1 12.8 Parand 2 6.6 4 5.1 1 8.6 1 34.5 Bahia 1 14.2 2 6.9 1 23.5 1 32.7 Santa Catarina 2 5.0 6 5.5 1 14.1 1 13.6 GoiBs 2 17.8 6 20.6 0 8.7 2 65.7 Amazonas 5 38.2 13 12.3 1 14.4 3 33.3 Mato Grosso 4 20.1 3 17.3 0 18.0 1 36.6 CearB 0 10.2 11 4.8 4 5.1 1 15.0 Paraiba 5 12.7 6 6.3 0 7.0 1 32.3 MaranhHo 1 4.1 6 7.3 0 19.0 2 7.1 Source: 2003 Brazil InvestmentClimate Survey. Figure 1.3. 1 Firms owing power generators by size Figure 1.3. 2 Share of firms' electricity from own and state generator by size and state (inpercent) L1 w Source: 2003 Brazil Investment Climate Survey. Source: 2003 Brazil InvestmentClimate Survey. 26 1.26. To a great extent, transportation costs reflect the poor conditions of Brazilian roads. The condition of the road network has significantly deteriorated over time. Brazil's federal paved-road network, totaling about 58,000 km, was built between the 1950s and the 1970s, representing assets of nearly US$10 billion. It carries most of the country's long-distance traffic and between 25 percent and 30 percent o f the network i s rated in poor condition. This state o f affairs adds about US$500 million each year to vehicle operational costs, mostly paid by the private sector (World Bank, 2004). 1.27. For export firms, ports and customs services are a key bottleneck. In spite of recent improvements, port tariffs and time delays caused by transit and inefficient customs services are still too high.For example, inSantos, once the costliest port inLatin America, container handling costs have dropped from over US$ 300 in 1997 to US$180 in2000. These costs remainhigher than those inRotterdam(US$ 100per container) and Buenos Aires (US$ 120 per container). Also, cargo transit time through Brazilian ports (Le., the total time needed for cargo to pass through the port, from ship call to the port exit gate) is still excessive when compared to other countries (see Table 1.3.6). In the near future, ports such as Purunagua' and Santos may face capacity constraints, even though the lack of reliable information and poor regulatory environment complicates the determination of how much capacity expansion i s really required (as compared to efficiency gains from better use of existing capital stock). Table 1.3.6 Brazil,China, IndiaandMalaysia:port transit times (indays) Country Imports Exports Average Longest Average Longest Brazil 13.8 32.4 8.4 16.9 China 7.5 12.2 5.5 8.1 India 10.4 21.6 5.1 9.3 Malaysia 3.4 1.4 2.6 5.1 Source; WorldBankilnvestmentClimate Assessments. 27 2. INFRASTRUCTUREINVESTMENTSINBRAZIL This chapter reviews the performance of infrastructure investments in Brazil, public and private. It addresses the reasons for the decline of public investments from 1980-2002, including public administrationand state-ownedenterprises, andwhy private investments were not enough to compensate for the retrenchment of the public sector, despite the large surge of private participation ininfrastructure inBrazil during the 1990s. 2.1. INFRASTRUCTUREINVESTMENTSINTHELAST TWO DECADES 2.1. Infrastructure investments in Brazil fell dramatically from 1980 to 2002, with the largest decline occurring in the late 1980s. Between 1980 and 1985, infrastructure investments in Brazil increased 5.17 percent on average - one of the best results in Latin America. However, investments in infrastructurehad plunged by the late 1990s, from an average of 5.22 percent of GDP from 1981-85 to 2.35 percent in 1996- 2000. In2001, levels of infrastructure investment correspondedto half of what they were in 1981.The drop of 2.87 percentagepoints betweenthe two periods is muchlarger than the regional fall of 1.2 percentage points, reflecting Brazil's larger initial investment levels in the early 1980s. This fall was concentrated during 1981-1996, when levels dropped 2.61 percentage points (or 93 percent of the total). Despite partial recovery in subsequent years, the steepest decline in infrastructure investment in Brazil occurred in 1987-89,when it droppedby roughly 50 percent (Figure2.1.1). 2.2. The fall in infrastructureinvestmentsin Brazil was caused by a substantial contractionof publicsector spending,a trendcommonto most countriesin the LAC region. Public investments in infrastructure in Brazil fell from 3.6 percent of GDP, on average, during 1981-85 to 1.0 percent of GDP in 1996-2000, a drop of 2.6 percentage points that accounts for 90 percent of the overall decline (2.87 percent). More thanhalf of this decrease occurred between the periods of 1986-90 and 1991-95. The rest of the decline in public investment could be split, almost equally, between the remaining two periods (before 1986and after 1995). 28 29 30 inelectricity dropped from 2.68 percentagepoints of GDPin 1980to 0.73 in 1990a 1.95 point reduction that corresponds to more than three-fourths of the total decline in the period. The contribution of the transport sector (road and railways) to the overall fall of total public investments comes second with approximately 19 percent, with a reduction of investment levels from 0.55 to 0.08 GDPpoints (Table 2.2.1). What factors explain these reductions? Table 2.2.1 Brazil: investments ininfrastructure by sector, 1981-90(percent GDP) Sector 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 Electricity 3.13 3.30 3.63 3.29 3.56 2.88 3.82 2.94 1.89 1.55 Transport 0.94 0.62 0.88 0.86 0.68 0.46 0.40 0.73 0.14 0.13 Telecoms 0.61 0.66 0.82 0.80 0.74 0.72 0.80 0.90 0.70 0.52 Water 0.45 0.37 0.30 0.18 0.27 0.24 0.30 0.35 0.17 0.19 Source: Own elaboration based on Calderon and Serven (2004). 2.8. One underlying cause of the reduction of public investments in electricity is the deterioration of the economic performance of state-owned enterprises (SOEs). By the late 1980s, Brazilian state-owned utilities provided highly inefficient services and contributed to the public sector's deficit (approximately half o f the overall debt during 1983-1988). This resulted from a mix of growing politicization of the administration; centralization of operational, regulatory and planning attributions in the SOEs and large investments. Cost-plus regulation, for instance, created all the well-known disincentives for good managementq7The centralization of the system and the volume of cash-flow mobilized by the companies induced all types of rent-seeking and capture: with the collapse of the public state banks, distribution companies took their place in lending to state governments.* In part for technical reasons, but also in part corresponding to the Aver-Johnson effect, projects prioritized scale and were capital-intensive, stressing the supply and the cost of finance for the sector. 2.9. Two other events affected the investment capacity of the electricity sector during the 1980s: worsening conditions in the international financial markets and decreasing real value of tariffs. External financing was a recurrent option o f the energy sector since the mid 1970s, when the government encouraged electricity companies to borrow internationally to finance Brazil's balance o f payment needs and mitigate the impacts of oil prices on the economy. With the 1982 Mexican collapse and successive balance of payments crisis, financial conditions o f energy companies deteriorated significantly while their access to international markets decreased s~bstantially.~ In addition, tariffs were systematically used as a price-stabilization mechanism. It is estimated that electricity prices dropped more than 20 percent inreal terms between 1980 and 1989. Combined with the increase of debt, the reduction of revenues further limited the availability of resources for new investments." Inthis context of financial distress, '*Law 5,655/71 established a minimumrate of returnof 10percent for investments inthe electricity sector. Inthis context, the sector accumulated credit with the Treasury: by December 1992,federal government's debt with the electricity companies corresponded to 2.5 years of sector revenues. See Ferreira (2000). loSeePires and Piccinini (1988). 31 Eletrobrhs' investments collapsed and the major part o f sector funds was employed to interest payments on debts (98 percent in 1989). 2.10. The decline in public investmentsin transport sector -is - the second most relevant linked to similar factors. Investments in the railway sector fell from 0.37 percent o f GDP to 0.02 percent of GDP between 1980 and 1990, corresponding to more than three-fourths the fall inthe sector during that period. As inthe case of the electricity sector, this was, inpart, caused by the impact of deteriorating macroeconomic conditions and political interference. Between 1985 and 1990, Rede Ferrovidria National's revenues per TKU declined 33 percent, as result of tariff control by federal government. Demand for railway services also declined together with large drop in economic activity, in a context where a fragmented operation network structurally affected its competitiveness with the road modal. On the cost side, payroll expenditures per TKU increased 10 percent during the same period, despite the reduction of 8,500 employees. Increasing debt cost and growing operational losses reduced its investment capacity significantly.' 2.11. In addition to the structural factors described above, the 1988 Constitution further discouragedpublic investmentsin infrastructurein at least two ways. First, it abolished the federal infrastmcture-specific taxes without compensating with any other clear source of funds. Second, it reduced the availability of budgetary sources for capital expenditure and increased budget rigidity. In a nutshell, federal revenues previously "earmarked" for infrastructure investments were transferred without "earmarking" for other tiers o f the administration, while other priorities were "earmarked" and increased over time. Between 1995 and 1998, current expenditures in the public administration increased almost 2 percentage points to 41.65 percent o f the GDP, while capital expenditures declined 1percentage point to 1.93 percent o f GDP (Figure 2.2.1). 2.12. The Constitution expanded the VAT basis by abolishing federal infrastructure-specific taxes without compensating for any other clear source of funds. The Imposto sobre Combustiveis e Lubrificantes Liquidos e Gasos (IUCLIG) and the Imposto Unico sobre Energia Elitrica (WEE), two transport and energy specific federal taxes, respectively, were incorporated to the state value added tax system. The Imposto sobre ServiCos de Transportes Rodovidrios was transferred to the states. Other railway and telecommunications taxes were also abolished. While part o f a necessary effort to rationalize Brazilian taxation system, these initiatives eliminated sources o f funding. Only in 2001 a similar tax - ContribuigCo Unica sobre Dominio Econbmico (CIDE) -was partially allocated for the transport sector. 2.13. The 1988 Constitutionalso reducedthe availability of budgetary sources for capital expenditure and increased budget rigidity. First, it promoted further decentralization of revenues among the government tiers o f the Brazilian federation, increasing the transfers o f federal funds to states and primarily municipalities. The municipal share of the federal income tax and industrial product taxes increased from 17 to 22.5 percent, while the municipal share of the state VAT grew from 20 to 25 percent. 11See Marques (1996),and Castro and Lamy (1994). 32 2.14. Over time, federal public ~ ~ ~ ewere f~rthercrowded out by rising ~ t ~ e ~ t ~ current e ~ p ~ ~ d i t uCentral g o ~ e ~ ~ e~~ ~t si '~ ex~~ndit~res r e ~ . ~ a r y rose from 13.7 percentof GDP in 199Ito 21.fipercent in2002 while fcderal p ~ b ~1i ci ~ ' e s ~ ~declined ~ e n ~ s from 3.45 ~ e r c ton0.75 percent of GDP. ~ ~ ~ o ~t ~~ ~ ~~ eiof-lthe~~increase~in~ ~ i ~ o ~~ p r i e~~ ~~e n~d ~(6.9rpercent out of 8.9 percent of. GDP) was due to pension ~ ~ e s benefits, ~~~1~~ employees9salaries an ~ ~regime ~ ~ ~ o ~ ~ ~ e iiicrease in these e I 1 3.50 50 1 I , 3.00 .; 4 45 * I 40 , 2 5 0 7 + 35 I 2.00 ; 30 I I 25 7.50 I I 20 100 G * $5 10 0.50 * 1 5 0.00 3 0 1995 1996 1997 1998 1999 2000 2001 2002 2.15. During the prj~r~ti2ati~nperiod of 1995-98, the net impact of p ~ i ~ ~ ~ t i ~ a t i o ~ may be ~ p p r ~ ~ i m ~byt ethe d j f f ~ r ~of~ thee i ~ ~ ~ ~ of m ~ ~ t $ d ~ t st~te-~wne~ enterp~ises~~~~~)at the b e ~ iand ~ ~ the ~endi of the ~ ~ ~ period, This ~ ~ I ~ ~ ~ indicates a red~ictionof 0.89 ~ e r ~ e ~ i ~points of GDP, cQ~iip~re~3 drop in 0.69 3 g e lo 33 percentage points of GDP in the public sector. Investments from public administration increased slightly avoiding a larger decline. Investments in telecommunications sector decreased 0.93 percentagepoints of GDP with the privatization of the Telebrds System. In July 1998, the 12 companies originated by the break-up of the Telebrds System - including the three fixed phoned companies, the international carrier (Embrutel) and 8 mobile phone providers were transferred to the private sector. All but 3 out of the 20 companies of the electric sector - 3 generators and 14 distribution companies, corresponding to 92 percent of the revenues from the transfer of assets of the electricity sector - were privatized in 1996-98. The effect of privatization is twofold. First, investments formally accounted as public investments that became private are not accounted as such, which causes an effect of reduction but represents indeed a simple substitutionof investmentsby the private sector. Second, it causes a wave of investments in the years immediately before privatization - as SOEs were prepared for sales--that artificially raises the basis for subsequent comparison(Figure2.2.2). 2.16. From 1998 to 2002, public investments in infrastructure declined 0.75 percentagepoint of GDP. This result, which includes the year immediately before the fiscal effort was initiated, could be seen as an "upper bound" estimate for the impact of the fiscal effort in public infrastructure. Investment from public sector companies declined 0.18 percentage points and 0.57 in the public administration. Interestingly enough, most of the contraction came from the sub-national governments, roughly two- thirds of the overall fall (0.5 percentageGDP point). This is hardly a surprise, as the bulk of investmentininfrastructure is concentratedinsub-national governments. Also, current expenditures increasedmore than two percent of GDP during this period, which coupled with a high level of rigidity inthe public budget made capital expenditures major targets for cuts. Political economy factors and the well-known incentive problems with the provisionof public goods added more complexity to the fiscal dilemma. 2.17. Therefore, except for specific cases maintenance - - as for example federal road it is hard to say that the fiscal effort was a major cause for the declineof the investmentsin infrastructurein Brazil.This impact may have increased a bit more in 2003-2005 due to a raise inthe primary surplus seek by the administration. Before that, however, the commonly alleged inverse relationship between the primary surplus and public investments in infrastructure (Figure 2.2.3) should be seen with caution, as it pictures only part of the story. The reduction of infrastructure investments (0.75 percentage points of GDP) contributed to less than one-fourth the change in the primary surplus between 1988 and 2002 (3.88 percentagepoints). The rest of the change inthe primary surplus (3.13) shouldhavecomefrom somewhereelse, notably an increase intaxation. Above all, most of the declineinpublic investmentsoccurredprior to 1998. 34 Figure 2.2.2 Public sector investment in Figure 2.2.3 Brazil: primary surplus and public infrastructure (percent of GDP), investment ininfrastructure (percent of GDP), 1995-2002 1998-2002 iI 3S% sc Public 3.0% 2.5% +Administration Public 2.0% 15% 2.0 I 1.5 1.0% 1.o OS% 0.5 0.0% , , 00 1995 1996 1997 1998 1999 2000 2001 2002 1998 1999 2000 2001 2002 o P u b l i c lnlsstmentin Inhastwlure+Primary Surplus Source: Own elaborationbased on Afonso et al. Source: Own elaboration based on Afonso et al. (2005). (2005). 2.3. WHERE DIDALL THEPRIVATEINVESTMENTGO? 2.18. Brazil was the largest beneficiary of private investments in infrastructure during the 1990s. Between 1994 and 2004, approximately US$164 billion were allocatedin investments with private participation in Brazil, corresponding to 21percent of the world's total accumulated in the period. This is close to the whole volume of East Asia and Pacific, a region that includes China, the Philippines, Malaysia and Thailand; and larger thanany other region inthe world (Table 2.3.1). 2.19. Telecommunications and energy sectors were the main recipients of these investments. Between 1994 and 2002, roughly half the private investments in infrastructure went to telecommunications while 34 percent went to the energy sector. Transport (14.5 percent) and water and sanitation were the sectors that received the least amount of investment (Figure 2.3.1). A significant volume started entering the country in 1996, culminating on the 1998's peak, year of the privatization of the telecommunication system. The flow decreases after 1998for all sectors. 2.20. Most important, however, the majority of private investments between 1994- 2002 were directed to asset acquisition and not greenfield investments. This lowered the impact of private investment on the expansion of the stock of infrastructure capital. Greenfield investments corresponded to less than one quarter of the total volume of investments. Most of private investment in Brazil was concentrated in divestiture of public assets (Figure 2.3.2). Divestitures are not a problem by themselves: as the case of the case of telecommunications demonstrated, if the concession contract provides the right incentives, private companies will invest. Inthis particular case, entry inthe market 35 3h exceptions, as the cases of Chile and Colombia indicate: in these two countries, energy and transport were the major beneficiaries. Malaysia and the Philippines illustrate cases where the share of private investments in water and sanitation was much larger, indicating the sector potentialities (Figure 2.3.4). Figure 2.3.3 Composition of private flows to Figure 2.3.4 Composition of private flows to infrastructureby investment modality for selected infrastructure by sector for selected countries countries (inpercent), 2003 (inpercent), 2003 mB8%ive I W i P W Source: World Bank/PPIdatabase. Source: World BanklPPIdatabase. 37 3. INFRASTRUCTUREAND GROWTH:HOWMUCHI S NEEDED? This chapter addresses the question of whether Brazil needs more investments in infrastructure and - if so - how much is needed. First, it discusses whether infrastructure i s a "binding constraint" for growth in the country. Second, it reviews arguments linking infrastructure, growth, and social indicators. It essentially argues that higher infrastructure investments favors pro-poor growth in Brazil. Third, the chapter presents preliminary estimates on how much infrastructure investment i s needed.l2 3.1. I S INFRASTRUCTUREA BINDING CONSTRAINTTO GROWTHINBRAZIL? 3.1. Since 1960,the Braziliangrowthperformancecan be dividedintwo different periods. An exceptionally good phase of high and steady growth in 1960-80, when annual rates averaged 8 percent; and a subsequent period when Brazil underperformed compared to the rest of the world or the L A C region, with an average growth of 2.2 percent in 1981-2005. Understanding the disappointing performance of Brazil GDP growth in the eighties and the persistence o f low growth rates since the macroeconomic stabilization in the last decade constitute one of the most relevant issues in the contemporaneous economic debate in Brazil. What are the reasons behind the slump of the growth rate inBrazil? Are they relatedto the decline of infrastructure investments? 3.2. Recent literature on growth accountingin Brazilsuggests that the decline in investment ratios is the underlyingcause for declining growth rates in the 1990s. Both capital accumulation and total factor productivity decreased in the 1980s' but the fall on capital accumulation was the main cause of the slowdown of the Brazilian economy in this decade. In the 199Os, a period of macroeconomic stabilization, low capital accumulation obstructed higher growth rates, despite the significant improvements in total factor prod~ctivity.'~ This does not imply a linear relation between investment ratios and economic performance: evidence show that the main reduction inthe per capita income gap between Brazil and the U.S. in the post-war period occurred during 1964- 1970 (the PAEG years), when capital deepening was not particularly intense but many institutional reforms were introduced. 3.3. By comparison, evidence suggests that labor force supply was not an impedimentfor economic growthinBrazil.Until 1980, the contribution of raw labor to l2 An assessment of the impacts of an increase in infrastructure stock on growth and income distribution in the Latin American Region was performed by Calderon and Serven (2004). A review of evidence o f the impacts of infrastructure on social indicators may be found in Brenneman and Kerf (2002). Estimates for infrastructureneeds were provided by M.Fay and T. Yepes. l3See, for example, Bonelli (2005), BachaandBonelli (2004), andPinheiro et al. (2001). 38 growth was about 20 percent and had a similar contribution to the change o f growth rates between periods. Given the strong fall of total factor productivity, its contribution to growth in the lost decade increased to 83 percent, but its contribution to the fall of GDP growth inthis period was low (6 percent). Pessoa et al. (2003), found that the importance of human capital dramatically increased (to 47 percent) in the nineties. Coincidently, Brazil experienced a continuous but slow improvement of educational indicators since the seventies that were accelerated inthe nineties. 3.4. I t is far less obvious, however, to claim that the fall in total investmentswas caused by a decline in the investment in the infrastructure sector. In principle, the decline in investments in the tradable sector may have been at least as important as in infrastructure (Table 3.1,l).l4 addition, the decline in the overall investment levels is In clearly related to the poor evolution of gross sa~ings.'~Also, the fall of private investments i s related to the decline inthe marginal productivity of capital, a period when trade and industrial policies increased substantially the costs of machinery and equipment.16 Inthis sense, the macroeconomic imbalances and the industrial policies of the 1980-90s rivals, to say the least, with the decline of infrastructure investments as the main explanation for the reduction ininvestment levels.l7 3.5. Rather, investments in Brazil seem to be severely constrained by the high level of current public expenditures. In a context where levels o f public indebtedness are already high, higher current expenditures implied low capital expenditures and higher levels o f required tax-revenues. Not only taxation i s high by its incidence i s cumbersome and distorts economic incentives for efficiency, as for example, inhibiting firm growth and formalization. High and cumbersome taxation, coupled with an "investment climate" that increases private risks and increases the cost of investments tends to reduce the returns to private investments." Curbing high public expenditures, as a pre-condition for the reduction of taxation levels, are an important condition for higher levels o f sustainable growth rates. l4Preliminary data indicate that overall investments declined approximately 7 percentage points of GDP with manufacturing investments declining 3.6 GDP points, compared to 2.84 GDP points in infrastructure in 1970-2003, '' l5 High domestic saving rates and increasing external indebtedness were responsible for the capital accumulation until the late 1970s, when savings rate fell to 20 percent of GDP to fluctuate between 18 ercent-20 percent henceforth. The price of capital goods compared with the GDP deflator increases continuously from 1970 to 1990 while capital productivity fell strongly since 1973 to stabilize at its lowest levels in the last fifty years in 2002. l7One different conclusion is reached by Cavalcanti and Araujo (2005). Combining impulse-response and integration analysis with counterfactual simulations, they concluded that the observed decrease in capital expenditure inthe recent past might have hurt growth and brought about high output costs. lSIndeed, the importance of taxation comes hardly as a surprise if one takes into account that Brazil's tax rate, reaching 37 percent of the GDP is too high when compared to its income level. In addition, a highly distortionary tax system, as in the Brazilian case, increases the cost (the deadweight-loss o f the taxation) of the public funds, which in turn reduces the number of projects that are worth finance through public resources. 39 Table 3.1. 1Brazil: gross fixed capital formation by economic sector (inpercent of GDP), 1971-2000 Economic sector 1971-80 1981-90 1990-94 1995-96 1997 1998 1999 2000 Manufacturing industry 6.85 4.20 2.61 2.65 3.06 3.08 2.95 3.25 Infrastructure 5.42 3.62 2.17 1.79 2.36 3.17 2.70 2.58 Energy 2.13 1.47 0.87 0.53 0.69 0.89 0.77 0.67 Telecommunications 0.80 0.43 0.53 0.66 0.78 1.18 1.17 1.07 Transport 2.03 1.48 0.70 0.48 0.61 0.75 0.56 0.63 Sewerage 0.46 0.24 0.07 0.13 0.28 0.35 0.20 0.21 Civil construction 4.95 4.71 4.03 3.99 4.23 4.25 3.97 3.60 Others 6.28 5.47 6.09 8.57 6.70 5.95 6.48 7.07 Total 23.50 18.00 14.90 17.00 16.35 16.45 16.10 16.50 Source: Based on Bielchovsky (2002) upud World Bank (2004). 3.6. Indeed, evidence suggests that the private sector ranks other factors of the business environment as higher priorities, notably the tax burden. In the 2003 Investment Climate Survey (ICs), entrepreneurs ranked electricity, telecommunications and transportation as minor obstacles for the growth o f their business, when compared with tax burden and other governance factors (policy uncertainty, macroeconomic instability corruption, and the cost of financing) as well (Figure 3.1)." Therefore it i s unclear whether infrastructure services associated to the current stock o f infrastructure are constraining private sector development and therefore economic growth. It is the service provided by the infrastructure stock that matters, not the stock itself. 2o 3.7. Comparing the returns of investment in infrastructure with other areas, for example education, does not necessarily favor infrastructure either. Rates of return to infrastructure depend on the existing level of assets as well as their quality. As a result, the productivity of infrastructure i s usually found to be higher among low and middle income countries, and decreases among high income countries. Estimates of social rates of return from cost-benefit analysis of World Bank projects inL A C vary from 11percent to 22 percent depending on the sector.21Camargo (2002) estimated private internal rates of return for an additional year o f education in 5 to 15 percent in Brazil with the social rate of return three to four times the private ones. Less than 22 percent of the firms reported infrastructure services as "a major or severe obstacle," which contrasts with the 84 percent of the firms that reported tax rate and cost of financing as "a major or severe obstacle." Tax administration, in addition to tax-rates, is mentioned a severe obstacle to almost 70 percent of the sample. The first problem imposed by this warning is that apparently congestion infrastructure assets may yield further returns if better managed and regulated. For example, analysts claim that port terminals in Brazil may generate higher outputs if consistent labor laws allow for enough flexibility in labor employment. In some cases, congestion may even be an misleading sign of investment needs: taking the port sector as an example, over-utilization of a port incity A may simply reflect the need of an investment in linking city B to C, where another port may be used, not the need to invest in the port in the city A. Congestion may reflect scarcity only imperfectly also because infrastructure assets are long-lived: to keep the example short, congestion inroad D main simply indicate the need to better maintenance and rehabilitation inthis road and not the need for its duplication. In addition to these factors, traditional problems of the provision of public goods - the free-riding and preference revelation - as well as the frequent use of distortive taxes to finance infrastructure investments should be also taken into account (Prudhomme, 2004). Briceiio-Garmendia et ul. (2004). 40 Figure 3.1.1. Perceptions of obstacles to growth inBrazil (inpercent) Source: 2003 Brazil InvestmentClimate Survey. 3.2. INFRASTRUCTURE, GROWTH AND SOCIAL PERFORMANCE Infrastructure and Growth 3.8. Evidence favors, however, the hypothesis that an increase in infrastructure investmentswould be beneficial to growth. This, naturally, does not imply a support to any possible infrastructure expansion in Brazil: the country's economic history has several examples of large infrastructure projects (the most notable one is the TrunsarnuzGnica)that resulted irrelevant from an economic point of view with sometimes significant environmental and social costs. Similarly, it does not imply that economically sound infrastructure projects should be predominantly supported by public funds. The case to be made i s that a carefully designed expansion of the stock of infrastructure - essentiallyfunded inprivatefunds (may provide a positive impact inthe growth potential of the Brazilian economy). Also importantly, it may be shown that infrastructure investmentsmay contribute to poverty andinequality reduction. 3.9. First, a growing body of economic research su ports the link between infrastructure stock and growth in Brazil and elsewhere?' For instance, Calderon et al. (2003) estimate that infrastructure compression in the 1990s reduced longer-term growth by about 3 percentage points a year in Argentina, Bolivia, and Brazil, and by 1% -2 percentage points a year in Chile, Mexico, and Peru. Using a large panel data set encompassing over 100 countries and spanning the years 1960-2000, Calderon and 22 Schauer (1989) found that the stock of public infrastructure capital is a significant determinant o f aggregate TFP in the US but his results were deemed implausibly large and not robust to the more sophisticated econometric techniques. Hulten (1996) finds that differences in the effective use of infrastructure resources explain one-quarter of the growth differential between Africa and East Asia, and more than 40 percent of the growth differential between low- and high-growth countries. Also, Gramlich (1994) provides an overview of this literature. One problem i s reverse causality: infrastructure may cause growth, but growth also causes firms and people to demand more infrastructure. Failure to take this into account would result in the over-estimation of the contribution of infrastructureto growth. 41 Serven (2004) find a robust impact of both infrastructure quantity and quality on economic growth: bringing the stock and quality of Brazil's infrastructure to East Asian tigers median would generate a 4.4 percent growth increase. Ferreira and Malliagros (1998) show that a 1percent increase ininfrastructure capital stock rises the GDP growth rate between 0.5 percent and 0.6 percent inthe long run.Estimation using sector specific capital stock show similar results with stronger impacts coming from transportation and electricity sectors.23 Glomm and Rioja (2004), using overlapping generations' model calibrated to Brazil concludes that infrastructure investments in Brazil would have to reach about 5 percent o f GDP to maximize impact on economic growth. 3.10. Micro-level evidence suggests that poor infrastructure services (transportation and energy) seem also to have a large and negative effect on a firm's performance (probability to export and productivity levels). Using ICA database for several L A C countries, the assessment of the impact of infrastructure on competitiveness, carried out by Escribano et al. (2005), shows that infrastructure services are major determinants of total factor productivity (TFP) in Brazil and other LAC selected countries. Figure 3.2.1 summarizes the average productivity impact (gains and losses) o f some I C A variables in Brazil. For example: (i)if the index created for the average duration of power interruptions increases by 1 percent, the average productivity (TFP) decreases by 25.8 percent, ceteris paribus; (ii) a firm has its own power generator, its if TFP increases by 11.5 percent everything else held constant; and (iii) the losses as a if share of total sales increase by 1percent, the firm's TFP decreases b y 7.0 percent, ceteris paribus. Figure 3.2.2 shows the impact of these variables on Brazilian firms' probability to export. Interestingly, the highest impact in absolute terms i s found to be the variable representing "customs services." In addition, both transport and energy deficiencies (measured as shipment losses and duration of power outages, respectively), negatively impact a Brazilian firm's probability of exporting. 3.11. A second reason is related to avoiding pure dilapidation of infrastructure stock through appropriate maintenance expenditures. The importance of public capital maintenance on the economic growth has been highlighted by Devarajan et al. (1996), Rioja (2003), Kalaitzidakis and Kalyvitis (2004). Although usually neglected in favor of building new infrastructure, these expenditures enhance the productivity o f the existing capital stock and tend to have a larger marginal effect on GDP growth. Therefore it is possible that by re-allocating public expenditure to favor capital maintenance in detriment of investment in new investments, governments are able to raise the long run growth rates of their economies. This result depends, of course, on a number o f circumstances, including the size of new investments as a share of existing public infrastructure. Based on data for the 1980s, it was estimated that a raise o f 1percent of GDP in maintenance activities could raise output by 2.52 percent, compared to 1.87 23Electricity has the largest coefficient, 0.68 percent increase in GDP for a 1 percent increase in the stock of capital, followed by transportation, 0.57 percent, and telecommunications 0.42 percent. Within the transportation sector, the highest impact on GDPcomes from investment inrailroads, followed by ports and roads, in this order. Ferreira and Araujo (2005), applying co-integration analysis confirms and extends these previous results and using standard capital series or physical measures of infrastructure, find that the association between output and infrastructure is strong and robust, as estimated long-run elasticities were statistically significant and close to or above one inmost cases. 42 percent in an average Latin American country.24Although data on maintenance and rehabilitation i s very difficult to gather an overview of the road sector may give a good illustration of the lack of maintenance expenditures inBrazil. 3 Figure 3.2.1 Brazil: average productivity (TIT) impact (gains and losses) of I C A variables (inpercent) Infrastructure Regulation O t h e ompetitiv 1 , B u r e a u c r a c y ntage 7 - 11. I I 4 . O5 - 2. 30 L_ io.0 I 4.1 35 1.1 1.2 1.3 1.4 1.5 1.6 2.1 3.1 3.2 3.3 .1 index lor avg duration 01 power outages 2.1 Index for avg time to clear Customs .2 Dummy lor lirm having own generator 3.1 Dummy lor lirms exporting more than 10% 01saies .3 index for avg duration of water outages 3.2 Fraction 01 capacity utilized by the lirm .4 Losses due to transport interruption (% saies) 3.3 Dummy lor firms having external auditors .5 Fraction of workers using computers in their job 4.1 index 01 competitive disadvantage in power .6 Dummy lor lirm using e-mail or website in conducting business Figure 3.2.2 Brazil: impact of I C A variables on firms' probability of exporting (inpercent) Other Competitive Ariuantage I 456.7 200 152.6 -200 ' I 1.1 1.2 1.3 2.1 2.2 3.1 3.2 4.1 4.2 4.3 6.3 1.1Indextor avg durationof power outages 1.2 Lossesinshipment (%sales) 4.1 Total Factor Productivity 1.3 Dummy for firm usinge-mailor website in conducting business 4.2 Dummy for firms having external auditors 2.1 index for fractiond time dealingwith bureaucracylreguiation 4.3 Age of the firm 2.2 Indexlor avg time to clear customs 6.3 Index lor comparative advantage in Labor 3.1 Dummy lor firmswith IS0 certification 3.2 Dummy lor flrms prwiding trening to employees I 1 1 Source:Escribano et al. (2005). 3.12. Only 25 percent of the Brazilian road network is consideredto be in good conditions. This implies an estimated US$300 million per year in additional vehicle 24 This result can be explained because Brazil has spent more (as a percentage of GDP) that the average L A C country on infrastructure and the infrastructure network inBrazil is, therefore, larger than an average L A C country. 43 operating costs to road users (principally from the private sector). The annual death toll caused by road accidents is estimatedat well over 30,000 persons per year -over 3 times that caused by AIDS inthe country: While many of these accidents should be attributed to other causes (speeding, drinking and vehicle conditions) road conditions are also consideredto be a significant contributing factor. The roadnetwork, with an extensionof 58,000 km and transporting over 70 percent of total physical output in the country, was built during the 1950-1970but for most of its extensionthe last rehabilitation activities happenedmore than 10 years ago. According to World Bank data, public investments in road transportation declined from 0.18 percent of GDP in 1980 to less than 1/3 (0.04 percent) in 1998.Inthe same period, public investmentsinChile raisedfrom 0.78 percent to 1.03 percent of GDP. Over the last 6 years, an annual averageof about RS$600million for rehabilitation and RS$150 million for maintenance have been spent on the paved federal road network. In order to increase the share of road in good conditions tcJ63 percent, World Bank (2006) estimates that roughly RS$1.2 billiona year will be required over at least the next 6 years. Such investments would allow the percentage of network traffic traveling on roads in good condition to increase from 25 percent in 2005 to 63 percentin2010. Infrastructure and Social Development 3.13. A growing body of empirical research confirms that improved access to infrastructure contributes to raise health and education levels of the poorest. For health, access to clean water and sanitation i s clearly fundamental. Diseases from drinking contaminated water and a lack of safe water and sanitation for household hygiene are among the main causes of child mortality (World Health Organization, 2002). InArgentina, a recent study by Galiani et al. (2005) finds that child mortality fell by 8 percent in areas which had privatized water utilities (and hence experienced improved coverage and quality), with most of the reduction occurring in low-income areas where the water network expandedthe most. More generally, Fay et al. (2005) find that allowing the poorest quintile in developing countries the same access to basic services as the richest would reduce child mortality by 8 percent and stunting by 14 percent. There are other, less obvious, linkages. Improved transport facilitates access to healthcare, as well as easier staffing and operation of clinics. For girls, enrollment i s also helped by greater access to piped water, which would otherwise have to be fetched. Electricity also allows more time for study, while the positive health impacts of clean water allow moretime inthe classroom. 3.14. Mendonca and Seroa da Motta (2005) show that access to water and sewage services -as well as education and health care -havean impact on child mortality. The incidence is concentrated on diseases that transmitted through untreated water and sewerage. They estimated that a marginal variation of 1 percent in the rate access to sewage services would reduce in 216 the number of infant deaths in the country. The same marginal variation in the rate of access to water services would reduce deaths by 108.By comparison, a reductionof 1percent inthe rate of illiteracy inwomen older than 15 years old would reducethis type of death in 162, while an increase in 1percent inthe expenditure with health care would reduce the number of deaths by 415. In order to increase in 1 percent the access to water and sanitation services, the authors estimated a 44 cost of RS$ 2.6 and 3.6 million, respectively. The cost to decrease illiteracy in 1percent among women 15 years old and older i s estimated in about RS$ 12.9 million while the cost of increasing 1percent government expenditures in health care corresponds to RS$ 1.95 million. In this context, a measure o f comparative costs of each alternative can be achieved by dividing the cost o f the 1percent variation inthe access to these services by the reduction in the number of infant deaths. The least-cost means to reduce infant death rate i s through a reduction in illiteracy among mothers 15 years old or older, with a cost of RS$ 63,000. Increased expenditure inthe public health system results inRS$ 282,000 by death reduced, followed by RS$241,000in access to sewage services and RS$ 168 in access to water. One major caveat to this result: no externalities o f the investment in water and sewerage are computed, for example, on the value of the property of the poor. 3.15. Preliminary evidence generated by this report also indicates that better access to infrastructure services is strongly correlated with superior educational performance. In a background paper for this report, Neri and Moura (2005) use data from the National Household survey (PNAD) generated by the Ministry of Education (for performance evaluation) to identify differences in three indicators of educational performance (enrollment, attendance and performance) among those that have and do not have access to infrastructure services (electricity, water and sanitation and telecommunication) and housing, after controlling for idiosyncratic characteristics of the population (such as gender, race, parent's education, household income, etc.). The research was motivated by the fact that bi-variate analysis o f these two groups (with vs. without access to infrastructure and good vs. bad housing conditions) indicated substantive differences: for example, students with access to water and electricity at home attained, on average, higher scores in the 1999, 2001 and 2003 proficiency exams. After controlling for other idiosyncratic characteristics, most o f the results hold as well when a multi-variate analysis i s performed. Although causality may seem intuitive, these results need to be understood as correlations only and no causality may be implied. Table 3.2 illustrates the differences in educational performance for afro-descendent women in the states of Slio Paulo when access to infrastructure i s taken into account. For example, for a ten-year old afro-descendent girl living with 7 other residents in a precarious neighborhood, the probability of being formally enrolled in 2001 increases by 36.2 percent when she has access to infrastructure. This same hypothetical girl would increase her probability of never missing a class during the school year (absent) by 44.7 percent and her scores in the proficiency exam would be 12.3 percent higher (assuming she was enrolled in the 4* grade) in the case she has access to infrastructure services. Holding everything else constant (including access to infrastructure), Table 3.3 illustrates the results associated with a change in housing conditions, again for the same hypothetical girl. For example, access to good housing conditions is associated with a performance 7.1 and 4.6 percent higher in the 4* and 8* grades, respectively in 2001. A summary o f the methodology and further illustration of the results may be found inAnnex 1. 45 Table 3.2. 1Silo Paulo state: change ineducationalperformanceassociatedwith access to infrastructure (in percent) Attribute First Year Last Year A B A B Proficiency 4" grade 11.4 12.3 10.9 11.8 8" grade 12.3 12.9 11.8 12.4 31dyear of high school 13.1 13.8 12.8 13.5 Attendance Never absent 49.9 44.7 -- 6+ days absent -20.8 -23.5 -- _--_ Enrollment 38.1 36.2 29.0 31.9 Notes: For proficiency: first year (1999) last year (2003). For attendance: only one year, considered first year, is available (2001). For enrollment: first year (2001) and last year (2004). "A" is associated with a household of four residents in the house. "B" is associated with a household with eight residents in the house. For attendance and enrollment, the household is located in a precarious neighborhood, built in a terrain provided by employer or other means. Estimated for an afro-descendentgirl living in an urban area, whose parent's education is inferior to 4" grade and attendingapublic school. Source: Neriand Moura (2005). Table 3.2. 2 Silo Paulo state: results for difference in percentage for scenarios with good housing conditions inrelation to those with bad conditions inthe presenceof infrastructure variables (inpercent) Attribute First Year Last Year Proficiency 4" grade 7.1 6.8 8" grade 4.6 4.4 31dyear of high school 4.6 4.5 Enrollment 3.9 3.5 Notes: For proficiency: first year (1999) last year (2003). For enrollment: first year (2001) and last year (2004). Good conditions are associated with 4 residents in the house, non-precarious neighborhood, and house built inan owned and paid terrain. Bad conditions are defined as 8 residents inthe house, precarious neighborhood, and house built in a terrain provided by employer or other means. Estimated for an afro- descendent girl living in an urban area, whose parent's education i s inferior to 4th grade and attending a public school. Source: Neri and Moura (2005). 3.16. The expansion of infrastructure stock may favor income distribution and poverty reduction,helpingto frame a pro-poor growth pattern. Using a large panel data set encompassing over 100 countries and spanning the years 1960-2000, Calderon and Serven (2004) find a robust impact of both infrastructure quantity and quality on income distribution in Latin America: bringing the stock and quality of Brazil's infrastructure to East Asian tigers medianwould reduce the Gini index by 9 percent. Ina paper examining whether growth has been pro-poor in Brazil, Menezes-Filho and Vasconcellos (2004) show that poverty is associated with poor access to infrastructure, lower levels of education; having children, being non-white, being unemployed or working in an informal agricultural job. The paper concluded that a 10 percent rise in income reduces extreme poverty by about 8 percent on average, the growth-elasticity of poverty depending positively on the initial level of income and negatively on initial inequality. Regression analysis particularly highlighted the importance of human capital and infrastructure for pro-poor growth. Results also indicated that inequality may have 46 positive effects on growth, which implies that trade-offs may exist between growth and pro-poor growth strategies but that no such trade-offs are associated with investments in human capital and infrastructure. In other words, this means that the "growth cost" of infrastructure investments as part o f a pro-poor growth initiatives may be negligible, reinforcing that carefully planed and ideally private1 funded infrastructure investments are a good component of a growth strategy for Brazil. & 3.3. HOWMUCHNEEDEDTHE LONG I S IN RUN26 3.17. Different approaches can be used to calculate Brazil's infrastructure investment needs, and the estimates vary depending on the objective set (Table 3.3.1). The goal can be set through benchmarking with other countries or relative to a particular policy objective. Simple costing exercises involve picking a particular objective (for example universal coverage of basic services) and pricing it, using either local or best practice or international prices. More sophisticated approaches rely on econometric or engineering/economic models to calculate the objective (e.g. what level of investment i s needed to ensure the integrity of an electricity system). Simple benchmarking 3.18. Benchmarkingcan be done through a simple costing exercise. For example, what would it cost for Brazil to achieve the infrastructure density o f a country deemed appropriate comparator or goal. Korea i s an interesting candidate since its infrastructure coverage was similar to Brazil's in the early 1980s and it i s at the median of East Asia's fast growing "tigers" (Calderon and Serven, 2004). One complication however i s that Korea has a much higher population density than Brazil so using the same road or rail density i s inappropriate. We therefore price the cost of reaching one-third the transport infrastructure density. Using average international prices, we estimate that it would cost between 72 percent of Brazil's GDP (if the road target includes overall - paved and unpaved- road density) and 194 percent (if the road target only includes paved roads) (Table 3.3.2). Assuming a reasonable rate o f population and GDP growth rate, this would amount to 3 percent to 7.5 percent annual investments over the next twenty yearsS2' Additional resources would be needed for maintenance and rehabilitation. 25These results are consistent with the interpretation that access to infrastructure services contributes for improving economic opportunities for the poor. Infrastructure development in poorer regions reduces production and transaction costs: poorer individuals and underdeveloped areas become connected to core economic activities; gaining access to additional productive opportunities improves (Gannon and Liu, 1997). Infrastructure access can raise the value of the assets of the poor as well: recent research links the asset value of poor farm areas as proxied by the net present value of the profits generated by their crops - - to the distance to agricultural markets. In this context, improvements in communication and road services imply capital gains for these poor farmers (Jacoby, 2000). It is difficult, however, generalize those conclusions as connectivity may also cause poverty - at least in the short-term - by increasing product competition. 26B y M.Fay and T. Yepes (methodology inAnnex 2). 27Assuming a 2.66 percent geometric growth rate for GDP and 1.24 percent population growth average per year. 47 Table 3.3.1 Different approaches can be use to estimate expenditure needs ininfrastructure "Benchmarking" Set target Simple costing - Physical Stock target: what would Ex: exercise it cost to get Brazil's infrastructure - MDGs - what would it cost for Brazil to (per capita; per unit of GDP; per achieve universal service coverage in water km2) to the level of the East Asia and sanitation, and electricity? median? - Flow target: how does Brazil's expenditures on infrastructure compare to peers. Sophisticated Econometric: Engineering-economic models: models -Supply: What level of infrastructure These are "set" targets inasmuch as the target coverage i s needed to achieve x i s a particular level o f coverage and quality percent level of growth and reduce as defined through engineering-economic inequality by z percent. For example models the model developed by Calderon - Power sector: well defined international and Serven (2004). methodology, which estimates the investment -Demand: What level of needed to maintain the integrity of the infrastructure coverage will be network and satisfy predicted expansion in demanded by firms and consumers, demand. for given growth projections? - Waterhanitation: financial model that estimates investment needed to attain the coverage goals set innational plans. - Roads: well defined methodology for rehabilitatiodmaintenance expenditures; combined with road sector expert opinion on definition of maior corridors and investment needs for their completion. Source: World Bank (2005b). 3.19. This exercise shows that substantial progressin infrastructure does not come cheap. The implication of this exercise is not of course to claim that Brazil could become Korea if it doubles or triples annual expenditures in infrastructure. Nevertheless, most countries that have grown fast over long periods of time have invested substantial amounts of resources in developing their infrastructure - as did Korea over the last decades and as China i s doing now, spending close to 7 percent per annum on infrastructure. Table 3.3.2 The cost for Brazil o f reaching the infrastructure density level of Korea (as percent of Brazil's GDP) Telephone Cellular Electricity All (paved Paved Rail Total (using Total (using mainlines phones and unpaved) roads all roads) paved roads) Road (Per worker) (per hZ> 3% 8% 38% 23% 145% 26% 72% 194% Notes: Inthe case of electricity, the infrastructure stock used is electricity generating capacity but the price includes a provision for associated network costs; for road and rail density the goal is to reach one third the road density of Korea. For a discussion of the methodology see World Bank (2005~). Source: Own calculations. 48 Costing set targets 3.20. Universal coverage of water sanitation and electricity could be achieved within 10 years, at a cost of 0.2 percent of GDP. The total cost would be 2 percent of GDP, of which 0.41 percent for electricity, 0.33 percent for water and 1.3 percent for sanitation. This amounts to 0.2 percent per annum between now and 2015. This estimate i s relatively modest partly because it relies on alternative technologies in circumstances where the price o f a connection to the grid or the network would become prohibitive.28A separate exercise prepared by the Brazilian government with the support of the World Bank estimates that RS$ 9 billion per year are needed for the universalization of water and sanitation services by 2015. Estimating derived demandfor infrastructure 3.21. This approach estimates the demand for infrastructure services associated with a particular economic growth rate. It i s based on the methodology developed by Fay and Yepes (2003), and is by no means normative - it does not refer to any socially optimal measure of need for infrastructure. Instead the investment needs refer to the additional investment necessary to satisfy consumer and producer demand based on predicted GDP growth. Finally, it does not include any estimates o f rehabilitation nor of urban transport, ports or airports infrastructure. As such these numbers should be considered a lower bound estimate. 3.22. Projections for different growth scenarios suggest increased demand for electricity generating capacity, potable water and sanitation, and a substantial increase in teledensity (mobile and fixed). However, the model predicts no further road or rail construction, so a constant road and rail density. The limitation of the model is that it cannot capture road improvements (such as enlargements) or urban roads and ring roads. The estimated cost o f the suggested additional demand for infrastructure would be an annual 1.5 percent of GDP for investment. This estimate is not very sensitive to the growth scenario selected. An alternative estimation method (Arellano-Bond) shown in the last row of Table 3.3.3 produces slightly higher expenditures. 3.23. Maintaining existing assets is also very costly. Fay and Yepes (2003) also develop a simple method to estimate maintenance needs based on best practice estimates of maintenance costs as a share of replacement values. Applying this to Brazil suggest maintenance would cost another 1.5 percent of GDPper annum. For electricity, the average price of US$ 1,000 per new connection implies that households too far from an existing network to be connected at a price inferior or equal to US$ 1,000, would be served by alternative off-grid technologies. Since there are still some households to be connected that are relatively close to existing grids and could be connected at lower prices (say US$500or so), the price that determines a switch to alternative off-grid technologies could be somewhat above US$ 1,000. Inthe case of water and sanitation, it also assumes that households in low density areas would not have access to sewerage connections but alternative sanitation systems (e.g., latrines) and that a proportion of households would have access to water but not necessarily in house connections. The prices used are US$ 400 for water and US$ 700 for sanitation per new household connected. The estimates we use are: 2.45 million households without electricity, 4.92 million without water, 11.2 million without sanitation, and GDP equal to US$ 605 billion. 49 3.24. Drawing these estimates together shows how dramatically "investment needs" vary according to the goal. Responding to increased demand (1.5 percent of GDP), achieving universalcoverage (0.2 percent of GDP) and maintaining existing assets (1.5 percent of GDP) adds up to 3.2 percent of GDP per annum and can be considered a lower boundestimate. A more ambitious investmentprogramthat would includecatching up to more aggressively investing countries, such as Korea over 20 years, as well as maintaining existing assets and achieving universal coverage varies between4.7 percent and 9 percentof GDPper year over 20 years. Table 3.3.3 Resource needsfor the infrastructure sector - a lower bound scenario Growth scenario Electricity Mainlines Mobile Paved Rail potable improved Total (annual growth rate) Generation Road water sanitation Capacity. Stock (MW per per 1,000 people (hn/km2) %households with capita) access 2000 0.40 140 136.6 0.015 33.4 87.0 76.0 2005 2.0 0.45 188 236.7 0.015 31.3 88.7 78.0 2010 2.0 0.50 247 455.6 0.015 29.4 90.2 79.7 2010 3.5 0.50 254 468.1 0.015 29.4 90.8 80.7 2010 5.0 0.51 261 472.0 0.015 29.4 91.5 81.6 Investment as percent of GDP 2000 0.74 0.20 0.31 0.00 0.00 0.04 0.07 1.36 2005 2.0 0.72 0.17 0.24 0.00 0.00 0.04 0.06 1.23 2010 2.0 0.74 0.21 0.51 0.00 0.00 0.03 0.06 1.55 2010 3.5 0.73 0.21 0.49 0.00 0.00 0.03 0.06 1.52 2010 5.0 0.71 0.21 0.46 0.00 0.00 0.03 0.06 1.47 Investment and maintenance as percent of GDP 2000 1.19 0.38 0.43 0.18 0.02 0.12 0.19 2.52 2005 2.0 1.20 0.41 0.46 0.16 0.02 0.11 0.17 2.52 2010 2.0 1.26 0.51 0.91 0.15 0.01 0.11 0.17 3.11 2010 3.5 1.20 0.49 0.86 0.14 0.01 0.10 0.16 2.97 2010 5.0 1.15 0.47 0.80 0.12 0.01 0.10 0.16 2.81 Arellano Bond estimates of investments 2010 3.5 1.21 NC NC 0.19 NC 0.14 NC Notes: Growth scenario refers to averageannual growth inthe preceding five years. NC: Not consistent estimations. Source: Own elaboration. 50 PART11:HOW TO REVITALIZEINFRASTRUCTURE INVESTMENTS 4. PRIVATEPARTICIPATIONSININFRASTRUCTURE INBRAZIL This chapter reviews the performance of private participation in infrastructure in Brazil between 1997 and 2003. Given simple standard financial results, the public in Brazil tends to assume that the private sector's returns on infrastructure investments are too high, while private investors are avoiding further investments inthe sector, alleging that returns are not high enough. In addition, there i s a general discontent with the privatization process, at least inpart due to unfulfilled expectations o f results. The current chapter addresses two issues: (a) whether private investments in infrastructure have earned adequate returns as compared to the opportunity cost o f capital, and (b) the outcomes of the privatization of state-owned firms. 4.1 WERE PRIVATEINVESTMENTSADEQUATELYPROFITABLE? 4.1. This section extends the results of Sirtaine et al. (2005) from a sample of 18 infrastructure concessions that involved private investors (12 in the power industry, 3 in fixed telecommunications and 3 inwater and sanitation). The study i s built on the Capital Asset Pricing Model (CAPM), which formalizes the observation that expected returns are related to risk. The study uses four measures o f effective returns: (a) shareholders' internal rate of return (Shareholder IRR), (b) return on equity (ROE),(c) project internal rate of return (Project IRR), and (d) return on capital employed (RoCE). The first two are measures o f the returns earned by equity investors, while the last two are measures of the profitability of concessions overall, independent of their financing structure. A summary of the methodology may be found inAnnex 2. Infrastructure Concessions'Returns 4.2. The measure of returns chosen defines which cost-of capital indicator needs to be employed. The Project IRR and the RoCE, which measure returns earned on the concession's overall capital structure, must be compared to the weighted average cost of capital (WACC), which represents the expected return on all o f a company's securities. 51 The ~ h ~ r e h o l dIRR~ and the ROE,both of which ~ e a s returns earned over equity e ~ r ~ capital, must be ~onipare~the a~pro~riate to cost of equity (CE),which is a ~ ~ e ~of~ u r e the r e t u ~~ n ~ e s ~ orequire on equity i n ~ y e s ~ ~ e ngiven the level of risk of such r s ~ s , 4.3. ~ e ~ aofrthe indicator used, data show that returns on ~ n r a ~ t r ~ c t ~ ~ e ~ ~ ~ ~ s ~ ~ n c ~in~Brazilohavenot beensufl s j ~ ~ the lowest in water and s a ~ ~ ~ a tResults for the kvater and sstn~tationsector were, in j o ~ . princ~p~e,~ ~ n e ~ ~ e cast eitd is the sector usually considered to have the highest level of , r e ~ ~ ~ au t~oi cr ~e i ~ a i ~ i ~ ~ ~ 4.4. The poor p e r f ~ r of ~ ~~ ~~ ~f ~ ~~ s t~ r u~in Brazil~~is~s ~ ~ ittol ~ ~~ ~ a ~ t~ ~ i ~ n ~ the ~ v results or~ LAC~countries during l P ~ ~ - a~~ t~h~~ ulthere are ~ ~ e ~, h differences in relative sector ~ e r f ~ r In ~ ~ c ~ all sectors,~ the net a~rera~a~~~~~ . e operating profitability of c o ~ ~ c e ~(RoCE) nin~ the region was not high ~ ~ ~ o n * ~ ~ c o ~ ~ ~ atoethe respective cost of capital, Historical returns were negative in all LAG r d countrie~, except in Colombia, where they were ~ ~ a r ~ ipositive. l G ~~ ~ ~ ~ e ~ s i o n ~ ~ a l s h a r ~ h o ~ don~average, earned negative returns on their ~ n v e s ~ (as~expressed by ~ s , ~ e ~ ~ s the difference between ROEaiid C,). ~ o n ~ e s s i oin~the water sector appeared to be s rela~i~relythe Least a t t ~ a ~ ~with i v e ~eoncessions in the ~ ~ ~ e ~ o n i ~ ~ isector a ~ ~ o n s n i ~ giving the i ~ ~ ~toebe the sriiost ~or o~~ ~ toverall,eregardless o f the ~ndica~orused. ~ ~ a ~ l 52 4.5. Other studies also showed negative returns on energy investments and positive results for telecommunications. Based on a sample of 25 electricity distribution concessionaires, accounting for 88.1 percent of the local distribution market, a recent PEA study found that return on equity for Brazilian electricity distribution companies was systematically negative from 1998 to 2003, recovering only in 2005.*' In telecommunications, calculations made by Campbell et al. (2004) found CEand WACC for fifteen Brazilian telecom operators in 2004 to be, on average, 16 and 14 percent, respectively. The market shares of the firms in the sample covered 100 percent of the Brazilian market and included both fixed and mobile telecommunication. The numbers were somewhat similar to the ones calculated for the period 1997-2003, especially for the case of WACC. 4.6. Thus, contrary to the public perception, private investors in infrastructure concession are not making excessive profits. The estimated average internal rate of return for projects with terminal value (TV) in 1997-2003 is negative for telecommunications (-26 percent) and energy (-5 percent) - indicating returns below their opportunity cost of capital - and positive for water (16 percent). The average return-on- equity for infrastructure services varies between 3 percent (in water) and 5 percent (in telecommunications). Negative perceptions of privatization may be due to misleading evidenceprovidedby standard(unadjusted) financial results. 4.7. More important, perhaps, returns indicate that infrastructure investment in Brazil is a relatively risky business. Returnson infrastructure investments are extremely volatile either over time or among firms within the same sector. Telecommunications i s the only sector that managedto yield positive rates of ROEin 1997-2003 with a relatively high level of variance (Figure 4.1.2a). The lower bound point on the scatter plot represents the RoEs from Light, the distribution company that operates inthe city of Rio de Janeiro, which allegedly has had high non-technical losses (Figure 4.1.2b). Without the Light data points, RoEs' variance has not significantly increased over time, with the distribution of points for 2003 shifting gradually upward inthe graph.30The level of risk of infrastructure investments inBrazil is consistent with other LAC countries. 29,Among the groups with greater profitability after the energy crisis in 2002 were EDP Brasil, CPFL Energia, Neoenergia, Cataguazes and, starting in 2004, Elektro and public-owned distributors. When compared to selected distribution companies in Argentina, Chile, and US.results show that Chile and the U.S. usually managed to reward shareholders according to the opportunity cost of capital. 30 The median ROEin Brazil in the energy sector, for example, was 6 percent in 1997, zero in 1999, -7 percent in 2002, and 8 percent in 2003. Interestingly, the standard deviation of the ROEsignificantly increases after the energy crisis of 2001, indicating that the crisis affected companies indistinct ways. The water sector is the least stable across time among the three sectors in the sample, but its performance i s more homogeneous across companies. Telecommunications present some variation in the ROEover time, but it does not show a median negative return between 1998 and 2003. All three sectors show an increase in the dispersion of the median return after year 2000. 53 4,lO. Despite these negative s h o ~ t ~ tp~osp~cts~~ a ~ ~ are capable iof~ n ~ e ~ ~ e ~ ~ ~ ~~ d e q~returns ~ the l r ~ u ~ ~ in ~For that, however, s ~r~ ~ ~~ r~n.~ ~~ ~~~~~~ t~ e r a ~ ~~ ~ rely both on various sources of r e ~ ~ ~ ~ n e~ri an ct ~l uo~~d ~i1 g~ ~~ ~~~ e a ~ e~~fecs~eandt , ~ n g i ~ n~ capital gains) and on cons~s~en~lyo u t ~ e ~ ~ho~ sr ~~o~r imarket ggrowth over the entire ~~ a~l ~ lengtf~of their ~ o ~ ~ c ~ AssshownI in Figure 4-13, while the IRR is ncgativc - and s i ~ ~ . therefore lowcr than the WACG - for all c ~ ~ n t r l cins the sample, the: IRR with TV is above the: average:%?ACCwhen future growth is at. Ieast.cqrtal to each country's average ~ ~ ~ s t . oer i ~ aol r ~growthi ~and the residual valuc added is taken into ~ ~ ~Tfiesc ~ u n ~ . o ~ ~ results suggest that ~o~1cess~~~1a~resoperate with long-tern1 p e r s p e ~ ~ ~and~rely on the ~ e s entire ~ o ~ ~ ~period ~ order~eo build an a ~ e ~ ureturn ~ e s in i o ~ t e e 7 ~~ 2QO5). at., i ~ ~ i ..,............................................................ Figure4.1.3 Avertzge cotlcesslonreturitby country ".................................. ................................ ............................. ................................ " .................",......,....."............................................. ................................................... ......................... I I I I d One ~ o ~ ~ ~ factor fori lower returns i ~ ~ t n g Index (IGP-M), used in almost all concession contracts (see next section in this chapter). It reveals that the index was an imperfect substitute for traditional financial-hedging, underlining the need for future concession contracts to appropriately address the issue of exchange rate risk. 4.13. Appropriate allocation of exchange rate risk is likely to continue to be a central issue in attracting private capital to infrastructure investments. Preliminary discussions within the federal administration in 2005 considered the possibility of replacing the IGP-M. One hypothesis under consideration was the use of sector-specific cost-related indexes. This may be a good strategy as it eliminates the implicit and imperfect hedging mechanism that existed in the contracts and forces a separate discussion about both operational costs and managing exchange rate risks. 4.14. A second factor contributing to low returns on infrastructure investmentsis poor enforcement of tariff policies. This can be due to weaknesses in sector laws or concession contracts. For example, telecommunications present better returns than the other sectors. This i s not surprising because telecommunications i s the sector with the most stable and consolidated regulatory environment - indeed, it was established before privatization occurred. Inthe water sector, an extreme case, the regulatory environment i s incomplete and contracts are subject to systematic political interference (see Chapter on Regulation and Contract Renegotiations). 4.15. A third factor may be the persistence of non-technical losses, particularly in the electricity and water sectors. Light, the electricity distribution company that operates in the city of Rio de Janeiro, reports that 30 percent of the energy that it distributes i s stolen through clandestine connections. The state-owned company of water and sanitation in S5o Paulo state (SABESP) had economic losses due to illegal network connections that reached RS$48 million in 2003 just for the metropolitan region of S5o Paulo city. This accounted for 1.2 percent of the company's profits or 45.7 percent of the investments made by the company in2004 (see next section of this chapter). 4.16. In the electricity sector, another contributing cause for reduced operational returns inBrazil may have been the 2001energy crisis. Insufficient water levels inthe Brazilian reservoirs led to a negative return on energy investments in that year. The solution to the crisis was the implementation of electricity consumption limits for households and penalty applications for those that exceeded the limits.32As a result, the consumption of electricity sharply decreased, leading to financial losses to the concessionaries. Water companies also had negative returns in 2002, probably due to a decrease in the consumption of water, as a consequence o f the measures taken for the energy crisis. 4.17. There are signs that Brazil's cost of capital may have declined in recent years. Despite their high levels, both cost of capital indicators (CE and WACC) show signs of recent decline. Brazil has one of the highest opportunity costs of investing in 32The government prepared an educational bulletin for consumers approaching the energy sector inBrazil, the reasons behind the crisis, and what could be done by consumers to save electricity in their daily activities (see http://www,idec.org.br/files/MMA-energia.pdf). 56 infrastructure among Latin American countries (9 percentage points higher than Mexico and 10 percentage point higher than Chile, in 2004). Capital costs for the three sectors analyzed remained stable from 1996 to 2001 and started to decline after 2001. Telecommunications presented the highest WACC and water the lowest until 2001, when both telecommunications and energy dropped below water. The WACC i s around 5 percentage points lower than the CEfor water and energy and around 8 percentage points lower than the CE for the telecommunications sector; while the WACC for the water sector remains constant over the whole sample period. 4.18. One important cause of the low profitability of infrastructure investments is the high investment levels in the initial years of concession. Also, as network expansion reaches more remote areas, the profitability of the privatized distribution companies tends to be further reduced. The local telecommunications companies in Brazil had ambitious expansion targets inthe first years after privatization: the number of fixed lines per 100habitants increased from 8.5 in 1994 to 27.8 in2003. Universalization targets also existed in the energy and water sectors but on a much lower scale.33In the case of the LAC region, average investment levels as a share of total revenues inthe first initial years varied from 21 percent in energy to 32 percent in water. As the investment levels stabilize over time, concession returns tend to increase. 4.19. The evolution of WACC in the three sectors illustrates the importance of the consistent design of infrastructure concessions. It i s interesting to note that only after 2001, when investment levels were already declining, the cost of capital in telecommunications fell to levels lower than those in water, a sector with a more cumbersome regulatory framework (Figure 4.1.4). Data for the U.S. during the 1990s also place the WACC for telecommunications among the highest for infrastructure sectors. This most likely reflects the investment boom associated with the expansion o f the internet and competition from internet-related services. In designing future concession projects, therefore, Brazil must achieve a careful balance among its coverage (and thus investment) needs, the expected return on the concession, and the opportunity costs of investors, if further private capital i s to be attracted. 4.20. A positive factor favoring the decline of the opportunity cost of capital for infrastructure concessionsis the reduction of the country risk premium. The country risk premium is the main discriminating factor for the CE and the WACC. There is a noticeable correlation betweenthe CEand the country risk premium. Not surprisingly, the CE in Brazilian infrastructure sectors has been falling since 2001. The cost of capital, estimated by the WACC, follows the same pattern as the CE. Inparticular, the decrease in the CE after 2002 can be explained by the mitigation of the political uncertainty associated with the 2002 presidential elections (Figure 4.1S).The reduction in country risk is a reason for optimism, as the country may find it easier to attract private capital to i t s infrastructureprojects. 4.21. As country risk reduces, Brazil can obtain significant gains by creating a credible and stable regulatory environment for infrastructure investments. Stability 33 For the case of LAC see Sirtaine et al. (2005). 57 Box 1The Case of the Water Sector: 1s There a Paradox? The watcr sector is normally c~~ns~dere~ the sector uith the least caprtcit) to attract private inve~t its r e ~ ~ ~e a~ ~~v i~r o) i ~ r~~u n~~p.r i ~ etlcaugh, ~ t l a ~ ~hasi sshown it tu have relatttcly itsw cost y e i n ~ l y s and highreturns.Hoib can this a p p a ~ ~paradox be e ~ p ~ ~ i ~ ~ e d ? i i ~ for the the medianRUEof the water *~cctoris worth noting, It nteaiis that the Rater sector makesan efficient use of the capitalempl(~~ed gencratcspoorer revenues to its ~har~h~l~ers. but 4.2 WASmIVATIZATI0NA FAILURE? 4.22. Comparisonsof performanceindicatorsof utility companies before andafter privatizationgeneral1 show a clear improvement after privatization in the LAC region and el~ewhere!~Inthe fixed telecommunications sector in Brazil, for example, the number o f subscribers per employee more than doubled; the percentage of incompleted telephone calls fell almost by half; and, the number o f subscribers per 100 inhabitants rose by 52 percent. Coverage of electricity distribution expanded by 5 percent, while distributional losses in the water sector decreased by almost 13 percent - more than twice the result for L A C (Table 4.2.1). But were these improvements just a continuation o f pre-existing trends or did pre-existingtrends really change after reforms? Addressing this question i s a key step toward better understanding the effects of privatization inthe provision of infrastructure services inBrazil. 4.23. This section employs the methodologyand the datasetavailable in Andres et al. (2006)to examinethe Brazilianprivatizationexperience.The focus is on four main variables: (a) labor productivity, (b) distributional losses, (c) service coverage, and (d) price. The series for each o f these variables were normalized based on observable data, defining the value 100for the first year when a firm was privately owned and computing results inlevels (and changes inlevels) and ingrowth rates (and changes ingrowth rates). Three periods were constructed: (a) the pre-privatization period, referring to the three years prior to the transition to private ownership, (b) the transition period, covering the years starting at the time when the reform was announced and ending one year after the concession or the privatization was awarded, and (c) the post-privatization period, referring to the three years after the transition. The sample consists o f an unbalanced panel data of 181 firms in the electricity distribution, fixed telecommunications, and water distribution sectors.35Methodological details and tables with complete results are presented in Annex 3.36 Results for each of the four variables are discussed in the remainder of this section. 34 See, for example, Boardman and Vining (1989), and Megginson, Nash and van Randenborgh (1994). 35 For electricity distribution, the sample consists of unbalanced panel data that includes 116 firms and 1,103 firm-year observations in Argentina, Bolivia, Brazil, Chile, Colombia, El Salvador, Guatemala, Nicaragua, Panama and Peru. For fixed telecommunications, the sample consists o f unbalanced panel data that includes 16 firms and 267 firm-year observations in Argentina, Bolivia, Brazil, Chile, El Salvador, Guatemala, Guyana, Jamaica, Mexico, Nicaragua, Panama, Peru, Trinidad and Tobago, and Venezuela. Finally, for water distribution, the sample consists of unbalanced panel data that includes 49 firms and 515 firm-year observations inArgentina, Bolivia, Brazil, Chile, Colombia, Mexico and Trinidad and Tobago. 36 Note that to avoid the problems that arise in cases when there are different companies across periods, first it was calculate the difference within each firm and, then, the average difference between periods. 59 Table 4.2.1 Brazil andLAC without Brazil: pre- and post-privatization comparisons per sector (inlevels) Variable Stats Brazil LAC withoutBrazil Pre Post %Change Pre Post %Change (1) (2) (2)-(1) (3) (4) (3)-(4) Electricity Distribution Laborproductivity Mean 432.9 642.8 62.6 377.8 586.5 65.3 (connectionsper employee) sd 173.6 218.4 45.2 151.2 261.4 62.7 N 33 34 32 31 41 29 Distributional losses Mean 0.16 0.15 -5.4 0.18 0.14 -15.7 sd 0.08 0.08 17.7 0.07 0.05 16.8 N 31 33 30 38 48 36 Coverage Mean 0.82 0.86 5.2 0.69 0.72 8.4 (residentialconnections per 100HHs) sd 0.12 0.11 4.3 0.19 0.18 10.5 N 32 31 31 44 46 44 Average pricesper MWH Mean 90.42 102.4 13.8 101.1 101.98 3.1 (in real local currency- 100) sd 6.08 3.26 8.5 11.2 8.27 17.6 N 39 40 39 42 47 42 Fixed Telecommunications Labor productivity Mean 196.0 424.3 118.1 65.6 93.2 43.2 (subscribers per employee) sd 25.5 28.3 20.3 29.9 44.4 44.4 N 3 3 3 13 12 12 Distributional losses Mean 46.50 40.62 -12.40 49.68 56.75 -32.70 (percentageof uncompletedcalls) sd 4.24 1.39 5.0 33.31 42.55 43.6 N 3 3 3 5 8 5 Coverage Mean 10.26 15.43 51.9 6.88 8.28 23.6 (subscribers per 100inhabitants) sd 4.12 5.66 5.4 4.67 5.52 15.2 N 3 3 3 13 13 13 Average prices (in real local currency - 100) For a three-minutecall Mean 63.74 100.33 57.4 99.12 81.83 -9.4 sd 0.00 4.81 7.6 25.99 26.56 49.3 N 3 3 3 7 7 7 Monthly chargesfor residentialconnection Mean 48.95 122.97 151.2 98.94 140.1 43.20 sd 0.00 8.27 16.9 14.02 98.3 96.8 N 3 3 3 9 9 9 For installinga residentialconnection Mean 432.86 22.12 -94.9 105.2 134.0 26.9 sd 2.57 1.25 0.3 27.1 90.6 78.4 N 3 7 3 I I I Water Distribution Labor productivity Mean 154.5 263.2 85.2 237.5 334.7 48.7 (connectionsper employee) sd 43.1 64.3 55.3 142.4 169.3 39.6 N 4 8 3 21 30 21 Distributional losses Mean 44.80 36.67 -12.7 45.40 42.49 -7.5 sd 12.26 11.23 6.5 13.76 13.57 13.7 N 4 6 3 19 22 19 Coverage Mean 69.04 64.54 5.7 73.90 74.29 6.1 (residentialconnections per 100HHs) sd 37.21 40.77 4.2 31.23 33.83 6.3 N 6 12 6 24 26 22 Averageprices per cubic meter Mean 106.90 86.94 -23.6 95.28 118.62 26.5 (inreallocalcurrency 100) - sd 11.43 8.79 1.9 8.15 29.90 42.0 N 3 2 2 11 11 11 Notes: To avoid the problems that arise when there are different companies across periods, first the difference within each firm was calculated and, then, the average difference between periods. sd = standard deviation. N = number of firms. H H s = households. Source: Own elaboration. 60 Labor Productivity 4.24. The rate of improvement in labor productivity rose for all three infrastructure sectors, particularly during the transition period (Figure 4.2.1). In electricity distribution, connections per employee increased 19.0 percent per year in the transition period, compared to 15.3 percent inthe pre-privatization and 4.6 percent during post-privatization. Infixed telecommunications, the annual growth rates inthe number o f subscribers per employee are the same for the pre- and post-privatization periods (18.8 percent), corresponding to almost half the annual growth rate during the transition (35.0 percent). A pattern similar to the electricity sector was found for water distribution: annual growth rates increased remarkably during the transition, from 3.1 percent to 28.4 percent, followed by a decrease after the change in ownership (6.3 percent per year). All sectors but electricity presented a relatively low standard deviation. Except for electricity, the rate of improvement of labor productivity during the post-privatization period i s at least as good as inthe pre-privatizationphase. 4.25. Improvements in labor productivity were related to both an increase in output and a sharp decline in employment. The number of connections of electricity and water distribution had important increases duringthe three periods, presenting a clear positive trend. Their respective annual growth rates were 4.4 percent and 6.6 percent during pre-privatization and 4.4 percent and 7.4 percent in the transition, decreasing to 3.6 percent and 2.4 percent in the post-privatization period. For fixed telecommunications, since the number o f operating lines was increasing, the average number of fixed lines inthe transition was 76.5 percent higher than the pre-privatization level. However, most of the annual increase in fixed telecommunications' output occurred during the transition, when this rate reached 19.4 percent per year (vs. 9.4 percent and 2.2 percent during pre- and post-privatization, respectively). These increases resulted from supplying service to long waiting lists accumulated when concessions were awarded. Since most state-owned enterprises were over-staffed, significant reductions in labor were observed during both the transition (when governments decided to reduce the number o f employees to increase the value o f the companies) and post-privatization. The results show that labor layoffs inthe electricity and water sectors were the largest during the transition, (-11.1 percent and -14.0 percent per year, respectively), while, for the fixed telecommunications sector, most of the reductions happened during post-privatization, with an average annual decreaseof 15.5 percent. 4.26. Changes in labor productivity caused by privatization in Brazil compare favorably with the results achieved within the LAC region. For electricity distribution, Brazilian annual growth rates were higher than those for L A C in the pre- privatization period (15.3 percent and 13.4 percent, respectively) and the transition (19.0 percent and 18.4 percent, respectively), but became lower in Brazil than in LAC during post-privatization (4.6 percent and 5.5 percent, respectively). Also, the acceleration o f labor productivity annual growth rates for energy during the transition and the deceleration of these rates during post-privatization were greater in Brazil than in LAC. These findings may be partially explained by the originally higher levels of productivity in Brazil that made permanent further improvements less likely. In fixed telecommunications, during the three periods, Brazilian annual growth rates were higher 61 than those for LAC, but they reached similar values in the post-privatization phase. In water distribution, even though the LAC original level was 54 percent higher than that of Brazil, trends in both groups were relatively close in pre-privatization (annual growth rates in Brazil and LAC were 3.1 percent and 5.5 percent, respectively). However, inthe transition, Brazil presented a higher annual growth rate than LAC (28.4 percent and 17.5 percent, respectively). This situation changed in post-privatization, when the annual growth rate of labor productivity in the water sector in Brazil was outpaced by that of L A C (6.3 percent and 7.3 percent, respectively). Figure 4.2.1 Brazil and LAC without Brazil: cross-period comparisons for labor productivity per sector (inlevels) (a) Electricity distribution (b) Fixedtelecommunications (c) Water distribution Electricity Distribution - 81 FixedTelecornmunlcatlons I f Waterand Seweraae Lsbor PrcducbiW:ConnWns peremployee Lsbor Prcductinty:Subacrberrperemployee 8 -LAC -Trend Withobi BI.*c Braal---- Trend belore-- -- I- LAC Without B U C Brazil---- Trend belo--- LAC wihobi Br+ Brazil---- Trend belo*-- an01 -nand aner -Trend aner Source: Own elaboration. Distributional Losses 4.27. Privatization seems to have had a positive effect on distributional losses as well, although the relevance of the impact changes with sector (Figure 4.2.2). Privatization appears to have "only" accentuated a pre-existing trend of reducing distributional losses inthe telecommunications and water sectors, while it seems to have reversed a trend of continuous losses in electricity distribution. Distributional losses in the telecommunications and water sectors were declining at the rates of 1.7 percent and 7.1 percent per year, respectively, during the pre-privatization period, compared to 6.6 percent and 21.2 percent per year, respectively, during transition. Losses in electricity distribution were increasing at 2.1 percent per year before privatization but declined at an average rate of 4.1 percent per year duringthe transition phase. 4.28. The timing and sustainability of privatization effects on distributional losses varied significantly among the three sectors. The pace of loss reductions for fixed telecommunications was higher inthe post-privatization phase (7.7 percent per year) than the other periods, while electricity distribution returned to loss-making - at a rate of 3.4 percent per year - inthat same period. Fixed telecommunications gains were accentuated after privatization, probably reflecting the learning process of privatized companies, as reductions are related to better oversight of employees and overall enhanced strategies to curb non-technical losses. The striking loss-making in the electricity sector is at least 62 partially related to the difficulties inholding back non-paid services. Standard deviations indicate a largeheterogeneity inthe electricity sector. 4.29. Trends in distributional losses presented in Brazil do not differ significantly from those in LAC, although some systematic discrepancies emerge. With the exception of water distribution, a shaper decline in technical losses seems to have occurred in the L A C region. Interestingly, technical losses in the region were only slightly higher than those of Brazil For instance, the percentage of uncompleted calls in L A C before privatization was only 3 percentage points higher than in Brazil. In LAC, technical losses in fixed telecommunications started rising during post-privatization, while results for electricity distribution continued to improve, a clear contrast to what happened in Brazil. Thus, differences in subsequent performance seem hardly to be attributable to decreasing marginal returns. Figure4.2.2 Brazil andLAC without Brazil: cross period comparisons for distributional lossesper sector (inlevels) (a) Electricity distribution (b) Fixed telecommunications (c) Water distribution Electricity Oistribution - FixedTelecommunications Water and Sewerage E- Percentageof uwmple(ed calls DlsmbutbnalLO- 8- X - ' 8 .I- -- 5 0 6 I -LAC -TrendWithovt B r * c Brazil---- Trend be(ore-- I- LAC Withom B u c Braz11---- Trendbelore- aher -Trend LAC Wnhovt B u c Braz11---- Trend befoe--- aher -Trend aner Source: Own elaboration. Service Coverage 4.30. Similarly to labor productivity, the rate of coverage expansion after privatization also improved in the three infrastructure sectors, particularly during the transition period (Figure 4.2.3). Inelectricity distribution, the number of residential connections per 100 households increased, on average, 1.7 percent per year in the transition period, 0.5 percentage points higher than inthe pre-privatizationphase. Infixed telecommunications, annual growth rates inthe number of subscribers per 100inhabitants varied from 9.4 percent before privatization to 17.4 percent during the transition period. Inwater distribution, the average annual growth rate on the number of water connections per 100 households reached 7 percent during the transition phase, compared to almost zero inthe previous period. 4.31. Despite being relatively small, these improvements in the expansion rates of service coverage significantly improved access. For instance, average coverage levels in fixed telecommunications before and after privatization show that coverage ratios increased 2.5 times between these two periods. Improvement in electricity was less 63 spectacular (13 percent) but was still significant, with coverage ratios among privatized f i r m s reaching 86 per 100 households. In water distribution, level improvements in post- privatization (11.2 percent) contributed to reverse a declining trend in the pre- privatization period. The superior performance o f telecommunications i s related to concession contracts that, by conditioning the entry o f concessionaires inmore profitable markets to pre-defined coverage targets, created an appropriate mechanism to induce the system expansion. 4.32. As for the other two discussed indicators (Le., labor productivity and distributional losses), privatization impacts on service coverage in Brazil compare favorably with the LAC average. Brazil's results were (a) much better than L A C in telecommunications, (b) similar to the region in water distribution, and (c) below the region in electricity distribution. This i s partially related to the fact that concessioned companies inBrazil were, on average, more than three times larger than those inthe rest of the L A C region. Brazilian companies' initial coverage levels were 20 percent higher than the average regional level, which increases the cost of marginal expansions. Figure 4.2.3 Brazil andLAC without Brazil: cross period comparisonsfor coverageper sector (inlevels) (a) Electricity distribution (b) Fixedtelecommunications (c) Water distribution Eiectnclly Distribution - FwedTelecommunications Waterand Sewerage Cowage: ResrlenUalC o n n m n per 100 HHS Cowage Sutacrbenper lo0Inhabtmnb Coverage'WaterConnnecUonper1o0HHs I I - LAC WllhoUt B e Brazil---- Trend belore-- - LACWithod B w r C B1a211---- Trend balorC--- LACwiihod B& B~~ZII---- Trend b e l o r - - -Trend aner -Trend aner -Trend aner Source: Own elaboration. Price Evolution 4.33. Prices of infrastructure services, measured in domestic currency (Real), increased in real terms in Brazil after privatization, a trend also found in the LAC region (Figure 4.2.4). In electricity distribution, real prices in Brazilian Real rose 40.4 percent between the pre- and the post-privatization periods. With the exception of residential installation services, prices of fixed telecommunications services measured in real local currency rose as well: the overall price changes for a three-minute call and monthly charges were 63.9 percent and 348.9 percent, respectively, while the fixed-cost for having access to the telecommunications network decreased by 97.5 percent. In domestic currency, the prices of electricity (per GWh) and water (per cubic meter of water) services increased by 40.4 percent and 30.3 percent, respectively. In the L A C region excluding Brazil, the overall price increases in domestic currency for electricity and water services were, respectively, 19.1 percent and 68.9 percent higher after 64 privatization, while the price of fixed telecommunications installation services decreased by an averageof 30.0 percent. 4.34. Prices in Reds sharply increased during the pre-privatizationperiod, but this trend decelerated in the post-privatizationphase. In electricity distribution, for instance, the largest rise occurred during pre-privatization (16.5 percent per year) and the lowest (3.6 percent per year) in the post-privatization. The price of a three-minute call accelerated more during the pre-privatization period (44.3 percent per year) as did the prices of monthly charges (88.4 percent per year). Interestingly, installation fees of fixed telecommunications decreased more once the privatizationprocess started (-64.4 percent per year during the transition) than before it (-27.3 percent per year in the pre- privatization). Price acceleration prior to privatization probably reflects a deliberate strategy of the government to recover tariffs and maximize the value of state-owned- companies. 4.35. Inaddition, the compositionof pricechangesmay haveharmedconsumersof fixed telecommunications services in the short run. The remarkable increase in the monthly charges for telecommunication services in Brazil and in LAC, as compared to the other services, had the clear purpose of increasing revenues of concessionaires - as demand for this "service" is relatively price-inelastic as compared to other services. This may allow the recovery of high-fixed costs incurred by telecommunications firms. One open question i s whether the extension of the network, given the existing cross-subsidies, i s consistent with consumers' willingness to pay. Given their inability to adjust price by reducing quantities, consumers have "voiced" their dissatisfaction and the regulator i s discussingthe introduction of some alternative modalities of monthly sub~cription.~~ 37 Inmid-2005, several parties criticized the monthly charges of fixed telecommunications companies. As a result, ANATEL developed an alternative plan so that companies would be able to charge lower monthly fees. This alternative plan, offered without the bonus of 100pulsos given with the regular monthly charges, was launched in July 2006. Besides, in December 2005, a new tariff plan was proposed by ANATEL. Through ANATEL's resolution No. 23, the tariff regime was changed in the following way: (i)the tariff unit changes from pulso (0-240 seconds) to 1/10of a minute; and, (ii) monthly charges remain, being non- cumulative and differentiated by the type of subscriber, giving the residential subscribers a bonus of 200 minutes (against the previous bonus of 100pulsos) and the commercial subscribers a bonus of 150minutes (90pulsos, before). This resolution stated that concessionaries should adopt the new tariff regime between March and July 2006; however, ANATEL's resolution No. 432 of January 2006 extended this deadline by 12 months. The tariff regime, which includes the use of minutes instead of pulsos, will certainly increase the cost of long time calls, though. To deal with this problem, another alternative plan i s under discussion by the regulatory team (Alternative Plan o f Mandatory Supplied Services, PASOO), which aims at maintaining the cost of long distance calls at the same level as of today. PASOO might be available by the end of 2006, when concessionaries will be required to offer the plan to every customer. After its formal approval, PASOO must be offered just as it was defined by ANATEL. Concessionaries will have the obligation to offer the plan, but its use will be optional, an alternative to the subscribers - this feature makes PASOO different of other alternative plans. With PASOO, ANATEL creates an option to be chosen by consumers, which will benefit specially those who frequently make long distance calls. 65 Figure 4.2.4. Brazil and LAC without Brazil: cross period comparisons for real price inlocal currency per sector (inlevels) (a) Electricity distribution (b) Water distribution (c.1) Fixed telecommunications (c.2) Fixed telecommunications (c.3) Fixed telecommunications (three-minute call) (monthly charges) (installation fee) FixedTelecommunications FixedTelecommunications Avo Prketor a wnhutecall(in~ e ~ocalCunency) I d RealLocalCunenc A-P hfw a re6WlUaIhstalamn(InRe4 W Cunency) 81 \ I- -- -----------. -I LAC WnhoYt 0- Brazil---- TrendDelOe-- -Trend aner Note: For the water sector, price results are not robust (sample size = 1). Source: Own elaboration. 4.36. Prices of infrastructure services measured in U.S. dollars decreased in real terms in Brazil after privatization, as opposed to what happenedin the rest of the LAC region(Figure 4.2.5). Inelectricity distribution, prices inU.S. dollars decreasedby 19.0percent inBrazil and increased by 5.1 percent inLAC excluding Brazilbetween the pre- and the post-privatization periods. The average U.S. dollar price of a cubic meter o f water decreased by 27.8 percent inBrazil, while it increased by 26.5 percent inthe rest of LAC region. With the exception of residential installation services, U.S. dollar prices of fixed telecommunication services also declined in real terms after privatization. The event behind these differences i s the devaluation o f the exchange rate inBrazil, following the change in the exchange rate regime from pegged to floating in February 1999.38The prevalence o f U.S. dollar price decreases in Brazil affected the RoCE of investments made in infrastructure sectors. As a result, it became more difficult to attract new private investments ininfrastructurewhen investors compared Brazil to other countries. 38In 1999,for example, the Brazilian Real depreciated by 48 percent againstthe US.dollar. 66 Figure 4.2.5 Brazil and LAC without Brazil: cross period comparisons for price inUS dollars per sector (inlevels) (a) Electricity distribution (b) Water distribution Electncitv Distribution - Water and Sewerage A* price per m3ofwasted mllected(inDollars) I -- I ! LAC Wnhout 0- 0ra211---- Trend belo--- - LAC WnhoYt Blare 01a111---- Trend bBfom--- -Trend alter - Trend aner j, ;.c j 1- (c.1) Fixed telecommunications (c.2) Fixed telecommunications (c.3) Fixed telecommunications (three-minute call) (monthly charges) (installation fee) FixedTelecommunications FixedTelecommun edTeboommun~catms A* ~ rforba 3-mhutacall(hDOIIO~S) lycharpesfora fesldendsl Dollars) _____----- sI ____----- 0 / ----- t w r y - 5 ? 5 ? -_I LAC Without 0- 0raz11---- Trendbelo%--- --LAC Without 0rs*c 01a211---- Trend befa%--- ---LAC Wnhout 0 w Y - 018111---- Trend belo%--- -Trend ansi -Trend aner -Trend sner Note: For the water sector, price results are not robust (sample size = 1). Source: Own elaboration. 4.37. Table 4.2.2 summarizesthe mainfindings by indicator: Labor productivity showed a significant change among the three infrastructure services, especially during the transition period. These improvements were associated with sharp reductions in the labor force (since most state-owned enterprises were overstaffed) and moderate improvements inoutput. Distributional losses were positively affected by privatization as well. Most improvements in the electricity and water sectors occurred during the transition, while they happened later on for fixed telecommunications. Data for coverage suggest that privatization effect was not as large as in the other two cases. Prices in real domestic currency (Real) significantly increased after privatization, especially during the transition, a common trend in the region. In contrast to the LAC region, real prices in US dollars declined after privatization. 67 .... Table 4.2.2 Brazil: summary of the effects of privatization onthe performance o f utility companies ..... Sector Labor Distributional Coverage Prices Productivity Losses (inReal) Electricity distribution 000 - 0 Fixedtelecommunications 0000 - 3-minute call: 0. monthly charges: 0.0 installation fee: 000 Water distribution 0.. 00 - N.A. Notes: = increase. o = decrease. N.A.= not available. Rice results for the water sector were not robust (sample size = 1). Source: Own elaboration. 4.38. An important lesson that emerges from this analysis is that the way the governmentundertakesreformsand privatizationscan significantlyaffect outcomes and service distribution. The gains from privatization were substantial but relatively concentrated in technical efficiency, as opposed to coverage expansion or efficiency improvements. Also, these possible windfall gains decreased significantly, at best, after the privatization process was concluded. It i s unclear to what extents these gains have been transferred to consumers. These results should hardly come as a surprise: motivated by fiscal concerns, most privatizations were implemented with the objective of maximizing the value of state-owned companies to be sold (therefore the pre-sale improvements) and eliminating the fiscal burden caused by operational efficiency (as effectively occurred in the case of labor productivity and technical performance). Better regulation i s needed to encourage permanent improvement and ensure that gains are at least partially transferred to consumers. 4.39. Negativeperceptionsof the privatizationprocessmay be due to factors other than the economic performance of firms. One possible explanation is that public discontent i s linked to disappointment that outcomes have not matched expectations. Also, privatizations have often been perceived as unfair, rightly or wrongly. Game theory shows that people would rather gain nothing than agree to a deal in which they feel they gain less than their fair share. The implication for governments i s that perceptions of fairness must be carefully managed. That means not only that transactions must be transparent and above board, but that the proceeds o f privatization be used in a way that offsets any possible sense o f injustice (Fay and Morrison, 2005). 68 5. REGULATORYAND POLICY ISSUES This section reviews the evolution of regulatory and policy issues affecting private investments inthe telecommunications; power; transport and water and sanitation sectors in Brazil, including some common topics such as the new Public-Private Partnership (PPP) law; the environmental regulation; the proposed Lei dus Ag2ncius. A stable and efficiency regulatory environment i s necessary for the attraction of private investments and for adequate provision o f infrastructure services. It reduces incentives for opportunistic behavior from the state and protects consumers from abuse of dominance from incumbent firms.39 Clear and stable policy choices are important as well. The provision of infrastructure services share several o f the problems classically attributed to pure public goods while economies of scale and scope and externalities distort price signals and limit ~ornpetition.~' 5.1 THETELECOMMUNICATIONS SECTOR 5.1, Telecommunications probably has the most complete regulatory environment of any sector in Brazil. One exception and possible pending issue is contract indexation. Telephony tariffs are indexed by the General Price Index - Domestic Availability (IGP-DI), which captures the effects of cambial shocks and international trade that have nothing to do with the provision o f telecommunications services. A new sector index (Telecommunications Services Index, IST) has been developed by the regulator (National Telecommunications Agency, ANATEL) aming at reflecting telecommunications costs more precisely. IST has been designed to be a measure of the main costs of the concessionaries, but it does not aim at reflecting exchange rate and international trade effects. ANATEL sets a schedule to define the main items o f the tariff readjustment policy, which will be implemented with the ratification of new contracts that extends the concessions of fixed telephony. 5.2. One regulatory challenge is to develop a strategy to dismantle cross- subsidization without putting the present level of service coverage at risk. Brazil is a country that suffers from inequality - there i s a large gap between the affluent part o f its population and the great majority of the lower middle and poorer classes. The expansion and universalization of telecom services were achieved in large part through the allowance of some level of cross-subsidization between services. In addition, due to the 39Investments in infrastructure industries have large sunk costs and high degree of asset specificity (Le., their assets cannot be easily transferred to other lines of business. There are also important economies of scale involved and a high political content as infrastructure investments involve a large amount of consumers, stakeholders and voters. Because investments are sunk and politically sensitive, politicians may see a chance to act opportunistically, requiring new targets or imposing extra-costs to regulated fiims after investments are made. Given that assets are not easily transferred to alternative activities, investors imperfectly adjust to this new business environment and the new conditions translate into lower returns to investment or, simply put, into capital expropriations. Faced with the risk o f administrative expropriation by future governments, firms are discouraged from participating in the privatization process in new markets, lower their bids in concession audits or delay technological modernization in existing markets (Henisz, 2002, and Henisz and Zelner, 2001). 40See, for example, Prudhomme (2004). 69 universal service goals, a large part o f the fixed telephony concession companies' plant is not lucrative. The provision of services for lower income consumers i s made possible, therefore, by cross subsidies from services provided to high income (mainly corporate) consumers and from interconnectionfees. 5.3. Operators (concession companies) are concerned with the possible adoption of measures to promote competition with no revision of the universal service goals. Operators believe that competitors (mirrors companies) will adopt "cream-skimming" strategies, "robbing" lucrative customers from their plants. As a result, operators maintain that it will not be possible to keep universal service goals. Many operators suggest reviewing the goals or substituting them with another set of goals. A popular option among operators is one that makes the telephony network the main provider of broadband service for digital inclusion, with financing from the Universalization Fund for Telecommunications Services (FUST). 5.4. In October 2006, a policy for the use of FUST's resources was proposed by the government. FUST contains resources to be used for the universalization of telecommunications services, as stated inthe General Program for Universalization Goals (Progrumu Gerul de Metus deUniversaZizugio,PMGU). However, so far the lack of an investment policy has blocked the use o f its financial resources. In an attempt to address this issue, in October 2006, the Ministry of Communications developed a proposal that seeks to promote the use of FUST's resources in the following years (to begin in 2007). In the recent past, government attempts towards FUST implementation have been stopped by the Brazilian Court o f Audit (Tribunal de Contus du Unilo, TCU) due to conflicts on rule interpretation. To clarify rules' terms, a coordinated effort betweenthese two bodies was launched, allowing for the creation of a draft decree that has already been submitted to the Civil Department (Casu Civil), a presidential body that advises the Brazilian President. 5.2 THEPOWER SECTOR Electricity: Waitingfor thefirst results of the new model 5.5. Starting in 2003, Brazil has tried to restructure its power sector in order to increase supply and avoid a crisis similar to the one observed in 2001.4' Over the last three years, Brazil has used a new conceptual model for the power sector with the aims o f increasing and securing the supply o f energy, and applying tariff threshold.42 The model adopted in 2004 centralizes distribution companies' purchase o f energy through unified auctions, aiming to increase stability of the supply and the bargaining power of the 41The electricity rationing program of 2001-02 was not the cause of the sector's model restructuring. 42In2003, the Brazilian government decided to reformulate the power sector model. InMarch 2004, Laws No. 10,847 and 10,848 were approved with the goals o f securing energy supply, and applying tariff threshold. The new model focuses on the long term planning with the creation of the Energy Research Company (EPE), which works as a pool in a market of regulated and free energy acquisition, considering that distributors must forecast and acquire enough power to supply 100percent of their demands. 70 distribution companies (Discos), which would generate better energy prices for consumers. Although it does not rely on a pure single-buyer model, which would be undesirable, one important characteristic of the new framework is that the progressive role of freely negotiatedcontract terms andprices foreseen inpre-crisis models will have a more limited scope and apply to the non-franchisedmarket only. Self-dealinganddirect negotiation have been banned in the captive market. Mandatory contracting levels for distribution companies were raised to 100 percent (from 80 percent in the pre-crisis model). Explicit penalties for under-contracting- non-existent in the pre-crisis model - were established. Despite the desirability of retail competition, there were delays in the reduction of the 3 M W thresholds, above which customers have the option to choose alternative suppliers. World Bank experts consider the new model to be "converging asymptotically to the post-crisis old model" but new investors appear to be less enthusiastic and are not so sure about the directions of change.43 5.6. Existing generators believe that changes in rules were executed without preserving some of the existing rights of hydro concessions previously granted via competitive bid processes. Under the previous regime, a concession was awardedbased on the highest fee offered for the rights to use the water. Under the current model, concessions are granted in tandem with a purchase power agreement to those who offer the lowest tariff for the assured energy to be potentially delivered to the grid. The transition has been problematic insome respects, as it may have created strandedcosts to some developers. It is difficult to determineprecisely what those strandedcosts might be, since it i s not known if the high premiums paidfor the use of water rights would indeed be recovered via freely negotiated purchase power agreements. However, investors feel that rules have been changed without proper compensation. This perception creates a sustainability problem for the model, as those new investors will make their decisions based on their appraisal of the current conditions, government reputation, and past behavior. If the current rules penalize old undertakings, potential investors might feel discouragedto enter the market.44 5.7. In particular, the differentiation between "old" and "new" energy, introduced by the new model, may affect the rate of return of some plants under construction. The new model introduced a differentiated treatment for "old" (existing or amortized) and "new" generationinvestments.The idea was to hold new energy auctions three and five years in advance of the delivery date, whereas auctions for old undertakings might be held just one year in advance of the forecast delivery date. Contracts for new projects are to be awardedfor periods of 15-35 years, as opposedto 3- 15 years for old undertakings. Existing generators believe that changes in the rules were carried without preserving some of the existing rights of plants under construction, such as the ability to sell energy at freely negotiated prices to the captive market.45If the current rules penalize existing undertakings, potential investors might be discouraged to enter the market. 43 See Maurer (2004). 44 The government has been working on means to tackle sustainability issues, such as the mandatory ''ayment by hydropower plants to pay for using the public good. Electricity generated by plants whose construction was finalized and are currently operating i s not regarded as "old" energy. 71 5.8. Brazil has a poor record regarding contract enforcement in the energy sector.46Contract design and enforcement are relatively new to the power sector. Only with the 1998 reform did contracts become the "glue" that binds together the multiple pieces of a vertically disintegrated system. Despite some progress since then, there have been many contract disputes. For instance, a large debt of the state-owned generators owed to the wholesale energy market (MAE) reducedthe liquidity of the market by late 2000. During the discussions about gains and losses prompted by the 2001 crisis, there were multipleregulations and contractual clauses being challenged, involving large sums of resources. Post-rationing, there were additional unilateral attempts to breach contracts of thermal power plants and energy imports, involving substantial amounts. Investors also feel that the enforcement mechanisms are weak, since courts are slow and not preparedto deal with the nuances of the power industry. A recent suggestionwas madeto create a special tribunal to deal with matters pertaining to the electric sector (Table 5.2.1).47 Table 5.2. 1Disputes pertaining to property rights on energy in2001 DisputeI Contenders Key Issues 1. Distributioncomuanies Concession Contracts - Acknowledgement of legitimate regulatory lag for - "Parcela A" pass-through 2. ItaipuExcess Energy - Eletrobrtis vs. Property rights on (parts) of existing capacity and Distributioncompanies generation 3. MAEAgreement - Generators vs. Distribution Discussions onbasic pillars -e.g., "recompra" Companies (new), capacity fee 4. Imports from Argentina - CIEN vs. Furnas + Should new MAErules modify agreed contractual Gerasul obligations? 5. IPP- Cuiabi Ivs. Furnas Ditto 6. Annex V - Generators vs. Distributioncompanies Should a particular contractual clause be applied during rationing? 7. Angra I1- Furnas - Generator vs. Distribution Efficacy of capacity contracts regulatedby Decree companies: taken to COEX for [frustrated] No. 2,65511998 arbitration Source: Maurer (2001). 5.9. The success of the new energy procurement model in the Brazilian industry is alsojeopardized by the fact that the government still controls 75 percent of power generation. (This includes hydro generation and ownership of "old energy"). Auctions 46Prior to the late 1990s reforms, there was not a tradition to establish purchase power agreements between buyers and sellers, since they were all state owned companies. Furthermore, any issues on the allocation o f property rights were resolved by an administrative forum based on collaborative rules and coordinated by Eletrobrds. These functions have been absorbed by the Operador Nacional do Sistema and Mercado Atacadista de Energia. However, those new entities have no role inaddressing contract disputes, except the ones emerging from the interpretationof GridCode and Market Rules. 47See, for example, Brown and de Paula (2004). 72 will be based on Discos' projections of future demand to serve their franchised customers. This captive market i s - and will continue to be for many years - the largest portion of the total electricity demand. Therefore, getting the rules right is of key importance to support capacity expansion. However, despite the effort inputtingtogether an efficient procurement model, there are still concerns that it i s not yet a level playing field able to attract new players. To start, there i s a structural issue, since state owned generation still controls a large part of the existing market and those are potentially active players to the expansion of the system. Second, concerns exist about the transparency of the auction rules, and the real independence of the Government given its double role as regulator and entrepreneur. And third, the bidding motivation of state-owned companies i s more in terms o f recovering cash costs, and not so much in terms of assuring a return higher than the cost of capital, on a full costing basis. 5.10. Another important regulatory flaw is that there are no incentives to incorporatedemandresponse inthe decision-makingof the industry.First, the model overlooked the possibility of including, ina very explicit way, demand-side response as a potential bidder to the energy auctions. Inmany countries, demand reductions and energy savings compete with capacity expansion on a level playing field and inmany cases have proved to be more economic alternatives to meet demand requirements. Second, by blending existing (cheap) and new (more expensive) energies to be sold to the distribution companies, the model "muted" the price signal and the real marginal cost faced by Discos when serving their markets. Now prices are conveyed as a weighted average between old and new energy prices as revealed in auctions, with no explicit linkage to the revenue or costs. 5.11. The first "new energy" auction took place in December 16, 2005 and achievedmixedresults.The objective was to contract electricity to supply the forecasted demand increase in2008, 2009, and 2010 for the franchised market. It represented a first test of the adequacy o f the new power sector model to induce efficient generation capacity expansion. Table 5.2.2 summarizes the auction results. The auction did not succeed in meeting the entire forecast load for 2008 and 2009. In addition, most private investors refrained from participating in the auction, where state enterprises played a big role. The results were not least-cost because the auction resulted in the purchase of thermal electricity at about RS$ 130 per MWh (plus fuel costs), while preventing the purchase of hydro electricity at prices above RS$ 116 per MWh- the ceiling imposed by the auctioneers. On a related issue, the studies on the least-cost options for the expansion of the system were not publicized. Therefore, it was not possible for the public to make sure that the plants corresponded to the optimal expansion for the system, which reduced the transparency o f the process. Overall, results suggest that the auction structure and rules should be reviewed so that the results approach a least-cost expansion and that all forecasted demand i s met. 5.12. The new model did too little to address market distortions, and missed the opportunityto introduce a more sophisticated nodal pricing system in conjunction with the auction of financial transmissionrights.48The new model practically muted 48 See Chandley andMorozowski-Filho (2004). 73 location signals. A limited ability to convey correct prices at the wholesale level intime and space was already a feature of the previous model, which opted for a zonal pricing based on four geographic zones (sub-markets) for the entire country. It is widely known that in a country the size of Brazil, four electric zones are not enough to capture the diversity and variability of costs of generation in the basic grid, in time and space. By conveying incorrect price signals, the market gets distorted responses interms of location of new plants, leading to an increase in congestion costs. Location is not usually a problem under a pure deterministic planning process, but i s an issue when planning becomes indicative and investors have some discretion on when and where to locate their plants. It is widely accepted that players in the power sector tend to advocate for fewer rather than more electric zones (and as a corollary reject location pri~ing).~'It is incumbent upon the government to stand up and define energy policies that make sense for the country in the long run. While it is always possible to give the correct price signals for new players, it becomes increasingly costly to compensateexisting generators for the strandedcosts of having made poor location decisions inthe past, when operating under differentrules. The power sector inBrazil is already suffering the consequences of poor congestion management. Transmission costs (use of the transmission system fees) have skyrocketed in the last few years without a corresponding reduction in the costs of congestion (system service charges). There seems to be room for improvements in this area through a complete revision of the existing pricing methodologies. Table 5.2.2 First new energy auction results, 2008-2010 2008 2009 2010 Total Power purchases 632 901 1,753 3,286 (average, MW) Demandforecast 1,090 1,900 1,670 4,660 (average, MW) Demand unmet by auction 458 999 -83 1,374 (average, MW) Total unmet demand 458 1,457 1,374 1,374 (average MW) Share of demand forecast 58 47 105 71 contracted at auction (%) Typeof Energy 2008 2009 2010 -- Hydro (inRS$ per MWh) 106.95 113.89 114.83 -- Source: CCEE and PSR. 5.13. The lack of demand response in the current model is inconsistent with the energy rationing program successfully adopted in 2001-02. Brazil showed that demand response can be decisive and can lead to less distorted outcome^.^^,^^ The 49To a large extent, all players in the power sector are potential energy marketers. In that capacity, they tend to prefer fewer electric zones to limit their risk exposure to difference in prices among sub-markets. Few sub markets socialize congestion costs (ESS) which are recovered through the entire customer base. 50A very simple and ingenious price scheme was put inplace whereby customers would be charged short runmarginal cost (or proxies thereof) for consumptionover and above their assigned quotas. The response to price was fast and effective. Higher prices induced conservation and gains inenergy efficiency which are 74 program was essentially based on quotas and financial incentives and penalties. The World Bank considers the Brazilian experience in handling power shortages and entertaining demand response an international best practice. No black-outs or brown-outs were necessary during the eight-month rationing period, with a 20 percent load reduction almost on a country-wide basis. However, the current reform did not incorporate any mechanism of demand response into the market model. It appears that the lesson learned with rationing-that price signals do work -has not beenput into practice. 5.14. Reforms preserved institutions created for the "market model," but centralizedthe planningof the industry.There is a trade-off between better planning and coordination and political and regulatory risk. Interms of institutional change, all of the previously created institutions, such as the regulator (ANEEL), the system operator (ONS), the market administrator (former MAE now somewhat reformed as the CCE) are preserved. However, the Ministry of Mines and Energy (MME)has a more central role in the industry, especially in planning and in terms of setting the guidelines for auctions. One issue that has beenraised with respect to the current institutional framework is how the functions of the MME will relate to the regulatory and monitoring functions of ANEEL. An innovation of President Lula's model was the establishment of a new company, Empresa de Pesquisa Energe`tica (EPE), with the mission of developing long- term planning for the energy sector in Brazil. Although the new model does not promote a paradigm change in terms of regulatory regimes and institutions, it does give the government a stronger role to play inareas that the reformers view as representing market failures and having contributed to the supply crisis, such as the lack of capacity expansion due to the absence of long-term contracting. The issue to consider here is the trade-off between addressing the so-called market failures via stronger planning and centralization, and increasing the uncertainty in the market, as a result of excessive politicization and centralization, which could scare private investors away. Natural Gas: lack of competition and policy definitions affect itsfill development 5.15. The development of the natural gas industry in Brazil has been historically hamperedby a combinationof supply and demand challenges.Until very recently, natural gas represented only two percent of the energy portfolio in Brazil. There were no large domestic supply sources and imports were nonexistent. With the construction of the Bolivia-Brazil gas pipeline and the end of Petrobras' monopoly in E&P, the situation started to change. Since then, Brazil has been able to tap into the large natural gas reserves in Bolivia, and important new offshore gas fields have been discovered in the Southeast region. The 2001 rationing episode underscored the importance of a solid thermal generation base to hedge the power system against the volatilities of hydro generation, which would entail a significant increase in the use of gas. Unfortunately, natural gas has not been available when needed, particularly inthe Northeast region, due still noticeabletoday. The anticipated GDP impact of the crisis was mostly avoided and the entire market, including residential consumers, was able to cut consumption significantly. 51It is important to keep in mind that electricity waste levels of households were higher in 2001-02 than they are now. 75 to both production and transportation bottleneck^.^^ The growth inthe industrial market, intandem with the thermal generationrequirements, has the potential to spur demandto unprecedentedlevels. 5.16. The challenge of optimizing hydro and natural gas resources for thermal generation has not been resolved yet. Power generation represents the greatest individual potential market for natural gas. However, natural gas plants operate as "swing" producers, supplementing the needs of the hydro system. Hydro production is volatile, as it depends on the vagaries of rainfall. As a consequence, demand for natural gas for power generation is equally volatile. A typical hydrology pattern consists of several consecutive years of good rains, followed by a few very dry periods of relatively short duration. As a swing producer, natural gas plants are idle most of the time and once ina while the entire supply, transportationandpower generationfacilities are strainedto a maximum to deliver the necessary power. Regardless of the capacity utilization, the extremely high fixed costs of developing the natural gas infrastructure (from exploration andproduction, E&P, to distribution networks) needto be amortized. That is the primary reason for high ship-or-pay and take-or-pay clauses in the gas transportation and supply agreements. According to existing contractualarrangements, the power sector has to bear most of those fixed costs. As a consequence, reliability of the power sector has been achieved, albeit at a very highprice. 5.17. The creation of a secondary market is important to consolidate the natural gas industry. One obvious solution to the intrinsic natural gas demand volatility for power generation is to encounter alternative uses for the gas when the infrastructure is idle (e.g. favorable hydrology). This would help defray fixed costs and enable the natural gas to achieve a higher level of reliability at lower costs. Finding viable alternative uses for the gas encompasses three dimensions. The first dimension is technical; that is, the need to have industrial, commercial or other customers with switching capability. In times when the gas i s not required by the power sector, those customers should benefit from an abundant, supposedly cheap resource.When power generationrequires more gas, those customers should be able to switch to other sources of energy - for example, oil, diesel, or others. The second dimension i s commercial and regulatory: power plants, when not using gas, shouldbe able to swap or trade the unused gas with other customers, or to have it bought back by distribution companies. So far, regulatory arrangements between Petrobrcis, distribution companies, and the IPPs have not been built with the necessary flexibility, an issue that needs to be addressed as a matter of energy policy.53 The third dimension is the size of the secondary market per se. Given the sheer gas volumes usually required by power generation, it is unlikely that all the unused gas can be diverted to a secondary market, which is still small in relative terms. This difficulty will be gradually addressed as the industrial, commercial and vehicular markets grow in size and importance. 52Three natural gas power plants were instructed by the system operator to generate and reduce the pace o f depletion of the reservoirs in the S5o Francisco River. However, when instructed to dispatch, there was not enough gas to feed all the plants simultaneously. This episode exposed a geographically broader problem in terms of gas availability: when asked to dispatch simultaneously, plants in the Southeast did not have the necessary gas either, despite the existence of firm gas supply agreements with Gaspetro. 53Some flexible arrangements, such as interruptible contracts, should also be considered. 76 5.18. Alternativesupply of naturalgas hasbeen obstructedby the lack of access to the Bolivia-Brazilgas-pipeline.Natural gas transportation inBrazil is also controlledby Petrobra's, through its subsidiary Transpetro (with the exeption of Gasbol, which is controlledby Petrobra's's subsidiary Gaspetro).Open-accessis establishedby articles 58 of Law 9,478/97 (Lei do Petrdleo) and further regulatedby Portaria No. 027/2005 from the sector regulator (Agtncia Nacional do Petrdleo, A M ) . Enforcement of these rules, however, has been extremely cumbersome and costly, with potential entrants issuing successive legal complaints or simply voicing them publicly. ANP was first called in to arbitrate a conflict over the access to the Gasbol pipeline in2000, when Enersil Ame'rcia do Sul formally submitted a complaint. The challenge for the regulator is how to deal with asymmetric information, as it is very difficult to reach a conclusion regarding whether minimum conditions under which open-access should be granted - namely available capacity - existed or not. This has increased questioning as to whether the simple creation of a full subsidiary was sufficient to change the incentives for efficient management of transport infrastructure by Petrobra's. As a result, a draft law is under discussioninCongress(see Box 2). 5.19. More generally,the marketpowerofPetrobrh is a centralobstacle to the full development of the natural gas industry in Brazil. Despite the end of the legal monopoly for Petrobra's since 1997, the company is still a defacto monopoly in several stages of the gas chain. It carries out most of the E&P activities, including in Bolivia, is the operator and owner of most pipelines, andhas stakes inseveral thermal plants and gas distribution companies. Gaspetro, Petrobra's' subsidiary, controls at least 25 percent of the capital of 20 out of 23 distribution companies in the country. The pipeline system does not yet operate as open access, as discussed above. In the absence of export possibilities,private producers still dependon Petrobra's to have access to the natural gas markets. Finally, the company has moved downstream and is now a major player in the production of electricity with natural gas. While many of these investmentshave helped spur the natural gas industry, they may also hinder the emergence of new investors and the establishmentof a level playing field scenario - given the lack of competition in the current model. 5.20. Overall, the sector still lacks a long term strategy based on a policy decision regarding the composition of energy generation in Brazil. This involves developing a source of energy that does not depend on climate conditions, as is the case of hydro-based power generation. It is also relatedto the environmental (and social) costs of a strategy to expand the system, given the areas where hydro plants would be installed. As a result, Congress has started to debate a new legal framework for the natural gas sector, which included PLS No. 226/2005 (see Box 2), PL No. 6,673/2006 proposed by the government, andPLNo. 6,666/2006. Box 2. The New Legal Framework for the Natural Gas Industry (PLS No. 226/200S) This Draft Bill is organized into 13 chapters and addresses aspects of the natural gas sector that were not considered by Law No. 9,478/1997. Chapter Iestablishes that the National Council for Energy Policy (CNPE) as the institution responsible for providing the President with programs specific for the use of natural gas and the directives for its imports and exports. Chapter 11, considering public monopoly, adds three activities to Law No. 9,478/1997, namely imports, exports, and transportation through pipelines. 77 Economic activities related to the natural gas sector will be regulated and overseen by the federal government - through ANP - and by states in cases of local distribution. Chapter IV creates a National Operational System for Natural Gas Transportation (ONGAS) and defines its attributes, which include: (i) setting operational procedures; (ii)overseeing and coordinating natural gas transport and storage; (iii) planning the usage of transport and storage systems; (iv) interacting with the Company for Energy Research (EPE) to formulate system expansion plans; and, (v) developing and publicizing performance indicators for the natural gas transport and storage systems. Additional attributes are given to the Ministry o f Mines and Energy (MME) include: (i) implementing a national policy for the natural gas sector; (ii)regulating and overseeing activities in the natural gas industry; (iii) promoting public contests related to supply and capacity allocation of new gas pipelines; (iv) establishing criteria for tariff administration for transport and storage; (v) authorizing imports, exports, processing, liquefaction, re-gasification, compression, and decompression of natural gas; (vi) developing public announcements and carrying public biddings for natural gas transport and storage concession, implementing contracts and overseeing their executions; (vii) authorizing the construction and operation o f gas pipelines for transferring and production; (viii) developing and publicizing annual sector reports; and, (ix) working with ANEEL and state regulatory agencies to make uniform norms applicable to the natural gas and energy sectors. With respect to natural gas transportation, it establishes that this activity will be done through concession contracts (after public biddings). The concession of gas pipelines follows M M E ' s determinations, based on technical or sector reports developed by EPE or any other party. Public biddings will follow transparency and equity principles, regulated by MME,which will consider criteria for project dimensions; pickup and delivery points; average capital costs; the maximum transportation tariff and the methodology used for its calculation; and conditions for changes in project components. The project makes it clear that concessions for natural gas transportation will be given to companies that have either developed this type of activity ardor operated in the storage segment but not to firms exclusively operating in distribution. Firms transporting natural gas could store the product but not buy or sell it. The project also details the principles behind natural gas transportation tariffs to be fixed by MME,which are: (i)principle of financial and economic equilibria of contracts; (ii) non-discriminatory treatment among carries; (iii) consideration of the type of service provided and its efficiency; (iv) guaranteed profitability to carriers, based on the risks associated with the activity; (v) guaranteed safety of the activity; (vi) provide incentives to the carrier to reduce costs and increase supply of transportation ability; and, (vii) consider tax incidence over natural gas activities. No subsidy can be considered when calculating payments to be made by public, private or public-private companies. Another important feature introduced to the natural gas market is that any carrier interested in natural gas transportation is allowed to develop this activity as long as tariff payments are made. Capacity offers will be publicly announced by the carrier in the presence of available transportation capacity. Inaddition, any carrier can request the expansion o f the pipeline to MME as long as it i s either part of the contract or necessary (justification must be submitted). With respect to interconnection, the carrier can allow connections to be made to be concessioned with other pipelines - but connections and tariffs associated with these developments have to be previously approved by MME. Considering the distribution of pipelined gas, each state will explore local services either directly or through concessions. Concessions can be granted only to firms exclusively performing this type of activity. In addition, states can provide distributors (in the respective concession areas) with exclusive rights on natural gas distribution and commercialization for a given period of time (with an expiration day or ending date). Finally, considering natural gas commercialization, when the expiration date of exclusive commercialization rights is reached, residential and non-residential users will be able to acquire the product from any seller that uses distributors' pipeline networks to transport the product to its facilities. Any fiirm (complying with MMEi's requisites) canbe authorizedby states to commercialize natural gas. Transporting, storing or distributing firms cannot directly commercialize natural gas, with the exception of distributors in their respective concession areas (as long as it is part of the contract). 78 5.3. THETRANSPORT SECTOR Ports: Reforms Stalled in Recent Years 5.21. The reform of the port sector was initiated with the Port Modernization Law of 1993, which defined a new institutional framework emphasizing privatization of operations, intra and inter-port competition, and decentralization. Significant progress was made in the late 1990s with the concessioning of the major terminals to private operators, the opening of many private (own account) terminals for public use, andthe establishment of key sector institutions: the local Port Authorities (APs) and Port Authority Councils (CAPs) andthe Labor ManagementOrganizations (OGMOs), to help eliminate excessive staff. Terminal operators in some ports have managed to negotiate agreements with labor unions to adjust the work rules and the gang sizes of for-hire stevedores, leadingup to a 40 percent increase of labor productivity inthose ports. These changes resulted in substantially reducedport costs: for example, in Santos, which was the costliest port inLatin America, container handling costs have droppedfrom well over US$ 300 in 1997 to US$ 180 in 2000. In 2001, the national water transportation regulatory agency (ANTAQ) was created. Overall, privatizations generated US$ 421 million in 1997-1999. 5.22. The reform of the ports sector has stalled in recent years because of the absence of clear policies and guidelines needed for its full completion. Although the two main legal pieces for the sector-the Port Modernization Law and the Transport Sector Restructuring Law-provide significant guidance for the new organization of the industry, doubts remain regarding several important issues. In particular, the responsibilities of the Council of Port Authorities (CAPs), Port Authorities (APs), dock companies, still controlled by the Federal Government, and the regulatory agency (ANTAQ) are ill-defined.54 Inter-port competition issues were neglected during the process because each public dock company leasedterminals independently, on somewhat differing terms and without concern for potential. In addition, most of the dock companies have debt obligations that are well beyondtheir financial capacities,inhibiting their capacity for further investments in physical port infrastructure. Port labor is still responsible for a large portion of the high costs of cargo handling in Brazil's ports and has not yet beenfully adjustedto account for the mechanizationof the ports: of the total workforce of about 35,000 stevedores, at least 20,000 may be redundant. In 2003, the Government established an inter-ministerial working group to analyze the problems in the sector and to propose an action plan. The working group preparedrecommendations for addressing some of the main issues, as well as an investment plan which was approvedandissuedas the Agenda Portos in2004. 5.23. The Agenda Portos identifies the critical bottlenecks in the ports system, and defines key investment priorities in 11 major ports. The private sector has been investingheavily interminal facilities and equipment inthe face of the rapidly expanding external anddomestic coastaltrades. Butprogress has beenslow inundertakingthe major 54 Ministe'rio dosTransportes has delegated five ports to states or municipalities, but the roles and responsibilities of each parties have not been clearly defined either. 79 public investments,which involve capital and maintenancedredging works inport access channels, due to slow environmental clearance processes and to the limited capacity of the heavily-regulateddredging industry. 5.24. There is no clear consensus yet on proceeding with the decentralization and/or privatization of the port administrations. Only two ports (Ituqui and Santunu) have recently been delegated to the corresponding state or municipality (respectively MaranhBoand MacapQ/ AmapQ),bringing the total of delegatedports to twelve. Twenty ports, including Santos andRio de Janeiro, remainunder federal control throughthe eight dock companies. Although the dock companies have been included in the privatization programsince 1996, the Governmenthas not initiated, nor is it preparingto initiate, such a process. Heavily indebted and faced with numerous judicial actions linked to labor disputes and with the restrictions resulting from their inclusion in the privatization program, the dock companiescontinueto beunableto effectively carry out their new Port Authority responsibilities. In addition, dock companies are in need of new staff with managerial skills to reinforce their capacity to operate as port authorities. A first hiring program is being implemented, which would benefit the 4 main public ports. Within such a policy framework, which excludes broad delegation and major privatizations, the Government may consider defining an effective strategy to improve dock companies' performance. The Government may also privatize a few selected secondary ports with adequate operationalconditions (such as SGo SebastiBo and SBo Francisco do Sul) which could effectively compete with major public ports and contribute to improving the efficiency of the ports systemas a whole. 5.25. ANTAQ made some progress in regulating port services by issuing a norm that specifies procedures and provides guidelines to port administrations for the leasing of port facilities and areas to private operators and for regulating and supervising such leases. ANTAQ i s also proceeding with the establishment of a ports statistics system, now operational in 19 ports, which will allow for monitoring and evaluating the performance of individual ports and of the entire system. In order to coordinate the implementationof the International Ship and Port Facility (ISPS) Code, the government established an inter-ministerial group in early 2003, The National Commission for Public Safety in Ports (CONPORTOS) issued a number of resolutions requiring each port to carry out a safety risk analysis and to prepare a safety plan for controlling port perimeters and accesses for employees, vehicles and ships. Special budgetshavebeenallocatedto ports (RS$ 60 million) and the police (RS$ 40 million) for this purpose. Progress was made in implementing these measures in the major ports, particularly Santos, which is expectedto be certified shortly by CONPORTOS. 5.26. Labor issues -including the size of gangs; the excessive, registered, for-hire work force and workers' compensations, which substantially impact port costs - have not yet been fully resolved. A strategy to deal with the remainingredundancy in labor is important improve port productivity and further reduce port costs: for example, about 65-70 stevedores are paid on a full-container ship at Santos, compared with 15-20 at Buenos Aires and 7-8 at Antwerp, which explains in part the fact that the cost of Santos are still higher than these ports. The pool of for-hire stevedores is now estimated at about 27,000 nationwide, down from about 35,000 in 2002. This reduction, together 80 with the substantial increase in port traffic, has alleviated the problem countrywide, and somewhat resolved it in a few ports such as Vitoria, Rio Grande, and Salvador, where operators and labor unions have agreed to adjust labor hiring to actual requirements. But for-hire labor costs continue to be a problem in the major ports, particularly Rio de Janeiro and Santos. A key next step therefore remains the negotiation of some form of collective agreement between terminal operators and labor unions, facilitated by the Government, possibly within the forum of the Port Sectoral Chamber (Camara Setorial Portudria) or a new national commission for port development proposed by the Labor Ministry. 5.27. Regulatory and competitionissues are also pendingtopics for private sector participation. Cross-ownership issues between terminals in the Santos and Rio de Janeiro Ports affect incentives for inter-port competition. Intra-port competition has been limited by the reduced number of private operators, verticalization and anti-competitive behavior by dominant firms. In 2005, the Brazilian antitrust tribunal (CADE) ruled unlawful the charging of "Terminal Handling Charge-2," a cost imposed on private customs operators by the private terminal in a classic strategy to raise rivals' costs in a downstream market, namely warehouse and dispatch services. The case illustrates a situation in which anti-competitive behavior was possible due to the lack of appropriate regulation: because no commercial relationship existed between the private terminal and the private custom operator - except for the competition in the warehouse services - no cost should be attributed to the private terminal. Considering that losses imposed on private customs operators jeopardized the financial sustainability o f the business, the case may also be seen as an illustration of the risks that abuse of dominance imposes for private investments ininfrastructure. Highways: Waitingfor the second batch of concessions and a long-term plan for public expendituresfor road maintenance 5.28. A broad policy and institutionalreformwas initiatedduringthe 1990s in the road transportation sector. Its main objectives were to address the elimination of the road fund (the main source of maintenance and rehabilitation funds); an inefficient maintenance policy (contracting out the engineering designs, rehabilitation and other works, and maintenance and supervision services through traditional, input-based contracts) and the institutional deterioration of the national highway agency (DNER). The objectives of this reform included the transfer of highway sections with sufficient traffic to private concessionaires, and the recovery of all or part of the costs of operation, maintenance, and upgrading directly from road users through tolls (about 15,000 km). Other two important components were: (a) the reclassification of the highway network with a view to maintaining under federal jurisdiction only the main interregional and interstate highways of national interest (about 45,000 km); the transferring to state jurisdictions o f highways of mainly local interest (about 20,000 km); and (b) contracting out rehabilitation works and maintenance services combined on entire routes of the remaining network through long-term CREMA, with the contractors responsible for achieving specified levels of service (about 30,000 km). 81 5.29. Some progress has been achieved, but reforms have slowed down during the last few years. DNER was eliminated and a new national department o f transport infrastructure (DNIT) i s responsible for executing public sector programs. DNIT inherited the organization and culture o f DNER and i s not yet effectively performing the planning, management, supervision and monitoring and evaluation functions for which it was created. Agreements for road transfers have been signed with 14 states, aiming at transferring up to 14,000 kmof federal roads by 2006. However, the agreements have not always been implemented. The proposed law, which would reclassify the highway networks and provide the legal basis for fully implementing the decentralization agreements - the so-called National Transport Infrastructure System (SNV) law - has stalled in Congress on issues that are unrelated to the reclassification. The federal government reopened negotiations with the States to proceed with the decentralization program but the key controversies are over the amount of federal funds that need to be transferred for road maintenance and the conditions o f roads to be transferred. With the SNV law approved, the initial target of transferring about 20,000 km to the States could be achieved inthe medium term. A total of 5,000 km of Federal highways has been under concession for several years. Optimistic traffic forecasts and related investment obligations and generous contract renegotiation rules have led to many contract modifications that may have transferred commercial risks to road users and have resulted in substantial tariff increases. As a result, the second phase of the federal program - for about 2,500 km - has been stalled for several years due to intense questioning from Government auditors. The recently approved Law of National Transportation System was a step forward in this transferring process.55 It stated that, in order to support this transferring process, DNIT i s authorized to provide federal financial resources for construction work to maintain, fix, build, and signalize transferred roads, as well as supervise and develop essential studies and engineering works. 5.30. The federal government has been gradually strengthening its maintenance management capacity. Output-based contracts now cover roughly 23 percent o f the remaining federal road network, and a concession bidding process will be launched on another 5 percent of the network in 2006. A national road policy (PNR) addressing these issues and related decentralization i s being prepared. A multi-year highway expenditure program, also under preparation, will optimize public expenditures on highways, and leverage private financing and management on the basis of experience gained with highway concessions and with performance-based contracting. 5.31. Despite recent improvements linked to the introduction of output-based contracts, the management and execution of contracts are still inefficient. The complex organization o f DNIT and the heavy administrative requirements for contract management result in significant delays in contract design and work contracts procurement. For input-based contracts, it i s common for the complete contracting process (from bidding of design contract to the start order of the works contract) to take over 2 years, by which time it is likely that a number of design features will need to be renegotiated due to changing road and traffic conditions or substantial changes in input prices (e.g., bitumen prices doubled over the last 2 years). Irregularity o f contract 55Law No. 11,314/2006. 82 payment levels and frequency are high. Within a given year, road contractors have a window of not more than 6 months to executetheir contracts, taking into account that: (i) contractors are generally not paid from December to April as a result of administrative and budgetary shortcomings; and, (ii) generally affect major road works during rains several months, depending on the region. Inaddition, in recent years, monthly payment limits have been imposed to enable the federal government to meet its quarterly fiscal targets. Faced with slow and uncertain payments, contractors had to sharply slow down contract execution. This strategy resulted in: (i) substantial contract extensions (one- to two-year extensions are quite common) and price adjustments (at a rate of about 12 percent per annum); (ii) demobilization and re-mobilization of contractors' teams; and (iii)negotiations over the payment of arrears. Railways: Readyfor a New Inflow of Private Capital? 5.32. Prior to privatization, the railway system in Brazil had an operational structure that was not economically or politically viable. Regulatory policymaking and financial functions were performed by a state-owned operator.56Inthe early 199Os, four companies, operating 29,000km of railroads, accounted for 95 percent of the total railroad shipment in Brazil. The largest one, Rede Ferrovidria Federal S.A. (RFFSA), was the holding company for six regionally dispersed railroad operators. Ferrovias Paulistas S.A. (FEPASA) belonged to the state of Si70 Paulo. A third operator, Companhia Vale de Rio Doce (CVRD) - by then a state-owned industrial holding - managed two specialized rail lines, Estrada de Ferro Vitdria a Minus (EFVM) and Estrada de Ferro Carajds (EFC), linking their mining sites to ports in the north and southeast of the country. Unique to Brazil, the operational structure of the Brazilian railway network was fragmented due to the employment of different track gauge sizes among operators and sometimes within a single operator. This characteristic hinderedthe development of long haul services and restricted railway services to short or medium distances, where it faced competition from low-cost road transport. Since the mid-l970s, when the National Railway Department (DNEF), the General Transportation Comptroller, and the Transportation and Tariff Council were extinguished, RFFSA was also responsible for the administration of network operation (i.e., mutual traffic), tariff setting, and the financial managementof the system. 5.33. Sector reform was implemented in two phases, with privatization taking place first, leaving several regulatory loopholes. RFFSA's network was separated into six vertically integrated monopolies. Rail services were awarded, but rolling stocks and other existing infrastructure were kept under property of the federal government.57Four of these railroads connected ports along the coast with their respective hinterlands. FEPASA was included in the privatization program in 1998. The two other railways, EFVM and EFC, owned by CVRD, were transferred to the private sector as part of the 56See Castro and Larry (1994). 57The infrastructure would be simultaneously leased by RFFSA to the private operator. The reason for this double concession-leasing method was that, according to the 1988 Constitution, the federal government had to remain the titular to the right of providing rail transport services in the country and, in addition, retain under its ownership the assets involved inthose services. 83 company's privatization in 1997.58 Concession contracts spelled out targets on output and safety for the first five years and defined maximum prices to be charged to consumers based on the length of the haul, type of product, and the geographic location. The concession contracts established general guidelines for tariff setting, mostly in line with the financial-economic equilibrium of the concession. When the sector regulator was created in 2001, a number of regulatory loopholes existed.59These included: the lack of methodology for tariff calculations; the absence o f provisions for penalties for non- compliance with contract targets; poor protection of captive users and unclear conditions for mutual traffic, access prices, and interconnection. 5.34. The structure of ownership that resulted from the privatization created conflicts of interest between shareholders and their companies. Because no restrictions were imposed on participation in different concession or on the participation of major rail users or suppliers as shareholders (except for a cap on 20 percent of total stock), many concession shareholders had direct or indirect interests in their own companies as final or intermediate users. Inpractice, the 20 percent limit applied only to nominal capital: when voting rightswere accounted for, several shareholders owned more than that value. This vertical integration structure favored abuse of dominance (through price discrimination, blocking access and service rationing) and inhibited connectivity. In this sense, the ownership structure only accentuated the economic problems caused by the physical fragmentation of the network, jeopardizing the economic feasibility o f concessions and the availability of new funds for investments. The case of CVRD was particularly worrisome: it exploited EFVM and EFC railroads, controlled access to several ports where it operated major terminals and participated in almost all bidding for rail concessions. While an integrated logistics structure i s essential for competition inthe international iron ore market, the importance o f proper access to the main Brazilian ports must also be taken into consideration. 5.35. In the last four years, ANTT addressed most of the regulatory bottlenecksof the system, including the ownership structure and operational fragmentation issues. Clearer rules defining captive user and establishing the appropriate price to be charged were introduced while conditions for investments by users were defined (ANTT- -Resolutionpassage" and"mutual traffic" -were addressedbyResolutionNo.433/2004 No. 350). After successive problems with bilateral negotiations, access issues "right of and complemented by Resolution No. 895/2005. They establish the principle of free negotiations between private operators and ruled on conditions for arbitrage by the regulator when a private agreement i s not reached. The principle was recently applied for the case involving MRS and FERROBAN(see Box 3). The ownership structure issues were consolidated and the operational restructuring was concluded through various resolutions. Currently, railway operators in Brazil are Ferrovia Novoeste S.A. (NOVOESTE), Ferrovia Centro- Atldntica S.A. (FCA),MRS Logistica S.A. (MRS), Ferrovia Tereza Cristina S.A. (FTC),America Latina Logistica S.A. (ALL), Companhia Ferroviaria do Nordeste S.A. (CFN), Ferrovias Bandeirantes S.A. (FERROBAN), Estrada de Ferro Parana` Oeste S.A. (FERROESTE), Ferrovias Norte do Brasil S.A. (FERRONORTE),Ferrovias Norte Sul S.A. (VALEC), Estrada de Ferro Vitdria a Minus (EFVM), and Estrada de Ferro Carajds (EFC). ''TheAgincia Nacional de TransportesTerrestres-ANTT was created by Law No. 10,233/2001. 84 5.36. Railway sector reorganization started in 2003. This process started with the ownership restructuring of three private operators, since all operators had their investment abilities exhaustedand, with the exception of America Latina Logistica S.A., presented negative net assets. The ownership restructure of Ferrovia Centro-Atlintica S.A. targeted increasing the firm's capital by RS$700 million and investing RS$400 million between2003 and 2007 in the company to increase its production targets by 40 percent. To accomplish its goal, the restructuring of FCA involved the ownership alienation of CSN, which was distributed among CVRD, Mineraczo TacumZ, Taquare ParticipacdesS.A, and ItabiraRio Doce Company Limited. The second company to have its ownership restructured was Companhia Ferrovihria do Nordeste S.A. (CFN). This process targeted eliminating shareholders' disagreements and was done with ownership alienationof CVRD (32.4 percent of CFN's capital), which was transferred to CSN and Taquari Participacdes S.A. It was concluded with the finance contracting of RS$200 million andthe presentationof investmentprograms for railway network rehabilitation.60 Regarding the third operator, Amdrica Latina Logistica do Brasil (ALL), a new foreign shareholder joined the firm and ALL opened its capital in Ssio Paulo Stock Exchange (BOVESPA), an operation that brought RS$5 88.2 million to the company. Finally, it is worth mentioned that, in 2005, shareholders of MRS Logistica S.A. had to sign an additive term to their legal agreement to assure the efficacy of decisions to be made by ANTT and CADE.61Background for this inclusion was based on the fact that ownership changes in mining and ironworks indirectly altered the ownership structure of economic agents inthe working capital of MRS and that Editul No. A-O5/1996/RFFSA prohibited a single shareholder of having more than 20 percent of the concessionaire's working capital.62 5.37. Financial and operational results for the Brazilian railway system have greatly improved after firms' restructuring processes, which started in 2003. For example, between2003 and 2004, total shipment increased9.6 percent (from 345 to 378 tons), transport transportation raised by 12.6 percent (from 183 to 206 TKU), and investments increasedby 66.2 percent inreal terms (from RS$ 1.18 to RS$ 1.95billion). Economic-financialresults for each operator are presentedin Table 5.2.3. Between2001 and 2004, all railway operators improved their gross revenues and, except for FERROBAN and CFN, their operational profits - remarkably, ALL, EFVM, EFC, and MRS. In addition, major loopholes were also resolved. However, two issues are still pending. First, a better integrationof different transportationmodes has to be developed. By doing this, transportation costs would decrease, favoring both domestic and international trade. A necessary condition to develop this is the second issue: interconnection tariffs, for which an improved mechanism needs to be developed. A mathematical formula has already been used by ANTT to arbitrate the conflict on rights of passage between MRS and FERROBAN, regarding the access to the Port of Santos. This process established a starting point on the technical conditions to regulate this type 6oAfter the conclusion of ownershiprestructuring process, shareholders with participation inthe company's working capital with voting rights were the most of 20 percent and 40 percent for FCA and CFW, respectively. `*ANTTmain 61 ResolutionNo. 856/2005. The shareholders of MRS are CVRD, CSN, and MRB, all with business either inthe mining or the ironworks sector (or both). 85 of situation. Once major structural and regulatory bottlenecks are resolved, the sector seems ready for a new inflow of private investments.For that, it is recommendedthat the government prepare a new group of "concessionable" projects according to the logistics needs of the country. Box 3. Regulating Essential Facilities inPractice: The Case ofPorto de Suntos (Resolution ANTT No. 945/2005) Inorder to reach Port of Santos, the main terminal inthe country, Ferroban, one of the largest carriers of agricultural grains from the Brazilian cerrudo,neededaccess to M R S tracks between Campinas and Santos. In2004, BrasilFerrovias (the controller of FERROBAN)submitted a complaint about the costs and delays imposed by MRS Logistica S.A. for the right of FERROBANto the access its network and the Port of Santos. ANTT's arbitration of the conflict, which ended with the enactment of Resolution No. 945/2005.ANTTYsmain decisions regarding this arbitration process were: (i) creating the rights of passage between PerequC and ConceiGSozinha and between PerequC and Valongo, which are part of the network concessioned to M R S ; (ii) authorizing M R S to operate rights of passage system in the access to the patios and terminals in Pedernerias and Campinas, which are part of malha o f FERROBAN; (iii)establishing a payment system on the rights o f passage using a parametric formula, based on distance and volume; and, (iv) authorizing the construction of a second line with mixed gauges between PerequC and Valongo in the areaconcession to MRS. Regarding this last decision, Brasil Ferrovias started, in January 2006, the construction of a new railway network to access the Port of Santos (a project that should cost RS$ 25 million). This construction was conceived after the signature of a contract between Brasil Ferrovias and MRS, which holds the concession and operated a large gauge railway between PerequC and Valongo. The operation of this new piece of the network will be done by MRS. Inexchange, M R S will give Brasil Ferrovias a discount of 20 percent inthe right of passagepayments. Table 5.2. 3 Brazilian railway operators: gross revenues and operational profits (RS$ million), 2001 and 2004 Operator Gross Revenues (RS$ million) Operational Profits (RS$ million) 2001 2004 2001-04(%) 2001 2004 2001-04 (%) ALL 436 769 76.4 19 67 252.6 FTC 23 30 30.4 -3 -4 33.3 FERROBAN 178 189 6.2 -104 -108 -3.8 FERRONORTE 133 437 228.6 -154 -232 50.6 NOVOESTE 45 61 35.6 -26 -74 -184.6 FCA 289 661 128.7 -114 -91 20.2 EFVM 1,167 2,240 91.9 153 515 236.6 EFC 738 1,423 92.8 -182 130 171.4 CFN 23 51 121.7 -33 -39 18.2 M R S 658 1,621 146.4 -161 222 237.9 Source: ANTT. 5.38. Between 2004 and 2005, the railway ownership restructuring process was complemented by operational reforms. The continuation of reforms was aimed at improvingshipment services provision and assuring that the necessary investmentswould be made to improve operational efficiency of the concessionaries involved inthe process. First, the transfer of the railway track between [change these to non italics]Araguari (Minus Gerais state) and Boa Vista Nova (S& Paulo state) from FERROBAN to FCA 86 was authorized.It allowedthe transferringof shares from three firms inthe social capital of FERROBAN (CVRD, CampmelissaAdministraciio Financeira e Consultoria Ltda., andLogistica BandeirantesParticipacoesLtda.) to FCA.63Second, the transferringof the network track between Mairinque and Bauru in Siio Paulo state from FERROBAN to NOVOESTEwas authorized, whichhadthe additionalgoal of optimizingshipmentflows in the corridor from Mato Grosso do Sul state to Santos Port.64Third, in a process involving Ferrovias Norte do Brasil SA. (FERRONORTE), FERROBAN, and Brasil Ferrovias, it was authorized: (i) the constitution of Ferrovias Novoeste S.A. (NOVOESTE), due to the ownership division process of Brasil Ferrovias; (ii) the entrance of BNDESPAR in the managerial group of Brasil Ferrovias, which controls both FERRONORTE and FERROBAN; and, (iii)BNDESpar, Fundaciio dos Economia`rios Federais (FUNCEF), and Caixa de PrevidZncia dos Funciona'rios do Banco do Brasil (PREVI) direct or indirect ownership of Brasil Ferr~vias.~~ 5.4 WATER AND SANITATION 5.39. One important legal bottleneck for private participation in water and sanitation (WSrS) relates to the uncertainty as to who has the power to award concessions -the "poder concedente" issue. The Brazilian Constitution (article 175) specifies that other public services (water and sanitation included) are to be providedby a public authority either directly or via a concessionor permit. The Constitutionalso gives the responsibility for the provision of public services of "local interest'' to the municipalities (article 30), and transfers to the States the jurisdiction over metropolitan regions and urban agglomerations with the objective of providing public services of "common interest". Inaddition, different aspects of the water and sanitation sector-from extraction of water to discharge of treated or untreated wastewater-are subject to legislative and executive powers assigned to different levels of government that often overlap.66 5.40. The controversy is related to the difference in the legal definitions of the concepts of "local" and "common" interests. "Local interest" has beenmost commonly characterized as an isolated water and sanitation systems. On this basis, a municipality 63ANTT Resolution No. 1,009/2005. ANTT Resolution No. 1,010/2005. 65ANTT Resolution No. 1,011/2005. 66 Article 23 (IX) of the Constitution specifically states that "the promotion of programs ... to improve basic sanitation" is a common responsibility of the Union, States, and Municipalities. This article also states that the norms of cooperation between levels of government should be established in a Lei Complementar. Article 21 (XX) establishes the competence of the Union to institute directives on "urban development, including ... basicsanitation", while article 22 (IV)specifies that the Union has the sole right to legislate on "water, though it can authorize States to legislate on these issues in a Lei Complementar. Finally, article 24 (VI) specifies that the Union and States have concurrent powers to legislate on "... protection of the environment and the control of pollution," where the responsibility o f the Union is limited to defining general norms, though under article 20 (111, IV)the Ministe`rio Pu'blicohas asserted the Union's right to protect federal property such as federal rivers and ocean beaches from pollution or other environmental damage. 87 that has its own source o f bulk water, its own storage reservoirs and treatment plants, its own distribution means, and its own collection and disposal facilities within its territory, would have the right to award a concession for water and sanitation services. Municipalities within metropolitan regions, for the opposite reason, could not hold concession rights, which should belong to the corresponding state. This view was challenged in recent years by municipalities, such as Vila Velha (Espirito Santo state), and SBo Paulo (SBo Paulo state), that approved local legislation granting concession rights over water and sanitation services but these initiatives were overruled by state level courts and interpreted as unconstitutional. 5.41. This controversy is amplified by changes in relative income that the ownership transfer of concession rights may cause. In acquiring concession rights, municipalities aim at sharing the present value o f future profits that may be generated from the use of the existing assets and by new investments in expanding networks net o f the any compensation for transferred assets that i s demanded by the existing operator. Despite improvement in selected companies, the financial situation of many operators i s still unclear particularly in terms of hidden liabilities and the quality of assets. Under these circumstances, municipalities served by State companies challenge the level of compensation that has to be paid or liabilities that are transferred to the new concession. 5.42. Some definition of the concession rights is better than no definitionbut, as in any transition, it mattershow it is implemented.Without clear guidance, either inthe form of legislation or a decision of the Supreme Court, decisions by lower courts will reinforce the combination of great uncertainty and perverse incentives for inter- jurisdictional disputes. But the allocation o f concession rights to municipalities or metropolitan regions may generate side-effects that need to be mitigated: as the rich municipalities would tend to leave the state-based system (and absorb the subsidies currently transferred to poorer cities), the system would be left with the loss-making municipalities with obvious implications for the states' public finances and for the sustainability of the system. The impact on the political equilibrium should not be underestimated either. 5.43. In theory, the impact of thepoder concedente issue should be restricted to areas served by integrated systems and should not totally obstruct private investmentin other regions.Approximately 800 municipalities are served by integrated systems. For the other 85 percent o f Brazilian cities covered by isolated systems, there should be no controversies, with the understanding being that municipalities' own the concession rights.67For these municipalities, reaching a minimum efficient scale and subsidizing the poor efficiently are key issues. 5.44. Optimal scale in water and sanitation varies according to topography and density conditions.World Bank studies indicate that economies of scale are stronger for utilities serving less than 125,000 people and are rather weak for companies serving more 67See Abicalil (2002). 88 than 250,000.68 InBrazil, despite a highdegree o f heterogeneity, private local companies serve on average 205,000 people (compared to 4.2 million for state-owned), totaling 4 percent of the market. In March 1997, Aguas de Paranagu6 (Paran6 state) reached 115,000 people while Aguas de Nitercii (Rio de Janeiro state) serviced 450,000. The problem, however, i s that approximately 90 percent of the Brazilian municipalities - corresponding to 35 percent of the overall population - have less than 50,000 inhabitants. On the other hand, less than 120 municipalities have more than 200,000 people. This indicates that a major effort to induce optimal scale through the agglomeration of water and sanitation provision i s needed, regardless of whether asset ownership i s public or private. 5.45. The costs associated with operating on relatively smaller scale may be compensated, at least in part, by better management, as the Brazilian experience illustrates. Despite their smaller size, private local companies presented better performances than large state companies inrecent years. In2003, for example, net profit margin(as a share of revenues) of smaller companies was on average 36 percent, while state companies' average was -13 percent; they covered 96 percent of the water services' consumers, as compared to 90 percent for state-owned firms; and their distribution-loss rates were 34 percent - better than the 48 percent rate for state-owned companies (Seroa da Motta and Moreira, 2004). The recently approved Consortium Law (Law No. 11,107/2005) provides an adequate framework for associations of municipalities seeking to achieve economies o f scale. However, municipal consortiums are not a sufficient condition to achieve efficiency inthe provision o f water and sanitation services. There i s a need to give service providers the appropriate incentives, which are closely related to regulation, such that they take full advantage of these potential economies o f scale.69But no incentives exist for municipalities with surplus to engage in consortiums with municipalities with deficit (i.e., those whose output do not allow them to achieve economies of scale). One way for the poder concedente with surplus to transfer resources to municipalities with deficits would be by establishing an average regional tariff. Inthe areas where the marginal cost i s lower, profits could be used to reduce losses o f municipalities with deficits, which have higher marginal costs. Finally, regarding consortia among municipalities with deficits, the lack of water and sanitation infrastructurenetworks can make such associations not viable. 5.46. Apart from the well-known effects in terms of misuse of resources, the effectiveness of cross-subsidies provided by state-owned enterprises has been increasinglyquestionedon the basis of its targeting. The loss of cross-subsidies from state-owned services may not have a substantial impact on the poorest. Data from the 2000 Census show that coverage for water and sewerage collection among household See World Bank (1999, 2005d). The first report claims that arrangements to serve at least 50,000 people would be economically feasible but that lower levels would imply severe cost penalties. 69 The Consortium Law is an important means for municipalities to provide water and sanitation services more efficiently - but it was not sufficient to align the incentives for the establishment and sustainability of municipal consortiums. This Law will certainly permit municipalities with budget surpluses to improve the provision of water and sanitation services through consortiums. Thus, mechanisms to penalize municipalities that do not comply with either the consortium or loan payments for credit used to acquire equipment and machinery are essential to stimulate these municipal associations. 89 incomes over the equivalent of 10 minimum wages i s 50 and 100 percent higher, respectively, than the rates for households whose income equals up to two minimum wages. The cross-subsidy was not enough to correct the inequality in services access, at least in part, because this type of untargeted strategy transfers subsidies from the poor living in rich municipalities to the rich living in the poor ones. The preferred solution i s well-targeted direct transfers to the poorest. Preliminary estimates indicate that roughly RS$ 500 million transferred annually to the sector for poverty alleviation would be enough to reach the 15 percent poorest consumers currently covered by water and sanitation service^.^' Overall, public resources have been channeled through different sources, not always consistent with strategic objectives or cost-effectiveness. For example, the Program Saneamento para Todos - which i s funded with FGTS resources and replaces the Prd-Saneamento- i s conceived under the "command and control" logic (as opposed to out-put based); most decisions on technology, operations and financial aspects to projects are centralized in the Ministry of Cities; and defined targets may be very difficult to achieve (Abicalil, 2002).71 5.47. Beyond optimal scale and cross-subsidies, the overall institutional environment in the sector requires better definition. This situation is well illustrated by the weak regulatory environment for the operation of the state companies. Most of them operate under precarious, if existent, concession contracts; key economic issues - such as investment levels and tariffs - are controlled by state governments and basic information-such as that needed for appropriate regulatory accountability-is missing. Not surprisingly, a study commissioned for this report, estimating aggregate levels of productivity for state, public-local and private-local operators using SNIS data for 1998 and 2002 and applying non-parametric techniques (DEA analysis), concluded that the levels of efficiency of state-companies are on average inferior to that of local ones.72 5.48. The lack of appropriate regulatory institutions not only reduces consumers' benefits but also increases the risk for private investors. For example, in the well- known Limeira case, the municipality refused to endorse any readjustment of tariffs, unless a larger revision of the whole contract was also implemented. This was despite the fact that annual readjustments were clearly provided for in the contract and the municipality should be limited to "homologate" the readjustment. The concessionaire failed to persuade the courts as well, tariffs fell approximately 40 percent in real terms and all investments were suspended. The case illustrates the weaknesses of concession design, its vulnerability to political interference, and the risks associated with counting on the judiciary as an arbiter of conflicts over contract enforcement, even over relatively standard commercial clauses. At least in part, the judiciary decision was motivated by a 70See also Seroa da Motta (forthcoming). 71 It i s important to acknowledge that the use of social tariff could be an improvemernt, compensating for losses inthe transferring of resources from the poor of municipalities operating with surpluses to the rich of municipalities with deficit. A social tariff would allow poor households of municipalities with surpluses to pay a lower price for the water and sanitation services. Thus, cross-subsidy policies combined with social tariffs could be increase efficiency inthe provision o f water and sanitation services. 72 There are of course exceptions like SABESP and COPASA. These firms are becoming models o f corporate governance in the public sector and, as a result, became able successfully enter the financial markets. 90 legal doctrine that rejects the idea of constraining the powers and discretion of the public authority based on contractual arrangements, an interpretation that touches on the core of the delegation principles o f the reforms aimed at increasing private investments in infrastructure. 5.49. Another aspect of the overall regulatory environment for water and sanitation in Brazil that may affect the incentives in the sector is that the new Consortium Law allows state-companies to be directly contracted by municipalities. There i s no reason u priori to think that the provision of water and sanitation services by state-owned companies will appropriately serve the public interest. Given the bargaining process for contract design, it i s likely that consumers' interest (either in the short- or in the long-term) will not be seconded by those of politicians and other stakeholders. More importantly, this process creates an asymmetry in service provision that tends to discourage private participation. The lack of competitive and transparent mechanisms for concession awards i s clearly contrary to the objectives of improving governance, public accountability and cost-effectiveness inthe water and sanitation sector inBrazil. 5.50. In sum, despite the emphasis on the "poder concedente" issue, the efficient provision of water and sanitation services requires more comprehensivereforms. At its center i s the need for a regulatory environment that creates the appropriate incentives for long-term efficiency and protects private investors from short-term political interferences. The traditional long list of principles of good regulation applies, including delegation over the relevant issues (including tariff revision and re-adjustment) to autonomous and well staffed bodies subject to public accountability, and overall design of institutional mechanisms that encourage the achievement o f minimum efficient scale by local providers. Ideally, the "poder concedente" issue should be treated within this broader context - in terms of the assignment of concession rights that i s most likely to promote effective private concessions with the capacity to finance new investments and to improve operational performance. A broader review of the water and sanitation sector i s provided inAnnex 4. 5.51. Two draft laws (Projetos de Lei) that address some of these issues are currently being reviewed by the Congress. A third version could be presented under the initiative o f the Congress. The Lula administration sponsors PL No. 5,29612005 while PLS No. 155/2005 was developed by Senator Gerson Camata. Neither draft clearly assigns concession rights in areas serviced by integrated systems. The law supported by Lula includes ownership by municipalities of the distribution o f water and the collection of sewerage (the "retail" section) in "all cases" while water and sewerage treatment plants (the "wholesale" section) would be the regulated and managed by all the parties involved. Camata's draft law separates services into "local" and "common" interest, which provides for transfer of rights to the states when integrated systems prevail. The Lula law limits the current structure of cross-subsidies for five years and provides for the creation of a sector fund among municipalities. Camata's draft law establishes that asset reversion has to be conditioned to prior repayment of the non-amortized investments. However, both draft laws lack an institutional design - particularly a solid regulatory framework - that induces efficiency gains through the adoption o f optimum efficient 91 scale or the establishment of tariff rules that penalize inefficiency and favor cost and tariff reduction. Table 5.4.1summarizes the main characteristics of both projects. Table 5.4. 1. Comparative analysis on the Brazilian sewerage policy Issue PLSNo. 155(Senate) PLNo. 5,296 (Government) Poder concedente Municipalities and states(metropolitan Municipalities (municipalconsortium) consortium) Functionalscope Water and sewerage Water, sewerage, solid disposal,and drainage Legalscope Generalguidelinesand guidelines for General guidelinesand detailednational nationalpolicy policy Service structure Step division (eachcan be granted to a Integral(holder is responsiblefor the whole differentholder) service) Definition of service Local: distribution andcollection; Local: infrastructureservices locatedand interest Integrated commonstepsbasedon mutual providedat the municipallevel. agreements. Possibilityof consortium. Common:infrastructureservicesprovided Public roledepends onother players. for two or more municipalities,createdby complementary state legislation. Planningbasedon: Water basin Municipality Planning Holder decides.It can be delegated. Made by holder for all services.It includes regulation, investment, and tariffs. Socialcontrol Holderdecides.It only created the Federal Public councilsat all government levels to SewerageCouncil participateon decisionsregardingregulation, tariffs, planning, and delegation Regulation Same structure regardlessregimenor Differsaccordingto provider's provider's characteristics.Agency oversees characteristics and serviceregimen. all services, with independentdecision- Regulatoryagency oversees delegated makingprocess;it alloweddelegating services only, without independentdecision- agencies' functionsto other bodies. makingprocess; it is forbiddendelegating agencies' functionsto other bodies. Federalpolicy Establishespolicy guidelines to be It creates SISNASA, resourcesconditioned implementedby the government to adhesion,submittingmembersto federal rules Federalfinancing Part of the federal investmentplan Dependingon choice to bepart of the federal investment plan Tariffs Definedby contractsand regulatory Submittedto citizencouncils;regional agencies. It foresees regionaltariffs, but tariffs, cross-subsidiesamong systems requiresthe identificationof localcosts. requireconsortiumcreationamongdifferent Readjustmentsare contractual with an open parties and amanagement fund for tariff methodology. Revisionsbasedoneconomic- resources.Readjustmentsare contractual financialequilibrium(every four years). with an open methodology (price index). Revisionsbasedon economic-financial equilibrium (every four years). Competition Privateconcessionor public consortium Privateconcessionor public consortium contracts.Contract with expirationdate and contracts. Contract with expirationdate and requiredpaymentsto controlexit. No requiredpaymentsto controlexit. Incentives incentives for de-verticalization. for de-verticalizationaccordingto local and common interests. Legal additions None 862 (of which433 are effective). It is being process with 5 other DraftLaws. 92 Currentsituation Reporterpresentedhis vote, which was Reporterpresentedhis preliminaryreport. approvedinthe ConstitutionandJustice Mainchanges: definition of commonservice Commission.Two writing additionswithout andownership; independentregulatory changingthe draft's contents. Projectis agencies; delegationfrom anotherparty to beingprocessedby the ConsumerRights the agency is possible; acknowledges Commission, ending its processingat the regionalservicesand tariffs, but requiresthe InfrastructureServicesCommission.After identificationof local costs; SISNASA that, it goes to the Lower House. integratesall parties involved inthe process; createdthe national seweragecouncil. Reportneedsto be formalizedand voted at the Special Commissionand, after that, voted inthe Lower House. Ifapproved, it goes to the Senate. 93 2 8 8R 8 * E? 0 2 a E 8 Y g .e * 0 rd 1 k E? .-SE8Y8 Y VI Y a E 1 0 0 3 2 i;j 4 g5 .;: 3P Y P 5.4 ENVIRONMENTAL REQUIREMENTS 5.52. Brazilhasprobablythe oldest andmost consolidatedenvironmentallicensing system among developing economies. It was launched in the early 1970s when industrial growth and the proliferation of cars in Rio de Janeiro and STlo Paul0 began to generate significant urban pollution. Two state environmental agencies (FEEMA in Rio and CETESB in Si0 Paulo) equipped and trained personnel, anticipating actions by the federal government. Later, the federal government started to focus on natural resources problems, such as biodiversity, soil, forests, water, and coastal zone management. Currently, it has qualified personnel and environmental licensing procedures, including complex environmental impact assessments, and standardized and widely disseminated procedures used for government clearance. The system includes all states, and municipal governments are beginning to play a role inenvironmental licensing. 5.53. Environmental licensing is a legal obligation that is part of the process of establishing any enterprise that could potentially pollute or cause environmental degradation. One of the main characteristics of the process is social participation inthe decision-making, by means of the Public Hearings. Environmental licensing responsibilities are shared by the State Environmental Agencies and the federal environmental agency, the IBAMA, as integrant parts o f the SISNAMA (National System of Environment). The IBAMA acts, mainly, in the licensing of large infrastructure projects that impact in more than one state and all the activities of the oil and gas sector inthe continental shelf. The main guidelines for the licensing process are expressed in Law No. 6,938/1981 and Resolutions CONAMA No. 001/1986 and No. 237/1997. Beyond these, the Ministry of the Environment recently issued Position No. 312, which explains the state and federal responsibility for the licensing, based on the size and type o f the impact. In the case of large infrastructure projects, the licensing process i s initiated with the request by the entrepreneur to start the process, submitted to the Office of Licensing and Environmental Quality (DILIQ)of IBAMA. 5.54. The process of Environmental Licensing has three distinct stages: prior licensing, installation licensing and operation licensing. Prior licensing must be requested of the IBAMA in the planning phase o f the project. This license does not authorize the installation of the project, but attests to its environmental viability and authorizes its location and the technological concept to be used. In particular, it establishes the conditions to be considered in the development of the executive project. Installation licensing i s required for the beginning of construction or the establishment of the enterprise. The expiration date of this license is based on the timetable of the project or activity, but cannot be longer than six years. Projects that will cause deforestation require the "Authorization for Suppression of Vegetation." Operation licensing must be requested before the enterprise actually starts operations. Its concession i s conditional to an inspection in order to verify if all the described requirements and technical details presented in the approved project had been developed and taken care of throughout its installation and if they are inaccordance with the other two licenses. It i s valid for at least 95 four and up to ten years.73Each stage o f the licensing process includes specific studies to be prepared: in the initial stage, if the project i s deemed to have significant potential impact, an Environmental Assessment Report Environmental Impact Assessment Report (EL4 / RIMA) is prepared. The RIMA is a public document and, at this stage, public hearings are carried out so that interested stakeholders are properly consulted. 5.55. While the system has achieved participation and transparency, it can be bureaucratic and adversarial to entrepreneurs. Government procedures still treat environmental issues outside the planning process and this encourages subsequent litigation. Also, in large infrastructure projects, the scope o f licensing remains unclear. For example, the indirect impacts of roadbuildinginfrontier areas are far more important than the direct impacts. But while many of these indirect impacts may be o f an environmental nature, such as increased deforestation, many are not, such as effects on indigenous communities or population growth. Inthis context, IBAMA has thus inflated its mandate to license many economic activities. Uncertainties inthe licensing system for private and public energy infrastructure projects (electricity, oil and gas) have led to delays ininvestment and operation of such projects, and thus to lessened growth (see Box 4). 5.56. The current challenge is to mainstreamenvironmentalmanagementinto the broader economic agenda. In particular, environmental scrutiny should be moved "upstream" in the awarding process. That would allow for early consideration of alternatives and inclusion of environmental improvements in planning and design. The govenunent of Brazil has now established that any new enterprise, such as a highway or hydroelectric unit, can only be offered to the private sector with valid prior licensing and with the process of obtaining the prior licensing carried out by the relevant sector regulator. This tends to reduce the environmental risk to the investor. For an existing enterprise that the government intends to privatize or sell as a concession to the private sector, it can only do so after all relevant legislation has been addressed, including environmental legislation. 73 During the licensing process, consultations of the relevant state agencies responsible for the environmental licensing (OEMAs) and the Federal Agencies for Historic Patrimony (IPHAN), for Indigenous Peoples (FUNAI), of Quilornbolus Communities (Palmares Foundation), and public health (FUNASA) are also undertaken. 96 Box 4. EnvironmentalBottlenecksinPractice A. EnvironmentalLicensingof the ItumirimHydropowerPlant The National Electric Energy Agency - ANEEL - offered the hydroelectric power plant of Itumirim for bidding based on its 50 MW potential in 1999. The Cornpanhiu Energe'ticu do Itumirim won the bid, offering a price seven times higher than the minimum price. The plant is expected to be built on the Corrente River. This river is formed by the junction of the Formoso and Jacuba Rivers, which have some o f their springs inside a National Park (Purque Nacionul das Ems). The area o f the dam will flood a small part of the park and most of the area of an important, relatively intact, ecological corridor, which is home to many endangered animal species, such as the cervo do pantanal (deer), onppintudu (panther), lobo-guard (wolf) and tamandud- bandeiru. Because the Corrente River i s a state river, the company submitted its environmental license to the state environmental agency. However, because of the possibility of affecting a National Park as well as endangered animal species, the Ministe'rio Pliblico together with IBAMA took legal action against the license to be issued by the state agency. The action was approved and environmental licensing was transferred to the federal agency, IBAMA, which considered the enterprise unfeasible in 2001 due to the aforementioned impacts. The entrepreneur requested the licensing of a new plant which would lower the dam's level from 680 to 678 meters, with an operational level of 675 meters. InFebruary o f 2002, IBAMA issued a final decision refusing the construction of the plant even under the new proposed conditions. B. TenderinginAbrolhos: UpstreamConsiderationsofEnvironmentalImpacts InMarchof 2003, ANP launched a tendering processfor the exploration of areas with oil and gas potential. The areas were selected based exclusively on their production potential. The terrestrial and marine blocs were located inten different states covering a total area of 195,000 square kilometers. InMay of 2003, a group of environmental NGOshanded areport to the Ministries of Environment, Energy and Mines and to ANP, indicating that 243 of the proposed blocs would have a significant impact on the Marine Park of Abrolhos, seriously threatening the existing coral reefs, algae, mangroves, beaches and marshes. The region i s the largest nesting area of the jubarte whale in the South Atlantic, in addition to being home to large colonies of marine birds, turtles and corals. IBAMA and ANP acted quickly to evaluate the issue. After an expeditious review, IBAMA recommended to ANP that 162 blocs originally contemplated for tendering be excluded, based on the potential environmental impacts. ANP fully endorsed the recommendation and excluded those blocs from the process before the tendering documents were published. In addition to ensuring the integrity o f the area, this prevented entrepreneurs from acquiring blocs with minimal chances of obtaining future environmental licenses. Most of the other areas were offered and bought for exploration. 5.5. OTHERREGULATORY ISSUES 5.57. The Bill concerning regulatory agencies (Lei das Agencias), submitted to Congress in 2003, proposes several changes in the legal framework of the regulatory agencies. Among the changes, four deserve attention: (i) the transfer to the Ministries the power to award concessions; (ii) creation o f management contracts between the agencies and the executive power; (iii)changing the mandates of the directors and the partial overlap with the mandate o f the country's president; (iv) creation of new mechanisms of external control; and (v) reassignment of planning policies to the Ministries. 97 5.58. The current institutional framework of the regulatory agencies gives the right to award and design infrastructure concessionsto the regulatory agencies. The new Lei dus Agencius transfers this power to the Ministries with the purpose of delegating the power to define and implement public policies inthe infrastructure sectors. However, this motivation i s questionable. The current framework already defines the elaboration and implementation of public policy to the executive power. The problem i s more the lack o f functionality of the bodies responsible for the policy than legal delegation. The question i s about the incentives o f the executive power to design contracts that are less susceptible to opportunistic behavior, since they will not be responsible for managing the contracts afterwards. Poorly designed contracts increase the complexity of the regulatory process and augment the risk o f expropriation. This adds to the regulatory risk and consequently the premium for investing with a higher risk. 5.59. The new Law establishes the possibility of using contracts to set targets for the regulatory agency. It also establishes that the contracts will have two basic goals: (a) to promote transparency and increase the control of the agency by the society, and (b) to improve the relationship between the regulatory agency and the direct administration in the implementation of public policies. The contracts should also specify: (a) targets o f performance for the agency, (b) budget, (c) obligation and responsibilities, (d) measures of evaluation, and (e) penalties for violatiws o f the contract. 5.60. The use of a contract to define the relationship between the regulatory agency and the direct administration is complex and risky. Different from other public services that produce more tangible goods and services (health, education), the regulatory service i s non-tangible and consequently, difficult to measure. This difficulty suggests that the purpose of the contract is to guarantee a minimum standard o f the administrative aspects (length of time of processes, etc.) and not a minimum standard in the regulatory decisions. The problem is the wide range of possible interpretations o f the Law. The existence of penalties can make the administrative characteristics of the service a priority to the regulator, and also increase the influence of the Ministry on the regulatory agency. The necessary independence of the regulatory agency makes the use of contracts a dangerous mechanism. 5.61. The new Law of the Agencies proposes changes to the term of the General Director of the Agencies. The Law proposes that: (a) the General Director i s appointed to a specific term, and (b) the term of the General Director ends at the first semester o f the second year o f the term of the country's president. Intransforming the term of the General Director, the Law eliminates the risk of the President removing the Director from his or her post before the end of the term. On this point, the Law of Agencies reinforces the independence of the regulatory agency. The separation of the terms of the General Director and the President decreases the regulatory risks associated with the political cycle. It is important to mention that the term of the General Director will not coincide with the terms of the other Directors, and that the final decisions of the Agency depend on the approval o f the board of directors. In this sense, this change will not modify the regulatory framework. 98 5.62. Finally, the proposedLei dusAgencius re-assigns sector planning attributions to the sector ministries. President Lula's administration started to reassess some o f these issues. One the regulatory side, one first issue i s the amount of power delegated to regulatory agencies. A sense of over-delegation (and the corresponding weakening o f the power o f the executive branch) i s due to the following, among other factors: (a) the proliferation of "executive" agencies as a means to bypass rigidities of the public administration, with the task of implementing government policies (ANVISA, ANS, ANCINE), (b) the lack of definition of important policy issues by other bodies - federal government or Congress - creating a sense of self-legislation by infrastructure agencies (e.g. asset price evaluation in the electricity sector), (c) the involvement of agencies in issues that do not necessarily affect the costs of capital nor the efficiency Qfthe regulation but rather have clear equity (political and social) aims (e.g. policies for the universalisation of services, granting concessions and conducting auctions), and (d) the lack of public planning (when appropriate) by the executive body. 5.63. A second issue affecting regulatory risk is the revision of regulatory decisions by the courts. At present, disputes between private operators and regulators must be heard inregularjudicial courts. This may not be an appropriate system for resolving such disputes because the courts lack the technical expertise; tend to focus on administrative procedures (examining whether the regulator follow the appropriate administrative process or due process), inducing the regulators to adopt unnecessary procedural rigidity; and often take too long to make a decision. Insome cases, regular judges were asked to decide on the appropriate rate of tariff readjustment for electricity distributions or road concessions. Results so far have been increasingly positive with the courts avoiding reforming the contents of the decision taken by the regulator. While court revisions of administrative decisions are needed to guarantee the appropriate accountability of regulatory decisions, alternate mechanisms may be less burdensome and risky for all the parties involved. The creation of a special appellate tribunal comprised o f technical, economic and legal experts to deal with such disputes on a fast-track basis-with further appeals narrowed to due process issues - could be an alternative approach to the problem, as inthe cases o f Australia, Chile, and New Zealand. 5.64. A final issue is collaboration among public institutions involved in the regulatory process. Relatively easy cases, such as those involving the water, power and environment regulation, involve federal institutions with relatively common mandates and objectives. Two far more difficult - but particularly relevant - cases are those o f the role of the Ministe`rio Pu'blico and the Tribunal de Contus du UniGo. These are legally independent institutions with mandates and objectives that are not necessarily aligned with those of the regulators and sector policy-makers: the first with a general mandate of law enforcement and the second with a mandate to fight corruption. Both have been actively involved in reviewing environmental licenses and concessions. While the surveillance of these institutions i s essential for the appropriate checks and balances to be inplace, some prudence is recommended so that regulatoryriskiskept under control. 99 6. CONTRACT RENEGOTIATION A complete legal framework, good governance and strong regulators are essential to reduce regulatory risk, but that may not be enough to facilitate private investment if concession contracts are poorly designed. Investments in infrastructure have high sunk costs, which cannot easily be recouped if the economic environment deteriorates. These sunk costs may tempt governments to behave opportunistically, taking regulatory actions that expropriate available quasi-rents once costs are sunk. That possibility i s the main source of regulatory risk, which has an impact on levels of investment, costs of capital and tariffs and public subsidies, since additional premiums are required to cover the risk. Exchange rate fluctuations and commercial and project risks further complicate concession design and are normally required to be managed by host governments; however, governments must be careful to neither take on too much risk nor to over- compensate for it.Compounding the problem i s the additional objective of increasing coverage - particularly for the poor - which often does not mesh well with the natural incentives of private operators. The provision of cross-subsidies to increase coverage makes the liberalization of the sector difficult. Basedon dataset from Guasch(2004), this chapter presentsa brief history of concessionaward and renegotiations in It is not only the government that may behave opportunistically. Once an enterprise has been granted a concession or franchise in an infrastructure sector, that enterprise may correspondingly be able to take actions that "hold up" the government, for example through insisting on renegotiating the regulatory contract ex post, or through regulatory capture to extract supernormal rents from the users, in detriment of efficiency. The extensive informational advantages that the enterprise possesses over the government regulator (as well as over other potential operators) are one reasonfor this. If those issues are not addressed properly, the result may be a regulatory arrangement that i s less effective than envisioned inprotecting customers from monopoly abuses. Inthis context, a complete legal framework and good governance strong regulators are essential to reduce regulatory risk, but they may not be enough to reduce risk and facilitate private investments if concessioncontracts are poorly designed. 6.1, Brazil had a greater proportion of contracts renegotiated when compared with the other countries in the region. (The Latin America region's proportionwas 30 percent; while Brazil was 41 percent.) The numbers also vary within sectors: from 100 percent of renegotiation inwater and sewerage, to 57 percent on the transportation sector and no renegotiation in telecommunications and energy. The more competitive nature of l4 Guasch, J.L. (2004). Granting and Renegotiating Infrastructure Concessions - Doing it Right. The World Bank Institute- DevelopmentStudies:Washington, D.C. 100 telecom and energy decreases the bargaining power o f the concessionaires and their incentive to engage in contract renegotiation. In fact, this i s the pattern found in almost all the countries in LAC; a higher incidence in transportation and water and sewerage thanontelecom and energy (Table 6.1). 6.2. The mean time of the first renegotiation (between the signature of the contract and the first renegotiation) was smaller in Brazil than in the rest of LAC (2.2 years in Brazil versus 2.7 in LAC). This result is heavily influenced by the transportation sector, which averages one year until the first renegotiation in Brazil, as opposed to three years inL A C (Table 6.2). 6.3. All concessions in Brazil were awarded through a competitive bidding process, while only 78 percent of concessions in the LAC region were awarded competitively. The energy sector in L A C i s remarkable in that only 40 percent of concessions were awarded through competitive bidding. After controlling for competitive bidding,the incidence of renegotiation inL A C was 49 percent, while inBrazil it was 42 percent (Table 6.3). 6.4. Brazil differs from the LAC region in terms of who initiated the renegotiation. Seventy-three percent of renegotiations in Brazil were initiated by the government, with the remainder initiated by the operator or both government and operator. In LAC, only 26 percent were initiated by the government, 61 percent b y the operator and 13 percent by both. Guash (2004) presents two explanations for political interference in concessions, such as initiating renegotiation: the first i s corruption, where a corrupt government tries to capture the quasi-rent of the concession. He finds empirical evidence that corruption increases the probability of renegotiation in LAC. The second explanation comes from the election cycle. New governments tend to re-evaluate actions taken by the previous administration (Table 6.4). 6.5. The type of regulation affects the probability of renegotiation through risk allocation. Rate of return regulation lowers the probability of renegotiation because the costs of potential adverse events are borne by the government. This contrasts with price cap regulation, where risk i s borne by the operator and i s more fragile to shocks, such as when adverse events trigger a demand to renegotiate by an operator seeking to restore financial equilibrium. This i s an important driver of renegotiations, since more than three- quarters of the concessions in L A C are regulated using price cap regulation, and the region has a rather volatile economic environment. 6.6. Brazil used more price cap regulation than the rest of LAC region. Infact, all regulated concessions in Brazil adopt a price cap regulation. The riskier nature o f the price cap regulation may in part explain the higher incidence o f renegotiation in Brazil than in LAC. Since price cap i s also the dominant mode of regulation in LAC, other factors also explain this difference (Table 6.5). 101 Table 6. 1.L A C and selected countries: fraction of contracts renegotiated Country Telecom Energy Transport Water and Sanitation Total Fraction Percentage Argentina 0117 0131 34140 11/14 451102 44.1 Bolivia 010 2/17 015 112 3/24 12.5 Brazil 0187 017 28/50 42/50 701194 36.0 Chile 0112 4181 6/27 013 101123 8.1 Colombia 010 010 28144 017 28151 55.0 Mexico 0163 2151 46191 46/58 941263 35.7 Total L A C 31273 251256 1511276 102/137 2811942 30.0 Source: Own elaboration basedon Guasch (2004) data. Table 6.2. L A C and selected countries: mean time untilthe f i s t renegotiation Country Telecommunications Transport Water Total Argentina n.a. 2.2 2.2 2.2 Brazil n.a. 1.o 3.0 2.2 Chile n.a. 3.0 n.a. 3.0 Colombia n.a. 5.1 n.a. 5.1 Mexico n.a. 3.2 1.6 2.3 Peru 5.0 2.0 n.a. 3.5 Total L A C 5.0 3.0 2.1 2.7 Note: n.a. = not available. Source: Own elaboration based on Guasch (2004). Table 6.3. LAC: nartv that initiated the renegotiation bv infrastructure sector Sector Concessionaire Government + Concessionaire Government Total Telecom 2 0 1 3 Energy 0 1 12 13 Transport 18 75 55 148 Water 0 36 67 103 Total 20 112 135 267 Source: Guash (2004). Table 6.4 Brazil: Dartv that initiated the renegotiation for transnortation and water sectors Sector Concessionaire Government Total Transport 0 32 32 Water 22 28 50 Total 22 60 82 Source: Own elaboration basedon Guasch (2004). Table 6.5. LAC: distribution of regulatory regimen Regulatory Regime Frequency Percentage Cumulative L A C Brazil L A C Brazil L A C Brazil No regulation 136 3 15.53 1.85 15.53 1.85 Price cap 608 159 69.41 98.15 84.93 100.00 Rate of return 50 -- 5.71 -- 90.64 -- Hybrid 82 -- 9.36 _- 100.00 _- Source: Guash (2004). 102 6.7. Renegotiation was far more likely inL A C (including Brazil) when concession contract awards were basedon the lowest proposedtariff rather than on the highest transfer fee. O f the contracts awarded by the lowest tariff in LAC, 60 percent were renegotiated, while inBrazil the proportion was 95 percent. However, while 11percent of the concessions awarded by the highest transfer fee inLAC were renegotiated, Brazilhad only 5 percent (4 out of 82) of the contracts renegotiated in this case. The first explanation for this i s that tariffs are subject to constant revision, and thus awards based on tariffs are more likely to be renegotiated than those based on transfer fees. As stated by Guash (2004), "it is foolish to think that [tariffs] will be maintainedfor the duration of a concession usingthe accorded adjustments." The second reason i s that basing the award on a minimum tariff imposes little sunken commitment on the concessionaire. Since the concessionaire does not pay anything upfront, it i s easy to walk out of the process, increasing the concessionaire's leverage. Finally, minimumtariffs are a proxy for optimal tariffs, and their use as an award criterion can lead to deviations from the optimal tariff, giving rise to renegotiations. 6.8. Renegotiation occurred more frequently when there was no regulatory body in the sector. Where there was no existing regulatory body in place, 61 percent of the contracts in LAC were renegotiated, in contrast to 17 percent where a regulatory body existed. For Brazil, these ratios are 100 percent and 20 percent, respectively. The existence of an independent agent that evaluates claims for renegotiation and rejects inappropriate claims decreases the incidence of renegotiation. 6.9. Renegotiation was more likely when the regulatory framework was embedded in the contract than when embedded in a decree or a law. InLAC, 40 percent of contracts that included the regulatory framework were renegotiated, as opposed to 41.5 percent in Brazil. When the regulatory framework was embedded in a decree, 28 percent of contracts in LAC were renegotiated, as opposed to 29 percent in Brazil. No cases o f contracts embedded inthe law were renegotiated inBrazil. A stronger legal grounding for regulation (embedded in a law rather than in a decree or contract) lessens the probability of renegotiation and increases the political cost of government-led renegotiation. 6.10. The main reasons for contract renegotiation in concessions in Brazil are tariff review (54 percent) and change in investment plans or needs (46 percent). In LAC, the leading driver of renegotiations are macroeconomic crises (30 percent) followed by changes in investment plans (23 percent) and tariff renegotiation (22 percent). Macroeconomic crises involve unexpected changes in the economic environment that lead to renegotiations to adjust possible imbalances, and therefore are welfare improving. They do not play a role in initiating contract renegotiations inBrazil. Investment plans and tariff renegotiations, on the other hand, may be due to opportunism from both the government and the concessionaire (Figure 6.1). 6.11. All renegotiations inBrazil resulted in changesin investment plans or tariffs. Incontrast, almost one-third of renegotiations inLAC resulted inthe government taking over, cancellation of the contract or a new concession; one-fourth resulted in changes in investment plans; and the remainder was evenly distributed across extensions of the 103 the current state government of the legal situation of the concession contracts signed by the previous admini~tration~~.decided in favor of the current administration, the toll If charges would be reduced. Two facts are of importance here: the first i s the political opportunism of trying to renegotiate contracts for political gain. The second i s the general perception of the high price of the tariff that gives rise to political interventions of this sort. 6.13. The high tolls might be due to over-investment by the concessionaire, also known in the economic literature as the Averch-Johnson effect. These contracts are cost-plus contracts, since they determine a tariff that i s sufficient to correctly remunerate the concessionaire's investment, as stated on the contract. The fact that the profit of the concessionaire depends on its capital base induces the concessionaire to over-invest, in order to increase its base and have higher profits. It i s one of the main drawbacks of cost- plus regulation. A less ambitious investment plan would involve lower toll charges and less room for political intervention. When possible, price cap regulation with quality targets instead of an investment schedule would provide better incentives to achieve more efficient investment levels. 6.14. If potential opportunism is not addressed properly in contract design, the result may be a regulatory arrangement that is less effective than envisioned in protecting customers from monopoly abuses. It is not only the government that may behave opportunistically: Once an enterprise has been granted a concession or franchise inan infrastructure sector, that enterprise may then be able to take actions that "hold up" the government, for example through insisting on renegotiating the regulatory contract ex post, or through regulatory capture to extract supernormal rents from the users. The extensive informational advantages that the enterprise possesses over the government regulator (as well as over other potential operators) are one reasonfor this. CaseStudy: Limeira Water Concessions 6.15. The municipality of Limeira in Brazil was a pioneer in concessioning the local water company, Aguas de Limeira, to the private sector. The concession was awarded to the bidder that proposed the lowest tariff and an ambitious investment plan was embedded on the contract. The subsequent administration questioned the concession process and the mainclauses of the contract, taking the matter to court to try to cancel the concession. During the four years of the legal process, no tariff adjustments were allowed, and as a responsethe concessionaire stopped all investmentsunder the contract. In 2001, both parties came to an agreement: the concessionaire obtained a tariff readjustment that partially compensated for the years of the price freeze and a longer time length for the implementation of the investments stated on the contract, while also assumingpart of the debt of the old state water company. 75The two governorsbelong to different political parties. 105 CaseStudy: ViaoesteConcessionaire 6.16. The Viaoesteconcessionaire leveragedits work under one part of the contract to pressure the government to renegotiate other sections. Given unexpectedly low traffic on a road where it had made significant investments, the concessionaire Viaoeste triedto renegotiate its concessioncontract to compensatefor the low profitability of some segments of the concession. The government was unwilling to renegotiate, since it considered the possibility of low-profitability a normal business risk, and had not stated inthe contract that renegotiation should arise inthis case. As aresult, the concessionaire stopped its work on the duplication of Raposo Tavares,a highly used highway, in order to pressurethe government to renegotiatethe contract. 106 7. REGULATORYGOVERNANCEAND EFFECTIVENESS This chapter reports the state of regulatory governance inBrazil based on the results of a survey of 21 regulatory agencies. The complete report can be obtained in Correa et al. (2006).767 Besides incompleteness of legal and policy framework, another component of 77 regulatory risk i s regulatory governance: weak governance leads to meager implementation outcomes. Governments have to solve the problem of credibly committing to secure property rights inter-temporally and one solution to this dilemma involves the delegation of authority to independent regulator^.^^ 7.1. The state of regulatory governance defines the conditions available for regulators to enforce existing concession contracts and legal framework. Governance i s the result of the combination of four distinct features: autonomy (to exercise effectively the powers that are granted by statute); adequacy of decision making processes; access to appropriate decision making tools (legal and regulatory instruments); and ac~ountability.~' The relationship among concession regulatory governance, effectiveness, and industry performance is illustratedby Figure 7.1. 7.2. Autonomy is of primary importance due to its direct effect on regulatory risk.By granting formal anddefacto autonomy to regulatory agencies, governments seek to reducethe regulatory risk of investors. Autonomy refers to different objects: the first is autonomy vis-&vis governments and the second is autonomy vis-&vis the regulated 76 The complete report can be obtained in Correa, P.G., C. Pereira, B. Mueller, and M. Melo (2006). "Regulatory Governance in Infrastructure Industries - Assessment and Measurement of Brazilian Regulators." PPIAF Trendand Policy OptionsNo. 3. 77 The 18-page questionnaire used in the survey was composed of 106 questions divided into the four subsections o f autonomy, decision-making, decision tools, and accountability. This wealth of information is greater than that used by previous studies to created governance indexes, such as those created by Holder and Stern (1999), who used 32 questions, and Gutierrez (2003), who used eight factors. It was substantially enriched by the analysis provided by IDEC (2002), A M C H A M (2003a, 2003b, 2004a, 2004b, 2005a, 2005b, and 2005c), and Brown and de Paula (2004).The questionnaires were often applied by the same two research team members during visits to the agencies or during the 2005 Conference of Brazilian Regulatory Agencies inManaus. Only four state regulatory agencies sent their answers by e-mail. High-level officials, most often either the president or directors, answered the survey. Also, all o f the questionnaires were applied during the same time period (from April to June 2005). A total of 21 agencies are represented in our sample - 6 federal-level and 15 state-level. All of these conditions are ,important for the quality (completeness and consistency) of the information provided. 78 B y delegating powers to independent regulatory agencies, the executive assures private investors that it will not be able to arbitrarily intervene in the market and expropriate rents, once investments are sunk (lowering interest rates or administratively expropriating investors after privatization through lower tariffs). Delegation is, therefore, a solution for an inter-temporal problem: by relinquishing short-term interests, political actors can minimize the risk of expropriation (regulatory risk) and its effects on cost and availability o f private capital, thus, benefiting inthe long run. See Levy and Spiller (1996), and Spiller and Tiller (1997). 79 Recent work on regulatory governance has defined it in different but largely complimentary ways. Levy and Spiller (1996) defines it as the mechanisms that society uses to constrain discretion and to resolve conflicts that arise inrelation to these constraints. industry interests. Political autonomy is reflected inthe degree of insulation of regulators from the political market. Legal status o f regulators can only be effective if enforcement i s isolatedfrom short-term political influence. Insulation from private sector interests also plays a role as decisions by regulators who have ties to the private sector tend not to be sustainable over time. 7.3. Four devices are essential for political autonomy: (a) tenure and staggered mandates for regulators (not-coincident with the executive), (b) legal means to enforce decisions, (c) financial autonomy, and (d) established process for appeals to the judiciary (and not to any executive body). "Autonomy requires that the agency's funds are not subject to impounding or appropriation for other purposes" (Smith, 1997). Institutional design should include mechanisms for appealing decisions that are neither excessively disruptive of the regulatory process (Le. when there too many sites [unless this is a technical word it does not make sense-"cites" perhaps?] of appeal by non- specialized agents) nor be weak and ineffective. An appeal through the executive branch - governors, presidents or line ministries - represents interference in the regulator's autonomy and should be prohibited. Appeals should normally be made on the grounds of procedure (not statutory or evidential grounds) and involve only the agency and the relevantjudicial institutions. The latter should have developed expertise inregulation and have designated courts for dealing with regulatory matters. 7.4. The use of clear administrative procedures affects the management of regulation. The adoption of administrative procedures induces compliance with existing rules and norms, as for example with the mandate of the institution. Compliance with due process also reduces the risk that regulator's decision will be reversed incourt, increasing the sustainability of the regulatory system. As some level o f arbitrariness is inherent in regulatory decisions, procedural requirements limit the range of feasible choices available to the regulators, protecting private investors from abuse and misuse o f discretionary power. Administrative procedures also encourage deliberative rationality, strengthening coherence and predictability of law and contract enforcement, which are important outputs of the regulatory process. 7.5. Among the decision-making rules, how decisions are reached simple majority or consensus -is - through a key aspect. The degree to which past decisions bind future ones - a factor that serves to anchor investors' expectations - is also of importance (Berg, 1998). Also important i s the extent which decisions are required to be explained in written documents, as it helps to establish a record that can form the foundation for consistent implementation of the law (Guasch and Spiller, 1999). On the issue o f stakeholder participation (and also fairness), instrumental factors are the existence of formal means of participation, the practice of taking submitted opinions into consideration and whether parties have equitable access to decision-makers. 7.6. Information is the most valuable resource in the regulatory process. Regulators normally lack information on cost and demand structure, and regulated firms have little incentive to reveal what they know (Guasch and Spiller, 1999). Incomplete information and limited ability for firms to observe the work o f the regulator are intrinsic to the regulatory process and create opportunities for strategic behavior on the part o f 108 both the regulator and the regulated firms (Baron, 1989). The informational problem is aggravated by evolving market structures and rapid technological progress that create frequent new developments (Lafont and Tirole, 2000). Access to information and resources to obtain and process it are therefore essential to the regulatory process and important for reducing the risk of mismanagement of contracts or ineffective regulation. 7.7. Four groups of inputsare extremely important for good management of the regulatory process: (a) legal means to collect information, (b) appropriate budget to manage and process this information, (c) qualified personnel, and (d) regulatory tools. Regulators neednot only the right to request information, but the effective legal power to request it, which usually requires the ability to issue warnings and impose fines. The regulator's technical staff should ideally have competitive pay scales and benefits and some degree of job stability and access to training programs. Selection should occur mainly through competitive exams, to avoid problems related to conflicts of interest or the influence of political factors. Regulatory tools include regulatory accounting systems, methodologies for tariff setting and instrumentsfor monitoring quality. 7.8. Once regulation is delegated, a legitimate question is: who regulates the regulator? Once regulators have the monopoly over regulation, how can society protect itself from the risks of a monopoly, notably inefficient outputs and excessive costs? If it i s supposed to be sustainable, delegation should not represent a blank check from the principals (the executive/legislative bodies) to their agents (the regulators). Autonomy needs to be reconciled with measures to ensure that the regulator i s accountable for its actions. Checks and balances are requiredto ensure that the regulator does not stray from its mandate, engage in corruption or simply become inefficient. 7.9. One important device of accountability is providing effective arrangements for appealing the regulator's decision. Appeals should normally be made (a) on the grounds of procedure (not statutory or evidential grounds), and (b) involve only the agency and the relevant judicial institutions (not the executive branch). Another factor is oversight mechanisms: (a) agencies should be subject to legislative oversight by specific legislative commissions and to mandatory provision of periodic reports on the effects of regulation, and (b) agencies should be monitored by the public prosecutor's office andthe corresponding audit office. Another important device of accountability i s transparency, including open decision-making (Smith, 1997). Transparency requires publicizing decisions and meetings by disclosing relevant information, announcing in advance the schedule of meetings and their respective agendas and making the minutes of meetings heldavailable to thirdparties. Brazil was a relative late-comer to the creation of regulatory agencies, with the first federal agency created only in December 1996 (ANEEL--electricity). This was followed in July and August of 1997 by the creation of ANATEL (telecommunications) and ANP (petroleum), respectively. The next infrastructure regulatory agency created was ANA (water) in July 2000, followed by ANTAQ (water transport) and ANTT (ground transportation), both in June 2001. State administrations created independent regulatory institutions approximately at the same time that the federal government set up 109 the federal institutions." Between 1997 and 2003, 21 agencies were created in 19 of Brazil's 27 states.81 Figure 7.1Regulatory governance, effectiveness and industry performance parliamentary), executive-legislative relations, independence of the judiciary, bureaucratic capabilities, electoral rules - 1 r Legal Framework I Sector laws and Concession legal environment Contracts Autonomy Decision Making Instruments Powers Regulatory Outputs 1 P Industry Performance Investment, innovation, prices, and quality Most of the federal regulatory agencies exhibit very similar institutional designs. Despite the high degree of organizational isomorphism among the agencies, however, the rationales for the creation of the agencies are very distinct. While in the case of infrastructure, the rationale for the creation of the agencies is related to the search for credibility, inother areas, such as health and sanitation, it relates to blame shifting, credit claiming and administrative flexibility. 81 Eighteen states set up institutions that are multi-sectoral (i.e., they regulate more than one industry), whereas two states (Le., Paraiba and Silo Paulo) created two sectoral agencies. Results presented below on the characteristics of regulatory governance were obtained from a survey of agency directors and are found inCorrea et al. (2006). 110 7.1 AUTONOMY 7.10. Results indicate that, on average, there are aspects of the agencies' governance that conform to what would be recommended according to theory, but there are also aspects that contradict those recommendations. On the positive side, we see that in over 60 percent of the cases, there are legal restrictions on the dismissal of the directors; however, inalmost 35 percent o f cases directors can be dismissed for a non- legal reason such as for "threatening the agency's integrity or autonomy." Moreover, in almost 86 percent of the agencies, directors' terms do not coincide with those of the executive. In order to mitigate political risk, 62 percent of the agencies need legislative approval for their appointed executives. This involves all federal agencies and 47 percent of the state agencies. In terms of the clarity o f rules, 16 agencies (76.2 percent) considered the legislation to be clear on their mandates and attributions, and 71percent of the agencies acknowledge to have been delegated the power to set andchange tariffs. 7.11. The survey shows that 81 percent of the agencies have financial autonomy: they control their own expenses throughout the year. There was with no statistical difference between federal and state agencies. In general, Brazilian regulatory agencies' budget revenues come from several sources: donations, fees and fines, (42.9 percent) or a combination of these sources with public support (47.6 percent). In only 2 out of the 21 sampled agencies the government exclusively provided revenues. On the other hand, 13 agencies have had their revenues impounded (contingenciados) by the Executive. The impact of impoundment on the a ency's operations was, in general, "very high" (46.2 percent) or "high" (30.8 percent).8KAlthough one might assume that the budget suggested by each regulatory agency and approved by Congress should be considered sufficient, the continuous executive impounding o f their budgets have undermined the capacity of the agencies to operate. 7.12. The existence of formal attributes for autonomy does not prevent other government bodies from attempting to influence regulatory decisions. The three most common mechanisms used by ministries or state governments to control regulators are budget increase/decrease (57.1 percent), nominating new directors (57.1 percent) and substitution of directors (28.6 percent). The survey shows that 62 percent o f the agencies classify these mechanisms as either "very highly" or "highly" effective. As a result, inthe same proportion (62 percent), ministries or state governments have formally or informally interfered in agencies' policymaking decision processes. A comparison of the replies o f the state and federal agencies regarding the use and efficacy of these means of control indicates a higher incidence inthe state agencies. That is, state governments have a higher propensity to intervene inregulatory agencies than does the federal government, which is perhaps due to the higher need by the latter to credibly signal to investors that opportunistic behavior inthe form of administrative expropriation will not take place. **Note that this is not a predicament exclusive to particular regulatory agencies, but to the entire public sector. Nevertheless, one should recognize that agencies, which are used to operating with higher budgets in the recent past (such as ANATEL, ANEEL and ANF') in order to accomplish their roles and commitments, have been more restricted in the present in comparison to agencies that have learned to operate with smaller budgets since their creations. 111 I Box 5. Meeting the Regulators: Where They Come from and Where They Go I ~~ _ _ _ _ _ _ _ ~ ~ ~ _ _ _ _ _ ~ The survey shows that before working for a regulatory agency, from a total of 93 directors, 28 were working at old public companies in the same sector, 11 came from universities, 10 came from political parties, and 10 were working as consultants inthe private sector. After the quarantine period (if applicable) or directly after finishing their terms o f employment, where does this highly skilled labor force go to continue their professional careers? Our survey showed that most directors (from a total of 42) went to either universities (10) or corresponding state or municipal secretariats (10) or to the private sector to work as consultants (7). The fewest number of former directors pursued career paths in private regulated companies (4 in the largest companies and 3 in others) or in old public companies in the same sector (2). Part of good governance is imposing a quarantine period for regulatory agencies' directors after their appointments are over. Of 21 agencies surveyed, 15 have quarantine rules for their directors; inall 15 cases, the quarantine was paid by the agency. 7.2. DECISION-MAKING 7.13. The majority of regulatory agencies reported that a qualified quorum was not needed for decision-making. Regulatory decisions are taken by the agencies' boards - a collegiate composed of directors or commissioners. Only 3 out of 21 surveyed regulators reportedto have limitations on the number of directors who must attend a valid deliberation meeting. The main voting rule used to make a decision i s simple majority. Only 2 agencies (ANATEL and ARTESP) rely mostly on qualified majority, while 2 other agencies (AGEPAN and CSPE) use each of the two rules for different types of decisions. Directors have veto power over the boards' decisions inonly 2 cases. 7.14. Case handling by commissioners or directors is not randomized, and it is relatively easy to obstruct a processonce it is assigned to a specific authority. Only 9 surveyed agencies distribute cases among directors in a random basis. Most agencies allocate cases among their directors on the basis of either their areas of specialization (14 agencies) or the decision of the director general (7 agencies). All of the latter cases are in state agencies, except for some decisions by ANATEL. Only in 2 instances neither a director nor the president could obstruct a case once it had been assigned to a given director. Inboth cases, the agencies are state regulators. 7.15. Although formal documentation of the decision and technical reasoning is needed, citing existent jurisprudence is not always required, undermining regulatory consistency over time. For 17 of the surveyed agencies, board action must be documented in writing, but in 3 agencies an oral discussion i s ~ u f f i c i e n tIn ~ . ~our sample of 21 Brazilian regulators, 18 require decisions made in the board meetings to be supported by technical analysis. Among the 18 agencies that are formally required to document their final decisions, only 7 regulators are additionally required to cite jurisprudence. 84 83Note that some of the questions refer to the legal requirement of certain actions, and others refer to actual behavior. The gap between the formal (de jure) rules and the actual (de facto) behaviors is addressed in section 4. 84 P-value for the difference inmeans was 0.10. 112 7.16. "Cheap talk" and "decision-rigging" are not always prohibited in the decision-making process. "Cheap talk" consists of informal exchanges of information between stakeholders and regulators. It weakens the regulatory process because it tends to bias the decision while reducing its procedural legitimacy. Only 3 (out of 18) agencies reported to possess a legal apparatus that prohibits informal meetings of directors with stakeholders. Decision-rigging consists of non-public agreements between board members--a process that undermines the legitimacy and transparency of formal decision meetings. Only in the case of 6 regulators are decisions taken without previous communication and discussion among boardmembers. 7.17. Survey results show a better picture in terms of public participation in regulatory decisions. In 17 agencies, external actors and those affected by the agencies' decisions are entitled to participate in the decision-making process. Participation varies: consumers, competitors, civil society, regulated firms, and other actors all actively take part. This participation is legally ensured in a majority of the agencies, mostly by the enabling regulation and by internal rules. It takes place mostly in public hearings (12 agencies) but also inboard meetings (8 agencies) and inthe case of technical analyses (7 agen~ies).~~ As an indicationof effectiveness, 15 regulators report that such participation has led to changes in the agencies' decisions. This is positive from the governance perspective, as participation is institutionalized and transparent. A summary of results i s presentedinTable 7.2.1, 7.18. Does Decision-Making Improve Over Time? Of the 21 questions in the decision-making section, only 2 exhibited a statistically significant difference between "older" and "younger" agencies. The first was the fact that "older" agencies tendedto use a qualified quorum for board decisions more frequently than the "younger" agencies. The second was that a greater share of "older" agencies publicly announces the issues to be discussed inboard meetings. The first result seems to indicate that less attention has been given to the decision-making rules when new agencies have been created. This is, however, an important issue, as the rules of decision-making strengthen technical coherence over time, consistency, and legitimacy. The second result may be interpreted as an indication that, over time, pressure to increase public participation in regulatory decisions - with obvious improvements in transparency - has generated positive outcomes. 85Note that these percentages need not add to 100 percent because more than one kind of participation occurs insome agencies. 113 b) 5 6 a1!! > ; cE ! E z F Y E F L: L I I; E Y 7.3. DECISION TOOLS 7.19. Out of the eight regulatory instruments considered, five were available for more than 60 percent of the sample: (a) tariff readjustment (90.5 percent), (b) methodologies for tariff revision (81.O percent), (c) monitoring quality (66.7 percent), (d) databases for regulatory accounting (66.7 percent), and (e) technical standards (61.9 percent). The other three instruments considered were available for fewer agencies inthe sample: (a) five-year revision (42.9 percent), (b) benchmarking (33.3 percent), and (c) methodology for defining interconnection tariffs (14.3 percent).86 After considering regulatory instruments, it i s important to check if agencies have the power and legal means (such as warning concessionaires, issuing fines and canceling concessions) to guarantee compliance to their decisions by regulated entities. The large majority (90.5 percent) has this legal apparatus available, of which 94.8 percent classify as "effective." 7.20. The picture presented by the survey regarding personnel-relatedissues is much less favorable. On average, only 20.7 percent of the agencies' workforce was admitted by public exams. Public exams are required for all civil servant positions in Brazil and represent stable employment. Temporary positions also exist, but these do not afford the same sort o f stability and benefits to the employee. Due to fiscal rigidities that limit the number of civil servants one can employ, several state organizations have to make do with contracting temporary staff, which in most cases has negative effects on governance. There i s a difference between the average percentage o f civil servants admitted through public exams infederal and state agencies: 26 percent inthe former and 18 percent inthe later, which i s statistically different with a p-value of 6 percent. 7.21. Agencies' salaries are also still considerablylower than those in comparable institutions.Considering the top technical and managerial positions, salaries offered by the agencies were considered by 90.4 percent o f the respondents to be at least 10 percent lower than the salary of the general attorney or the state finance secretary. In 2004, the government approved Law No. 10,871, which creates a career development plan and a salary structure specific to the staff of regulatory agencies. Even though the law represents progress in itself, increasing wage differentials with comparable careers within the Brazilian public sector (e.g. Gestor Pu'blico), seems to have caused a high level of desistance among approvedcandidates. 7.22. Not surprisingly,staff qualificationneeds significant improvement, given the complexity of functions performed. Out of the 14 agencies that provided data on this issue, 9 employed personnel with masters degrees, in which case they represented, on average, 4 percent o f the staff members in that category. Only 6 agencies had employees with PhDs, representing 2 percent of their staff. Approximately 95 percent o f the agencies' staff had undertaken short-term courses. On average, federal agencies have a better-educated staff than state agencies. And perhaps as an additional indication of the relatively low staff qualifications, the survey indicates that despite the fact that practically 86With respect to the instrument "defining interconnectiontariff," it is important to take into account that not every agency regulates sectors where interconnection is an issue. 115 all technical personnel (95.2 percent) had received some training, the overall turnover in the regulatory agencies i s extremely low. 7.23. Do agencies have the appropriate legal means to enforce the existing laws or contracts?The large majority (19 regulators) has this legal apparatus available, which 17 agencies classify as "effective." Inaddition, two-thirds o f the surveyed agencies have the power to solve conflicts between existing firms and otential entrants and to establish consumers' rights legislation in the regulated market. 8 Moreover, power was delegated to 15 agencies so that they would promote competition inthe regulated markets. Finally, 19 out of the 21 surveyed agencies have the power to financially punishthe actors in the regulated market that do not comply with contracts. 7.4. ACCOUNTABILITY 7.24. From a long list of institutions to which affected parties can appeal against a decision by a regulatory agency, the two main ones are the agency itself (95.2 percent) and the judiciary (85.7 percent). In 23.8 percent of the agencies, a case has reached the Supreme Court-not such a small number if one considers the short period of existence of most agencies and the relative probability of such an event. The results indicate that the Attorney General's Office (Ministe`rioPu'blico) and General Accounting Offices (Tribuniais de Conta du UnilZo/Estudo) do not play a significant role in the control of the agencies, although it i s possible that they may have a deterrent effect against opportunistic behavior by the agencies that creates little need for interference. The legislatures, on the other hand, exert some control over the agencies in 81.O percent of the cases. Of the four dimension analyzed, accountability was the one that demonstrated greater difference betweenolder and newer agencies. Additional Issues 7.25. A comparison between the federal and state regulatory agencies shows that, on average, federal regulatory agencies have scores 22 percent higher than state regulatory agencies in the autonomy component. This indicates that federal agencies are more insulated from the interests of regulated industries and show higher degrees o f financial and political autonomy. One of the factors behind this difference between federal and state regulatory agencies in the autonomy component is the fact that most state regulators oversee several industries - whereas federal regulators oversee a specific industry. 7.26. In addition, on average, federal regulatory agencies present scores in the decision-making component25 percent higher than state regulatory agencies.This is an indication that federal agencies are better endowed in the adoption of standard administrative procedures than state agencies. This mightbe related to the fact that, while state agencies differ significantly intheir institutional design, federal regulators present a highdegree of institutionalhomogeneity. *'Regulatoryagencies with powers associated with these two issues are not necessarily the same. 116 7.27. With respect to the decision tools component, on average, federal regulatory agencies present scores 26 percent higher than state regulatory agencies.This reveals that federal agencies have more developed and appropriate tools available for their use in the regulatory process. Tools include well developed regulatory accounting systems, adequate methodologies for tariff setting, the availability o f an appropriate mechanism for quality monitoring and more qualified and selected personnel. 7.28. Stakeholders affected by the regulatory process can appeal a decision made by the agency to the judiciary. In this sense, the judiciary reduces the agency's discretion, since any action that negatively affects one o f the parties involved in the process, consumers or the industry ingeneral, can be addressed by the court. Onthe other hand, judicial review of regulatory decisions takes place at the first level of the judiciary, which lacks relevant expertise and training, and can have a significant negative impact on the regulatory outcome. This interference by the judiciary i s important for the accountability and transparency of the regulatory process, but can also damage the credibility of the agency and increase regulatory risk if used improperly. 7.5. SUMMARY OFTHE RESULTS 7.29. Overall, the data show that considerable care was taken to follow "best- practices" in terms of good governance, but several aspects have been neglected.For instance, regulators are, in most of the cases, endowed with mechanisms that provide autonomy, formal delegation of relevant powers (such as the right to impose fines) and enabling legislation that defines the channels through which affected parties can participate in the regulatory process. One important neglected factor relates to the consolidation of rules and procedures that govern the decision-making process. The fact that the decision-making process must be formally documented in most agencies should weigh against the fact that jurisprudence does not have to be cited as part of the basis for the decision, increasing the degree of arbitrariness. Moreover, there are no legal impediments for board members to rig decisions prior to decision meetings inmost of the cases, and nothing precludes directors from participating in informal meetings or engaging inundocumented exchanges of information with stakeholders. 7.30. Another key aspect is the availability of appropriate means for decision- making, particularly well-qualified staff and regulatory tools. Most of the agencies reported salary levels that may be interpreted as non-competitive, and entrance through public exams for permanent positions was rather rare. This may create an adverse selection problem, with agencies unable to attract the most qualified personal, as indicated by the low share of staff with graduate studies. More sophisticated instruments, especially those related to economic regulation - such as benchmarking instruments and methodologies for the establishment of interconnection tariffs - are available for a much smaller number o f regulators. None of these problems seem to have been reduced over time. 117 8. CONCLUSIONSAND POLICY IMPLICATIONS Amid a shifting policymaking environment, this report addresses the question of how policymakers can revitalize infrastructure investment in Brazil. Given Brazil's infrastructure needs and its current fiscal dilemmas, the report focuses on how public policies may attract more and better private investment. It argues that Brazil needs to articulate a long-term strategy to transform infrastructure needs into private opportunities, which involves reducing the cost of capital and raising long-term returns o f concessions. The report, then, articulates a set of public initiatives aiming at revitalizing infrastructure investments inBrazil. The main conclusions andpolicy recommendations are as follows. 1. Infrastructure investmentsfell mainly due to the collapse of the institutional framework of the 1970sand not due to the sector reforms. Infrastructure investments in Brazil plunged dramatically in the 1980s and 199Os, a decline twice as large as that of the Latin American and the Caribbean (LAC) region. Most of this fall occurred during 1986-95, when the institutional framework for public investments in infrastructure completely deteriorated, and rising federal current expenditures further crowded out capital expenditures. Public financing for infrastructure investment, a central piece of that institutional framework, collapsed with a sequence of events that included the impact of the 1982 Mexican default on indebtedness levels of state-owned enterprises; the undervaluation of tariffs for inflation control purposes; the elimination of sector-specific federal taxes by the 1988 Constitution; and new priorities for public spending established by the same Constitution. Infrastructure reforms and fiscal adjustment followed these events and had relatively little impact on the overall decline. On the other hand, the 1990s raise of private financing for infrastructure was basically directed to asset transfer (divestiture) and not to the expansion of infrastructure stock. In other countries, such as Indonesia for example, most of the resources were applied to new (green-field) projects. While the fall in infrastructure investments was more or less common in the LAC region, the striking difference i s the role played by the private sector: in Chile, instance, private financing more than compensated for the fall in public expenditures since 1989. 2. Infrastructure investment needs to increase.. . Brazil's infrastructure stock compares well within the L A C region, but poorly with international peers from East Asia, for example. Access improved in the last decade, but 118 there were major gaps in rural areas and in access by the poor. Small and medium firms inlabor-intensiveindustries arethe most affectedby inappropriateinfrastructureservices. Although it i s not possible to claim that infrastructure i s a bidding constraint to higher sustainable growth rates in Brazil - especially when compared to high current expenditures and high levels and incidence of taxation - evidence shows that higher infrastructure investments may lead to higher growth rates and better social indicators. For instance, growth accounting exercises show that low capital deepening i s the main cause o f low growth rates while estimates of long-run elasticity of output to infrastructure investments vary between 0.5 and 0.6. Estimates for the elasticity of output to expenditures in maintenance are much higher (2.52). Also, this report estimates that the probability of an afro-descendant female child being enrolled in school in S6u Paulu - a state with one of the best infrastructure endowments in Brazil - increases by 38 percent when she has access to electricity, water, sanitation and telecommunication services. 3. ...andprivatefinancing should play a leadingrole. The report estimates annual expenditures of 3.2 percent of GDP per annum as a lower bound scenario for Brazil in 2010 (a cumulative figure that includes responding to increased demand; universal coverage and maintenance). A much higher amount (up to 9.0 percent of GDP) would be required to bring Brazil to the current levels of coverage in Korea. While ambitious, this effort - which would add more than 4 percentage points to the Brazil's GDP growth - i s not unrealistic: similar increases in infrastructure investments have been achieved by East Asian countries inthe 1970s and 1980s. Given the magnitudes of infrastructure needs and considering the fiscal dilemmas currently in place, focusing on improving the conditions for private participation in infrastructure in Brazil seems to be more appropriate than counting mainly on public financing. Creating fiscal space for public investments is, naturally very important, but the use of public funds should be restricted to those circumstances where social returns tend to be superior to private investments. In revitalizing infrastructure investments in Brazil, the primary objective of public authorities should be to enable more and better private investments in infrastructure. The country should avoid swinging the pendulum back to public financing. 4. Yet,private concessionshave not been adequatelyprofitable.. . This report estimated that the average internal rate of return for projects with terminal value in 1997-2003 was negative for telecommunications (-26 percent) and energy (-5 percent) - indicating returns below their opportunity cost o f capital - and positive for water (16 percent). The average return-on-equity for infrastructure services varied between3 percent (inwater) to 5 percent (intelecommunications). Inaddition, infrastructure investment inBrazil is a risky business. The level of risk of infrastructure investments in Brazil i s consistent with other L A C countries but it i s at odds with the OECD countries, where infrastructure investments are long-term, low- riskllow-return alternative for conservative investors. The level of risk becomes a 119 particularly important issue if Brazil intends to attract equity capital and illustrates the difficulties associated with attracting institutional investors, particularly pensionfunds. 5. ...exceptin the long-run, when thefull concessionperiod is taken into account. Infrastructure concessions inBrazil are capable of generating adequate returns only when the full concession period i s taken into account. Thus, enabling more and better private investments in infrastructure requires allowing investors to collect adequate dividends in the long-run. This implies, in turn, enforcing a stable and credible regulatory environment, turning private infrastructure concessions into a low-risk / low return business. In order to revitalize infrastructure investments, Brazil needs to enable private participation by diminishing the cost of private capital and raising the long-term returns of infrastructure concessions. More concretely, attracting back the private capital will require a coherent set of public initiatives aiming at reducing the risk-return ratio of projects, curbing regulatory risk, and improving cost recovery. 6. With the goal of attracting back theprivate sector, onefirst step is to create a stable and credibleregulatory environment. One challenge i s to eliminate the main legal and policy bottlenecks in infrastructure. In the power sector, unsolved legal loopholes, efficiency-related problems induced by the new regulation, and the stranded costs generated by recurrent changes in sector legislation are hindering private investment. The current legal framework for natural gas i s unable to generate supply contracts suitable for thermal plant operation in a predominantly hydro system and it was ineffective in curbing anticompetitive behavior, two needed steps to unleash opportunities for private investments inthe sector. In logistics, the five year delay of the second phase of the federal road concession program; the paralyzing of the non-truck highways decentralization process and the interruption of the port reform are factors deterring private investment in the sector. In water and sanitation, regardless of the solution for the "poder concedente" problem, Brazil will have to address the issues of economies of scale and agglomeration, appropriate regulatory design and the adequate level of tariffs if the vast majority of municipalities to be efficiently served, at least inpart, by the private sector. Despite recent progress, the remaining obstacles in environmental licensing should be reduced; the impact of regulatory revision by the courts on regulatory risk should be mitigated; and collaboration among public institutions involved in the concessioningl regulatory process (especially the roles o f Ministbrio Pu'blicoand Tribunal de Contas da Unilio)should be improved. 120 7. As a second step, Brazil needs to improve contract design to avoid "excessive" contract renegotiations. Recent history o f contract renegotiation of infrastructure concessions in Brazil shows evidence of "excessive" government-generated renegotiations. Brazil had a greater proportion of concession contracts renegotiated (41 percent) than the L A C region (30 percent). Roughly three-fourths of the renegotiations were initiated by the government, compared to one quarter in LAC. The average time until the first renegotiation occurs is also lower inBrazil. Renegotiations were initiated because of tariff revisions and changes ininvestment plans or needs, as opposed to the rest of the region, where renegotiations were largely due to macroeconomic crises. Contributing factors for renegotiations in Brazil included the inexistence o f an independent regulator, the fact that the regulatory framework i s embedded in a contract and not ina sector law, the use of price-cap as the tariff policy and the use at lowest tariff as the concession awarding criteria. Although contract renegotiations cannot be considered negative per se, the phenomenon in Brazil may be a symptom of poorly designed concession contracts and a cause o f inadequate risk-premium rates for infrastructure projects. The design of concession contracts i s additionally complicated by the definition o f concession objectives andrisk allocation. The 1999 exchange rate devaluation contributed for lower returns on infrastructure services in 1998-2003 as prices of telecommunications, energy, and water services in Brazil declined in real terms (when measured in U.S. dollars), contrary to most L A C countries. A second factor was poor enforcement of tariff policies. A third factor may be the persistence o f non-technical losses, particularly in the electricity and water sectors. Another important cause of the low profitability of infrastructure concessions was the highinvestmentlevels inthe initial years. Appropriate definitions of coverage objectives and allocation o f risk, particularly exchange rate and regulatory risks, are likely to continue to be a central issue inattracting private capital to infrastructure investments. The new Public-Private Partnership (PPP) Law provides for guarantees against the federal government's failure to comply with the financial obligations established by any PPP contract but not against regulatory risk. Preliminary discussions within the federal administration in 2005, considering the possibility of replacing the IGP-M by sector-specific cost-related indexes may be a good strategy to eliminate the implicit and imperfect hedging mechanism that existed in the contracts but does not solve the exchange rate risks problem - such strategy has already been implemented in the telecommunications sector in 2006, with the use of the sector- specific index (Telecommunications Services Index, The second phase of highway concessions will probably include awarding mechanisms that discourage InJanuary 2006, tariffs of telecommunications services began to be indexedby the IST, discounting the productivity factor calculated based on the rules set by regulatory agency (ANATEL), and price increases were effective inJuly 2006. The IST's methodology was designed to reflect the telecommunications sector- specific cost-related structure, being monthly published by ANATEL and its composition should be revised inevery two-year period 121 strategic bidding by possible operators but will not protect investors against regulatory risk and opportunistic behavior bythe public sector. 8. As a third step, Brazil needs to enhancethe effectiveness of regulators(improve regulatorygovernance). A survey of 21 regulatory agencies inBrazil, prepared for this report, shows that most of the general elements for good governance transferable by law seem to have been put in place. The challenge, then, i s related to the development of those more detailed attributes that could not be transferred by law, as well as to their effective enforcement. For instance, despite the fact that regulatory independency i s granted by almost all sector laws, more than half the regulators reported that the executive branch interfered in their final decision. Also, while regulators are formally required to document their decision- making process, few agencies are required to cite jurisprudence in support of their decisions or provide for legal sanctions against informal meetings between directors and stakeholders. Most of the general elements for good governance transferable by law were put in place and the challenge is related to the development of those more detailed attributes that could not be transferred by law, as well as to their effective enforcement. Initiatives to improve the state of regulatory governance include the draft Bill concerning regulatory agencies (Lei dus Agencius) and the Career Development law. The draft Lei dus Ag2ncius includes some important achievements, such as the transfer of the power to award concessions and the reassignment of planning policies to sector Ministries.*' The proposed introduction of management contracts between the agencies and the executive power may threaten regulators' autonomy and no major change was proposed interms of improving the detailed attributes of regulator^.^^ The approval of a career development law has enabled some public exams for new hiring but the existing salary structure and benefits seem to be inferior to other comparable careers within the public sector. As in many other countries, infrastructure regulators were created in Brazil with the objective of increasing the credibility of the government's long term commitment to respect concession rights. However, this objective can be fully accomplished only when the appropriate regulatory governance i s in place. Weak regulatory governance leads to inadequate implementation of regulatory rules and concession contracts, thereby influencing the return-risk ratio of infrastructure projects. 9. In sum, Brazil needs to forge a coherent strategy to enable more and better private investmentsin infrastructure. Given the magnitude o f the needs, as well as the constraints on reallocation o f public expenditures and expansion of public debt, the private sector must play a critical role. This does not mean that the role of the government should decrease, but rather that the country should avoid swinging the pendulum back to the pure public financing option. The challenge Brazil currently faces i s how to bring back the private sector - how to *'The draft Bill concerning regulatory agencies (Lei dus Agencius) was submitted to the House of Representatives inApril 2004 and has not been voted yet. For further details see Volume 11,Section 5.5. For a comprehensive review of the draft Lei dus Agencius see Correa and Pereira Net0 (2005). 122 translate infrastructure opportunities into projects with competitive rates of return. This requires curbing regulatory risk and raising projects' revenues, two tasks for which the role o f the public sector i s central: governments are ultimately responsible for regulatory risks, and through tariff policies, subsidies and related mechanisms can, directly or indirectly, influence projects' revenue. A second challenge i s to make sure that the benefits of private participation are transferred to consumers and the economy as a whole. 10. The strategy to reinvigorate the infrastructure sector in Brazil should encompassesthefollowing short and medium term components. A. Short-Term: e Strengthen thefundamentals for private participation in infrastructure. This will require completing the legal framework, improve regulatory governance and design better concession contracts. Major legal bottlenecks still exist inthe energy (natural gas); transport (port) and water and sanitation but more emphasis should be place on improving regulatory governance and contract design in other to avoid unnecessary renegotiations and reduce the risk o f capital expropriation. e Improve the quality of public expenditure in infrastructure. In particular, given the possibility open by the PPP contracts, it i s very important that a clear framework for the use of public resources ininfrastructure projects be articulated. When and how should be public money be used? The need to institutionalize the planning process; establish priorities and select. Clearly define the sector policies. Sector policies are an important part of the enabling environment for private participation in infrastructure. Sector policies also affect relative prices and may - or may not - enable private investments, as for example in the case of natural gas. Sector policies are also important to improve public expenditures, as for example inthe case o f the national road policy and its decentralization strategy. Properly allocate funds for road maintenance. First, it is important to establish a stable source of funds for the sector. Lack of funds for road maintenance impose major economic and fiscal burden, in this last case related to exponentially higher costs of (re)construction. In addition, renegotiations of contracts due to delays in disbursement, inevitably leads to higher costs. Second, it i s important to invest inpersonal and training, as output-based contract are cost-effectivebut difficult to prepare and implement. Look for creative mechanisms of public support to private investments. Besides improving the fundamentals for private participation and the quality of public expenditures, Brazil could consider the use of creative public mechanisms that support private investment. One example provided by this report is the use o f federal partial risk guarantees covering (municipal) regulatory risks in water and sanitation sectors, mimicking the partial risk guarantee mechanism that the World Bank offers, for example inPeru(see Box 6 for further discussion). 123 B. Medium-Tern: e Planning Brazil infrastructure needs. Defining an infrastructure strategy for Brazil at some point will involve discussing long-term projects, which i s not necessary consistent with political incentives. Also importantly, it will require consolidating a planning process that is still fragmented, either among ministries or between federal entities. This last aspect is particularly important given the decentralization of public funds to state andspecially municipalities. e Protecting infrastructure investments. The causes of the decline of infrastructure investments in Brazil are clearly associated to the classic problems of the provision of public good, where free-riding and coordination issues play a central role. These are certainly aggravatedby a political institutionalism that favors short-term local objectives in detriment of long-term national ones. The most obvious example is the impact "emendas" favoring parochial investments on the overall public budget. Another example i s capital expropriation: while it may increase votes in the short-term in one state it will almost surely raise the cost of capital inthe whole country inthe long-term. e Targeting the poor: a tool to reduce inequality. As part of this planning exercise, particular emphasis should be placed on the programs target to improve the access to the poor. Some consolidation of the existing programs may be also recommended, especially when the benefits of combined access to infrastructure on poverty and inequality reduction seems to behigh. Box 6. "Buying Credibility" throughPartialRiskGuarantees(PRG) against RegulatoryRisk Delegating regulatory functions to an independent institution addresses only partially the problem of creating the "credibility" that concession contracts will be enforced in the long-term and that the government will refrain from act opportunistically once investments are sunk. The problem is that regulators have to be credible themselves which may takes time to occur. The credibility o f regulators depend essentially on the state of regulatory governance - the degree of independence, adoption of appropriate decision-making rules; access to regulatory tools and staff; and accountability. In addition, incomplete legal and policy frameworks and poorly designed concession contracts may also affect the reputation of the regulator. One way governments could "buy credibility" in the context of stimulating private investments in infrastructure is offering a guarantee of risks under the control of government. On example of this class of public support to private investment i s the PPP provision of guarantees for non-payment of federal government's financial obligations under this type of contract. But even under pure concessionable projects, there are risks under the control of the government, broadly speaking: from a change insector law; to the disrespect of a contractual clause or a change insector regulation that clearly affects the returns to the investment. In order to cover for regulatory risk, all needed is a set of incentives in which who ultimately absorbers the risk o f regulatory failure is capable of sufficiently punish who causes it. By definition, the risk is limited to acertain amount of well-known states of nature and is therefore "partial." This structure of incentives may or may not involve an international institution such as the World Bank. Consider a potential role for the federal government in a hypothetical PPP by a Brazilian municipality. The private investor faces at least three risks: (i)the commercial risk of the project; (ii)the risk of non-payment from the municipality; and (iii)the regulatory risk, associated with the poder concedente issue and other explicitly expressed in the concession contract (such as tariff re-adjustment, revision rules; quality targets, investment plans, etc). One alternative to reduce the regulatory risk in this case is by federal guarantee for the contractually agreed regulatory commitments - which are fully 124 controlled by the municipality (and not for the poder concedeiite issue, which still requires further consolidation by the judiciary). The guarantee would be issued against the right to suspend federal funds once its use is triggered by the private investor. The fact that the municipality volunteers itself to such a deal would be, therefore, a powerful sign of regulatory commitment. Note that what is needed is a hierarchical financial relation between the institutions which takes the risk the institution that may affect it. A similar product is offered by the World Bank. The design and implementation of public infrastructure programs have traditionally been highly centralized inBrazil. Federal and state programs are generally prepared sector by sector and managed by federal ministries or state secretariats. As a result, there i s a multiplicity of public programs with little coordination among the three levels of government and even among the sector programs o f the same government. And a lack of spatial coordination among public sector programs has led to poor results on the ground and inefficient use of public resources. The impact of infrastructure reforms on that situation, except intelecom, was marginal. Inelectricity sector, the important coverage deficit is inrural areas. A range of different initiatives to promote rural electrification i s currently in place, with the federal government supporting the two major programs Luz no Campo, through Eletrobrds and Prodeem (by the Ministry of Mines and Energy). The Ministry of Agriculture has its own separate initiatives and a number of states have their rural electrification programs. Emphasis was put on grid expansion with very limited experience on off-grid electrification. Off grid electrification programs have failed, among other reasons, due to the absence of mechanisms for maintaining the systems. Regulatory reform (Law No. 10,438/2002) passed after privatization, together with other regulations developed by ANEEL, has introduced specific targets for distribution concessionaires with the aim of reaching universal coverage. Cross-subsidization was suggested as the financing instrument. Recently publishedtargets by ANEEL raised concerns inthe private sector as it would imply roughly US$ 1.8 billion of extra-expenses per year. This strategy has clear effects on tariff levels and on regulatory risk, as new obligations to distribution concessionaires are created after the contract i s signed (a possible example of government opportunism). Intelecommunications, the TelecomsSewices General UniversalizationPlan established by ANATEL previous to privatization, established targets for number of telephones installed, public phones etc, for the companies holding fix-phone service concessions. By meeting their targets before the schedule time, firms were free to enter the markets o f future competitors. Not surprisingly, most of targets established prior to privatization were met two years before the deadline set in the concession contracts: inJune 2002, the number of installed fixed lines reached 45.1 million, against 15.3 million in 1998. As a trade-off, excess capacity (unused lines) is close to 15% and the expansion of incumbents' market-share prior to competition may increase the likelihood o f abuse o f dominance, with possible impacts on future market performance. Moreover, the universalization fund (FUST) is completely ~nderutilized.~' 91See Pinheiro (2003). 125 Recent analytical work inBrazil and elsewhere emphasize the need to address these issues thro~gh:'~ 0 The consolidation and integration of the many federal and state programs for local and regional development, harmonizing objectives, criteria and incentives; An integrated, multi-sector strategy to improve the delivery of basic infrastructure services while also strengthening ecosystem and water resources management, improving education, establishing effective social safety nets, and reforming land, labor, and capital markets; 0 Greater spatial focus and coordination o f programs, concentrating infrastructure investments in areas with clear potential for environmentally sustainable development and building on rural-urban and public-private synergies; Decentralization coupled with institutional strengthening at the local level and appropriate mechanisms for community participation, all aimed at ensuring effective delivery o f public infrastructure services; and, 0 Output-based rather than input-based funding of local infrastructure programs, to provide incentives for efficiency and results on the ground. Improving infrastructure services for poor people requires urgent attention. A sectoral approach may be warranted, to ensure a coherent strategy for addressing the issues of subsidies, cost recovery, rural coverage, access, willingness to pay, service delivery mechanisms, regulation, and, most important, sustainability. Options to be explored include introducing subsidies (cross-subsidies or targeted subsidies), using competition based on the lowest required subsidy in allocating rights and obligations to provide new coverage, segmenting services by quality and type, and allowing exclusivity rights inun- served areas. 0 A rolefor competition? Government should also consider a role for competition in infrastructure sectors. Risks associated with horizontal concentration exist in power generation (electricity and natural gas) and railways. Most importantly, perhaps, abuse o f dominance due to verticalization (control of an "essential facility") i s becoming frequent in telecommunications, natural gas, port, and railways sectors. Sometimes the legal means for mitigation are inplace. Inall cases, however, enforcement i s cumbersome. Economic research also suggests that the performance of infrastructure industries i s positively affected by the extent to which competition is introduced in vor) the market.93 This may be particularly true indeveloping countries where institutional endowments are less conducive to institution-intensive arrangements (as regulation). Although recommendable in principle, competition in developing countries may be less feasible due to small market size in a context where minimum efficient scale is still large. Yet, 92See Estache, Foster, and Wodon (2002). 93See Bouin and Michalet (1991); Kwoka (1996); Kleit and Terrell(2001); and Banget al. (2002). 126 competition i s not naturally born after privatization but depend on design of concessiodprivatization design; and on the enforcement o f specific regulation (access to natural monopoly facilities and antitrust law) which inturn are also institution-intensive. I s open access legislation present and effective (are regulators prepared to enforce it)?I s there enough competition in(or for) the market to induce efficient-outcomes? I s abuse of dominance likely to reduce competition? I s antitrust law applicable? I s the antitrust agency effective? These are some of the questions to be raised. Despite significant progress inprivatization, competition infixed telecommunications has shown slower progress. Innone of the Latin American countries, new entrants were able to gain more than 15 percent of the market: even in Chile, usually considered the main reformer in the region, new entrant's share after almost 20 years o f market liberalization was slightly less than 15 percent. Results in developed economies are not much better, reflecting difficulties inregulating access to the local loop ("last" mile) and market power from incumbent firms. Data on fixed telecommunications, however, do not reflect competition among technologies: thanks to technological change and convergence, not only private participation i s greater in related segments (such as mobile telecommunications, cable and internet), but substitution for these segments i s increasingly possible. By 1999, a large number o f Latin American countries - including Panama, Paraguay, and Venezuela - had the number o f mobile subscribers being larger thanthat of fixed-line subscribers. Unlike fixed-line services, the mobile industry tended to face competition since its early periods of liberalization: half of liberalizing countries in the region had licensed a second mobile operator by 1999. Despite lower required fixed investments, competition in the mobile segment also depends on institutional features, such as mobile number portability and technical standards. 127 REFERENCES Abicalil, M.T. (2002). "Setor de Abastecimento de Agua e Esgotamento Sanitdrio no Brasil: Fontes de Financiamentode Investimentoe outros Temas Relacionados." Mimeo. Afonso, J.R.R., E.A. Aratijo, and G. Biassoto Junior (2005). "Fiscal Space and Public Sector InvestmentsinInfrastructures: A Brazilian Case-Study." IPEA, Textopara DiscusslioNo. 1139.Rio de Janeiro: IPEA. AMCHAM (American Chamber of Commerce-Sio Paulo Charter). 2003a. Relatdrio sobre a ag2ncianacional de telecomunicaC6es.Mimeo. .2003b.Relatdrio sobreaAg2nciaNacional deEnergia Ele`trica.Mimeo. .2004a. Relatdrio sobre aAggnciaNacional de Telecomunicapjes.Mimeo. , 2004b. Relatdrio sobre aAg2ncia Nacional de Energia Ele`trica.Mimeo. , 2005a. Relatdrio sobre aAg2nciaNacional de Telecomunicag6es.Mimeo. .2005b. Relatdrio sobre a Ag2nciaNacional de Energia Ele`trica.Mimeo. .200%. Relatdrio sobre aAg2nciaNacional deVvigildncia Sanita`ria.Mimeo. Andres, L.,V. Foster and J.L. Guasch (2006). "The Impact of Privatization in Firms in the InfrastructureSector inLatin American Countries." World Bank, Washington D.C. Processed. Bacha, E.L. and R. Bonelli (2004). "Accounting for Brazil's growth experience- 1940- 2002." IPEA, TextoparaDiscusslioNo. 1018.Rio de Janeiro: IPEA.. Berg, S. (1998). "Lessons in Electricity Market Reforms: Regulatory Processes and Performance." Electricity Journal (June), pp. 13-20. Boardman, A.E. and A.R. Vining (1989). "Ownership and Performance in Competitive Environments: A Comparison of Private, Mixed and State-Owned Enterprises." Journal of Law and Economics 32, pp.1-33. Bonelli, R. (2005). "Economic growth and productivity change in Brazil." Document prepared for the World Bank (2005). "Brazil Investment Climate Assessment." Mimeo. Bouin, 0. and C.A. Michalet (1991). Rebalancing the Public and Private Sectors: Developing CountriesExperience,OECD: Paris. 128 Brenneman, A. and M. Kerf (2002). "Infrastructure and Poverty Linkages: A Literature Review." Mimeo. Briceiio-Garmendia, C., A. Estache, and N. Shafik (2004). "Infrastructure services in developing countries: Access, quality, costs and policy reform." World Bank PolicyResearchWorking Paper, WPS3468. Brown, A. and E. de Paula (2004). Strengthening of the Institutional and Regulatory Structure of the Brazilian Power Sector. PPIAF. The World Bank: Washington D.C. Calderon, C., W. Easterly, and, L.Serven(2003). "Infrastructure CompressionandPublic Sector Solvency." In: Easterly, W. and L. Serven, The Limits of Stabilization: Infrastructure, Public Deficits, and Growth in Latin America. Calderon, C. and L. Serven (2004). "The Effects of Infrastructure Development on Growthand IncomeDistribution.'' World BankPolicy ResearchPaper, WPS3400. Camargo, J.M.A.G. (2002). "Missing the Target: Assessing Social Expenditures in Brazil." TheBrown Journal of WorldAfsairs, vol. 3(2). Castro, N. and P. Lamy (1994): A Reforma e a ModernizaGGo do Setor de Transporte FerroviAriode Carga. IPEA Text0para Discuss60 No. 339. Rio de Janeiro: PEA. Chandley, J.; W. Hogan and M. Morozowski-Filho (2004). "Alternative Approaches to Transmission Constraints And Congestion Pricing." PPIAF. The World Bank: WashingtonD.C. Correa, P.G., C. Pereira, B. Mueller, and M. Melo (2006). "Regulatory Governance in Infrastructure Industries- Assessment and Measurement of BrazilianRegulators." PPIAF Trend and Policy Options No. 3. Devarajan, S., V. Swaroop, and H.Zou (1996). "The Compositionof Public Expenditure andEconomicGrowth." Journal of MonetaryEconomics, Vol. 37, pp.313-344; Escribano, A., J.L. Guasch, L. Garrido, N.Peltier, and H. Singh (2005). "The Impact of Infrastructure on Competitiveness: A firm Level Analysis Bases on ICA Surveys." Washington, DC. June 6-7, 205. The World Bank and The Inter- AmericanDevelopmentBank. Estache, A., V. Foster, and Q. Wodon (2002). "Accounting for Poverty in Infrastructure Reform: Learning from Latin America's Experience." World Bank Institute DevelopmentStudies, Washington, D.C. Fay, M.and T. Yepes (2003). "Investing in Infrastructure: What is needed from 2000 to 2010?" World BankPolicyResearchWorking Paper, WPS3102. 129 Fay, M., L.Cohan, and K. (2005). "Public Social Safety Nets and the Urban Poor." In: M.Fay (ed.), The Urban Poor in Latin America. Washington, D.C.: The World Bank, pp. 239-66. Fay, M.andM.Morrison (2005). "Infrastructure inLatinAmerica: RecentDevelopments and Key Challenges." The World Bank, WashingtonD.C. Ferreira, C.K.L. (2000). "Privatizing the Electric Power Sector in Brazil." In Pinheiro, A.C. and K. Fukasaku (2000). Privatization in Brazil: The Case of Public Utilities.Rio de Janeiro: BNDEWOECD. Ferreira, P.C. and T.G. Malliagros (1998). "Impactos Produtivos da Infra-estrutura no Brasil: 1950-95." Pesquisae Planejamento Economico, 28(2), August. Fonseca,V.I. andF.D.Por (2005). "Ecossistemas." Available at the Brazilian Ministry of ForeignAffairs website http://www.mre.gov.br. Galiani, S., P. Gertler, and E. Schargrodsky (2005). "Water for Life: The impact of Privatization of Water Services on Child Mortality." Journal of Political Economy 113, pp.83-120. Gannon, C. and Z. Liu (1997). "Poverty and Transport." TWU Discussion Paper 30. World Bank, WashingtonDC. Glomm, G. andF.Rioja (2004)."Fiscal Policy andLong-RunGrowth inBrazil." Mimeo. Gramlich, E. M. (1994). "Infrastructure Investment: A Review Essay." Journal of Economic Literature,no. 32, pp. 1176-96. Guasch, J.L. (2004). Grantingand Renegotiating Infrastructure Concessions-Doing it Right. The World Bank Institute-DevelopmentStudies:Washington, D.C. Guasch, J.L. and P. Spiller (1999). Managing the Regulatory Process: Design, Concepts, Issues and the Latin American and Caribbean Story. Washington DC: World Bank. Gutierrez, L.H. 2003. "Regulatory Governance in the Latin American Telecommunications.'' UtilitiesPolicy 11(4), pp. 225-40. Henisz, W.J. (2002). "The Institutional Environment for Infrastructure Investment." Industrial and Corporate Change 11(2), pp.355-89. Henisz, W.J. and B.A. Zelner (2001). "The Institutional Environment for Telecommunications Investment." Journal of Economics and Management Strategy 10(l), pp. 123-47. Holder, S. and J. Stern (1999). "Regulatory Governance: Criteria for Assessing the Performanceof RegulatorySystems." Utility Policy 8(l), 33-50. pp. 130 Hulten,C.R. (2006). "Infrastructure CapitalandEconomicGrowth: How Well you Use it may be more Important than how much you Have." Working Paper No. 5847, National Bureauof Economic Research(December 1996). IDEC (Instituto Brasileiro de Defesado Consumidor) (2002). A Prote@o ao Consumidor de Sewigos Pu'blicos. SZo Paulo, SP: Max Limonad. Jacoby, H.G. (2000) "Access to Markets andthe Benefits of Rural Roads." The Economic Journal 110, pp. 713-37. Kalaitzidakis, P. and S. Kalyvitis (2004). "On the Macroeconomic Implications of Maintenance in Public Capital." Journal of Public Economics, vol. 88(3-4), pp.695-712. Kleit, A. andD.Terrell(2001). "Measuring PotentialEfficiency Gains from Deregulation of Electricity Generation: A Bayesian Approach." Review of Economics and Statistics 83, 523-30. Kwoka, J.E. (1996). Power Structure, Ownership, Integration and Competition in the US Electricity Industry. Boston: Kluwer. Lafont, J. and J. Tirole (2000). Competition in Telecommunications. Cambridge, MA: MIT Press. Levy, B. and P.T. Spiller (1994). "The Institutional Foundations of Regulatory Commitment: A Comparative Analysis of Telecommunications Regulation." Journal of Law, Economics, and Organization 10(2), pp. 201-46. Marques,S.A. (1996). "PrivatizaGBo do SistemaFerroviirio Brasileiro." IPEA Textopara Discuss60 No. 434. Rio de Janeiro: PEA. Maurer, L.(2001). The Virtuous Circle of Contracting. Rio de Janeiro:Abraceelhstitute of the Americas. Megginson, W.L., R.C. Nash, and M. van Randenborgh (1994). "The Financial and Operating Performance of Newly Privatized Firms: An International Empirical Analysis." Journal of Finance 49 (2), pp. 403-52. Mendonca, M.J.C. and R. Seroa da Motta (2005). "Sabde e Saneamento no Brasil." IPEA, Textopara DiscussczbNo. 1081.Rio de Janeiro: IPEA. Neri, M. (2001). "Gatos: de inimigos a s6cios do racionamento." Rio de Janeiro: Valor EconGmico.June 12*. Neri, M.andR. Moura (2005). "Infrastructure andEducationalProgression." Mimeo. 131 Pessoa, S.A.; V. Gomes; and F. Veloso (2003). "Evoluqiio da produtividade total dos fatores na economia brasileira: Uma antilise comparativa." Ensaios Econdmicos No. 483. Rio de Janeiro:FundaqiioGettilio Vargas. Pinheiro, A.C. (2003). "Is Institutional Reform the Key to Brazil's Accelerated Development?"Mimeo. Pinheiro, A.C., I.Gill, L.Serven, and M.Thomas (2001). "Brazilian Economic Growth, 1900-2000:Lessons and Policy Implications." In:World Bank (2001). "The New GrowthAgenda." Pires, J.C. and M.S. Piccinini (1988). "Mecanismos de Regulamentaqiio Tarifiiria do Setor ElCtrico: a Experiencia Internacional e o Caso Brasileiro." Texto Para Discussiio do BNDES No. 64.Rio de Janeiro: BNDES. Prudhomme, R. (2004). "Infrastructure and Development." Washington DC, Paper preparedfor the ABCDE(Annua1Bank Conference on Development Economics), May 3-5. Rioja, F. (2003) "The Penalties of Inefficient Public Infrastructure." Review of Development Economics, February2003, Vol. 7(1), pp.127-37. Rocha, K., G.F. Braganqa, and F. Camacho (2006). "Remuneraqiio de Capital das Distribuidorasde EnergiaElCtrica:Uma Antilise Comparativa." IPEA Textopara Discussiio No. 1153.Rio de Janeiro: IPEA. Schauer, D. (1989). "Is Public ExpenditureProductive?"Journal of Monetary Economics 23, pp. 177-200; and, Holtz-Eakin, D. 1994. "Public Sector Capital and the ProductivityPuzzle." TheReview of Economics and Statistics 12-21. Seroa da Motta, R. "As Opq6es de MarcoRegulat6riode Saneamento no Brasil." Revista Plenarium No. 3. Forthcoming. Seroa da Motta, R. and A.R.B. Moreira (2004). "Efficiency and Regulation in the Sanitation Sector in Brazil." IPEA, Texto para Discussiio No. 1059. Rio de Janeiro: IPEA. Sirtaine, S., M.E. Pinglo, J.L. Guasch, and V. Foster (2005). "How Profitable are InfrastructureConcessionsinLatinAmerica?Empirical Evidence and Regulatory Implications." The World Bank:WashingtonD.C. Processed. Smith, W. (1997). "Utility Regulators- The IndependenceDebate." InThePrivate Sector in Infrastructure: Strategy, Regulation and Risks. Washington, DC: The World Bank/The InternationalForumfor Utility Regulation. Spiller, P.T. and E. Tiller (1997). "Decision Costs and the Strategic Design of Administrative Process and Judicial Review." Journal of Legal Studies 26(2), pp. 347-70. 132 World Bank (1999). "Brazil - Regulations for Better Water and Sanitation Services." ReportNo. 19568-BR. World Bank (2004) "Brazil First Programmatic Loan for Sustainable and Equitable Growth." ReportNo. 27507. World Bank (2005a). "Brazil Investment Climate Assessment - Volume 11: Background Documents."Mimeo. World Bank (2005b). "Mexico - Infrastructure Public ExpenditureReview." Report No. P089103. World Bank (2005~)."Infrastructure in Latin America: Recent Developments and Key Challenges." ReportNo. 32640-LCR. World Bank (2005d). "Models of Aggregation for Water and Sanitation Provision." ReportNo. 32030. World Bank (2006) "Project Appraisal Documentfor the BrazilRoadSwap." Report No. 32005-BR. Zhang, Y., D. Parker, and C. Kirkpatrick (2002). "Competition, Regulation and Privatisation of Electricity Generation in Developing Countries: Does the Sequencing of the Reforms Matter?" Centre on Regulation and Competition, Working Paper No. 62. 133 ANNEX 1 The Impact of Access to Infrastructure on EducationPerformance inBrazil94 1.OBJETIVO Este trabalho tem como objetivo dar umprimeiro passo no sentido de avaliar impactos educacionais de mudanqas no acesso B infra-estrutura a partir das bases de microdados oriundas da PNAD/IBGEe do SAEB/MEC.A dificuldade central destas anhlises correlaqdes brutas e as parciais entre as varihveis de infraestrutura e as de desempenho escolar para fins de desenho de politicas C a determinaqiio da direGiio de causalidade entre varihveis ex6genas e as end6genas. Discutimos ao fim extensdes em torno de anhlises de diferenqa em diferenqa deste impactos tomando como base a ocorrencia de experiencias onde algumas comunidades siio beneficiadascom aumento da oferta de infraestrutura e outras niio. 0presenteprojeto propdedesenvolver sistema de provisiio de informaqiio interativos e amighveis, tais como simuladores de probabilidades desenvolvidos a partir de modelos logisticos (e.g., freqiiencia escolar), alCm de simuladores de regressdes de varihveis em nivel (e.g., proficiencia escolar) e banco de dados disponibilizados na internet. Os indicadores sociais, portanto, a serem avaliados siio: matricula, freqiiencia (assiduidade), progress0 escolar (repet2ncia) e proficiencia. As varihveis explicativas a serem consideradas siio de infra-estrutura, tais como: 1. Acesso a serviqos ptiblicos: abastecimento de hgua, esgotamento sanithrio e iluminaqiio elCtrica e comunicaqdes. 2. Infra-estrutura privada: ncimero de pessoas no domicilio, direito de propriedade. 2. METODOLOGIA Os principais elementos desterelat6rio consistem na avaliaqiio das correlaqties entre acesso B infra-estrutura ptiblica e privada conforme indicadores detalhados anteriormente e outros indicadoresde desempenho escolar tais como matricula, frequencia, repetencia e proficiencia escolar a partir de tCcnicas micro-economktricas. A metodologiada pesquisa consiste na geraqiio, descriqiio e anhlise de umconjunto de base de dados. Utilizar-se-h de pesquisa domiciliar tradicional como a PNAD, que nos permite uma avaliaqiio detalhada das correla$des entre as condiqdes de acesso B infra- estrutura domCstica em algumas varihveis sociais. Os efeitos sociais da provisiio de infra-estrutura tambCm sera captado atravks de informaqdes do MinistCrio da Educaciio por meio de avaliaqb de desempenho do SAEB. Nesse caso, estimar-se-h a equaGiio de notas dos alunos, enfatizando os efeitos de varidveis ligadas ao acesso B infra-estrutura. Os resultados das regressdesseriio apresentadasem formato que permita a identificaGiio imediata do impacto de cada varihvel. Todos os modelos empiricos gerados seriio convertidos em ferramentas interativas e amighveis a fim de aumentar a capacidade de difusiio dos resultados da pesquisa. 94Background paper by Marcel0 Neri (Centro de Politicas Sociais do IBRE e da EPGE) and Rodrigo Moura(Doutorandoda EPGE). 134 2.1. TCcnicas An6lise Bivariada 0 objetivo da anhlise bivariada C trapr um perfil da estrutura de correla@es entre as varihveis de infra-estrutura e as de desempenho escolar, analisando o papel de cada atributo tomado isoladamente. Isto C, desconsideramos possiveis inter-relagdes das varihveis de desempenho com outras "varihveis explicativas." AnhliseMultivariada A anhlise multivariada visa proporcionar um experiment0 melhor controlado que a anhlise bivariada. Seu objetivo C captar o paddo de correlap5es parciais entre as varihveis de interesse e as varihveis explicativas. N a anhlise multivariada captamos as correlap3es das varihveis de acesso B infra-estrutura com varihveis de retorno social citadas mantendo as demais varihveis constantes. Trabalharemos com duas variantes do modelo de regress50 multivariada: a) Regressgo em minimos quadrhticos ordiniuios para varihveis continuas; b) Regressgo logistica binomial multinomialenvolvendo como end6genas duas ou mais categorias de varihveis discretas. 2.2. BasesdeDados SistemaNacional de Avalia@o da Educa@o B6sica(SAEB) 0 Sistema Nacional de AvaliagBo da Educaggo Bhsica (SAEB) avalia os alunos brasileiros da 4a. e da 8a. sCrie do ensino fundamental e da 3a. sCrie do ensino mCdio em Lingua Portuguesa e Matemhtica. AlCm das provas das duas disciplinas, questionhios coletaram informagies sobre as caracteristicas da escola, do diretor, do professor, da turma e dos alunos. Infra-estrutura est6 incluida em varihveis de backgound familiar. Especificamente, estudaremos nessa base de dados, o desempenho escolar mCdio dos alunos da 4" e 8" sCrie do ensino fundamental e 3" sCrie do ensino mCdio medido atraves de proficiencia e de atraso escolar fuq5o das seguintes varihveis: tem eletricidade, tern hgua encanada. PesquisaNacional por Amostra de Domicilios(PNAD) A pesquisa abrange todo o Brasil, exceto as iueas rurais de Acre, Amaph, Amazonas, Rondhia, Roraima e Par& ContCm informaqdes sobre diversas caracteristicas demogrhficas e s6cio-econ8micas da popula$io. Especificamente, utilizaremos as seguintes varihveis: i. Educacionais: sabe ler e escrever, frequenta escola ou creche, sCrie e espCcie de curso que frequenta; ii. ServiFos pablicos: hgua proveniente de rede geral de distribui@o, esgotamento sanitiuio (rede geral), iluminaqgo elCtrica. 2.3 Modelo de anllise 0 desempenho dos alunos pode ser afetado pel0 acesso a serviGos pdblicos como luz, hgua etc. Melhor infra-estrutura, tanto na casa como na escola, deve melhorar a produtividade dos estudantes, reduzindo portanto, a repetencia. Mas a questgo que se 135 levanta aqui C que outros fatores podem afetar estas varihveis de interesse. Assim, os controles assumidos ajudam a isolar de forma mais precisa o efeito da infra-estrutura no desempenho escolar. Assim, por exemplo, a educaggo dos pais pode afetar positivamente a produtividade de seus filhos, mesmo que haja certa deficiencia em infra-estrutura na escola. AlCm disso, o n6mero de moradores na casa do aluno pode tambCm influenciar. Temos observado que hh geralmente um n6mero 6timo de moradores que ajudam a melhorar a performance escolar do estudante. AlCm disso, controlamos tambCm por sexo, cor e Unidade Federativa (UF), visto que as caracteristicas escolares podem divergir entre esses grupos. Assim, os modelos a serem estimados se baseiam naequasgo geral: Y I =W P ) em que, y~C a varihvel dependente, que sgo os indicadores de desempenho escolar, para os quais serh estimada regressaes separadamente; x C o vetor dos controles, tal que x=(IND; FAM;ESC; INF;ENV).INDC o sub-vetor que contCm as caracteristicas dos individuos, FAMas caracteristicas da familia, ESC da escola, INFas varihveis de infra-estruturareferenteamoradia, eENV inclui alguma externalidade (por exemplo, favela). A seguir, discutimos brevemente o modelo para cada umdos indicadores, apontando para potenciais e problemas e suas possiveis solugies. Profeiincia As regressaes para a equasgo de proficiencia foram estimadas por sCrie. 0 modelo estimado foi um modelo linear nas co-varihveis. Estimamos a regressgo por minimos quadrados ordinkios, mas consideramos para a hip6tese de heterocedasticidade dos erros. Assim, estimamos a matriz de variiincia atraves do estimador de White (1980) produzindo erros padraes robustos a heterocedasticidade. ReprovaGcZb e assiduidade Para as variiveis independentes reprovasgo e assiduidade utilizamos um modelo logit multinomial, sendo que o valor base de cornparasgo C nunca ter reprovado e nunca ter faltado B escola. 0 modelo estimado por maxima verossimilhansaC o que segue: P(y = j x) =exp(xbj)/ 1 Cexp(xflj) I [ + 1 J ,j =1,2, J =2 em que, y C a varihvel reprovasgo, o vetor pj C o conjunto de parhetros para j=1 h=l (reprovou uma vez) e j=2(reprovou 2 ou mais vezes). Como as probabilidades devem somar um, temos que ter: Assim, sempre comparamos em relaggo ?i base (y=O) que C nunca reprovou. Podemos simplificar a notasgo da probabilidade de resposta como: Pj (x,P>= P(Y= j I P&P) = JYY =01x1 136 Deve-se ressaltar que a interpretaqlo da magnitude dos parametros estimados deste modelo n l o C direta. 0 efeito marginal decorrente de uma mudanqa em uma variivel controle continua 6: aPj ("9P>= Pjk exp(xPj +ci=lexp(xPh1)- exp(xPj +ci=lPhk exp(xPh1) (lci=lexp("Ph + )y AlCm disso, atravCs da raziio das probabilidadesemrelaqlo B basetemos: ou ainda: l o g bj ("9 P>/ Po P)l="Pj ("7 Ou seja, temos uma interpretaqlo mais direta de uma variaqlo de uma unidade em x, que mostra o quanto varia o log da razgo das probabilidades (log-odds), atravCs do parametro estimado. Matricula e atraso Para a variivel matricula e atraso, estimamos um logit por mixima verossimilhanqa. 0 modelo estimado aqui C um cas0 particular do logit multinomial para quando J=2, ou seja: Pr(y =1 x)= I exP("P) em que, y C a matricula, x C o vetor dos controles e p o vetor dos parhetros. 0 efeito 1+exp(xP> parcial deste modelo para umamudanqa em uma variivel continua 6: aR(Y =1 x) - exp(xP> I axj - (1 exp(xP>)2 + Pj=Pr(y=lIx) 1 exp(xP> + Assim, o sinal do efeito parcial C dado pel0 sinal do parhetro. Se xk for discreta, o efeito parcial C obtido mudando xk de um valor para outro, ceteris paribus, e calculando a diferenqa da probabilidade. N o cas0 de x ser varitivel biniriateriamos: - exp(P1+Pzx2+*..+Pk-1"k-I) 1 exp(P1 +P2x2+ +P +exp(P1+P2"2 + * * * + f l k " k > I * - k X k +exp(P, +P2.2 +"'+ flk-l'kil) E conhecendo o sinal de Pk, C suficiente para determinar o sinal do efeito parcial. Mas deve-se ressaltar que para encontrar a magnitude do efeito dever-se-ia estimar as equaqaes acima. 2.4 Apresentaqfio dos dados Nesta seqlo apresentamos os dados que i r b alimentar o modelo e tambCm na anilise bivariada. Os dados foram retirados da Pesquisa Nacional por Amostra e Domicilios (PNAD) e do Sistema Nacional de AvaliaqBo da EducaqBo Bisica (SAEB). Emrelaqso ao SAEB, os dados estlo agrupados para alunos, professores, diretores e escola. 137 Utilizamos informaqjes das bases dos alunos e da escola, e unimos estes dados no nivel do aluno. A PNAD divide sua base para pessoas e domicilios. As variiveis de infra- estrutura sfio extraidas da parte de domicilios. Assim, juntamos os dados no nivel das pessoas. Para a proficihcia e reprova@o, as variiveis independentes utilizadas, respectivamente, foram, segundo os grupos definidos: - IND: sex0 (0 = homem e 1= mulher); dummies para o atributo (d-cor1 =branca d_cor2=parda d_cor3=negra d_cor4=amarela d_cor5=indigena) e idade. - FAM:educmae (l=nunca estudou 2=<4serie 3=4s 4=<8s 5=8s 6=serie =4 ......................................... Quando voci? comeqou a estudar? N o maternal N a Prt-escola N a l a sCrie N a 2a strie N a 3a strie -> serie = 8 --_-__-___----___-_-____________________- Quando voce comeqou a estudar? 147 2509924 619,627 29,040.50 12,557.60 ->serie = 11 1646444 423,162 19,859.50 7,693.49 Reprovapio 0 indice de reprova@o para escolas urbanas C bem menor do que as rurais, principalmente para as rurais. Vale ressaltar que, em 2003, existem apenas escolas rurais da 4" sCrie do EF. 2003 local urbana rural 0 indice de reprova@o para todas as skries C bem maior nas escolas p6blicas do que privadas, apesar de que, de 2001 para 2003 as reprovapjes cairam bastante na rede p~blicade ensino. 1999 rede mtdia(reprova) N(reprova) 2001 rede mtdia(reprova) N(reprova) 148 mCdia(reprova) -------_-_---_-- N(reprova) _----_------_ 0.5136786 7855772 0,1906985 1114717 A mensagem da tabela abaixo diz que quanto mais cedo as criangas comeprem a estudar menor serii o indice de repetencia. 0 tinico resultado ambiguo apresentado C que quando se pergunta para alunos da 8" sCrie do EF e da 3" sCrie do EM, o indice de reprovaG8omaior para os que comepram na 2" sCrie do EFe n80 na 3" sCrie. -> serie =4 Quando VOCEcomeCoua estudar? -_--__---------- mtdia(reprova) N(reprova) ____--_-__-__ 0.2830364 970,206 0.3943322 1821329 0.5875055 800,552 0.6996055 41,956.50 0.7794674 39,017.10 3.2 Aniilise Bivariada de Matricula, Atraso e Assiduidade A aniilise bivariada estende-se agora h matricula, atraso e assiduidade. As variiiveis de infra-estrutura encontradas na PNAD diferem um pouco das encontradas no SAEB, vistas na aniilise precedente. Aqui, restringimos a popula@o de interesse h faixa etkia de 5 h 17 anos que est80 na "idade escolar ativa". N3o consideramos todas pessoas, pois 149 estimativas iniciais se mostraram desconformes ou pouco robustas devido ao viCs amostral"'. Os valores para estas variiveis siio: matricula e atraso (0= niio e l=sim), e assiduidade (O=nunca faltou B escola, l=faltou pouco - 1a 5 dias, 2= faltou muito -6 ou mais dias). ServiGosPdblicos Matricula Emrela@o aos serviGos pdblicos, notamos que o acesso B igua gera maiores indices de matricula. E a luz recebida na forma elktrica tambem incentiva B freqiiencia escolar. AlCm disso, no context0 dinhico, a mCdia de matricula escolar tem crescido entre aqueles que tem acesso B igua e luz elCtrica, ao longo de todos anos. mCdia(frequenta) _-__---_----_-__ ____---____--_- N(frequenta) 0.8176481 4959570 0.8898703 22000000 mCdia(frequenta) _---___-------_- N(frequenta) --_-_________-- 0.8820219 25200000 0.8047988 1682280 0.701266 90522 N(frequenta) 7790659 35100000 luz N(frequenta) --_-_----___-- elCtrica 40400000 deo, queroseneou gis de botijzo 2287583 outra forma 182286 elCtrica deo, querosene ou g6s de botijb lo'Nas anilises bivariadas iniciais observou-se diferenga praticamente nula da frequhcia mCdia entre aqueles que, por exemplo, tinham acesso a Bgua em reIagBo aos que nlo tinham; ou ainda uma media maior para aqueles que n5o tinham este acesso. Este viis provavelmente foi causado devido 8 incorporag8o muito maior de pessoas que tem acesso 8 igua e nlo estHo mais frequentando a escola, ou seja, fora da idade escolar ativa. 150 outra forma 0.715452 144717 Atraso Para a variivel atraso, temos a mesma d i n h i c a que a apresentada para a matricula, em favorecimento a quem tem acesso B Bguae luz elCtrica. E tem tambCm decaido ao longo dos anos. 1999 agua para pelo menos 1comodo? N(atraso) ----_-_--- 6495717 32000000 luz _____---_-__----- mCdia(atras0) N(atraso) -_--_-_-__ elttrica 0.5023136 36500000 deo, querosene ou gis de botijZo 0.7267892 1848840 outra forma 0.6219706 138021 2003 agua mCdia(atras0) N(atraso) luz 151 elCtrica 0.4746866 37600000 61e0, querosene ou gfis de botijb 0.6942151 1395713 outra forma 0.578348 103538 mtdia(atras0) ---_-----__-_-_- -------_--_ N(atraso) 0.6189638 5961184 0.4451128 34900000 mCdia(atras0) _-_-_---_------- N(atraso) ___-------- Elttrica 0.46152 39200000 61e0, queroseneou g6s de botijgo 0.6914837 1534269 outra forma 0.5715523 123546 Assiduidade Relembrando que para assiduidade, um valor maior implica em maiores falta, ou seja, menor assiduidade, observamos um resultado abaixo contradit6rio hquelas discutidos mais acima, no sentido de que quem tem acesso a estes servigos ptiblicos a freqiiencia escolar C menor.. ComunicagGes Matricula Continuando a analise das conseqiihcias do us0 do cornputador na educaggo, verificamos, a priori, que o fato de possuir este bem em casa gera maiores incentivos 5 matricula escolar, mas que a diferenga em relaggo a quem ngo tem, vem decrescendo ao longo dos anos, mas permanece significativa. Em relaggo ao us0 da internet, observamos o mesmo comportamento e trajet6ria. 2001 Tem computador em casa? mCdia(frequenta) N(freqiienta) Este computador tem acesso h internet? mCdia(frequenta) N(freqiienta) 152 mtdia(frequenta) _-_--_---__-_--- N(frequenta) _-------------- 0.89803 37600000 0.980128 5447262 mCdia(frequenta) __-_-_---------_ 0.9003832 0.9786938 Estecomputador tem acesso ?i internet? mCdia(frequenta) --_------------- Nb 0.9032069 Sim 0.9827948 N a relaGZo entre telefone e matricula escolar, nota-se que os agentes que tern maior acesso a telefone fixo, em todos anos, bem como para o celular. 1999 telefone mtdia(frequenta) N(frequenta) 153 Atraso Para a variiivel atraso, observamos que a mCdiade atraso C maior entre aqueles que nZo temcomputadorem casa e nem acesso B internet. 2001 CPU _---_--_--------- media(atraso) N(atraso) -___----_- 0.5276967 34000000 0.40422 4409740 ---__-_---_--__-- mCdia(atraso) N(atraso) -_--_---_- 0.5219815 35700000 0.4072516 2841511 N(atraso) -_-__--_--- 33800000 5339015 N(atraso) _--__--_--- 35300000 3815008 154 0 mesmo cornportamento observamos para aqueles que niio tern telefone fix0 e/ou celular. 1999 telefone mtdia(atraso) N(atraso) 2001 telefone mCdia(atras0) N(atraso) _--_----_-- 0.5327811 21100000 0.42425 18000000 mCdia(atras0) N(atraso) -_--_-_--_- 0 0.5080833 25300000 1 0.4366927 13900000 mCdia(atras0) _----_---------_ N(atraso) 0.5124188 23300000 0.4150193 17600000 mCdia(atras0) --_----_-------- N(atraso) 0 0.5034124 22000000 1 0.4320223 18800000 Assiduidade EmrelaGiio B assiduidade, observamos que quem tem cornputador e internet, bem corno telefonia fixa siio mais asslduos. Apenas quem tern celular parece faltar A escola. 2001 CPU mCdia(assiduidade) N(assiduidade) 155 0 0.5428345 35600000 1 0.4258538 2840346 2001 telefone m&dia(assiduidade)N(assiduidade) _-----_----------- ----_-----_-- 0.548281 20100000 0.5179874 18300000 mCdia(assiduidade) _--_--_------___-- N(assiduidade) ------_------ 0.5023479 907423 0.5346315 37500000 Infra-estrutura privada Matricula Emrelaqiio, primeiramente, ao ndmero de cdmodos nota-se urnrelaqiio crescente com a freqiiencia. Ou seja, famflias com maior background de infra-estrutura em sua casa, gera maiores incentivos ti frequentar escolas. Jii em relaqiio a variiivel moradores, observamos que existe uma quantidade 6tima de moradores que seria em torno de quatro ou cinco moradores. Ap6s isso, provavelmente, as crianqas e adolescentes deixam de ir B escola, afim de trabalharem para contribulrem na renda de casa afim de expandir suas restriq6es orpmentiirias. mCdia(frequenta) N(frequenta) _-_--_-_-____--_ --------------- 0.7420627 209109 0.7766051 733732 0.8255257 2208692 0.8454088 4261058 0.8800182 7071854 0.8844594 5550128 0.901104 3096041 0.9250169 1735431 0.9246513 889505 0.9383388 1245855 mCdia(frequenta) _--_---__---___- -_--___-------- N(frequenta) 1 0.859843 530355 2 0.8190097 1205308 3 0.862588 3238524 4 0.9083726 7094408 5 0.8942238 5936724 6 0.8760253 3668047 7 0.8580535 2196025 8 0.8403407 1331679 9 0.8274537 746252 10ou mais 0.8021911 1054083 2001 156 Numero de comodosno domicilio mtdia(frequenta) ___--_---_---____ N(frequenta) _-_-__-_- 0.7491831 292552 0.7964185 1196209 0.8357358 3410463 0.8662214 6571449 0.8984153 11000000 0.9115904 8569825 0.9197211 5209386 0.9349114 2939378 0.9443455 1524514 0.9620727 2249458 mtdia(frequenta) -_-----_--------- N(frequenta) -----_-_--_-__ 0.6523692 20703 0.8086602 1118450 0.8965989 5162548 0.9288501 11900000 0.9099749 10300000 0.8939955 5979800 0.8710197 3381795 0.8586481 2152883 0.8485966 1210349 0.8152452 1762098 0.841241 47676 0.744 243762 0.812194 895104 0.8521 3076917 0.879945 6283230 0.906596 11400000 0.917944 8939524 0.930738 5297075 0.938857 3138058 0.946062 1557226 0.97316 2185717 mbdia(frequenta) _--_-_---___--_- N(frequenta) -_-____------__ 1 0.567391 17265 2 0.836081 1221426 3 0.912097 5737593 4 0.936871 12600000 5 0.924139 9803992 6 0.891362 5722288 7 0.877789 3332775 8 0.859879 1964053 9 0.880947 1113120 10 0.83317 1568211 2004 Numero de comodos no domicilio mtdia(frequenta) N(frequenta) 157 .--_-_--------- 1 0.7391282 275599 2 0.8246477 1028752 3 0.8641669 3166054 4 0.8846498 7035590 5 0.910362 12000000 6 0.9205015 9011824 7 0.9279163 5405774 8 0.9422122 3178438 9 0.9519637 1532985 10 0.9682235 2232684 moradores mkdia(frequenta) N(frequenta) .-____---__--_- .------------- 1 0.6716926 24672 2 0.8521075 1274047 3 0.9099234 6057666 4 0.9368107 13200000 5 0.9268593 10300000 6 0.9006416 5847590 7 0.8866862 3256621 8 0.8697135 2037426 9 0.857598 1082176 10 0.8270499 1755084 Emrela@o B titularidade, observamos umresultado interessante: quem tem casapr6pria e ainda esta pagando, apresenta maiores taxas de matrlcula. E quem mora em terreno pr6prio segue nesta mesma dire@o. Ou seja, quem j i tem seu terreno, mas ainda esta pagando a moradia pr6pria parece gerar maiores incentivos a seus filhos (5 a 17 anos) a irem B escola. --_----_-__----- mkdia(frequenta) N(frequenta) _--_-__--___-_- 0.8852199 17300000 0.8581356 1609051 0.9285918 1385472 0.9038534 153266 0.867522 3171266 0.8272851 3384416 mCdia(frequenta) ___-----_----___- _--_-------_-- N(frequenta) 0.9089996 27500000 0.8633094 2721768 0.9444129 1898014 0.9134105 153633 0.8824748 5117237 0.8529015 5450261 158 0.918274 28000000 0.88398 2349075 0.943581 1845245 0.91549 203148 0.893723 5391861 0.869656 5306483 2004 Prop mCdia(frequenta) _---_-_--------- N(frequenta) -----------_-_- 0.9198748 29100000 0.8872339 2442444 0.9495703 1759597 0.918296 196955 0.9035383 5958344 0.8677395 5340431 Atraso Para a varihvel atraso, nota-se uma rela@o decrescente entre o ndmero de camodos e o atraso, ou seja, quanto maior a infra-estrutura da casa, menor o indice de repetencia. E para o ndmero de moradores, observamos o mesmo comportamento que as taxas de matricula, tendo umvalor 6timo perto de 3 ou 4 moradores. mtdia(atras0) --_-----_---_-_- N(atraso) ---------_ 0.6227348 155172 0.5905918 569820 0.578396 1823894 0.5628356 3604557 0.5578312 6223749 0.5712692 4909673 0.543408 2791260 0.5184385 1605803 0.4742307 823015 0.4361044 1171199 mtdia(atras0) -_--_---__--_--- N(atraso) _---_--_--_ 1 0.5963089 456022 2 0.6080293 988255 3 0.5164934 2794868 4 0.4659683 6446053 5 0.5216406 5309554 6 0.6023767 3215307 7 0.6598758 1885405 8 0.6686874 1119064 9 0.6859588 617489 10ou mais 0.7106066 846125 2001 159 mCdia(atras0) -----_---_------_ N(atraso) .--_--_--- 0.5648956 219175 0.5440521 953043 0.5 162625 2850246 0.535893 5694906 0.5199415 9844665 0.5268003 7812170 0.5107299 4791748 0.4838089 2748058 0.4733567 1440267 0.4255622 2164142 _-_-_------_---__ mtdia(atras0) N(atraso) -_--__---- 1 0.5730046 13506 2 0.5514182 904446 3 0.458287 4629095 4 0.4356106 11000000 5 0.4980779 9343145 6 0.5631033 5347712 7 0.6144245 2945610 8 0.652915 1849135 9 0.6836563 1027098 10 0.6722355 1436754 mCdia(atras0) N(atraso) 0.5654374 40107 0.5119184 181359 0.5245351 726998 0.4785532 2621842 0.4987277 5529380 0.4863301 10400000 0.4969668 8206789 0.477922 4930187 0.4740823 2946187 0.4312281 1473232 0.4170035 2127052 mCdia(atras0) N(atraso) 0.9010821 9796 0.5456659 1021211 0.4175749 5233725 0.4159759 11800000 0.4739346 9060464 0.5399646 5100631 0.585992 2925692 0.6029721 1689448 0.6381359 980600 0.6343412 1306587 2004 Numero de comodos no domicilio mCdia(atras0) N(atraso) 160 ----_-_-_--__-_-- --_-__ 0.5412242 203703 0.5088241 848358 0.4628386 2735999 0.491811 6224033 0.4690473 11000000 0.4840782 8295398 0.4747882 5016106 0.4472367 2994763 0.4309081 1459346 0.3989033 2161737 mbdia(atras0) _-_-_---_---_-_- N(atraso) _--_-__-_-- 0.6056601 16572 0.5210621 1085625 0.4139354 5512012 0.4110319 12400000 0.4625213 9578557 0.513551 5266583 0.5549115 2887601 0.6036179 1771977 0.6148833 928072 0.6242383 1451542 Observamos para a titularidade do terreno do domicflio, o mesmo comportamento anterior. Mas com um detalhe a mais: quem j A tem seu domicilio quitado e construido em terreno pr6prio apresenta um atraso maior em rela@o aos demais. Ou seja, os filhos apresentam umcomportamento mais c6modo em rela@o a sua situa@io. 1999 Prop mtdia(atraso) -------_-------- N(atraso) ----_------ 0.561671 15300000 0.5966313 1380784 0.4669941 1287663 0.5162853 138530 0.5 121284 2751677 0.548158 2799877 2001 Prop mtdia(atras0) N(atraso) 2003 Prop mCdia(atras0) N(atraso) 161 3 0.3626804 1741139 4 0.3581353 185980 5 0.4365447 4819037 6 0.4702275 4614817 2004 Prop -----_--_-_----- ----------- mCdia(atras0) N(atraso) 1 0.4860532 26800000 2 0.496411 2167019 3 0.3828523 1670861 4 0.3467818 180863 5 0.4198767 5383592 6 0.463526 4634103 Assiduidade Para assiduidade, novamente maior quantidade de c6modos leva a maior freqiiCncia. Jh o ndmero de moradores ngo apresenta exatamente a mesma rela@o anterior. Mas familias numerosas apresentam uma assiduidade elevada. mCdia(assiduidade) N(assiduidade) _--_--__-_-_-- 0,578784 219175 0.6085991 952683 0.5993659 2849054 0.566748 5691316 0.5406826 9833923 0.5268086 7809627 0.5221883 4789330 0.4988635 2746511 0.4550192 1438537 0.4458025 2164142 mCdia(assiduidade) ----__--_--__-__--N(assiduidade) _-_--_-___-__ 1 0.1849548 13506 2 0.613258 903486 3 0.5648911 4625350 4 0.5288798 11000000 5 0.5389259 9338072 6 0.5346892 5343383 7 0.5272954 2943553 8 0.4910912 1848569 9 0.4903841 1025851 10 0.4965716 1436542 Jh a varihvel assiduidade apresenta uma rela@o oposta h varihvel atraso, visto que quem mora em moradias quitadas e terreno pr6prio apresenta maior assiduidade. 200I Prop mCdia(assiduidade) N(assiduidade) 162 _------------__-_- --__--___-_-- 0.5090867 25000000 0.5563452 2348896 0.556349 1791665 0.5240861 140330 0.5985276 4510955 0.5844589 4648121 Outros fatores Matricula A variivel "local" 6 especifica para o local onde os individuos moram. E corrobora-se os esperado, o qual, pessoas que moram em centros urbanos tem um maior indice de matricula em relaqiio aos que moram na zona rural. Em relaqiio ivariivel "setor", que diz se a residgncia do morador est6 situada em um setor niio especial ou em um aglomerado subnormal (favela). Nota-se que, mesmo em condiG6es habitacionais adversas, os individuos tendem a buscar mais educaqiio, niio muito abaixo em relaqiio a quem mora em setores niio especiais (normais). 1999 local mCdia(frequenta) --------_____--- N(frequenta) --_------------ 0.8892497 21000000 0.8323189 5997361 mCdia(frequenta) ---------------- N(frequenta) -_--__-__------ 0.8769016 25800000 0.8701541 1188232 0 1 2004 local mtdia(frequenta) N(frequenta) 163 0 Atraso Nota-se abaixo que quem mora na zona rural apresenta altos indices de atraso escolar em rela@o aos habitantes urbanos. TambCm quem mora em favelas apresenta dificuldades na escola e consequentemente gerando maiores atrasos. Mas nota-se para todos os casos que o atraso escolar tern decaido ao longo dos anos. 1999 local mCdia(atraso) N(atraso) __-_---_--_ 18700000 4993333 N(atraso) _----_--_-- 22600000 1034507 mCdia(atraso) -_--_--_-__------ ------_-_- N(atraso) 0.4898743 31800000 0.6246523 6752832 mCdia(atras0) N(atraso) ---------- 0 0.5119646 36800000 1 0.546982 1692030 mtdia(atras0) N(atraso) ---------------- _-_--__---- 0.4631112 32400000 0.5784101 6713097 mCdia(atras0) --_------__----_ N(atraso) _--_---_--_ 0 0.4819192 37400000 1 0.5035926 1733721 2004 local mCdia(atras0) N(atraso) .------ 33300000 7626019 N(atraso) 164 0 0.469429 39100000 1 0.4903176 1838496 Assiduidade Para assiduidade, observamos que quem mora na zona rural C um pouco mais assiduo. Mas quando olha-se para os habitantes de favela, estes apresentam menor freqiiencia escolar. 0 1 165 Anilise Multivariada Prof cihacia Das tabelas"* abaixo das regressdes por minimos quadrados ordinfirios de variiveis continuas, observa-se que a falta de eletricidade tem um efeito mais perverso que a figua, na grande maioria dos casos. 0 impacto da infra-estrutura escolar (conservaqiio das instalaqdes hidrhlicas e elCtricas e iluminaqgo na sala de aula) tambCm afetam positivamente a proficihcia escolar. 0 acesso a computadores pode desenvolver no aluno uma melhor "aptidiio escolar", gerando bons resultados na escola. 0 acesso B internet em casa niio apresenta uma direqgo Clara, mas parece que em alguns casos pode gerar uma queda da produtividade do aluno. Quando comparamos o impacto de se ter computador em casa com o impacto de se ter computador na escola disponivel para o aluno, o primeiro fator tem peso, em termos de magnitude, bem maior que o segundo. Isso provavelmente C causado pel0 melhor ambiente encontrado em casa frente B escola, e pel0 fato do aluno ter uma menor competigiio pel0 us0 em casa em relaqiio B escola. Logo, em termos gerais, a infra-estrutura pode impactar positivamente na proficigncia escolar, podendo at6 chegar a um aumento em torno de 12pontos na nota (no cas0 de se ter computador em casa, em relaqgo a quem niio tem, ceterisparibus). Emrelagiio B infra-estrutura privada, observamos o comportamento esperado: um efeito positivo em nivel, e efeito negativo no termo quadrfitico. lo'A dummy omitida paracor foi a indigena, e a omitida para a UFfoi Rondhia. N b nos preocupamos em omitir determinada dummy, visto que os parametros de interesse siio os relacionados ao acesso 9 serviGospfiblicos (infra-estrutura). 166 3 ReprovagGo A titulo de exemplificaqiio, interpretando a variivel nres no an0 de 1999, temos que, um morador a mais no domicflio aumenta o log da raziio por 0.141289 (comparando reprovado urna vez corn nenhuma vez), ou seja, reduz-se a probabilidade de niio ser reprovado e/ou aumenta-se a probabilidade de ser reprovado uma vez. Das variiveis de infra-estrutura, observamos que a maioria C estatisticamente niio significativa. Somente cpu, cpunet, e igua, nres e nres2 em alguns casos siio significativas. Destas somente o acesso a computador e internet apresentam o sinal negativo e de acordo com o esperado (com exceGiio de 1999),ou seja, quem tem computador e internet em casa, tem uma maior chance de niio ser reprovado e uma menor chance de ser reprovado. As interaGbes na grande rnaioria niio siio estatisticamente vilidas e sua inclusiio reduz o efeito das outras e/ou tomando-as tambCm niio significativas, principalmente na 8" sCrie e na 3" sCrie. 170 3 N N Q\ 0 0 m 9 9 Matricula Nota-se que as variiiveis de infra-estrutura s50 significativas e aumentam a probabilidade de frequentar escola, na maioria das regressdes. Emrelaggo ao acesso do servigo p~blico,iigua e luz elttrica aumentam a probabilidade de matricula. Para a comunica@o, verificamos o efeito na mesma direggo, sendo o telefone o de maior impacto, na maioria dos casos. A infra-estmtura privada gera impactos importantes. Casas maiores levam a maior frequencia escolar, principalmente devido ao maior nivel de renda que esta varihvel est6 correlacionada. Outro aspect0 t ndmero de moradores. Considerando todos, nos leva a um resultado n5o esperado, em que quanto maior o n ~ m e r ode moradores, em nivel e marginalmente, gera uma reduggo na probabilidade de interesse. Restringindo a moradores com idade superior ou igual a 10 anos, o efeito se torna semelhante 21s aniilises multivariadas anteriores, em que, quanto um aumento eleva a probabilidade, mas marginalmente este acrCscimo vai se reduzindo. Para a titularidade, observamos que quem tem terreno prbprio, mas ainda paga, tem uma probabilidade maior de estar empregado, jii quem mora de aluguel ou domicilio cedido tem menor probabilidade. Emrelaggo aos outros fatores, quem mora na &ea urbana tem mais chance de estar matriculado e quem moram em aglomerados subnormais tem menor chance (para 2001 e 2004 que t estatisticamente significativo). Vale por fim ressaltar a variiivel renda per capita (rendapc) que 6 estatisticamente igual a zero, ou seja, diferengas na probabilidade de matricula n5o sgo explicadas por diferengas no nivel de renda. 177 Atraso Para o atraso escolar observamos que a maioria das varigveis de infra-estrutura s80 significativas. E apresentam a correlaqgo parcial esperada de reduzir a probabilidade de acesso em favor de quem tem acesso. Eas interaSdes s80 em sua maioria n8o significativas. 179 Assiduidade Em relagiio B assiduidade, observamos que a maioria das variiveis de infra-estrutura siio niio significativas, com excegiio do ndmero de moradores, proppago (comparando pouco faltoso com o sem falta). Quando considerado a segunda regressiio, comparando muito faltoso com o sem falta as variiveis se tornam significativas em maior ndmero. Como esperado, elas apresentamo sinal econamico robusto. Por exemplo, quanto maior o ndmero de moradores ou quemj i tem domicflio prbprio pago emterreno prbprio, aumenta a chance de ser mais assiduo na escola em relagiio aos moradores com caracteristicas diferentes dessas, 181 0 2 8 0 2 9 0 \D W 0 8 W P 0 8 0 Centro de Politicas Sociais da FGV disponibiliza site com um software amighvel chamado Espelho baseado dessas regressdes multivariadas, onde cada pessoa pode inserir a combinaqiio de dados, como sexo, idade, local de moradia entre outros e em particular varihveis de infraestrutura e cenarizar por exemplo qual a probabilidade de se estar matriculado ou atrasado na escola ou a sua respectiva proficiencia escolar. Posteriormente, rodamos outros rnodelos mais restritos que empilham todas as pesquisas e permitem que a varihvel ano atue apenas no coeficiente linear das regressdes estes modelos apresentam tipicamente resultados mais consistentes para efeito de simulaqgo, reduzindo as flutuaqdes das estimativas entre anos. Na anhlise empirica empreendida neste trabalho estudamos as correlaqdes brutas e as parciais entre as varihveis de infraestrutura e as de desempenho escolar. A dificuldade central destas anitlises para fins de desenho de politicas C a questgo da direqgo de causalidade entre variitveis ex6genas e as end6genas. Neste sentido, uma extensgo natural da anitlise C uma anitlise de diferenqa em diferenqa tomando como base a ocorrencia de experibcias onde algumas comunidades s20 beneficiadas com aumento da oferta de infraestrutura e outras ngo. Uma primeira candidata a experimento seria o programa Favela-Bairro no Rio onde algumas comunidades de baixa renda sgo beneficiados por melhoras maciqas de infraestrutura privada (acesso a esgoto, luz, dgua e melhora da constmq8o) e pfiblica (pavimentaqgo das ruas, iluminaqgo pfiblica, construqgo de praqas) enquanto os criticos do programa enfatizam o fato de que outros aspectos como investimento em capital humano ngo foi diretamente beneficiados. Neste sentido o Favela-Bairro constituiria num experimento fitil na identificaqgo dos impactos de investimentos em infraestrutura sobre as varihveis de desempenho escolar. Esta andlise seria feita atravCs da comparaqgo antes e depois do programa de comunidades afetadas e as ngo afetadas a partir dos dados do Cesno 1991 e 2000. A dataqiio do Censo e a possibilidade de abertura inframunicipal dos dados permtiriam uma anhlise similar as da PNAD. 183 Conclusijes Este relatcirio avalia as correlaqdes entre acesso B infra-estruturae indicadores de desempenho escolar tais como matricula, frequencia, repetencia e proficiencia escolar. As variiveis de infra-estruturautilizadas siio acesso a serviqos pdblicos (abastecimento de hgua, esgotamento sanitirio e iluminaqiio elCtrica e comunicaqdes.) e infra-estruturaprivada (ntimero de pessoas no domicflio, direito de propriedade). A pesquisa consiste na geraqiio, descriqiio e anilise de umconjunto de padrdes de correlaqdes brutas e parciais a partir de duas bases de microdados. Utilizamos pesquisa domiciliar a PNAD para uma avaliaqiio das correlaqdes entre as condiqdes de acesso B infra-estrutura domCstica e algumas variiveis sociais. Os efeitos sociais da provisiio de infra-estrutura em casa e aqui tambCm nas escolas seriio captados atravts de informaqdes do MinistCrio da Educaqiio por meio de avaliaqiio de desempenho do SAEB. Resultadosda Anillise Bivariada ProficiEncia e ReprovaSCo - SAEB Apresentamos a relaqiio das variiveis de infra-estruturatomadas isoladamente na proficiencia - exame de Matemitica - e reprovaqiio. Vale destacar que, na anilise da variivel reprovaqiio (referente B questiio "Voce jifoi reprovado?"), esta apresenta os seguintes valores (nenhuma reprovaqiio, uma reprovaqiio e duas ou mais reprovaqdes). Segue a anilise por grupos de variiveis de infra-estrutura: comunicaqdes (acesso B internet, computador etc), serviqos plStblicos(acesso B luz, igua etc). 184 Comunicaqijes As conseqiihcias do acesso a computadores em casa siio relativamente desconhecidas, e particularmente, a literatura dos impactos educacionais de computadores em casa C escassa. A importfincia dessa an6lise reside na evidhcia encontrada na literatura que seu us0 em casa pode compensar a falta da tecnologia da informagiio na escola. Assim observa-se uma proficihcia maior para os alunos que tern acesso B Internet em casa, via cornputador, para todos os anos, tendo um diferencial maior em relagiio aos que niio tern, no ano rnais recente. Discriminado por sCrie este impacto nota-se um diferencial crescente em relaqiio a series mais altas. Ou seja, a diferenga da nota entre quem tem acesso e niio tem 6 maior para quem est6 na 3" sCrie do EnsinoMCdio, do quem est6 na 8" skrie do Ensino Fundamental, que 6 maior em relagiio a quem est6 na 4" serie do Ensino Fundamental. Mesmo considerando os que tem computador em casa, sem discriminar se tem acesso B internet ou niio, o desempenho foi melhor para aqueles que possuem. Observa-se que quanto maior o ntimero de computadores disponiveis para os alunos melhor o desempenho dos mesmos na escola. Uma estatistica exclusiva para o an0 de 2001 que inferimos C se o aluno usa o computador para fazer a ligiio de casa. 0 que notou-se C que niio h6 uma relaqiio muito direta entre o us0 do computador para fazer a liqiio e a proficiencia dos alunos. Apenas no cas0 do aluno nunca utilizar, este tEm uma nota, em media, menor em relaqiio aos que usam, principalmente para a 8" sCrie e a 3" serie. Corroborando o resultado apresentado para a proficihcia, observa-se uma maior repetsncia para quem niio tem computador com acesso B internet. Em todas series, observa-se o mesmo resultado. Bem como para an6lise na escola, o maior acesso a cornputadores reduz a mCdia de repetencia entre os alunos, principalmente em 2003. Em relaqiio ao us0 de computadores na realizaqiio de tarefas escolares, evidenciou-se que o us0 do computador para fazer tarefas escolares pode reduzir a reprova@o e at6 melhorar o desempenho do aluno, mas desde que o mesmo niio se utilize deste artificio sempre. Serviqos Piiblicos 0 acesso B infra-estrutura niio tem sido objeto de estudo explicit0 na literatura. Apenas tem servido como controle para se avaliar o efeito de algum fator em alguma varihvel educacional (performance, matricula etc). Entiio torna-se imprescindivel focar a an6lise em aspectos de acesso B infra-estrutura e o desempenho escolar. A questiio que surge (e sera dirimida na an6lise multivariada) C o quanto que o acesso B luz, 6gua etc e melhora a proficiCncia escolar, tendo-se controlado para outros fatores (como educagiio dos pais, unidade federativa, ntimero de residentes na mesma casa etc). Observa-se que o us0 da eletricidade e da 6gua tem um correlagiio positiva com a proficihcia e ha um peso mais negativo sobre quem niio tem acesso B eletricidade do que B Bgua. Analisamos tarnbCm algumas vari6veis de infra-estrutura ligadas B escola. Observamos que a conservagiio das instalaqtjes hidrhulicas e elhricas tem urna correlaqiio positiva com o desempenho dos alunos. Outro aspect0 importante C a iluminaqiio no local de estudo, mostrando um melhor desempenho educacional nas salas iluminadas. 0 acesso B infra-estrutura tambCm reduz o indice de reprovaqiio, mas ao contr6rio do desempenho escolar, a eletricidade niio gera um impacto mais negativo do que o acesso B 6gua, em 2001 e 2003. Apesar da falta de robustez em relaq5o B infra-estrutura escolar, geralmente col6gios com instalagdes mais adequadas apresentam menor indice de repetsncia. 185 Outrosfatores Umdos elementos candidatos a explicar este diferencial de desempenho seria a diferenga de acesso a infraestrutura entre ireas rural e urbana. Observa-se uma pequena evoluggo de 2001 para 2003 nas notas dos dois tipos de escola, sendo 1.83% da rural e 1.2% da urbana. 0 indice de reprovagZo para escolas urbanas 6 bem menor do que as rurais, principalmente para as rurais. Outra deficiencia que se apresenta no pais que citamos brevemente C a da escola ptiblica em relaggo B particular, como 6 notado nos valores de proficiencia observado abaixo. As duas redes de ensino evoluiram pouco de 2001 para 2003 (em torno de 0.8%). 0 indice de reprovagfio para todas as series C bem maior nas escolas pdblicas do que privadas, apesar de que, de 2001 para 2003 as reprova@es cairam bastante na rede pdblica de ensino. Matricula, Atrasoe Assiduidade PNAD- A aniilise bivariada estende-se agora 9 matricula, atraso e assiduidade. As variiveis de infra-estrutura encontradas na PNAD diferem umpouco das encontradas no SAEB, vistas na anhlise precedente. Aqui, restringimos a populaggo de interesse B faixa ethia de 5 B 17 anos que estZo na "idade escolar ativa". Servicos Piiblicos Emrelaggo aos servigos pdblicos, notamos que o acesso B igua gera maiores indices de matricula. E a luz recebida na forma elCtrica tambCm incentiva B frequencia escolar. AlCm disso, no context0 dinamico, a media de matricula escolar tem crescido entre aqueles que tem acesso B igua e luz elktrica, ao longo de todos anos. Para a variiivel atraso, temos a mesma dintimica que a apresentada para a matricula, em favorecimento a quem tem acesso B igua e luz eletrica. E tern tambCm decaido ao longo dos anos. Relembrando que para assiduidade, observamos um resultado abaixo contraditbrio Bquelas discutidos mais acima, no sentido de que quem tem acesso a estes servigos pdblicos a freqiiencia escolar C menor.. Comunicaqijes Continuando a anilise das consequhcias do us0 do computador na educaggo, verificamos, a priori, que o fato de possuir este bem em casa gera maiores incentivos B matricula escolar, mas que a diferenga em relaggo a quem nZo tern, vem decrescendo ao longo dos anos, mas permanece significativa. Em relaggo ao us0 da internet, observamos o mesmo comportamento e trajet6ria.Para a variiivel atraso, observamos que a media de atraso 6 maior entre aqueles que nZo tem computador em casa e nem acesso B internet. 0 mesmo comportamento observamos para aqueles que nZo tem telefone fix0 e/ou celular. Em relaGZo B assiduidade, observamos que quem tem computador e internet, bem como telefonia fixa sZo mais assiduos. Apenas quemtem celular parece faltar mais B escola. Infra-estruturaprivadae outrosfatores Em relaggo, primeiramente, ao ndmero de c8modos nota-se um relaqZo crescente com matricula, assiduidade e atraso escolar. A variivel "local" 6 especifica para o local onde os individuos moram. Ecorrobora-se os esperado, o qual, pessoas que moram em centros 186 urbanos tem um maior indice de matricula, de assiduidade e menores indices de atraso escolar em relaqiio aos que moram na zona rural. Resultadosda Anhlise Multivariada 0 desempenho dos alunos pode ser afetado pel0 acesso a serviqos pdblicos como luz, hgua etc. Melhor infra-estrutura, tanto na casa como na escola, deve melhorar a produtividade dos estudantes, reduzindo portanto, a repethcia. Mas a questiio que se levanta aqui C que outros fatores podem tambCm afetar estas varihveis de interesse. Assim, os controles assumidos ajudam a isolar de forma mais precisa o efeito da infra- estrutura no desempenho escolar. Assim, por exemplo, a educagiio dos pais pode afetar positivamente a produtividade de seus filhos, mesmo que haja certa deficigncia em infra- estrutura na escola. AlCm disso, o ntimero de moradores na casa do aluno pode tambCm influenciar. Temos observado que hh geralmente um ndmero 6timo de moradores que ajudam a melhorar a performance escolar do estudante. AlCm disso, controlamos tambCm por sexo, cor e Unidade Federativa (UF), visto que as caracteristicas escolares podem divergir entre esses grupos. Profici6ncia As regressdes para a equaqiio de proficiencia foram estimadas por sCrie. 0 modelo estimado foi um modelo linear nas co-varihveis. Estimamos a regress50 por minimos quadrados ordinhrios, mas consideramos para a hip6tese de heterocedasticidade dos erros. Assim, estimamos a matriz de variiincia atravCs do estimador de White (1980) produzindo erros padr8es robustos a heterocedasticidade. A partir das regressdes por minimos quadrados ordinhrios de varitiveis continuas, observa-se que a falta de eletricidade tem um efeito mais perverso que a hgua, na grande maioria dos casos. 0 impacto da infra-estrutura escolar (conservaqiio das instalaqdes hidrtiulicas e elCtricas e iluminaqiio na sala de aula) tambCm afetam positivamente a proficiCncia escolar. 0 acesso a computadores pode desenvolver no aluno uma melhor "aptidiio escolar", gerando bons resultados na escola. 0 acesso B internet em casa niio apresenta uma direq5o Clara, mas parece que em alguns casos pode gerar uma queda da produtividade do aluno. Quando comparamos o impacto de se ter computador em casa com o impacto de se ter computador na escola disponivel para o aluno, o primeiro fator tem peso, em termos de magnitude, bem maior que o segundo. Isso provavelmente C causado pel0 melhor ambiente encontrado em casa frente B escola, e pel0 fato do aluno ter uma menor competiqiio pel0 us0 em casa em relaqiio B escola. Logo, em termos gerais, a infra-estrutura pode impactar positivamente na proficiencia escolar, podendo at6 chegar a um aumento em torno de 12 pontos na nota (no cas0 de se ter computador em casa, em relagiio a quem niio tem, ceteris paribus). Em relagiio B infra-estrutura privada, observamos o comportamento esperado: umefeito positivo em nivel, e efeito negativo no termo quadrhtico. Reprovapio A titulo de exemplificaqiio, interpretando a varihvel nres no an0 de 1999, temos que, um morador a mais no domicilio aumenta o log da raz5o por 0.141289 (comparando reprovado uma vez com nenhuma vez), ou seja, reduz-se a probabilidade de niio ser reprovado e/ou 187 aumenta-se a probabilidade de ser reprovado uma vez. Das variiveis de infra-estrutura, observamos que a maioria C estatisticamente n i o significativa. Somente cpu, cpunet, e igua, nres e nres2 em alguns casos siio significativas. Destas somente o acesso a computador e internet apresentam o sinal negativo e de acordo com o esperado (com exceGio de 1999), ou seja, quem tem computador e internet em casa, tem urna maior chance de n i o ser reprovado e uma menor chance de ser reprovado. As interaG6es na grande maioria n i o siio estatisticamente vilidas e sua inclusio reduz o efeito das outras e/ou tornando-as tambCm niio significativas, principalmente na 8" sCrie e na 3" sCrie. Matricula Nota-se que as variiveis de infra-estrutura s i 0 significativas e aumentam a probabilidade de frequentar escola, na maioria das regressties. EmrelaGio ao acesso do serviqo pdblico, igua e luz elCtrica aumentam a probabilidade de matricula. Para a comunicaGio, verificamos o efeito na mesma direGio, sendo o telefone o de maior impacto, na maioria dos casos. A infra-estrutura privada gera impactos importantes. Casas maiores levam a maior frequCncia escolar, principalmente devido ao maior nivel de renda que esta variivel esti correlacionada. Outro aspect0 C ndmero de moradores. Considerando todos, nos leva a um resultado n i o esperado, em que quanto maior o ndmero de moradores, em nivel e marginalmente, gera uma reduqio na probabilidade de interesse. Restringindo a moradores com idade superior ou igual a 10 anos, o efeito se torna semelhante Bs anilises multivariadas anteriores, em que, quanto um aumento eleva a probabilidade, mas marginalmente este acrCscimo vai se reduzindo. Para a titularidade, observamos que quem tem terreno prbprio, mas ainda paga, tem urna probabilidade maior de estar empregado, j i quem mora de aluguel ou domicflio cedido tem menor probabilidade. Em relaqio aos outros fatores, quem mora na ire, urbana tem mais chance de estar matriculado e quem moram em aglomerados subnormais tem menor chance (para 2001 e 2004 que C estatisticamente significativo). Vale por fim ressaltar a variivel renda per capita (rendapc) que C estatisticamente igual a zero, ou seja, difereqas na probabilidade de matricula n i o s i 0 explicadas por difereqas no nivel de renda. Atraso Para o atraso escolar observamos que a maioria das variiveis de infra-estrutura s i 0 significativas. E apresentam a correlaqio parcial esperada de reduzir a probabilidade de acesso em favor de quem tem acesso. E as interaG6es s i 0 em sua maioria n i o significativas. Assiduidade EmrelaGio B assiduidade, observamos que a maioria das variiveis de infra-estrutura s i 0 n i o significativas, com exceqio do ndmero de moradores, proppago (comparando pouco faltoso com o sem falta). Quando considerado a segunda regressio, comparando muito faltoso com o sem falta as variiveis se tornam significativas em maior ndmero. Como esperado, elas apresentam o sinal econ8mico robusto. Por exemplo, quanto maior o n6mero de moradores ou quem j i tem domicflio pr6prio pago em terreno prbprio, aumenta a chance de ser mais assiduo na escola em relaGio aos moradores com caracteristicas diferentes dessas. Simuladore extensijes 0 Centro de Politicas Sociais da FGV disponibiliza site com um software amigivel chamado Espelho baseado dessas regressties multivariadas, onde cada pessoa 188 pode inserir a combinaqiio de dados, como sexo, idade, local de moradia entre outros e em particular variiveis de infraestrutura e cenarizar por exemplo qual a probabilidade de se estar matriculado ou atrasado na escola ou a sua respectiva proficihcia escolar. Posteriormente, rodamos outros modelos mais restritos que empilham todas as pesquisas e permitem que a variivel an0 atue apenas no coeficiente linear das regressdes estes modelos apresentam tipicamente resultados mais consistentes para efeito de simulaqiio, reduzindo as flutuaqdes das estimativas entre anos. Na anilise empirica empreendida neste trabalho estudamos as correlaqdes brutas e as parciais entre as variiveis de infraestrutura e as de desempenho escolar. A dificuldade central destas anilises para fins de desenho de politicas C a determinaqiio da direqiio de causalidade entre variiveis ex6genas e as end6genas. Neste sentido, uma extensiio natural C a anilise de diferenqa em diferenqa tomando como base a ocorrCncia de experihcias onde algumas comunidades siio beneficiadas com aumento da oferta de infraestrutura e outras niio. Uma primeira candidata a experimento seria o programa Favela-Bairro no Rio onde algumas comunidades de baixa renda foram beneficiadas por melhoras maciqas localizadas de infra-estrutura privada e ptiblica (acesso a esgoto, luz, igua e melhora da construqiio, pavimentaqiio das ruas, iluminaqiio p6blica e construqiio de praqas) enquanto os criticos do programa enfatizam o fato de que outros aspectos como investimento em capital humano niio foi diretamente beneficiados. Neste sentido o Favela-Bairro constituiria num experimento 6til na identificaqiio dos impactos de investimentos em infraestrutura sobre as variiveis de desempenho escolar. Esta anilise seria feita atraves da comparaqiio antes e depois do programa de comunidades afetadas e as niio afetadas a partir dos dados do Censo 1991e 2000. A dataqiio do Censo e a possibilidade de abertura inframunicipal dos dados permtiriam uma anilise similar as da PNAD, mas com a possibilidade de identificaqiio da relaqiio de causalidade entre a provisiio de infra- estrutura e a obtenqiio de avanqos nos indicadores educacionais. ReferhciasBibliogrhficas Agresti, Alan An Introductionto CategoricalData Analysis, New York :Wiley Series in Probability and Statistics, 1996. Fairlie, Robert W. The Effects of Home Computers on School Enrollment 2003 Heckman, J. The Lessons from The Technology of Skill Formation, Working Paper 11142, Fev 2005 Selwyn, NeilThe effect of usinga home computer on students' educationaluse of IT Computers & Education 31(1998) 211-227 White, H.(1980), "A Heteroskedasticity-ConsistentCovariance Matrix Estimator and a Direct Test for Heteroskedasticity", Econometrica 48, 817-838. Wooldridge, Jeffrey M.Introductory econometrics: a modern approach. Cincinnati: South-Western College Publishing,2003 189 Apendice 2001 ->serie =4 621,62 1 2734229 466,185 1554201 2003 ->serie =4 471,070 3179690 71,494.60 -> serie = 8 na sua casa tern computador corn internet? Sim 567,048 Nao 2591274 nZo sei 22,599.40 190 na sua casa tem computador com internet? sim 502,285 nIo 1596230 nIo sei 7,884.80 mCdia(profic) N(profic) ---------------- ---------- 244.5079 2919663 206.4126 46,910.60 191 1961199 88,5 67.80 2003 ->serie=4 ->serie = 8 3119823 62,049.90 ->serie= I 1 ->serie =4 3428302 320.539 ->serie = 8 2981482 191,287 192 -> serie = 11 2017144 89,681.30 193 ANNEX^^'^ Methodology for the Calculation of ConcessionReturns A. Cost of Equity (CE) The cost of equity i s calculated with the following formula: CE=rf D + * (r, -rf) Crp + (1) where rf i s the risk free rate, I3 i s the beta o f the project, rmi s the expected stock market return, and Crp i s the country risk premium. The risk-free rate i s a theoretical interest rate that would be returned on an investment completely free of risk. Interest rates on government bonds are often used as proxies for the risk free rate. At the end o f December 2003, the risk free rate was 1.8 percent. Beta is a quantitative measure o f the volatility of a given stock relative to the overall market. The S&P 500 index is usually used as reference for the overall stock market. Beta i s defined using an index of 1:l (ie., a beta above 1means that the stock is more volatile than the market, and a betabelow 1means it i s less volatile thanthe market). The calculated leveraged betas are summarized inTable 1. Table 1.Leveragebetasby infrastructuresector, 1996-2003 LeverageBeta 1996 1997 1998 1999 2000 2001 2002 2003 Energy -_ 0.85 0.85 0.85 0.71 0.72 0.66 1.02 1.02 Telecommunications 1.62 1.51 1.47 1.25 1.62 1.75 1.47 1.74 Water 0.59 0.61 0.63 0.56 0.79 0.56 0.57 0.51 The market risk premium represents the additional return that investors will require to hold a risky investment (shares) rather than a risk free asset. It i s generally measured as the historical average annual excess return on the US stock market (using returns on the S&P500) above the risk free rate. We used the geometric average of these excess returns over the period from 1963 and 2003. We obtain a market risk premium o f 4.74 percent. The country risk premium corresponds to the extra return investors require to invest in stocks of companies in a country deemed riskier than a less risky country used as benchmark (often the US). They reflect the potential volatility o f investments in a given country due to defaults associated with political or other events inthat country. lo3 Based on Sirtaine, S., M.E. Pinglo, J.L. Guasch, and V. Foster (2005). "How Profitable are Infrastructure Concessions in Latin America? Empirical Evidence and Regulatory Implications." The World Bank: Washington D.C. Processed. 194 B. Returnon Equity(ROE) The Return on Equity (ROE)is a measure of the profits a company is able to generate given the resources provided by its shareholders. It i s the ratio of the concession's net income divided by the shareholders' equity investment inthe concession: ROE= Net Income/ Shareholders' Equity (2) where net income i s the after-tax profit, and shareholders' equity i s the total assets minus total liabilities, which are the book value of common and preferred stocks. C. Weighted Average Cost of Capital(WACC) The weighted average cost of capital (WACC) is derived from the following formula: WACC = E/ (D + E) * CE+D / (D E) (1-T)" CD + * (3) where E is the book value of equity, D i s the long-term debt, CE i s the cost of equity, CD i s the cost of debt, and T i s the corporate income tax rate. The required parameters have been estimated as described below. Book value of equity and long-term debt. We have used a typical capital structure for each sector, rather than concession-specific leverages. Cost of equity. We used the estimated cost of equity using the methodology described above. Corporate income tux rate. We used the nominal corporate tax rates of each country. Cost of debt. We used a typical cost of debt. We estimated the latter on the basis of the cost at which a hypothetical corporate issuer could issue local currency bonds in Brazil, given its country's rating. This cost we estimated to be equal to the sum of the risk free rate, the country risk premium and a 20bp premium for corporate issues over sovereign is~ues.''~Note also that we did not try to estimate the cost of potential debt renegotiation / restructuring. D. Returnon CapitalEmployed(RoCE) The Return on Capital Employed (RoCE) i s a measure of the returns a company i s realizing from its capital at large. It i s calculated as the ratio of profits before interests divided by the difference between total assets and current liabilities. The denominator i s infact a measure of the long-term financing structure of the company (Le. its capital and long-term liabilities). The numerator is a measure of the income generated b y the company independently of its financing structure. The resulting ratio therefore represents lo4 Note that we used the same country risk premium (a historical country risk premium) than to compute the cost of equity. The country risk premium relevant to compute the cost of debt may however be different since the relevant horizon is usually shorter (it would be higher if the risk of investing in a given country is perceivedas higher inthe short term than inthe long term, and vice a versa). 195 the efficiency with which long-term financing i s being utilized to generate net operating revenues. RoCE =EBIT (1- T) / Capital Employed * (4) where EBIT i s the earnings before interest and taxes, T i s the nominal corporate income tax rate, and capital employed is the sum o f fixed and current assets minus current liabilities, which are the shareholders' funds plus long-term liabilities. E. InternalRateof Return(IRR) The project internal rate of return (IRR) i s a measure o f how effectively a company uses the funds invested in its operations, independently of the nature of these funds. It i s the rate that makes the net present value of the net financial flows generated by the concession before financing equal to zero. The net financial flows generated by a concession are calculated as its earnings before interest, taxes, depreciation and amortization (EBITDA) - a measure of the financial flows generated by the concession's operations, independently o f its financial structure-, minus the investments, increases in working capital, and taxes financed by these flows, and minus the price initially paid for the concession. where CF i s the net cash flow generated by the concession (ie., EBITDA - Investiments -AWorkingCapital-Taxes-BidPrice), t=0isthefirstyearofoperations, and t =n is the last year of historical data available (2003). As for the terminal value (TV), it is equal to the estimated future value of the net flows generated by the concession. We assumed that they would grow at a constant rate: 5 f=y CFt +TV (1 ZRR)` + = O where CF i s the last historical net cash flow generated by concession, g i s the growth rate of future cash flows, r i s the weighted average cost of capital, and n is the concession's last year of operation. 196 ANNEX3 The Impactof Privatizationon InfrastructureIndustries:MethodologicalAppendix A. EmpiricalApproach The empirical approach relies on the methodology suggested by Andres et al. (2006) and we adapted it to show andcompare Brazilian features incontrast to the rest of the region. Ideally, to evaluate the impact of privatization, the performance of utilities under private operation should be compared to comparable firms in a similar environment that are still operated by the government, assuming that such firms are the contra-factual of the privatized ones. Inmost cases, it i s hard to identify a comparable firm; hence, most of the literature compares the evolution o f selected indicators before and after the change in ownership. Most of the literature employs two different strategies in order to estimate the effect of the privatization: first, since Megginson et al. (1994), there have been several studies using means and medians of the periods before and after the event of privatization, and testing the significance of the change. Some literature has been developed that considers different samples of SOEs among countries and evaluates indicators. Another branch of literature has assumed these policies to be treatments and, following the literature of program evaluation (see Heckman and Robb 1985),proposed a dummy for those periods where the SOE was privately owned, and checked the significance o f this dummy, as well as some interactions with characteristics specific to each paper (for example, Boardman and Vining, 1989;and Ros 1999). Section 2 of Chapter 3 followed both methodologies. For the first approach, the analysis basically spitted into three main periods: First, the transitional period, covering those years starting at the time when the reform was announced and ending one year after the concession or privatization was awarded; And, second, the periods before and after the transition. The changes in average levels and growths were analyzed, and the average changes between these periods were computed. Then, we concluded by testing the significances of these changes inmeans and medians. For the second approach, we started with a simple version of the model, as specified below: where yijr was the variables of interest (outputs, inputs, labor productivity, efficiency, quality, coverage and prices). The main coefficient inthis model was the dummy, PRIKj,, that was equal to one if the firm i of country j had private owners at time t . Given the 197 fact that there were several variables not observable to the econometrician, fixed effects were included to capture the characteristics of the firm, such as management, initial conditions, size, density of the network, etc., which we assumed to be constant for each firm across time. Hence, Po captured the effect on the outcome of interest, given by the privatization or the concession. As was pointed out by Frydman et al. (1999), adding the firmfixed effects may reduce the selection bias of the firms inthe sample. A second version of equation (1) was also estimated, introducing a firm-specific time trend: Equation (2) used the same dependent variables and the dummy used inthe static model, PRIVY,,The third coefficient captured the time trend of the variable of interest. Several factors may affect this, like the initial conditions. Hence, it was important to control for the firm's specific value. Finally, as inthe previous model, we included firm fixed effects to capture the characteristics o f the firm not observed by the econometrician. Again, the relevant coefficient was Po,which captured the effect on the outcome of interest caused by the privatizationor the concession. Moreover, there are some important effects during the transition. Most of the literature that was reviewed focused on basic approaches like equation (l), and in some cases used specifications like (2). Inother words, they evaluated the change exactly before and after the change in ownership. In this case, in order to identify the outcomes during the transitional years, specific dummies were defined for these years. Inparticular, as for the first approach, we defined a transition period as well as the years before and after it. Given the fact that it was not clear for all the cases when the process was announced, we assumed that it started two years previous to the moment it was awarded.lo5Therefore, we could define a dummy for the transition and a dummy for the after-transition period, so that Equations (1) and (2) become: ln(y,, =dTDUMMY-TRAN,, dPDUMMY-P O S ~ , ) + +C @ijD, vu, + (1') ij ij ij where DUMMY -TRAN,, 1 if -2I sijt I + 1 0 otherwise 'OsA review of this arbitrary period definitionwith severalcountry analysts was performed and this criteria seemed to respondto most of the cases. 198 and DUMMY -POSqj,, 1 ifs,,22 0 otherwise where sotwas a time trend that had a value equal to zero for the year when the privatization was awarded. Inthis sense, the first dummy identified the average change in the dependent variable during the transition with respect to the average level previous to these years. The second dummy identified the average change o f the dependent variable after the transition with respect to the first period. The first basic specification were equations (1') and (2') using the log level o f the indicators. In particular, this helped us to identify most of the conclusions. For those variables that presented trends (for example, number of connections), equation (2') was more enlightening. However, it relied on the assumption that trends between the three periods of analysis are the same. In order to relax this assumption, we run a third set of equations (l'), but using the annual growth in each indicator. In this case, the first equation identified the average changes ingrowth betweenthe periods. Given the fact that we were using a semi-logarithmic functional form o f these models for each o f the indicators, when interpreting the coefficient estimates of the dummy, it should be remembered that the %age impact in each indicator is given by e' -1 (Halvorsen and Palmquist, 1980). In order to correct for potential non-spherical errors a Generalized Least Square (GLS) approach was more adequate. However, the GLS estimation requires the knowledge o f the unconditional variance matrix of vUr, R ,up to scale. Hence, we must be able to write R=02C,where Cis a known GxG positive definite matrix. But, in this case, as this matrix was not known, we followed a Feasible GLS (FGLS) approach that replaces the unknown matrix R with a consistent estimator. Finally, in order to identify the specific effects for Brazilian firms, we will interact these dummies with a dummy with value one for these firms and zero for the companies inthe rest of the region. Hence, the equation to estimate will be: where p' and p' will identify the effects above (or below) the average effect of all the firms inthe region. These average effects inthe regionfor all the firms (included Brazil) are identified with 6' and Sp,for the transition and after that, respectively. 199 B. CompleteTables Table Al.1. Brazil and LAC without Brazil: pre- andpost-privatization comparisons for electricity distribution (inlevels) - Variable stats Brazil i Pre privatization Post privatization % Change Pre privatization Post privatization YOChange (1) (2) (3) (4) (5) (6) OUtDUtS Residential mean 887,475 981,046 10.5% 263,912 291,111 10.3% Connections p50 626,530 678,638 8.3% 168,856 186,458 10.4% sd 813,331 937,996 368,908 383,736 N 34 34 52 54 MWH sold mean 5,844,224 6,658,088 13.9% 1,143,165 1,317,301 15.2% per year ~ 5 0 2,394,593 3,201,487 33.7% 568,284 681,942 20.0% sd 7,577,295 8,197,025 1,718,651 1,995,268 N 35 34 50 54 lnDuts Number of mean 2,803 1,871 -33.3% 1,290 638 -50.5% Employees p50 1,688 1,247 -26.1% 483 354 -26.7% sd 3,182 2,283 2,088 889 N 34 36 31 43 Efficiency Connections mean 432.9 642.8 48.5% 377.8 586.5 55.2% per employee p50 388.6 637.3 64.0% 337.7 517.7 53.3% sd 173.6 218.4 151.2 261.4 N 33 34 31 41 GWH per mean 2,111.2 3,278.8 55.3% 1,260.3 2,088.7 65.7% employee ~ 5 0 1,730.5 2,863.9 65.5% 1,150.4 1,898.9 65.1% sd 1,176.4 1,693.1 519.0 1,077.9 N 33 34 30 41 Distributional mean 0.16 0.15 -8.7% 0.18 0.14 -21.O% losses P50 0.14 0.12 -13.7% 0.17 0.14 -17.2% sd 0.08 0.08 0.07 0.05 N 31 33 38 48 Qualitv Durationof mean 31.75 22.96 -27.7% 54.48 18.31 -66.4% Interruptions p50 25.90 20.08 -22.5% 25.59 12.88 -49.6% per year sd 23.63 15.00 52.15 15.80 per consumer N 38 39 8 17 Frequencyof mean 30.86 21.49 -30.4% 43.82 13.87 -68.4% Interruptions p50 22.34 14.85 -33.5% 13.30 9.31 -30.0% per year sd 26.79 14.81 60.72 12.27 per consumer N 38 39 8 17 Coveraae Residential mean 0.82 0.86 5.0% 0.69 0.72 4.6% Connections p50 0.83 0.87 5.0% 0.71 0.75 5.9% per 100HHs sd 0.12 0.11 0.19 0.18 - N 32 31 44 46 Prices Avg price per mean 90.54 74.55 -17.7% 175.54 173.51 -1.2% residentialMWH p50 93.19 71.61 -23.2% 108.24 102.62 -5.2% (indollars) sd 14.90 16.98 182.96 188.98 N 40 40 42 50 Avg price per mean 90.64 102.46 13.0% 101.06 101.98 0.9% residentialMWH p50 90.17 102.17 13.3% 103.01 100.73 -2.2% (in real local sd 6.18 3.24 11.23 8.27 200 Table A1.2. Brazil and LAC without Brazil: pre- and post-privatization comparisons for fixed telecommunications (inlevels) Variable stats Brazil- LACwithout Brazil Pre privatization Post privatization % Change Pre privatization Post privatization % Change (1) (2) (3) (4) (5) (6) OUtDUtS Residential mean 4,328,333 6,768,279 56.4% 901,431 1,160,272 28.7% Subscribers p50 4,910,000 7,304,838 48.8% 301,544 441,000 46.2% sd 1,570,480 2,297,481 1,273,434 1,706,629 N 3 3 13 13 Million of mean 58,200 63,800 9.6% 6,380 6,830 7.1% Minutes per p50 70,000 71,700 2.4% 6,380 4,700 -26.3% year sd 26,000 23,700 4,730 8,030 !gg& Number of mean 23,358 16,396 -29.8% 12,478 11,674 -6.4% Employees p50 24,084 16,758 -30.4% 5,855 6,441 10.0% sd 10,751 5,868 14,022 13,439 N 3 3 13 12 Efflciency Connections mean 196.0 424.3 116.5% 65.6 93.2 42.1% per employee p50 204.6 424.4 107.4% 62.7 99.0 57.9% sd 25.5 28.3 29.9 44.4 N 3 3 13 12 Thousandsof mean 2530.87 3929.65 55.3% 687.83 464.59 -32.5% Minutes per p50 2388.51 3736.82 56.4% 343.11 463.40 35.1% Employee sd 335.55 382.82 898.04 200.25 N 3 3 7 5 Qual[tv Percentageof mean 46.50 40.62 -12.7% 49.68 56.75 14.2% uncempleted p50 44.66 40.05 -10.3% 52.73 60.45 14.7% calls sd 4.24 1.39 33.31 42.55 N 3 3 5 8 Percentageof mean 49.14 78.93 60.6% 41.16 50.51 22.7% digitalized p50 45.95 80.00 74.1% 21.93 42.11 92.1Oh network sd 8.38 1.93 36.78 36.19 N 3 3 11 12 Coverage Subscribers mean 10.26 15.43 50.4% 6.88 8.28 20.4% per 100 ~ 5 0 9.78 15.00 53.5% 5.87 7.20 27.0% Inhabitants sd 4.12 5.66 4.67 5.52 - N 3 3 13 13 Prices Avg price for a mean 0.03 0.04 35.9% 0.19 0.13 -33.1% 3-minute call p50 0.03 0.04 36.0% 0.07 0.09 37.2% (in dollars) sd 0.00 0.00 0.40 0.19 N 3 3 9 11 Monthly charges mean 3.38 7.13 111.1% 5.80 6.46 11.2% for a residential p50 3.38 7.02 107.8% 5.90 6.98 18.3% conexion sd 0.00 0.38 4.73 4.22 Jn dollars) N 3 3 10 11 Price for the mean 1062.87 50.96 -95.2% 559.62 371.57 -33.6% installation of a p50 1059.16 49.66 -95.3% 229.87 189.24 -17.7% residential line sd 6.43 2.25 859.19 465.96 (in dollars) N 3 3 10 11 Avg price for a mean 63.74 100.33 57.4% 99.12 81.83 -17.4% 3-minute call p50 63.74 100.44 57.6% 100.00 85.71 -14.3% (in real local sd 0.00 4.81 25.99 26.56 currency) N 3 3 7 7 Monthly'charges mean 48.95 122.97 151.2% 98.94 140.09 41.6% for a residential p50 48.95 120.54 146.2% 101.77 112.19 10.2% conexion (in sd 0.00 8.27 14.02 98.31 real local curr.) N 3 3 9 9 Price for the mean 432.86 22.12 -94.9% 105.20 134.01 27.4% installationof a p50 431.37 21.39 -95.0% 101.77 97.24 -4.5% residentialline sd 2.57 1.25 27.08 90.64 201 Table A1.3. Brazil and LAC without Brazil: pre- and post-privatization comparisons for - -- water (and sewerage) distribution (inlevels) Variable stats Brazil -D LAC without Brazil Pre privatization Post privatization % Change Pre privatization Post privatization % Change (1) (2) (3) (4) (5) (6) OUtDUtS Water mean 90,253 70,485 -21.9% 174,538 176,657 1.2% Connections p50 50,016 50,788 1.5% 77,185 83,595 8.3% sd 89,763 77,677 255,649 261,577 N 7 13 27 33 Sewerage mean 46,115 37,111 -19.5% 413,090 347,835 -15.8% Connections p50 37,404 29,299 -21.7% 66,183 46,767 -29.3% sd 37,407 36,625 1,400,850 1,180,083 N 7 12 25 30 Cubic Meters mean 42,382 21,018 -50.4% 161,000,000 124,000,000 -23.0% per year 27,616 20,198 -26.9% 52,600,000 46,200,000 -12.2% sd p50 49,383 15,147 308,000,000 248,000,000 lnDuts N 6 11 20 31 Numberof mean 650 238 -63.4% 1,117 634 -43.2% Employees p50 455 234 -48.6% 549 310 -43.6% sd 464 146 1,922 909 N 4 8 21 30 Efflciency Water mean 154.5 263.2 70.3% 237.5 334.7 40.9% Connections p50 155.4 256.5 65.1% 217.1 335.6 54.6% per employee sd 43.1 64.3 142.4 169.3 N 4 8 21 30 Cubic Meters mean 65.0 95.0 46.2% 123,578.4 166,064.3 34.4% per employee p50 57.5 85.8 49.1% 130,140.0 155,972.0 19.8% sd 30.7 32.4 63,004.7 86,337.4 N 4 8 18 30 Distributional mean 44.80 36.67 -18.2% 45.40 42.49 -6.4% losses ~ 5 0 43.93 32.61 -25.8% 40.27 38.45 -4.5% sd 12.26 11.23 13.76 13.57 N 4 6 19 22 Qualitv Continuityof mean 16.05 19.86 23.7% the service p50 18.00 23.02 27.9% sd 8.80 5.03 N 0 0 15 18 Potability mean 81.14 90.40 11.4% ~ 5 0 98.78 98.50 -0.3% sd 35.30 22.59 N 0 0 16 21 Coveraae Water connectionmean 69.04 64.54 -6.5% 73.90 74.29 0.5% per 100 HHs p50 80.52 78.96 -1.9% 81.74 88.55 8.3% sd 37.21 40.77 31.23 33.83 N 6 12 24 26 Sewerage mean 62.67 61.09 -2.5% 66.39 68.68 3.4% Connections p50 83.09 71.24 -14.3% 75.00 79.91 6.5% per 100 HHs sd 37.25 33.26 22.09 24.59 - N 5 7 19 20 Prices Avg price per mean 0.62 0.45 -26.4% 0.50 0.52 5.0% residentialMWH p50 0.60 0.49 -17.8% 0.59 0.53 -10.0% (in dollars) sd 0.15 0.15 0.25 0.18 N 4 7 11 14 Avg price per mean 0.64 0.39 -39.1% 0.04 0.08 86.8% residentialMWH p50 0.54 0.30 -43.3% 0.04 0.08 86.8% (in dollars) sd 0.32 0.19 N 4 7 1 1 Avg price per mean 106.90 83.94 -21.5% 95.28 118.62 24.5% residentialMWH p50 100.60 83.94 -16.6% 99.03 112.00 13.1% (in dollars) sd 11.43 8.79 8.15 29.90 N 3 2 11 11 Avg price per mean 109.21 85.27 -21.9% 95.92 145.97 52.2% residentialMWH p50 109.05 85.27 -21.8% 95.92 145.97 52.2% (in real local sd 9.30 11.95 currency - 100) N 3 2 1 1 - 202 Table A1.4. Brazil: before and after estimations inlevels for electricity distribution Variable stats Mean Diff in Levels T-stat (2-stat)for difference in means (medians)in levels Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(I) (+(I) (3)-(2) (3)-(1) (1) (2) (3) (4) (5) (8) (7) (8) (9) OUtDUtS Residential mean 87.19 101.89 119.08 15.41 16.09 31.28 -21.235"' -16.268"' -18.920"' Connections p50 86.58 101.96 118.22 15.95 15.88 31.03 -4.937"' -4.197"' -4.197"' sd 3.79 1.78 5.58 4.1 1 4.74 7.93 N 32 44 23 32 23 23 MWH sold mean 80.95 101.33 105.75 20.85 5.1 1 24.92 -8.392"' -2.954"' -8.002"' per year p50 82.87 100.67 103.60 19.12 4.57 28.53 -4.918"' -2.457" -4.042"' sd 14.53 3.08 9.71 14.06 8.48 14.61 N 32 44 24 32 24 22 Number of mean 172.95 102.68 92.58 -69.27 -11SI -89.25 9.045"' 2.317" 7.896"' Employees p50 168.23 100.05 88.17 -65.42 -13.08 -81.95 4.782"' 2.616"' 4.107"' sd 45.30 6.87 28.26 41.95 23.81 53.02 N 30 44 23 30 23 22 Efflciency 'Connections mean 54.88 101.59 137.73 47.27 34.87 86.34 -19.812"' -5.860"' -10.969"' per employee p50 56.02 100.32 135.78 48.79 31.50 89.98 -4.782"' -3.863"' -4.107"' sd 12.98 6.38 32.18 13.07 28.44 36.92 N 30 44 23 30 23 22 GWH per mean 53.18 100.89 119.90 47.74 19.82 72.38 -17.369"' -4.334"' -11.031*" employee p50 54.11 100.00 122.20 47.80 19.59 66.94 -4.782"' -3.498"' -4.107"' sd 16.07 5.66 26.42 15.05 21.93 30.78 N 30 44 23 30 23 22 Distributional mean 106.77 99.42 96.68 -5.47 -1.87 -9.44 1.441' 0.592 1.542' losses p50 100.31 100.00 93.36 -3.37 -2.53 -3.35 0.901 0.852 1.643 sd 20.24 8.48 19.27 18.98 15.14 27.39 N 25 44 23 25 23 20 Qualitv Durationof mean 139.29 102.51 71.69 -35.66 -27.65 -49.90 3.722"' 3.219"' 3.603"' Interruptions p50 127.79 98.48 67.51 -24.11 -29.39 -34.37 3.725"' 3.441"' 3.323"' per year sd 69.57 27.78 27.32 55.05 43.80 61.94 perconsumer N 33 44 26 33 26 20 Frequencyof mean 134.18 96.76 69.79 -37.85 -25.39 -53.49 4.224"' 5.500"' 3.843"' Interruptions p50 119.54 97.35 88.25 -22.96 -25.88 -32.86 3.743"' 4.026'" 3.883"' per year sd 60.31 12.32 19.98 51.48 23.53 62.25 perconsumer N 33 44 26 33 28 20 Coveraae Residential mean 98.33 101.05 110.54 5.14 8.78 12.97 -8.189"' -8759"' -6.133"' Connections p50 95.99 100.59 110.78 5.08 8.37 14.14 -4.180"' -3.998"' -3.821"' per100HHs sd 4.16 1.42 8.95 4.33 5.95 8.72 N 27 44 21 27 21 17 Avg Tariff per mean 113.08 94.62 78.58 -19.27 -13.79 -22.60 5.615"' 6.203"' 12.043"' residentialGWH p50 124.42 92.17 72.64 -20.99 -18.28 -21.38 3.881"' 4.015"' 4.461"' (in dollars) sd 25.62 8.36 15.13 21.16 11.55 9.57 N 38 44 27 38 27 26 Avg Tariff per mean 83.69 100.80 115.00 17.19 13.27 33.80 -18.092"' -7.071"' -15.430"' residentialGWH p50 83.68 100.14 115.77 19.22 11.81 35.19 -5.375"' -4.328"' -4.461"' (in real local sd 6.21 2.45 10.97 6.59 9.75 11.17 currency) N 38 44 27 38 27 28 'significantat10%;**significantat5%;***significantat1% 203 TableA1.5. Brazil: beforeand after estimations ingrowth for electricity distribution Variable stats Avg. Annual Growth Annual Diff in growth T-stat (2-stat) for difference in means (medians) in growth Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(I) (2)-(I) (3)-(2) (3)-(1) (1) (2) (3) (4) (5) (6) (7) (8) (9) OUtDUtS Residential mean 4.4% 4.4% 3.6% 0.1% -1.0% -0.5% -0.174 3.541"' 1.340' Connections p50 4.4% 4.3% 3.4% -0.1% -1.4% -0.9% -0.216 2.940"' 1.207 sd 1.5% 1.4% 1.6% 1.7% 1.2% 1.4% N 31 32 18 31 16 17 MWH sold mean 6.6% 4.6% 2.1% -2.0% -4.0% -3.7% 2.078" 3.022"' 4.401"' per year p50 6.2% 4.4% 2.6% -1.5% -2.7% -2.7% 2.195" 3.018"' 3.385"' sd 3.0% 4.0% 4.1% 5.4% 5.7% 3.4% N 31 34 19 31 19 17 &ggg Number of mean -8.7% -11.1% -0.8% -2.6% 14.5% 3.8% 1.017 -6.562"" -2.582" Employees p50 -7.1% -9.4% -1.5% -1.6% 13.9% 4.7% 1.I14 -3.680"' -2.275" sd 7.7% 8.2% 5.3% 13.6% 9.4% 5.8% N 29 34 18 29 18 16 Efficiency Connections mean 15.3% 19.0% 4.6% 3.9% -19.7% -5.0% -1.012 6.544"' 2.997"' per employee p50 12.7% 15.2% 6.0% 3.4% -16.2% -6.0% -1.481 3.724"' 2.534" sd 13.2% 12.0% 4.7% 20.9% 12.8% 6.6% N 29 32 18 29 18 16 GWH per mean 18.0% 19.0% 3.0% 1.0% -23.6% -8.7% -0.244 6.163"' 4.409"' employee p50 16.1% 14.7% 3.5% -1.6% -17.2% -9.9% -0.182 3.724"' 3.051"' sd 13.5% 14.8% 5.5% 23.1% 16.2% 7.9% N 29 32 18 29 18 16 Distributional mean 2.1% -4.1% 3.4% -4.8% 10.8% 2.7% 2.275" -4.225"' -1.070 losses p50 2.7% -3.9% 7.1% -7.7% 9.9% -0.2% 2.159" -3.337"' -1.013 sd 5.0% 11.4% 9.4% 10.1% 10.5% 9.0% Qualitv N 23 32 17 23 17 13 Durationof mean 4.1% -8.2% -5.4% -10.9% 2.2% -10.5% 1.331' -0.263 1.126 Interruptions p50 -5.4% -12.1% -2.8% -8.3% 9.4% -5.1% 1.979" -1.006 0.711 .~ wr Year sd 32.2% 19.8% 19.3% 45.4% 35.9% 30.9% perconsumer N 31 39 I 9 31 19 11 Frequencyof mean 2.3% -9.1% -7.2% -10.5% 2.4% -17.8% 1.528" -0.523 3.093"' Interruptions p50 -6.1% -6.7% -5.8% -2.4% -0.2% -14.4% 0.142 -0.523 2.490" per year sd 29.4% 13.1% 15.3% 38.4% 20.2% 19.1% wrconsumer N 31 39 19 31 I 9 11 Coveraae Residential mean 1.2% 1.7% 1.8% 0.5% 0.3% 0.4% -1.359' -0.724 -0.772 Connections p50 1.4% 1.7% 1.8% 0.3% 0.0% -0.1% -1.009 -0.427 -0.178 per 100HHs sd 1.2% 1.5% 1.6% 1.9% 1.4% 1.7% N 25 31 17 25 17 11 Avg Tariff per mean 12.6% -8.4% 8.7% -22.7% 15.1% -16.9% 7.481"' -2.818"' 2.523" residential GWH p50 13.9% -10.0% 1.5% -24.4% 11.3% -24.8% 5.045"' -2.243" 1.874' (in dollars) sd 18.3% 6.5% 19.9% 18.2% 23.9% 27.6% N 36 39 20 36 20 17 Avg Tariff per mean 16.5% 4.2% 3.6% -13.1% -1.7% -23.3% 7.080"' 1.411' 8.589"' residential GWH p50 15.0% 3.8% 5.2% -11.7% -1.9% -21.7% 5.155"' 1.645' 3.630"' (in real iocai sd 11.5% 3.3% 4.3% 11.1% 5.5% 11.2% currenc 36 39 20 36 20 17 significant at 10%; '* significant at 5%; **' significantat 1% 204 Table A1.6. Brazil: before and after estimations inlevels for fixed telecommunications - Variable stats Mean Diff in Levels T-stat (2-stat) for difference in means(medians)in levels Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(1) (2)-(1) (3)-(2) (3)-(1) (1) (2) (3) (4) (5) (6) (7) (8) (9) OUtDUta Total number mean 67.89 119.84 211.03 51.95 91.20 143.14 -13.413"' -6.006" -7.514"' of lines p50 66.24 120.76 195.49 48.82 77.34 126.16 -1.604 -1.604 -1.604 sd 4.89 4.38 29.64 6.71 26.30 33.00 N 3 3 3 3 3 3 Total number mean 92.07 109.80 152.78 17.72 43.25 52.06 -2.043' of minutes p50 94.05 109.52 152.78 9.29 43.25 52.06 -1.604 sd 9.79 6.59 15.03 N 3 3 1 3 1 1 !!u& Number of mean 133.27 94.00 47.77 -39.27 -46.23 -85.50 5.548" 5.631" 5.599" employee p50 128.52 91.98 47.38 -36.54 -44.60 -81.14 -1.604 -1.604 -1.604 sd 24.28 12.02 2.28 12.26 14.22 26.45 N 3 3 3 3 3 3 Efficienci Total number mean 52.43 138.06 439.48 85.64 301.42 387.05 -12.248"' -14.076"' -14.444"' of linesper p50 57.15 145.93 424.46 88.78 281.52 367.31 -1.604 -1.604 -1.604 employee sd 8.92 20.87 52.71 12.11 37.09 46.41 N 3 3 3 3 3 3 Total number mean 70.56 124.29 334.91 53.72 206.69 256.27 -6.060" of minutesper p50 73.42 128.22 334.91 49.58 206.69 256.27 -1.604 employee sd 9.82 22.09 15.35 N 3 3 1 3 1 1 Percentageof mean 111.56 94.19 69.91 -17.23 -24.27 -37.37 14.791" 5.788" 482.726"' IncompleteCalls p50 111.56 93.90 73.87 -17.23 -21.38 -37.37 1.342 1.604 1.342 sd 0.35 0.95 7.41 1.65 7.26 0.11 N 2 3 3 2 3 2 !&&y Percentageof mean 61.15 105.44 132.26 44.30 26.82 71.11 -9.360"' -7.993"' -12.108"' Digitalized p50 59.97 106.10 131.37 41.82 25.27 75.06 -1.604 -1.604 -1.604 Network sd 9.14 3.38 2.45 8.20 5.81 10.17 N 3 3 3 3 3 3 Coverapg Number of Lines mean 70.10 115.03 179.96 44.93 64.93 109.86 -50.511"' -8.586"' -13.011"' per 100HHs p50 69.15 113.33 177.39 44.41 58.51 102.93 -1.604 -1.604 -1.604 sd 3.99 3.34 12.29 1.54 13.10 14.62 N 3 3 3 3 Prices 3 3 Avg Pricefor a mean 46.26 80.97 43.62 34.71 -37.35 -2.64 -17.580"' 23.007"' 1.111 3-minute call p50 46.26 80.27 43.03 34.01 -36.69 -3.23 -1.604 1.069 1.604 (in dollars) sd 0.00 3.42 4.11 3.42 2.81 4.11 N 3 3 3 3 3 3 Avg monthlycharge mean 24.35 79.99 61.61 55.64 -18.38 37.25 -23.376"' 6.891" -10.115"' for residential p50 24.35 78.62 61.99 54.27 -16.84 37.64 -1.604 1.604 -1.604 Service (indollars) sd 0.00 4.12 6.38 4.12 4.62 6.38 N 3 3 3 3 3 3 Avg Charge for the mean 1921.20 150.87 22.99 -1770.32 -123.87 -1891.03 336.085"' - installationof a p50 1914.02 150.82 22.99 -1767.16 -123.87 -1891.03 1.604 residential line sd 12.44 4.05 9.12 (in dollars) N 3 3 1 3 1 1 Avg Pricefor a mean 49.62 93.74 81.33 44.1 1 -12.40 31.71 -15.497"' 3.782" -5.206" 3-minute call p50 49.62 92.75 77.85 43.13 -14.91 28.22 -1.604 1.604 -1.604 (in real local sd 0.00 4.93 10.55 4.93 5.68 10.55 currencv) N 3 3 3 3 3 3 Avg monthlycharge mean 25.64 94.26 115.09 68.62 20.83 89.45 -19.890'" -3.859" -10.187"' for residential p50 25.64 92.29 107.56 66.65 15.60 81.92 -1.604 -1.604 -1.604 Service (inreal sd 0.00 5.98 15.21 5.98 9.35 15.21 localcurrency) N 3 3 3 3 3 3 Avg Chargefor the mean 2104.46 157.79 44.19 -1946.68 -107.87 -2052.61 390.437"' - - installationof a p50 2096.80 157.63 44.19 -1944.74 -107.87 -2052.61 1.604 residentialline (in sd 13.27 5.81 8.64 real local currenc N 3 1 3 1 1 significant at 10%: '* significant at 5%; **' significant at 1% 205 Table A1.7. Brazil: before and after estimations ingrowth for fixed telecommunications Variable stats Avg. Annual Growth fi T-stat Z-stat or di erence in means (medians) inarowth Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(1) (2)-(1) (3)-(2) (3)-(1) (1) (2) (3) (4) (5) (6) (7) (8) (9) OUtPUtS Total number mean 9.4% 19.4% 2.2% 10.0% -17.1% -7.1% -6.162" 16.174"' 2.694' of lines p50 9.4% 19.4% 1.2% 9.3% -18.2% -9.0% -1.604 1.604 1.604 sd 0.7% 2.4% 4.1% 2.8% 1.8% 4.6% N 3 3 3 3 3 3 Total number mean 3.1% 5.0% 3.8% 1.9% 3.2% -0.8% -0.426 of minutes p50 3.7% 2.9% 3.8% -0.8% 3.2% -0.8% 0.000 sd 1.8% 5.8% 7.7% N 3 3 1 3 1 1 Number of mean -7.9% -11.4% -15.5% -3.5% -4.0% -7.6% employee p50 -9.2% -11.3% -14.4% -1.2% -3.4% -4.4% sd 3.1% 2.3% 3.4% 5.1% 1.2% 6.3% N 3 3 3 3 3 3 Efficiency Total number mean 18.8% 35.0% 18.8% 16.1% -16.1% 0.0% -5.483" 4.856" 0.001 of lines per p50 19.8% 34.5% 22.3% 14.5% -13.8% 2.6% -1.604 1.604 0.000 employee sd 4.2% 1.0% 6.1% 5.1% 5.8% 9.6% N 3 3 3 3 3 3 Total number mean 12.0% 18.5% 28.6% 6.5% 11.9% 19.1% -2.028 of minutes per p50 11.4% 16.7% 28.6% 7.2% 11.9% 19.1% -1.604 employee sd 3.0% 3.9% 5.5% N 3 3 1 3 1 1 Percentage of mean -1.7% -6.6% -7.7% -5.1% -1.1% -5.2% 1.419 0.460 18.825" Incomplete Calls p50 -1.7% -6.2% -8.3% -5.1% -3.1% -5.2% 1.342 0.535 1.342 sd 1.5% 2.8% 1.9% 5.1% 4.2% 0.4% Oualltv Percentage of mean 25.2% 16.2% 0.7% -9.1% -15.4% -24.5% 0.882 3.144" 3.886" Digitalized p50 23.6% 13.0% 0.9% -13.6% -12.0% -22.8% 1.069 1.604 1.604 Network sd 11.2% 8.2% 0.3% 17.8% 8.5% 10.9% N 3 3 3 3 3 3 Coveraag Number of Lines mean 9.4% 17.4% 6.8% 8.1% -10.6% -2.5% -12.015"' 1.867 0.419 per 100 HHs p50 9.5% 17.6% 4.3% 8.4% -13.3% -6.5% -1.604 1.069 0.535 sd 1.6% 1.2% 11.0% 1.2% 9.8% 10.4% - N 3 3 3 3 3 3 Prlces Avg Price for a mean 51.0% -2.8% -25.1% -53.8% -24.0% -76.2% 28.743"' 4.953' 22.847" 3-minute call p50 51.0% -2.6% -25.1% -53.7% -24.0% -76.2% 1.604 1.342 1.342 (in dollars) sd 0.0% 3.2% 4.7% 3.2% 6.9% 4.7% N 3 3 2 3 2 2 Avg monthly charge mean 97.2% 28.1% -21.3% -69.1% -51.9% -118.5% 20.831 15.326 205.372"' for residential p50 97.2% 26.6% -21.3% -70.6% -51.9% -118.5% 1.604 1.342 1.32 Service (in dollars) sd 0.0% 5.7% 0.8% 5.7% 4.8% 0.8% N 3 3 2 3 2 2 Avg Charge for the mean -23.9% -68.1% -37.5% -44.2% 31.3% -13.0% 425.658"' - installation of a p50 -24.5% -68.8% -37.5% -44.3% 31.3% -13.0% 1.833 residential line sd 1.0% 1.1% 0.2% (in dollars) N 3 3 1 3 1 1 Avg Price for a mean 44.3% 8.6% 0.9% -35.7% -9.6% -43.4% 17.057"' 1.547 9.661" 3-minute call p50 44.3% 8.8% 0.9% -35.5% -9.6% -43.4% 1.604 1.342 1.342 (in real local sd 0.0% 3.6% 6.4% 3.6% 8.7% 6.4% currency) N 3 3 2 3 2 2 Avg monthly charge mean 88.4% 43.1% 6.0% -45.2% -39.8% -82.4% 12.214"' 10.917" 105.937"' for residential p50 88.4% 41.4% 6.0% -47.0% -39.8% -82.4% 1.604 1.342 1.342 Service (in real sd 0.0% 6.4% 1.1% 6.4% 5.2% 1.1% local currency) N 3 3 2 3 2 2 Avg Charge for the mean -27.3% -64.4% -15.8% -37.0% 49.3% 12.1% 179.735"' - installation of a p50 -27.9% -65.1% -15.8% -37.2% 49.3% 12.1% 1.833 residential line (in sd 0.9% 1.3% 0.4% real local currency) N 3 3 1 3 1 1 ~ significant at 10%; *' significant at 5%; "'significant at 1% 206 Table A1.8. Brazil: beforeand after estimationsinlevelsfor water (andsewerage) distribution - Variable stats T-stat (2-stat)for difference Mean Diff in Levels in means (medians) in levels Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(I) (2)-(1) (3)-(2) (3)-(I) (1) (2) (3) (4) (5) (6) (7) (8) (9) OUtDUtS ResidentialWater mean 66.22 103.70 121.41 15.67 16.28 26.90 -9.797"' -4.199"' -13.710"' Connections p50 86.78 102.60 114.07 15.29 11.17 26.35 -1.826' -2.366" -1.604 sd 2.45 4.75 14.17 3.20 10.26 3.40 N 4 14 7 4 7 3 ResidentialSewer mean 87.06 101.85 113.40 19.31 12.50 24.58 -3.216" -5.329"' -16.769"' Connections p50 85.48 100.37 112.05 15.87 11.48 25.97 -1.626' -2.366" -1.604 sd 4.77 5.95 5.43 12.01 6.21 2.54 N 4 14 7 4 7 3 Cubic Meter of mean 100.12 99.82 9@0 -0.44 -3.17 -3.99 0.051 1.223 0.318 producedwater p50 100.12 100.00 98.05 -0.44 -3.15 -3.99 0.447 0.944 0.447 sd 12.80 4.38 4.78 12.30 5.80 17.76 !w& Number of mean 162.53 103.48 100.01 -41.36 -12.16 -56.87 2.077' 2.211' 2.181' Employees p50 173.91 100.00 103.08 -48.60 -10.42 -63.04 1.604 1.826' 1.604 sd 52.43 12.56 13.07 34.48 11.00 45.15 N 3 14 4 3 4 3 Efficiency Water connectionsmean 56.70 100.56 117.82 33.41 15.87 50.89 -4.997" -6.971"' -5.723" per employee p50 47.26 100.00 109.88 38.88 14.88 55.14 -1.604 -1.828' -1.804 sd 21.00 12.00 21.46 11.58 4.55 15.34 N 3 14 4 3 4 3 Distributional mean 113.47 101.04 67.29 -18.87 -36.35 -44.03 6.456" 1.743' 1.015 losses p50 113.47 100.00 65.14 -18.67 -47.34 -44.03 1.342 1.461 1.342 Sd 7.32 6.09 40.17 4.09 41.71 61.37 N 2 14 4 2 4 2 Coveraae ResidentialWater mean 95.48 102.92 117.30 5.79 11.58 11.44 -1.887' -1.795' -1.258 Connections p50 95.42 100.00 107.97 4.43 3.78 3.90 -1.826' -1.905' -1.604 per 100 HHs sd 3.31 5.58 22.59 6.14 14.40 15.75 N 4 14 5 4 5 3 ResidentialSewer mean 95.62 101.53 108.99 9.29 7.17 12.56 -1.632 -2.822" -1.693 Connections p50 95.47 100.00 109.45 8.05 8.01 8.98 -1.095 -1826' -1.E04 per 100 HHs sd 6.43 3.78 6.20 11.38 5.08 12.87 N 4 14 4 4 4 3 Avg price per cub. mean 146.85 100.55 69.83 -52.48 -33.86 -25.78 0.904 3.879" meter of water p50 146.85 100.00 64.96 -52.48 -34.45 -25.78 0.447 1.626' (indollars) sd 79.88 12.11 20.30 82.06 17.47 N 2 14 . 4 2 4 1 Avg price per cub. mean 110.97 106.42 1s.39 2.20 2.17 33.59 -0.064 -0.173 meter of water p50 110.97 100.00 118.57 2.20 8.44 33.59 -0.447 0.000 (in real local sd 36.66 16.36 41.47 46.32 25.03 currency) N 2 14 4 2 4 1 Avg price per cub. mean 148.02 102.25 95.33 -52.57 -16.81 -8.77 0.762 2.075' meter of sewer p50 148.02 100.00 92.87 -52.57 -24.58 -8.77 0.447 1.604 (indollars) sd 97.00 13.09 26.00 97.62 14.03 N 2 14 3 2 3 1 - Avg price per cub. mean 110.04 107.02 166.24 4.43 36.00 54.77 -0.105 -2.538' meter of sewer p50 110.04 100.00 142.34 4.43 45.24 54.77 -0.447 -1.804 (in real local sd 49.90 19.72 52.89 59.65 24.57 currency)) N 2 14 2 3 1 'significantat10%;**significantat5%;'*' significantat1% 207 Table A1.9. Brazil: beforeand after estimations ingrowthfor water (andsewerage) distribution Variable Stats Avg. Annual Growth Annual Diff ingrowth T-stat (Z-stat)for difference in means(medians)in growth Preprivat Transition Postprivat (2)-(1) (3)-(2) (3)-(I) (2)-(1) (3)-(2) (3)-(I) si!&!&ResidentialWater (1) (2) (3) (4) (5) (6) (7) (8) (9) mean 6.6% 7.4% 2.4% -2.6% -5.7% -2.0% 1.051 1.633' Connections p50 5.4% 4.6% 3.4% -0.2% -6.3% -2.0% 1.604 1.483 sd 3.3% 7.2% 4.2% 4.2% 7.8% N 3 10 5 3 5 1 ResidentialSewer mean -0.2% 4.0% 11.6% 9.5% 9.8% -1.7% -0.900 -0.995 Connections p50 4.9% 4.4% 2.9% -0.6% -0.2% -1.7% 0.000 -0.405 sd 10.0% 7.9% 21.0% 18.3% 22.0% N 3 10 5 3 5 1 Cubic Meter of mean -0.3% -4.2% 1.3% -0.7% 0.7% 8.2% 0.545 -0.156 producedwater p50 -0.3% 1.O% 1.5% -0.7% 1.1% 8.2% 0.447 -0.535 sd 4.5% 16.5% 3.6% 1.8% 7.3% N 2 7 3 2 3 1 l!l6!ub Numberof mean 1.9% -14.0% -3.3% -12.8% 12.9% -3.7% 2.555 -14.257" - Employees p50 1.9% -16.3% -3.3% -12.6% 12.9% -3.7% 1.342 -1.342 sd 2.7% 7.0% 0.6% 7.0% 1.3% N 2 4 2 2 2 1 Bfficlency Water Connectionsmean 3.1% 28.4% 6.3% 15.2% -21.8% 1.3% -2.072 43.603"' - per employee p50 3.1% 28.1% 6.3% 15.2% -21.8% 1.3% -1.342 1.342 sd 3.3% 16.4% 0.6% 10.4% 0.7% N 2 4 2 2 2 1 Distributional mean -7.1% 21.2% 0.1% 4.7% 6.7% 6.0% -0.983 -10.157" - losses p50 -7.1% -2.3% 0.1% 4.7% 6.7% 6.0% -0.447 -1.342 sd 5.8% 52.5% 4.2% 6.8% 0.9% N 2 4 2 2 2 1 Bveraag ResidentialWater mean 0.1% 7.2% 3.5% 2.5% -7.1% 2.5% -1.367 1.470 Connections p50 0.5% 4.5% 3.0% 2.4% -3.3% 2.5% -1.069 1.604 per 100HHs sd 0.7% 8.3% 3.7% 3.2% 8.3% N 3 8 3 3 3 1 Residential Sewer mean 2.2% 5.3% 2.9% 5.4% -0.3% -3.6% -0.816 0.282 Connections p50 0.6% 2.5% 2.9% -0.4% -0.3% -3.6% 0.000 0.447 per 100HHs sd 4.3% 7.1% 0.9% 11.4% 1.8% N 3 5 2 3 2 1 Avg priceper cub. mean 1.1% -11.9% -26.2% -25.5% -4.9% meter of water p50 1.1% -16.1% -29.8% -25.5% -8.1% (in dollars) sd 24.9% 7.7% 10.4% N 1 5 3 1 3 Avg priceper cub. mean 3.1% 6.8% 0.6% -15.1% -4.8% meterof water p50 3.1% 0.1% -4.3% -15.1% -6.5% (in real local sd 20.3% 10.4% 10.9% currency) N 1 5 3 1 3 Avg priceper cub. mean -12.7% -11.7% -13.7% -5.4% -0.8% meterof Sewer p50 -12.7% -12.9% -13.7% -5.4% -0.8% (in dollars) sd 7.2% 8.2% 14.1% N 1 4 2 1 2 Avg priceper cub. mean -10.9% 5.6% 17.7% 6.3% 2.9% meterof Sewer p50 -10.9% -0.5% 17.7% 6.3% 2.9% (in real local sd 15.2% 11.1% 30.1% currenc N 1 4 1 2 significantat 10%; ** significantat 5%; **' significantat 1% 208 ANNEX 4 Recent Evolutions of Water and Sanitation Sector inBrazillo6 A. Abstract Regulation inthe sanitation sector inBrazil i s facing a great deal of debate regarding the government level in which conceding authority should reside and how private operators can fulfill social objectives. The main objective o f this study i s to show that these issues are not the crucial barriers to the development of the sector when one looks at the productivity performance of the operators. Therefore, we elaborate a detailed analysis of the productivity performance of the current structure of the sanitation sector in Brazil. In doing so, we are able to analyze how jurisdiction o f operators as well as the nature of the management, either private or public, has affected performance of the current operators. Moreover, we investigate how the absence of tariff regulation has dissipated efficiencies and allowed the practice of monopolistic tariffs. B. Introduction The sanitation sector presents scale and scope economy characteristics that justify the presence of natural monopolies, particularly in its distribution operations. Indivisible large scale and long-maturity investments thus characterize the sector, that is, activities with "sunk costs". However, in contrast with other regulated traditional sectors, the sanitation sector is also characterized by low technological dynamism both in production and distribution. However, a wide variation in the adoption of technologies can be observed in each firm for technical (topography, water abundance, etc.) and for management reasons. Therefore, besides returns to scale, the sector's inefficiency sources should be more dependent on movements toward the frontier ("catch up effects") than properly on frontier movements. Besides those technologic features, the political organization of the sanitation sector in Brazil i s singular both in relation to its similar in the rest of the world and in relation to other regulated sectors inBrazil. For example, in the Brazilian case, although the conceding authority i s municipal, more than 80% of the population i s served by state public operators through concessions. Given the precariousness of the regulatory framework, the contracts for these concessions are incomplete interms of goals, tariff setting and transparency o f crossed subsidies. `06Background paper by Ronaldo Seroa da Motta and Ajax R. B. Moreira from PEA - Brazil. They thank trainees Enrique F. P. Morais, Katcha Poloponsky, and Leonard0 Moura for the support on organizing and screening the data. 209 Additionally, the sector's financing sources are dependent on federal resources. If a latent rejection to private capital insertion with a fragile concessions regime i s added to those characteristics, the design and implementation o f a regulatory framework becomes far from trivial in this superposition of competences and interests. Not surprisingly, the country does not have a clear and stable regulation for the sector yet. Once public funds were restricted from the 1990's onward, the sector failed to keep up with its impressive growth that made it possible to expand treated water services to more than 100 million people in 20 years. Therefore, investments have dropped by 30% as an average since 1998. The absence of a regulatory framework creates barriers and increases costs to the potential private investments inthe sector. Privately managed services reach less than4% of the served population in the country. This is crucial since a considerable investment effort estimated inUS$60 billion i s needed to reach reasonable service coverage goals in the next 20 years.lo7 Apart from the severe fiscal restrictions in the latest years, which have made public resources scarce, the sector had been showing loss of financing capacity since the 1980's when state public operators faced the deterioration o f their financing capacity due to the highinflationary processes. From the mid-l990's, after the Plan0 Real, firms failed to recover their financing capacity and tariff policies mostly due to their inefficient management practices protected by a weak regulatory framework lacking compatible incentives for productivity improvements. The private sector, in turn, could not find clear and stable regulatory mechanisms to accelerate its investments. In2001 the federal government sent to the NationalCongress a draft bill proposinga new regulatory framework. According to that proposal, the state would be the conceding authority inmetropolitan areas, instead of municipalities. This interpretation was a way to assure the financial viability of state firms by the scale gains to be captured in this territorial base and supposedly used as crossed subsidies to poor municipalities within the area coveredby the firm.'08 To promote the efficiency of services and the participation o f private capital, the bill also introduced incentive-based tariffs for monopolistic regulation, such as "price caps" and "yardstick competition". Nevertheless, this bill was weak in terms o f governance since it did not set up a specific regulatory agency to apply these instruments. The bill did not advance due to controversies in its interpretation of the Federal Constitution about the conceding authority of states. Questioning also arose about the role of the private sector, whose fate was believed to be the practice o f tariff abuse and the inability to keep an investment patternthat reached poorer areas. lo'As revealed by the Federal Sanitation Autority inits National Sanitation Polict documents. lo*Another controversial point was the asset pricing in case of termination of concession contracts with no specific clause. 210 With these concerns in mind, the federal administration that started in 2003 initiated the preparation of a new regulatory framework for the sector in which, among other issues, conceding authority i s kept to municipalities when the service i s of local interest, and pricing as well as concession procedures are regulated by autonomous authorities. Not surprisingly, the same controversies have stalled the process of drafting that proposal, confronting those supporting municipal conceding power against those willing to preserve the cross-subsidy system generated within the state public operators. Inaddition to that, the aversion to private capital participation is also presented inthe debate. The main objective o f this study i s to show that these issues are not the crucial barriers to the development of the sector when one looks at the productivity performance o f the operators. Therefore, we will generate a detailed analysis of the productivity performance o f the current structure o f the sanitation sector in Brazil. In doing so, we will be able to analyze how jurisdiction of operators as well as the nature of the management, either private or public, have affected performance of the current operators. Moreover, we wish to investigate how the absence o f tariff regulationhas dissipated efficiencies and allowed the practice of monopolistic tariffs. For that purpose, a summary of the evolution o f the sector's regulatory framework i s initially presented. Next the literature inproductivity measurement inthe sanitation sector i s overviewed. Then we describe and apply our models to analyze the current performance of water and sanitation services in Brazil considering their territorial scope (state or municipal) and the nature o f its management (public or private). Based on that we identify sources o f technical and scale efficiencies and how those efficiency indicators have affected effective tariff levels. C. The PresentRegulatoryFramework The sanitation services coverage inBrazil has made great advances inwater supply and is already above levels observed in countries of equivalent per capita income. Inthe sewer services case, on the other hand, we are still below those economies. Almost 90% of urban households in the country are today served with a treated water system against 61% in 1970. This increased coverage included almost 100million people into the system in30 years. On the other hand, the water supply coverage in rural areas is only 9%. The national coverage of the sewer system doubled inthe same period, but it still does not cover more than half o f the urban population. It should be added that only 27% o f the collected wastewater i s treated. Regional differences incoverage, inall services, are also persistent. When services coverage i s analyzed by income classes, a rather regressive pattern i s observed. The population with an income lower than 2 minimum wages (SM) presents a coverage rate below national average. The higher classes with more than 10 SM, in turn, present a coverage that is 25% higher in water supply and more than 40% higher in sewerage than the lower classes with up to 2 SM. 211 As can be observed, the sector's investment pattern in the last decades was unable to universalize services either in regional or distributive terms. Institutional and regulatory aspects help to understand these results, as will be next summarized. Until the 1970's water supply and sanitation services in Brazil were supplied by municipalities under the supervision of the National Health Foundation (Funasa), which was, in turn, supervised by the Health Ministry. In 1971 the National Sanitation Plan (Planasa) was created to take charge of the entire sector's investment planning as well as tariff and credit policy and other related norms. The quality standard of water and sewerage collection and treatment were kept in the health and environmental legislations, and enforced by their regulators. Planasa's main aim was to promote the creation of state firms for water supply and sanitation, encouraging municipalities to make long-term concessions to these firms in exchange for investments granted b y the Housing National Bank (BNH), Planasa's financial arm. About 3.200 municipalities among 4.100 adhered to the Planasa. This centralization was justified by the need to reduce the transaction costs of planning and to reach economies of scale, particularly inthe fast-growing metropolitan areas. At the end of the 1980's, the Planasa highly centralized system started to present low performance levels. The tariff regime was no longer appropriate due to a hyperinflationary environment, and the funds for investment were financially weakened by default. The 1988 constitutional reform and its emphasis on decentralization made the Planasa scheme obsolete. Today, around 80% o f the population are assisted by 25 state regional firms. The rest of the market i s composed by 255 local operators. Intotal, the private sector i s responsible for only 3.4% of all the served population. The Federal Constitution of 1988 granted to municipalities the right to make concessions for public services o f "local" interest, while acknowledged that federal and state governments should guarantee an efficient and well-regulated water and sanitation services supply. In addition to that, the Constitution requested states to legislate on metropolitan areas. Those constitutional requirements opened debate about what are services of local interest and whether the conceding authority on metropolitan areas should be under municipal or state governments. In February 1995 the new Concession Law was approved, challenging state firms monopoly, particularly in metropolitan areas. A great majority of the contracts in force were signed in the beginning of the 1970's with the support of Planasa. Therefore, some of them would have their term expired, and others are precarious with practically no specification about important aspects as rules for assets return, performance goals and even investment plans. That precarious aspect of those contracts hinders negotiated solutions for the return of concessions, generating a high level ofjudicial litigation. 212 Privatization was also affectedby rules about tariff criteria. The ConcessionLaw stated, as well, that the approval of tariff adjustments would be at the conceding party's discretion in the carrying out of contracts. Consequently, it establishes risks for any concessionary when there is no clear and well-defined tariff policy that guarantees contract rights,lo9 Inthe attemptto settle all those issues, the federal governmentprepareda new regulatory structure within a law of the National Sanitation Policy. Within this framework management and financial autonomy was established for sanitation firms, as well as behaviorrules, tariff policy principles, and concessioncriteria. It stipulated, however, that states had concession power in metropolitan areas to assure economies of scale and promote inter-municipal cross-subsidies, Although it creates incentives for private concessions and reckons on sophisticated pricing mechanisms, the bill was not clear about governanceandno regulatoryagency was createdfor this purpose. The bill was, however, strongly contestedby municipal sponsors of sanitation sectors and due to a great number of questions about privatization related to universal coverage and the possibility of breakingup the vertical structure of privatized services."' The bill did not succeed to be voted and the resulting regulatory weakness creates uncertainties for private investors and discourages investments. With these concerns in mind the federal administration that started in 2003 initiated the preparation of a new proposal for the regulatory framework of the sector. Among other issues, this proposal addresses the definition of services of local interest, tariff policy, regulatory authorities and concession procedures. The main economic regulation characteristics of that proposition's final version are the following: Besides water supply and sewerage services, solid waste and drainage services are also included, all under the common sector now named environmental sanitation; As statedinCF 88, municipalities will have concedingauthority inservices of local interest, to be defined as those of water distribution, sewer collection, solid waste collection andmicro drainage; Other services, such as water capture, water treatment, sewerage, and solid waste treatment and macro drainage, are only of local interest in case of exclusive use of the municipality. Otherwisethey are underjurisdiction of the benefited municipalities. In these cases of multiple uses (more than one municipality) a new approach of integrated management is adopted with consortiatedconcedingauthority; In case of municipalities that do not operate their own systems, concessions will have to be bided out, either for public or private operators, with clear and logA very well known case is the concession of the municipality of Limeira that had the contracted tariff readjustment reduced in 40% by the Mayor's initiative, and sustained in a Federal Court [see World Bank (1999 and 2000)l. 'loAnother point of dispute has been whether municipalities should reimburse the assets of state firms at the end of concession. In the bill this indemnity would be agreed between the parties, or in the absence o f an agreement it would be equivalent to the gross invoicing of three years for later judicial decision. 213 specific structure o f tariff setting and review, transparent plan o f investments and cross-subsidies, and targets for universalization; and, e Concessions, either to public or private firms, will be regulated by an autonomous agency. There are two main gaps in the current proposal for the sector's regulatory framework, namely: 4 There are no specific guidelines for tariff policy and, consequently, pricing rules affecting efficiency and reducing the share o f efficiency gains with users may dominate. b) Although consortiums will carry out integrated management of local and multiple use services, the text does not mention that these arrangements will be also controlled by regulatory agencies. If these gaps persist, the aim of introducing the so much needed incentives to efficiency and fair tariff levels may not be fully accomplished since such gaps may create room for uncontrolledmonopolist power as the one today enjoyed by regional operators. D. An Overview of the Literature As previously mentioned the sanitation sector in Brazil is quite singular. Sanitation service operators are monopolistic concessionaries intheir operation area. InBrazil they can be either a state public operator (regional) serving a group of municipalities within a state,"' or a municipal-based operator serving one (local) or more municipalities (multilocal) that can be either directly managed by the municipality or by a private operator. That is, the debate in Brazil i s not only on private against public management, but also, if not mainly, about municipal against state levels. The literature on productivity performance in the sanitation sector,"' however, emphasizes the distinction between private and public management and ownership. For example, this debate in the United States has provided a set of studies [Crain and Zardkooki (1978), Feigenbaum and Teeples (1983), Byrnes, Grosskopf and Hayes (1986); Fox and Hofler (1986), Bhattacharyya, Parker and Raffiee (1994)l developed with data from the American Water Works Association (AWWA). Apart from methodological divergences, results inthe United States have not been able to generalize the dominance of private operators over public ones regarding efficiency. In fact, the latest studies advanced the hypothesis that the tariff policy adopted inthe United States based on rate of return to capital has created inefficiencies that dominate the ones expected from public managerial practices. That could be expected since, as shown in Few major state-owned regional operators have gone public without loosing state-control management. '12As well as in other regulated sectors, see, for example, Kumbhakar and Hjalmarsson (1998), Forsund and Kittelsen (1998) and Bagdadioglu, Price and Weyman-Jones (1996) for the electricity sector and Coelli and Perelman (2000) and Cowie and Riddington (1996) for railways and Resende (2000) for the telecommunications. 214 Laffont and Tirole (1993), this pricing regime minimizes incentives to efficiency, once it assures that operators will recover any incurred costs. Estache and Rossi (2002) analyze how public and private water operators perform inthe Asia and Pacific Region, and have also found that both show similar efficiency levels. Although they do not apply any quantitative analysis, they indicate that the well- consolidated inefficient tariff setting based on cost recovery could be biasing results. InAfrica, where regulation is weak and politically driven andpricing is only tied inthe case of concession contracts, Estache and Kouassi (2002) identify that private operators perform at higher efficiency levels than public ones. Ashton (2000) shows that privatization increased efficient performance of water operators in the United Kingdom. The same results were found by Estache, Guasch and Trujillo (2003)for Argentina. Tupper and Resende (2004) is the only study of our knowledge estimating productivity for the Brazilian sanitation sector. The authors, however, only covered state operators, and found that they show differences in productivity levels that could be reduced if proper tariff setting mechanisms were inplace. Therefore, our study will fill the gap in the literature for Brazil, since we will cover the whole sector. We will also contribute by modeling how tariff setting i s affected by the absence of monopolistic price control in this special case of the Brazilian sanitation sector, where regulation i s absent. E. TheModels So, for the purpose of our study and limited by our database (a point that will be later discussed), operators will be classified as follows: Regional operators (R):state-ownedfirms covering several municipalities;' l3 Autarchic operators (A): services directly managed by the municipal administration (in various formats, such as, autarchies, departments, foundations, and micro-regional); Private operators (P): services operated by private firms offered to one or more municipalities through concessions. Our main assumption i s that there are no sound market mechanisms and incentives in place to regulate pricing, service coverage and investments in the W&S sector in Brazil. Therefore, pressure from users, size of the operator and its institutional format may play a role interms of productive performance. Ifso, one could expect that: '13Due to data availability, the Tocantins private operator was taken out of the sample. 215 a) Movements towards the frontier must prevail over and above frontier movements; b) Catchup effects will be stronger inP types that were concededinprocurement mechanisms through tariff bidding; c) Types R will enjoy economies of scale and greater degree of investment capacity associated to their larger sizes but they are less sensitive to political pressurefrom users due to their broadstate-controlmanagement; d) Types A instead face stricter political pressure and perceive local lower prices for labor, land and other inp~ts,''~ and consequently they tend to be more efficient with lower tariff levels than R types; and, e) Although effective tariff levels practiced by operators may keep a close correlation to their productive levels, the absence of regulation will induce that efficiency gains will bemostly capturedby operators andnot by users. Inthe following subsectionswe will describethe modelingto test these hypotheses. E.1. The Measurement ofEfficient Scores Operators of W&S serve multiple outputs, such as water production, water supply (in volume and connections), sewerage collection (in volume and connections), and sewerage treatment. Operators serve these outputs in distinct proportions and costs, and usually at one single price, so direct outputlinput ratios are not applied for efficiency measurement of each output. To deal with that, both regression analysis and data envelopment analysis (DEA) are widely usedinthe literature. The former measures a stochastic frontier following specific functional forms, whereas the latter adopts non-parametric techni ues. Eachapproachhas its own advantages and disadvantages, as discussedelsewhere."' We will, however, o t for DEA since it allows for direct measures of changes of productivity levels overtime. ' E The efficiency measures are then given by the ratios of the weighted sum of outputs to the weighted sum of inputs (weights are counterpartsof the parameters inthe regression analysis). The primal problem is to choose these weights subject to the constraint that no firm score is higherthan 1.00. Prices of the inputs paid by each operator are not known, so it is assumed that relative prices are the same for all operators, or, that the input proportions are optimal given the input prices. Here we will consider a special case in which inputs are aggregated by their prices with the aim of deriving overall operationcost efficiency for multiple outputs. Then the efficiency measure is the distance between the overall operation cost and the output transformation surface. Formally, let yk = (ylk,.., ydk) be the six outputs, as l4Sometimesthey also access morefavorablecredit schemes. '15See Cubbin and Tzanidakis (1998),for example, for a detailed and applied review. '16 Estimating the same productivity measures with parametric functions, we found a quite similar pattern of distribution of operator types indicating the same conclusionsobtained with DEA estimates. 216 previously mentioned (all measured in volume levels and also in connections for water supply and sewage collection) producedby the operatorsk at the total operationalcostxk. The primal form of DEA would be to identify the weights (pi) determining the output surface for each operator m at the time t, as shown inthe expression(1)below. The dual approach, instead, as indicatedinthe expression(2), chooses weights (hk)that lead to an explicit efficient measure (em) and with higher degree of freedom since the number of operatorsis higher than outputs (p