Report No. 20236-ME Mexico: Fiscal Sustainability (In Two Volumes) Volume II: Background Papers June 13, 2001 Mexico Country Management Unit PREM Sector Management Unit Latin America and the Caribbean Region u Document of the World Bank CURRENCY EQUIVALENTS Currency Unit - Mexican Peso (mxp$) EXCHANGE RATE MARCH 17, 2000 9.35 MXP / 1 USD WEIGHTS AND MEASURES Metric System FISCAL YEAR July I - June 30 ABBREVIATIONS AND ACRONYMS ADE Acuerdo de Apoyo Inmediato a Deudores de la Banca ADEFAS Adeudos de Ejercicios Fiscales Anteriores ASA Aeropuertos y Servicios Auxiliares BANCOMEXT Banco Nacional de Comercio Exterior, S.N.C. BANJERCITO Banco Nacional del Ejercito, Fuerza Aerea y Armada, S.N.C. BANOBRAS Banco Naciona] de Obras y Servicios Publicos, S.N.C. BANRURAL Banca Nacional de Credito Rural, S.N.C. BoM Banco de Mexico CAPUFE Caminos y Puentes Federales de Ingresos y Servicios Conexos CFE Comisi6n Federal de Electricidad CNBV Comisi6n Nacional Bancaria y de Valores CONASUPO Compafdia Nacional de Subsistencias Populares EMBI Emerging Market Bond Index FAMEVAL Fondo de Apoyo al Mercado de Valores FARAC Fideicomiso de Apoyo a] Rescate de Autopistas FIDEC Fondo para el Desarrollo Comercial FIDELIQ Fideicomiso Liquidario de Instituciones y Organizaciones Auxiliares del Credito FINA Financiera Nacional Azucarera FINAPE Programa para el Financiamiento del sector Agropecuario y Pesquero FIRA Fideicomisos Instituidos en Relaci6n con la Agricultura FNM Ferrocarriles Nacionales de Mexico FOBAPROA Fondo Bancario de Protecci6n al Ahorro FOPYME Programa de Apoyo Financiero y Fomento a la Micro, Pequeina y Mediana Empresa FOVI Fondo de Operaci6n y Financiamiento Bancario a la Vivienda GDP Gross Domestic Product IMF International Monetary Fund IMSS Instituo Mexicano del Seguro Social INEGI Instituto Nacional de Estadistica, Geografia e Informatica IPAB Instituto de Protecci6n al Ahorro Bancario ISSSTE Instituto de Seguridad y Servicios Sociales de los Trabajadores del Estado LFC Luz y Fuerza del Centro LOTENAL Loteria Nacional para la Asistencia Publica MXP Mexican Pesos NAFINSA Nacional Financiera, S.N.C. NIPA National Income Products Account OECD Organization for Economic Co-operation and Development PEMEX Petr6leos Mexicanos PIPSA Productora e Importadora de Papel SCNM Sistema de Cuentas Nacionales de Mexico SCT Secretaria de Comunicaciones y Transporte SHCP Secretaria de Hacienda y Credito Publico VAT Value Added Tax The Bank team that produced this report was headed by Stephen Everhart (LCSPE)-task manager, under the guidance of Marcelo Giugale (LCCIC)-program team leader. Members of the Bank team include: Craig Bumside (DECRG), Joost Draaisma (LCCIC), Robert Duval (LCCIC), Andrew Feltenstein (IMF, Virginia Tech), Russ Murphy (Virginia Tech), Claudia Sepulveda (LCSPR), and Aaron Schwartzman (Emst & Young-Mexico). Production assistance was provided by Michael Geller and Elizabeth Toxtle (LCC IC). The Bank appreciates the invaluable support and advice of Eliana Cardoso (LCSPE) and Vicente Fretes-Cibils (LCC4C). This study was undertaken under the general direction of Mr. Olivier Lafourcade (Director, LCCIC). Peer reviewers are: Messrs. Luis Serven (Lead Specialist - Regional Studies, LCSPR) and Anwar M. Shah (Principal Evaluation Officer, OEDCR). VOLUME I: EXECUTIVE SUMMARY TABLE OF CONTENTS Fiscal Sustainability-Mexico: A Synthesis Rationale for the Study .....................................................1l Issues and Focus ......................................................2 Fiscal Policy, Business Cycles, and Growth in Mexico ......................................................2 Infrastructure, Extemal Shocks, and Mexico's Fiscal Accounts ..........................3...........................3 Infrastructure, Private Costs, and Payoffs from Additions to Infrastructure ...................................................5 Fiscal Impact of Contingent Liabilities ......................................................6 Fiscal Deficit, Public Debt, and Fiscal Sustainability in Mexico ..................................................... 10 An Extension: Balance Sheet Approach and Quality of Fiscal Adjustments .......................I ........................ 16 The Mexican Case ..................................................... 19 Implications of the Balance Sheet Approach ..................................................... 23 Conclusions: The Link Between Fiscal Sustainability and Fiscal Reform ................................................... 25 References ..................................................... 27 List of Tables Table E. 1 Estimate of the Overall Cost of the Financial Rescue, June 1999 Table E.2 Contingent Liabilities Recognized by the Federal Government Table E.3 Mexico Federal Debt as a Percentage of GDP List of Figures Figure E. 1 Concentration and Growth of Subnational Debt, 1994-1998: Selected States Figure E.2 Gross Federal Debt as Percent of GDP: Intemational Benchmnarks Figure E.3 Selected Latin American Eurobond Spreads Figure E.4 Mexico Budget Indicators: 1980 - 1998 Figure E.5 Tax Revenue as Percent of GDP, Selected Countries 1992 - 1998 Figure E.6 Primary Deficit vs. Public Investment (percent of GDP) Figure E.7 Primary Deficit vs. Public Investment (percent of GDP) 5 UMI Countries Figure E.8 Primary Deficit vs. Public Investment (percent of GDP) 6 LMI Countries Figure E.9 Primary Deficit vs. Privatization Revenues (percent of GDP) Figure E. 10 Primary Deficit vs. Public Investment (percent of GDP) Mexico Figure E. 11 Primary Deficit vs. General Government Consumption (percent of GDP) Mexico Figure E. 12 Public Investment vs. Oil Prices Mexico Figure E. 13 Prograrnmable Expenditure Decomposition (percent of GDP) Mexico. Figure E. 14 Primary Deficit vs. Privatization Revenues (percent of GDP) Mexico Figure E. 15 Deficit Reduction and Oil Dependence Figure E.16 Components of Tax Revenues as a Percent of GDP, Mexico 1980-1998 iii VOLUME II: BACKGROUND PAPERS TABLE OF CONTENTS Chapter 1. Fiscal Policy, Business Cycles, and Growth in Mexico Perspectives on Mexico's Fiscal Accounts from 1980-98 .............................................. 2 The Business Cycle in Mexico ............................................. 3 Trends and Cycles in Mexico's Fiscal Accounts ............................................. 6 The Cyclically Adjusted Budget Surplus in Mexico ............................................. 16 Methods for Computing the Cyclically Adjusted Budget Surplus ................. ............................. 17 Budget Surplus Estimates for Mexico .............................................. 19 How Fiscal Policy and Output Affect Each Other in Mexico .............................................. 23 A Small VAR Model of the Mexican Economy ............................................. 24 How Does the Fiscal Surplus Affect Output? ............................................. 25 Dynamnic Behavior of the Fiscal Surplus ............................................. 27 Policy Conclusions ............................................. 29 References ............................................. 31 Appendix ......................................... 33 Chapter 2. Infrastructure, External Shocks, and Mexico's Fiscal Accounts Background ...................................... 43 Model Structure ...................................... 45 Production ...................................... 46 Ba.ank.ing ...................................... 48 Consumption ...................................... 48 The Government ...................................... 49 The Foreign Sector and Exchange Rate Determination ...................................... 49 Money Supply ...................................... 50 Data Sources, Calibration, and Simulation ...................................... 50 Simulations ...................................... 55 The Benchrnark Case ...................................... 55 A Shock to Confidence in the Banking System ...................................... 55 Trade Shock ...................................... 57 Conclusion ...................................... 60 References ...................................... 61 Appendix ...................................... 62 Chapter 3. Infrastructure, Private Costs, and Payoffs from Additions to Infrastructure Background ...................................... 68 The Model ...................................... 72 The Data ...................................... 74 Sanple Horizon ...................................... 75 Limitations ...................................... 75 Estimation of the Model ...................................... 76 Methods .......................................76 Infreastructure ...................................... 76 Estimation Limitations ...................................... 77 Results .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Overview ...................................... 79 iv Payoffs from Additions to Infrastructure ........................................................................ 84 Static Costs and Benefits of Increased Infrastructure ........................................................................ 85 Optimal Infrastructure Stocks ........................................................................ 86 Conclusions ....... ................................................................. 87 References ....... ................................................................. 88 Appendices ....... ................................................................. 90 Chapter 4. Fiscal Impact of Contingent Liabilities: The Case of Mexico Coverage of the Study ........................................................................ 107 Government Accounting Issues ........................................................................ 110 Methodology ........................................................................ 111 Deposit Insurance Scheme for Private Banks ......................... ............................................... 112 The 1995 Banking Crisis ........................................................................ 113 Expected Fiscal Costs of Resolution of the Crisis ........................................................................ 114 Government Credit Assistance Programs ........................................................................ 120 Characteristics of Direct Loans and Loan Guarantees ........................................................................ 121 Expected Fiscal Cost of Government Credit Programs and the Budget ......................... ............................ 126 Liabilities Related to Social Security Programs .............................. .......................................... 127 The Financial Condition of IMSS ........................................................................ 127 The Financial Condition of ISSSTE ........................................................................ 128 Expected Fiscal Costs of Social Security Programs and the Budget ................................. ......................... 129 Private Provision of Infrastructure and Government Guarantees ........................................... .................... 129 Power Plants ....... ................................................................. 129 Highways ........................................................................ 130 The Fiscal Cost of Government Insurance Programs ................................................................. ....... 130 Policy Implications ........................................................................ 132 References ........................................................................ 134 Chapter 5. Fiscal Deficit, Public Debt and Fiscal Sustainability in Mexico The Mexican Fiscal Accounts: Stylized Facts ........................... ............................................. 138 Debt Management ........................................................................ 140 Budget Indicators ........................................................................ 142 Tax System ....... ................................................................. 143 Government Expenditure Comnposition ........................................................................ 146 Is Mexican Fiscal Policy Sustainable? ........................................................................ 148 Accounting Approach to Fiscal Solvency ........................................................................ 148 Pricing Approach to Fiscal Solvency ........................................................................ 149 Intertemporal Approach to Fiscal Solvency: The Medium and Long Term ................. ............................. 151 The Short and Medium Term ........................................................................ 152 The Long Term: A Time Series Analysis 1980:01-1999:05 ....................................................................... 154 Testing the Interternporal Budget Constraint: Unit Roots ........................................................................ 155 The Case of a Stochastic Discount Rate ........................................................................ 156 Testing for a Change in Regime ........................................................................ 157 Testing the Intertemporal Budget Constraint: A Co-integration Approach .................. ............................. 160 Testing Long-Run Relationship Between Government Spending Inclusive of Interest Payments and Revenue ....... ................................................................. 161 Testing Long-Run Relationship Between Govermnent Spending Exclusive of Interest Payments, Interest Payments and Revenue ........................................................................ 163 Policy Conclusions ........................................................................ 165 References ........................................................................ 167 Appendix ........................................................................ 169 v List of Tables Table 1.1 Sunmmary Budget Figures, 1980-81 1997-98 Table 1.2 Cyclical Properties of Public sector Revenue and Expenditure Table 1.3 Main Components of Public Sector Revenue and Expenditure, 1980-81 and 1997-98 Table 1.4 Estimnates of Revenue and Expenditure Elasticities Table 1.5 Impulse Response Functions from the VAR Table 1.6 Variance Decomposition of Output Table 1.7 Variance Decomposition of the Unadjusted Primary Fiscal Surplus Table l.Al Estimates of a Piecewise Linear Trend in the Logarithm of Seasonally Adjusted Real GDP Table 2.1 Real GDP, 1980-97 Table 2.2 Stocks of Infrastructure, 1970-90 Table 2.3 Cost Elasticities by Sector and Infrastructure Type Table 2.4 A Benchmark Simulation, 1995-2000 Table 2.5 Reduction in the Interest Elasticity of Money Demand, 1995-2000 Table 2.6 Interest Elasticity Decline Combined with an Infrastructure Increase 1995-200 Table 2.7 Trade Shock: Real World Income Stagnates, 1995-2000 Table 2.8 Trade Shock Combined with an Infrastructure Increase, 1995-2000 Table 2.9 Infrastructure Elasticities = 0 Table 3.1 Compound Annual Growth Rates 1960-93 Table 3.2 Infrastructure Compound Annual Growth Rates, 1960-93 and 1983-93 Table 3.3 Physical Infrastructure, Average Annual Growth Rates Table 3.4 Correlations: Physical and Financial Infrastructure Measures Table 3.5 Primary Data: Means and Standard Deviations Table 3.6 Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Table 3.7 Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Table 3.8 Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Table 3.9 Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Table 3.10 Mean Elasticities Table 3.11 Static Cost and Benefits of Increased Infrastructure Table 3.12 Optimal Infrastructure Stocks Table 3.13 OLS Panel Elasticities Table 3.14 OLS Random Effects Elasticities Table 3.15 FGLS Elasticities - Hetero, ARI Table 3.16 Growth Rates Used for Nominal Electricity Infrastructure Table 3.17 Coefficient Estimates Table 3.18 Coefficient Estimates (se's in (s), Dunmy Variables Excluded Table 4.1 Fiscal Risk Matrix Table 4.2 Federal Insurance Programs with Major Fiscal Risks Table 4.3 State Government Debt Table 4.4 State Government Pension Liabilities: 1997 Table 4.5 Pro Forma Balance Sheet of FOBAPROA, February 1998 Table 4.6 Estimation of the Cost of the Financial Rescue, June 1999 Table 4.7 Executed Fiscal Cost on Programs of Financial and Debtors Rescue Table 4.8 Estimates of Total Losses of Resolving Major Bank Insolvenses Table 4.9 Government Loans by Major Program Area, Fiscal 1997 Table 4.10 IMSS Retirement System Actuarial Deficit, December of 1994 Table 4.11 Government Net Liability as a Result of the 1995 Pension Reform Table 4.12 Pro Forma Balance Sheet of FARAC as of November 1998 Table 4.13 Contingent Liabilities and Fiscal Deficit Adjustments Table 4.14 Contingent Liabilities Recognized by the Federal Government Table 5.1 Accounting Approach Mexico in 1998 Table 5.2 Short and Medium-Term Indicators of Fiscal Sustainability as a percentage of GDP Table 5.3 Testing for Nonstationarity in Undiscounted and Discounted Net Public Debt, 1980:01- 1999:05 vi Table 5.4 The Zivot Andrews Unit Root Test for Undiscounted and Discounted Public Debt Table 5.5 Testing for Nonstationarity in Real Interest Rates, 1980:1-1998:07 Table 5.6 Testing for Nonstationarity in Real Government Spending Inclusive Interest Payments and Government Revenues, 1980:1-1999:05 Table 5.7 Results of Co-integration Government Spending Inclusive Interest Payments and Government Revenue, 1980:01-1999:05 Table 5.8 Testing for Nonstationarity in Real Government Spending, Interest Payments and Government Revenues, 1980:1-1999:05 Table 5.9 Results of Co-integration Noninterest Government Spending, Interest Payments and Government Revenue, 1980:01-1999:05 List of Figures Figure 1.1 Real GDP in Mexico, 1980-98 Figure 1.2 Seasonally Adjusted Real GDP, 1980-98 Figure 1.3 (a) Trends and Cycles in Real GDP: HP Trend Figure 1.3 (b) Trends and Cycles in Real GDP: Deviations from Trend Figure 1.4 Trends in Public Sector Revenues, 1980-98 Figure 1.5 Cyclical Components of Revenue, 1980-98 Figure 1.6 Trends in Public Sector Expenditure, 1980-98 Figure 1.7 Cyclical Components of Expenditure, 1980-98 Figure 1.8 The Budget Surplus and the Fiscal Imnpulse, 1980-98 Figure 1.9 Cyclical Fluctuation in Output Caused by Fiscal Shocks Figure l.Al Trends in Real GDP Figure l .A2 Cyclical comnponents of Real GDP Figure 3.1 Changes in Electric, Transport, and Communications Infrastructure Figure 3.2 Education Infrastructure Index Figure 4.1 Contingent Liabilities related to Potential Crisis of the Banking Sector Figure 5.1 Mexico Public Net Debt and Primary Deficit (+) as a percentage of GDP, 1980 - 1998 Figure 5.2 Mexico Overall and Primary Deficit (+) as a percentage of GDP, 1980-1998 Figure 5.3 Mexico Domestic and Foreign Public Net Debt as a percentage of GDP, 1980-1998 Figure 5.4 Public Sector Domestic Debt as a percentage of GDP, 1982-1998 Figure 5.5 Mexico Public Foreign Debt: Term Structure, 1982-1998 Figure 5.6 Mexico Budget Indicators, 1980-1998 Figure 5.7 Total Tax Revenues as a percentage of GDP-Selected Countries Figure 5.8 Oil Revenues as a percentage of Total Revenues Figure 5.9 Seignorage as a Source of Government Revenue Figure 5.10 Federal Government Revenue Tax Mix Figure 5.11 Total Expenditure of the Central Government as a percentage of GDP-Selected Countries Figure 5.12 Total Governnent Expenditure (million p$1994) Figure 5.13 Expenditure Composition Figure 5.14 Real Annualized Interest Rate CETES 28 days Figure 5.15 Brady Bonds Discounts Figure 5.16 EMBI Spread Rate Figure 5.17 Credit Rating for Mexico (100 lowest chance of default) Figure 5.18 Mexican Net Public Debt, 1980:1 - 1999:5. Undiscounted at Market Value (in Bill. P$ 1994) Figure 5.19 Sequential Zivot-Andrews Unit Root Test for the Mexican Undiscounted and Discounted Public Net Debt, 1980:1-1999:5 Figure 5.20 Government Spending Inclusive Interest and Govermment Revenues, 1980:1-1999:5 Figure 5.21 Governrment Spending, Interest Payments and Government Revenues: 1980:01-1999:05 vii 1 FISCAL POLICY, BUSINESS CYCLES, AND GROWTH IN MEXICO 1.1 This chapter investigates one aspect of the sustainability of fiscal policy in Mexico. It focuses on the role fiscal policy plays in detennining output in the short and medium term. It also looks at how fiscal policy, in turn, responds to the business cycle. Finally, it investigates the persistence of fiscal policy and how the authorities can use this persistence to forecast the government's financing needs. 1.2 The chapter looks at these particular issues for a number of reasons. A traditional role fiscal policy plays in industrial economies is that of a cyclical stabilizer. Fiscal policy is typically designed to "lean against the wind." That is, it is usually designed to stimulate output when the economy moves into recession and to be contractionary when an expansion broadens. This is usually accomplished in two ways. The first way is by having components in the budget that respond automatically to the business cycle, such as tax revenues (which respond positively) or unemployment benefits (an expenditure item that responds negatively). The second way is by using discretionary components in the budget to provide a stimulus during bad times. A fiscal policy designed in this way leads to a strongly procyclical budget surplus. 1.3 Mexico's fiscal policy does not lean against the wind. The analysis in this chapter will show that the budget surplus is quite strongly countercyclical, so that fiscal policy leans with the wind. The automatic stabilizers in place are weak, and are further weakened by the tendency of another automatic component of the budget, oil-based revenue, which responds sensitively to exogenous world oil prices, to move countercyclically. Furthermore, the discretionary component of the budget surplus also tends to move countercyclically. 1.4 If fiscal policy simply did not matter, then whether or not it leaned with or against the wind would be of little consequence. However, in Mexico, as in many other countries, fiscal policy does matter. The analysis suggests that an increase in the discretionary surplus of 1 percent of gross domestic product (GDP) causes GDP to decline by 0.6 percent in less than a year. Because in Mexico such increases typically occur during contractions, and these contractions are relatively short-lived (typically less than two years), this implies that discretionary policy exacerbates the cycle. 1.5 The results imply that Mexico's fiscal policy lacks a design that makes it a stabilizing feature of the economy. Furthermore, it has not been designed to render itself more sustainable. With procyclical fiscal policy (a countercyclical fiscal surplus), deficits cause debt to accumulate during economic expansions, but when the economic expansion inevitably ends, this debt suddenly becomes extremely costly. To finance it, the government must either take drastic discretionary fiscal measures, or it must finance the debt by borrowing at high real interest rates, Chapter I or by printing money and inducing rapid inflation. No matter which action the government takes, the implications are similar: a worsening of the economic downturn. This chapter is intended to encourage policymakers to forrnulate fiscal policies that can smooth, rather than exacerbate, real cycles, and that are therefore more readily sustained in the medium term. 1.6 The chapter begins by looking at the data. The sample period studied here-1980 through mid-1998-spans several episodes in recent Mexican economic history. The choice of period was largely driven by the availability of data. Quarterly national accounts data for Mexico are available from 1980 onward, while monthly fiscal accounts are available from 1977 onward. Trends and cycles in national accounts measures of real GDP are identified. Similarly, with GDP-based definitions of the business cycle in mind, this section describes the trends and cyclical fluctuations observed in various components of the public sector's fiscal accounts. 1.7 The second section of the chapter introduces the concept of the cyclically adjusted budget surplus. Historically, the main role of cyclically adjusted budget surplus measures has been as tools in policy analysis in the knowledge that cyclical movements in output affect the public sector's budget surplus. Cyclically adjusted budget surplus measures attempt to factor cyclical effects out of conventional measures. Once this is done, the adjusted measures are taken to be indicators of the stance of fiscal policy. This section examines a preferred definition of the cyclically adjusted budget surplus for Mexico. A question that naturally arises is whether this, or any other, adjusted budget measures are of particular use in policy analysis. 1.8 The third section of the chapter moves on to a more complex analysis of the data. Rather than working with simple indicators of the stance of fiscal policy, this section builds a simple structural model of the Mexican economy that isolates several important features, namely: * The nature of the feedback rule that implicitly determines fiscal policy, including the effects of economic activity on the budget - The exogenous shocks to the budget - The short- and medium-run effects of these shocks on economic activity * The extent to which forecasting the public sector's financing requirement is possible. 1.9 The main purpose of such a model is that the summary measures presented in the second section are typically useful in the context of a narrowly defined economic model. Furthermore, those summary measures are generally used to describe the effects of current policy on current activity. As such, given the lags with which fiscal policy is implemented and its effects are felt, the more forward-looking analysis of the third section should be of greater use to policymakers. Perspectives on Mexico's Fiscal Accounts from 1980-98 1.10 This section examines quarterly data on Mexico's fiscal accounts from 1980 through mid- 1998. While monthly budget data are available dating back to 1977, high-frequency data on GDP are only available from 1980 on. This section starts by defining the business cycle in Mexico with reference to quarterly data on real GDP from the national accounts. It then divides the fiscal accounts into their revenue and expenditure components and looks at trends in revenue and expenditure. 2 Chapter I The Business Cycle in Mexico 1.11 Figure 1.1 illustrates the behavior of real GDP in Mexico from 1980 through 1998. The raw data show a clear pattern of seasonality. Overlying the general upward trend and cycles is a pattern that indicates relatively low production in the first and third quarters and relatively high production in the second and fourth quarters. To identify these underlying features in the data, a seasonal adjustment filter was applied to the data.' Figure 1.1 Real GDP in Mexico, 1980-98 1500 - 1400 - 1300- 1200- 1100 1000 900 800 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 Source: National accounts. 1.12 Figure 1.2 shows the seasonally adjusted figures. This figure also delineates recessions using shading.2 Several episodes are worth noting namely: * The recession of 1982 through mid-1983 associated with the debt crisis * The period of slow growth thereafter, followed by the recession of late 1985 and 1986 * The implementation of the stabilization program in 1988, with an initial, slightly recessionary, year * The long expansion of 1989-94 * The short and intense recession of 1995 associated with the peso crisis and the subsequent recovery. 1. The adjustment procedure mimics the X - 11 seasonal adjustment procedure used by the U.S. Bureau of the Census. 2. The following definitions of expansions and recessions were used to generate the shading in figure 1.2. If the economy was not already deemed to be in a recession and seasonally adjusted real GDP fell for two successive quarters, these quarters were marked as the beginning of a recession. Until seasonally adjusted real GDP rose for two successive quarters, all subsequent quarters were also deemed to be part of the same recession. A similar, but opposite, definition was used to define an expansion, with the further assumnption that at the beginning of the sample (first quarter of 1980), the economy was in an expansion. 3 Chapter 1 Figure 1.2 Seasonally Adjusted Real GDP, 1980-98 1600 1500 ~1300 1200- 1100 1000 900 800 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 Source: Author's calculations. 1.13 One way to describe the cyclical properties of fiscal policy would involve comparing the behavior of revenue and expenditure during recessions to their behavior during expansions. However, with reference to figure 1.2, such an approach would clearly be unsatisfactory, because all expansions and all contractions are not alike. For example, the downturn after the peso crisis of December 1994 was much sharper and deeper than the ones experienced during the previous recessions. It involved a cumulative decline in output of 9.7 percent, versus 6.8, 4.7, and 0.8 percent in the previous recessions, and lasted two quarters, compared with six, five, and three quarters in the previous recessions. Similarly, the expansion since that recession has been more rapid than any of the previous expansions. 1.14 Furthermore, whether revenue and expenditure will differ according to whether output is rising or falling, rather than differing according to whether output is high or low is not obvious. The extent to which output, Y,, is high or low is typically measured with regard to some benchmark, Y,<. That is, the business cycle is typically defined as Y,' = Y, / Y, . The literature uses a number of benchmarks that can be the basis of a measure of the business cycle: * The level of potential output * The trend in output as defined by a linear, possibly piecewise, trend in its logarithm * The trend in output as defined by the Hodrick and Prescott (HP) (1997) filter * The permanent component in output as defined by the Beveridge and Nelson (1981) decomposition * The trend in output as defined by a peak-to-peak trend line. 4 Chapter 1 1.15 The next two sections focus on the third definition of the trend in output, the one defined by the HP filter with its parameter, A, set equal to 1,600 as is conventional for quarterly data. 1.16 Figure 1.3(a) illustrates the trend defined by the HP filter. Figure 1.3(b) illustrates the cyclical component as defined by the HP filter. Note that all the recessions marked in figure 1.2 correspond to cyclical downturns as defined by the filter. 1.17 As the technical appendix shows, the cycle defined by the HP filter is similar to the cycle defined by the piecewise linear trend. In addition, the cyclically adjusted fiscal surplus is not very sensitive to the definition of the cycle that is used. Similar measures result from the HP filter, a simple linear trend and a piecewise linear trend. The technical appendix discusses the other techniques. Figure 1.3(a) Trends and Cycles in Real GDP (a) Hodrick-Prescott Trend 1500 - X 1400- 1300- 1200- 1100- 1000 I 900 - 80 82 84 86 88 90 92 94 96 98 Source: Author's calculations. 5 Chapter 1 Figure 1.3(b) (continued) Trends and Cycles in Real GDP (b) Deviations from Trend 8- 6 - 0 - -4- -6- 80 82 84 86 88 90 92 94 96 98 Source: Author's calculations. Trends and Cycles in Mexico 's Fiscal Accounts 1.18 This subsection examines Mexico's fiscal accounts using a similar approach to the one used in the previous subsection. As in the analysis of output, the definition of trend and cycle used to analyze the fiscal accounts was the one given by the HP filter. Data from the fiscal accounts were treated similarly to the GDP data; i.e. they were converted into real terms by dividing by the GDP deflator and because many of the resulting series displayed seasonal patterns, they were further processed to remove the seasonal components. 1.19 The accounts used here are those of the public sector provided by the Mexican authorities. Included in the definition of the public sector is the federal government and public sector enterprises. State governments are only considered to the extent that federal government transfers to them are included as expenditure items. 1.20 Table 1.1 presents summary figures. These show that Mexico moved from a position of large fiscal deficits in the early 1980s to a position of near balance in the late 1990s. The primary deficit was narrowed in the mid-1980s in response to the debt crisis. This mostly occurred through a reduction in noninterest expenditure, although government revenue also showed a modest tendency to decline relative to GDP. Noninterest spending now represents about 20 percent of GDP, in contrast to the close to 30 percent of GDP it accounted for in the early 1980s. Total expenditure did not begin to decline until after the 1988 stabilization, when interest expenditure began to fall with the decline in inflation. 6 Chapter I Table 1.1 Summary Budget Figures, 1980-81 and 1997-98 (percent) GDP Budgeted Revenue Budget category: 1980-81 1997-98 Average 1980-81 1997-98 Average Economic surplus -8.6 -1.3 -5.3 -34.9 -6.1 -19.4 Primary surplus -4.9 2.5 3.3 -20.0 11.5 12.6 Totalbudgetedrevenue 24.6 21.6 25.6 100.0 100.0 100.0 Total budgeted expenditure 32.2 22.8 30.4 130.9 105.8 117.6 Interest 3.4 3.7 8.3 13.7 17.1 30.8 Noninterestexpenditure 28.8 19.1 22.1 117.2 88.6 86.7 Extrabudgetary surplus -1.0 0.0 -0.3 -4.0 0.0 -1.2 Extrabudgetary primary surplus -0.7 0.0 -0.2 -2.8 0.1 -0.7 Difference due to financing sources 0.0 -0.1 -0.2 0.0 -0.3 -0.6 Source: Public sector accounts. 1.21 On the revenue side of the accounts, a fairly detailed breakdown is available. Several of the available series are displayed in figure 1.4. Figure 1.4(a) shows total budgeted revenue, which displays a very different pattern than GDP. Unlike GDP, public sector revenue grew rapidly through 1985, declined in the 1986 recession, and then rose more slowly in the 1990s, with a marked decline during the 1995 recession. Yet despite the different trend behavior, revenue in 1997-98 (July-June) represented 21.6 percent of GDP, only a little less than the 24.6 percent of GDP it represented in 1980-81. Figure 1.5(a) displays the deviations of total budgeted revenue and of GDP from trend. As table 1.2 indicates, the cyclical movements of total revenue are not that highly correlated with output, with a correlation of only 0.2. Revenue was also somewhat more volatile than output, with a standard deviation of 4.4 percent (1.1 percent of GDP), as opposed to 2.6 percent for GDP. A different picture emerges once components of revenue are considered. Federal government revenue can be divided into tax and nontax components, displayed in figures 1.4(b) and 1.4(c) respectively, while other public sector revenue is displayed in figure 1.4(d). Federal tax revenue grew substantially during the period under review, as it has in many industrializing countries, but not as a percentage of GDP (see table 1.3). Notably, tax revenue fell sharply-by about one-third--during the 1995 recession. 1.22 A different picture emerges once components of revenue are considered. Federal government revenue can be divided into tax and nontax components, displayed in figures 1.4(b) and 1.4(c) respectively, while other public sector revenue is displayed in figure 1.4(d). Federal tax revenue grew substantially during the period under review, as it has in many industrializing countries, but not as a percentage of GDP (see table 1.3). Notably, tax revenue fell sharply-by about one-third---during the 1995 recession. 7 Chapter I Figure 1.4. Trends in Public Sector Revenue, 1980-98 (a) Total Budgeted Revenue (b) Tax Revenue ss________________ . __________ 40 75 355 65 3 45 20 90 8Z 84 86 S8 90 92 94 96 98 80 82 84 86 98 90 92 94 96 98 (c) Non-Tax Federal Revenue (d) Other Public Sector Revenue 25 145 220 235 so 30 ~~~~~~~~t5~~~~~~~~~~~~2 20 80 92 94 86 98 90 92 94 96 98 S0 S2 S4 96 SS 90 92 94 96 98 (e) Income Tax Revenue (h Revenue from Goods & Services Taxes 2 0 - _2___.5 is to0 o t oo 90 92 84 96 99 90 92 94 96 9S go 92 94 36 99 90 92 94 96 99 (g) Revenue from Trade Taxes (h) Revenue from Hydrocarbon Fees 22 ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~2 .14 90 9 2 94 86 99 90 92 94 96 99 to 92 84 36 99 90 92 94 96 98 (i) Revenue ofPEMEX (j) OtherNon-GovtPublic Sector Ravcoue 24- 24 - 20 2 96~~~~~~~~~~~~~~~~~2 012 16- 12 4 90 92 94 86 99 90 92 94 96 9n so 92 94 96 98 90 92 94 96 98 Source: Public sector- accounts. 8 Chapter I Figure 1.5. Cyclical Components of Revenue, 1980-98 (percent) (a) Total Budgeted Revenue (b) Tax Revenue 14 - . .................... . 8 20 ;. Y fd -4 0 -4 S0 82 84 86 88 90 92 94 96 98 S0 82 84 86 B8 90 92 94 96 98 (c) N on-Tax Federal Revenue (d) Other Public Sector Revenue 60 S8 0 8 (e) Inom Tax 4eeu ceu rmGos&Srie ae -30 is 60 1 - 8 230 __ -. 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (e) Income Tax Revenue (t Revenue from Goods & Services Taxes -60 -8 20 -8 80 82 84 86 88 90 92 94 96 98 80 82 94 86 88 90 92 94 96 98 (i) Revenue from Trade Taxe Othe ) Hydo carbon'Pbi FecoRes eu 60 5° 1 4 30 4~~~~~~~~~~~~~~~~~~- -30 -8 -30 80 82 84 86 85 90 92 94 96 98 90 82 84 86 88 90 92 94 96 98 (i) Revenue from Trad X Tax O he Nutonoo Pbl SctorReven 60 9 20- 9 30 4 30 [ - 0 . 0 0 - -30- -4 tO0 -4 60.3 -60------.- _________________________ -8 S0 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 Source: Reeucfo PM X(j thTN n-o' Public Sector acontendathrnuacuaios 60 20- 9~~~~~~~~ Chapter I Table 1.2. Cyclical Properties of Public Sector Revenue and Expenditure Standard deviation Correlation with Cyclical component of Percent Percentage of GDP GDP Total budgeted revenue 4.4 1.1 0.20 Memo item: petroleum revenue 11.2 1.1 -0.28 Federal government revenue 6.1 1.0 0.21 Tax Tevenue 6.8 0.7 0.60 Income tax 9.9 0.4 0.61 Taxes on domestic goods and services 7.4 0.3 0.11 Value added tax 9.1 0.3 0.03 Excise tax 16.8 0.3 0.14 Taxes on international trade 22.0 0.2 0.58 Other taxes 21.0 0.1 0.21 Nontax revenue 19.5 1.0 -0.21 Hydrocarbon fees 21.6 0.8 -0.23 Other public sector revenue 8.4 0.8 0.03 PEMEX 19.6 0.8 -0.11 Other 6.6 0.4 0.30 Total budgeted expenditure 7.6 2.3 0.25 Memo Item: noninterest expenditure 8.8 1.9 0.63 Current expenditure 7.7 2.0 0.08 Salaries and wages 9.3 0.4 0.49 Interest 24.6 2.0 -0.44 Transfers 20.4 0.6 0.08 Revenue shared with state governments 8.7 0.2 0.41 Materials and supplies 11.1 0.3 0.15 Other 22.7 1.0 0.35 Capital expenditure 16.5 0.7 0.45 Memo item: GDP 2.6 2.6 1.00 Source: Author's calculations. 1.23 While nontax revenues showed a considerable upward trend, they were extremely volatile. Other public sector revenues rose sharply in the early 1980s but declined sharply thereafter. Figures 1.5(b)-l.5(d) show that the deviations from trend of tax revenue were quite procyclical, whereas nontax and other public sector income were both approximately acyclical. This is confirmed by table 1.2 which indicates that the correlation between the cyclical component of tax revenue and the cyclical component of GDP was 0.60, whereas it was -0.21 for nontax revenue and 0.03 for public sector revenue. 1.24 The procyclical nature of tax revenue is not surprising, given that most taxes are in some way proportional to economic activity. This procyclical behavior may become more important to the overall budget in the future, because taxes are becoming a more significant portion of public sector revenue. For example, in 1980-81 taxes represented 41 percent of all public sector revenue, but by 1997-98 had risen to represent 48 percent of budgeted revenue (see table 1.3). 10 Chapter I Table 1.3. Main Components of Public Sector Revenue and Expenditure, 1980-81 and 1997-98 (percent) GDP Budgeted revenue Budget category: 1980-81 1997-98 Average 1980-81 1997-98 Average Total budgeted revenue 24.6 21.6 25.6 100.0 100.0 100.0 Memo item: petroleum revenue 8.2 7.4 9.5 33.3 34.1 36.3 Federalgovernmentrevenue 14.6 15.1 15.6 59.3 69.8 61.5 Taxrevenue 10.1 10.3 10.4 41.2 47.7 41.0 Income tax 5.1 4.3 4.5 20.8 20.0 17.9 Taxes on domestic goods and services 3.5 4.9 4.7 14.4 22.7 18.4 Value added tax 2.6 3.1 2.9 10.4 14.5 11.5 Excise tax 1.0 1.8 1.8 4.1 8.2 6.9 Taxes on intemational trade 1.0 0.6 0.8 4.3 2.6 3.0 Other taxes 0.4 0.5 0.4 1.7 2.4 1.6 Nontax revenue 4.5 4.7 5.2 18.2 21.9 20.4 Hydrocarbon fees 3.7 3.1 3.8 15.0 14.2 14.8 Other public sector revenue 10.0 6.5 10.0 40.7 30.1 38.5 PEMEX 4.1 2.3 4.1 16.8 10.8 15.3 Other 5.9 4.2 5.9 23.9 19.4 23.1 Totalbudgeted expenditure 32.2 22.8 30.4 130.9 105.8 117.6 Memo item: noninterest expenditure 28.8 19.1 22.1 117.2 88.6 86.7 Current expenditure 23.5 20.0 26.0 95.8 92.6 100.2 Salaries and wages 5.8 3.4 4.7 23.5 15.8 18.4 Interest 3.4 3.7 8.3 13.7 17.1 30.8 Transfers 2.9 5.0 2.8 11.7 23.2 11.5 Revenue shared with state governments 2.2 3.0 2.7 9.1 14.0 10.6 Materials and supplies 3.5 1.6 3.0 14.2 7.2 11.5 Other 5.9 3.3 4.4 23.9 15.1 17.2 Capital expenditure *8.8 2.9 4.4 35.9 13.3 17.5 Source: Public Sector accounts. 1.25 Despite their overall procyclical nature, the trends and cycles across various tax categories exhibit some interesting differences. Figures 1.4(e)-1.4(g) show the most important components of tax revenue: income tax; taxes on domestic goods and services, namely, value added tax (VAT) and excise tax; and taxes on international trade, which are dominated by import taxes. 1.26 Table 1.3 shows that domestic goods and services taxes have been an increasingly important part of revenue. In 1997-98 these taxes represented 48 percent of all tax revenues, versus 35 percent in 1980-81. Most of this increase has come from taxes of petroleum products: more than half of the increase in VAT receipts has come from PEMEX. Excise taxes on gasoline have risen sharply, while other excise taxes have fallen. Taxes on domestic goods and services are the only tax component that appears to have been rising significantly during the study period. Income taxes have risen somewhat, while taxes from international trade are roughly at the same levels now as they were in 1980, thus both have fallen as a share of revenue and GDP. The declining reliance on import duties, the slow expansion of income taxation, and the expansion of VAT are typical of countries at Mexico's stage of development. However, scope exists for the further expansion of VAT on commodities other than petroleum. 11 Chapter l 1.27 As concerns cyclical properties, figures 1.5(e)-1.5(g) show the cyclical components of the three types of taxes. Income taxes are clearly highly procyclical (table 1.2 indicates that the correlation with the cyclical component of GDP is 0.61), though clearly more volatile than the business cycle itself. Taxes on domestic goods and services are not particularly procyclical (the correlation with GDP is just 0.11), but this pattern appears to have been changing since 1994. Revenue from trade taxes is highly procyclical (the correlation with GDP is 0.58), reflecting the highly procyclical nature of imports. 1.28 The sole large component of nontax revenue is hydrocarbon fees, which are included in federal government revenue under one classification scheme, although they are sometimes classified as PEMEX revenue.3 They are plotted in figure 1.4(h). They display little overall trend, currently represent about 14 percent of all budgeted revenue, and have been somewhat countercyclical (the correlation with GDP is -0.23), as indicated by Figure 1.5(h). 1.29 The single largest component of rest of public sector revenue is the revenue of PEMEX, which is displayed in figure 1.4(i). Even when hydrocarbon fees are classified as government revenue PEMEX still represents 11 percent of all budgeted revenue and more than a third of the revenue generated by the nongovemrnment public sector. Although PEMEX's contribution has declined in significance from about 17 percent of revenue in 1980-81, this figure is misleading. If all petroleum-based revenue is aggregated across the public sector, it currently represents 34 percent of revenue as opposed to 33 percent of revenue in 1980-81 (see table 1.3). As figure 1.5(i) indicates, and table 1.2 confirms, PEMEX's contribution to revenue is somewhat countercyclical: the correlation with GDP is _0.11 .4 1.30 The rest of the public sector has declined slightly as a source of revenue (see table 1.3) and has somewhat procyclical income as indicated by figure 1.5(j) and table 1.2. Among the important contributors are the electrical utilities and railways. 1.31 On the expenditure side of the public sector flow of funds, a distinction will be made between interest on public debt and other forms of spending. Noninterest expenditure fell in the early 1980s, as indicated in figure 1.6(a). This occurred just as interest expenditure (figure 1.6h) accelerated with the inflation rate. As table 1.3 indicates, noninterest expenditure has fallen sharply as a fraction of GDP, from 29 percent in 1980-81 to just 19 percent in 1997-98. At the same time, the public sector has moved to a substantial primary surplus position on budget, with budgeted noninterest expenditure now representing 89 percent of budgeted revenue, as opposed to 117 percent in 1980-81. Interestingly, noninterest expenditure is highly procyclical, as indicated by figure 1.7(a) and table 1.2. The correlation of its cyclical component with the cyclical component of GDP is 0.63, and it is substantially more volatile than GDP with a standard deviation of 8.8 percent (1.9 percent of GDP), compared with 2.6 percent for GDP (table 1.2). This procyclical behavior of expenditure offsets the procyclical behavior of revenue and tends to make the primary budget close to neutral with respect to the cycle. As a recent study by the Inter-American Development Bank indicated (Gavin and others 1996), the public 3. In the "B" accounts of the Secretaria de Hacienda y Credito Puiblico the revenues from hydrocarbons are classified as government revenue, but in the "G" accounts they are classified as revenues of PEMEX. Consolidation across the entire public sector makes this accounting difference irrelevant. 4. Looking more narrowly, the export revenues of PEMEX, which are more closely tied to the world oil price, are more strongly countercyclical: the correlation with GDP is -0.29. 12 Chapter I sector thus acts much less as a stabilizer than it does in other economies of the Organization for Economic Cooperation and Development (OECD). Figure 1.6. Trends in Public Sector Expenditure, 1980-98 (a) Total Non-Interest Expenditure (b) Salaries & Wages 85 20 - o75- 1 -1 O65-~ 14- 55 1 45i 8 S0 82 84 j6 88 90 92 94 96 98 S0 82 84 86 88 90 92 94 96 98 (c) Transfers (d) Revenue Shared with the States 25- 12 20- 10 O .05 80 82 84 86 88 90 92 94 96 9S S0 82 84 86 8 90 92 94 96 98 (e) Materials & SuppEes (f) Other Current Expenditure 14- 25- 09 ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~0 5- (g) C apital Expenditure (h) Interest Expenditure 30 ~ ~ ~ ~ ~ ~ ~ ~~~~~~3 24 0. 19~ ~~~~~~~~~~~~~~1 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 Source: Public sector accounts. 13 Chapter I Figure 1.7. Cyclical Components of Expenditure, 1980-98 (percent) (a) TotalNon-Interest Expenditure (b) Salaries & Wages 24- _ 8 40- 12- 4 20~ 4 0 k0 0 ~ -12- -4 -20- -24 -8 -40 -8 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (c) Transfers (d) Revenue Shared with the States 30- 4 15- 4 0 0 0 0 -30 ~ 4 1 -4 -60 - -- -30 -S 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (e) Materials & Supplies () Other Current Expenditure 30-- 8 7018 is V 4 35- 4 0 - m A 0 0 - A ~~~~ 19.0A -45 -8 7 0 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (g) Capital Expendiure (h) Interest Expenditure 50 --8 80 ------- - ----------------- ---- 8 25- 4 40 0- 0W A N 0 -25 -4 -40 - -5 ----5__._-- _ 8-0 ----_............................................ _ ___ .. 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 Source: Public sector accounts and author's calculations. 1.32 Once expenditure is divided into its components, an even clearer picture emerges. Salaries and wages in the public sector, depicted in figure 1.6(b), have been declining in real terms almost steadily since 1980. Table 1.3 indicates that they now represent just 16 percent of budgeted revenue, compared with 24 percent in 1980-81. Most of the decline in the public sector wage bill has been within the federal government, which is perhaps surprising given the 14 Chapter I declining importance of the nongovernment component of the public sector. While the wage bill has fallen as the economy has expanded (figure 1.6b), figure 1.7(b) and table 1.2 indicate that the movements around this trend have been quite procyclical. Most of this procyclical behavior is not due to the federal government wage bill, but is largely determined by the behavior of wages within the corporate public sector.5 1.33 Transfers have become an increasingly significant component of the overall public sector budget, as indicated by figure 1.6(c). By 1997-98 they represented 23 percent of budgeted revenue as opposed to just 12 percent in 1980-81 (table 1.3). Because government transfers to public sector corporations are netted out of this figure, this represents a substantial increase in the importance of social programs within the economy. However, these transfers have not been countercyclical as one might have expected. If anything, transfer payments appear to be procyclical: figure 1.7(c) shows that spending on transfers rose rapidly during the early l990s and then declined sharply during the 1995 recession. On the other hand, table 1.2 indicates that, overall, transfers have been approximately acyclical. 1.34 The lack of cyclical variation in transfer payments deserves some further explanation. The figures used in this chapter for "transfer payments" correspond to that portion of the government's spending category "Aid programs, subsidies and transfers" that is classified as current expenditure. In 1997 this amounted to about 200 billion pesos. Economic classification of expenditure is available for the entire "Aid programs, subsidies and transfers" category with capital expenditures included. In 1997 this amounted to about 230 billion pesos. So current expenditures make up the bulk of this category. The portion of this overall budget dedicated to aid programs or social assistance spending was only about 32 billion pesos or about 15 percent of current transfers. Most of the rest of the transfer budget represents revenue sharing, subsidies or financing to public sector entities. Given these facts, it is not surprising that transfer spending is approximately acyclical. 1.35 The federal government shares a substantial portion of its revenue with states, and this revenue sharing has increased in importance, as indicated by figure 1.6(d). Revenue distributed to the states rose from 9 percent of budgeted revenue in 1980-81 to 14 percent in 1997-98 (table 1.3). It has also been procyclical, its cyclical component having a correlation of 0.41 with the cyclical component of GDP (table 1.2). This is similar in magnitude to the correlation of overall tax revenue with GDP. 1.36 Because the nongovernment public sector is responsible for most of the expenditure on materials and supplies, the decline of this component of the public sector is responsible for the decline in this expenditure relative to public sector revenue from 14 percent in 1980-81 to 7 percent in 1997-98.6 This pattern is confirmed by figure 1.6(e). Materials and supplies expenditures are only slightly procyclical, as indicated by figure 1.7(e). The correlation of their cyclical component with GDP is just 0.15 (table 1.2). 5. Disaggregated figures not presented in Table 1.2 indicate that the federal governnent's wage bill has a cyclical correlation with GDP of just 0. 18, while the figure for the rest of the public sector is 0.58. 6. Figures not presented in table 1.3 indicate that, on average, more than 90 percent of materials and supplies expenditure is accounted for in the nongovernment sector. 15 Chapter 1 1.37 Other current expenditure, illustrated in figure 1.6(f), consists of several items including payments for services rendered by the private sector. It has declined significantly as a share of revenue from 24 percent in 1980-81 to just 15 percent in 1997-98 (table 1.3). This is largely due to a decline in this type of spending by the federal government. It is somewhat procyclical as indicated by Figure 1.7(f), and has a correlation with GDP of 0.35 (table 1.2). 1.38 Capital expenditure is the last category of noninterest expenditure. This has declined significantly since its peak in 1982. Capital expenditure has declined from 36 percent of budgeted revenue in 1980-81 to just 13 percent in 1997-98 (table 1.3). The federal government has halved its capital spending from 1 4 to less than 7 percent of budgeted revenue, while the rest of the public sector has seen an enormous decline in capital spending from 22 percent of revenue to less than 7 percent. Capital spending has been quite procyclical, as indicated by figure 1.7(g). Its correlation with GDP was 0.45 (table 1.2) during the entire study period. Indeed, during severe recessions such as those in 1982 and 1995, capital spending was cut sharply. 1.39 The final item is interest expenditure, illustrated in figure 1.6(h). Inflation effects are the driving force behind changes in the size of interest flows. Interest expenditure shot up in 1982 and 1986, not only because public sector debt increased, but mainly because inflation accelerated dramatically. High real interest rates in the stabilization period after 1988 kept interest expenditure at high lev-els, but declining debt eventually brought interest spending down. It again rose in significance in 1995 as inflation accelerated during the peso crisis, but by the time the public sector's overall level of indebtedness was by then much lower than in the early 1980s. Interest expenditure is strongly countercyclical, its correlation with GDP being -0.45 (table 1.2), largely because in Mexico inflation has tended to be highest during periods of recession. 1.40 One way to deal with the inflation effects that dominate interest flows would be to compute real rather than nominal interest flows. Even though the interest flows pictured in figure 1.6(h) are expressed in real terms (that is, they are expressed in constant peso termns), they are nominal interest flows in the sense that they represent some average nominal interest rate times the nominal level of debt divided by the price level. Real interest flows, instead, would be computed by calculating some average real interest rate times the nominal level of debt, divided by the price level. Making accurate adjustments of this sort is quite difficult when a significant fraction of the debt is held outside Mexico and is denominated in many different foreign currencies. Furthermore, at times, significant portions of domestic debt have been indexed or issued in foreign currency. Hence no attempt at making inflation adjustments is made here. This is of little significance, because later sections argue that the focus should be on the primary budget surplus, which excludes all interest expenditure. The Cyclically Adjusted Budget Surplus in Mexico 1.41 This section outlines a methodological approach for computing cyclically adjusted budget surplus measures for Mexico. It examines several approaches that a variety of intemational organizations use and computes historical surplus figures using a preferred method. It then asks whether the adjusted surplus is a useful concept, that is, can it help policymakers make policy decisions? 16 Chapter I Methods for Computing the Cyclically Adjusted Budget Surplus 1.42 Economists have long recognized that budget surplus figures tend to be procyclical. In particular, budget surpluses are procyclical in most OECD countries for a number of reasons that will be elaborated upon later. In the context of Keynesian macroeconomic theory, when the public sector runs a larger budget surplus than previously, the government is said to have a contractionary fiscal policy stance. However, if the budget surplus is larger simply because the economy is going through an expansionary phase of the business cycle, thinking of fiscal policy as contractionary may be inappropriate. Thus, many economists have proposed that budget surplus figures should somehow be adjusted to allow for the effects of the business cycle. 1.43 The literature on adjusted budget surplus measures can be traced back to the paper by Brown (1956), where he argued that to measure the stance of fiscal policy correctly one had to distinguish between "automatic" and "discretionary" policies. Brown's paper did not propose an adjusted measure of the budget surplus, because he explicitly argued in favor of the differential treatment of the various components of revenue and expenditure with reference to an explicit Keynesian model of the economy. 1.44 Since Brown's paper economists have sought a single indicator of the stance of fiscal policy, similar to the budget surplus as a percentage of GDP, but adjusted for the business cycle. A number of government and international agencies produce these sorts of measures including the OECD, the World Bank, the International Monetary Fund (IMF), the European Community (EC), and their various member goverunents. A number of indicators have been suggested. Chouraqui, Hagemann, and Sartor (1990) and Price and Muller (1994) present good discussions of the various indicators. Blanchard (1990) and Buiter (1993) provide arguments against using single indicators. 1.45 Cyclical adjustment of the budget usually begins with the decomposition of output into some trend, or potential, component and its cyclical component. The technical appendix describes several methods including the one used here: the same method that the EC uses (see EC 1995 for even greater detail regarding its method of estimating structural budget deficits), which adopts the HP filter-based trend in GDP as the measure of potential output. 1.46 To compute cyclically adjusted surplus measures the EC, IMF, and OECD estimate the elasticities of various components of revenue and expenditure with respect to output. They use the estimated elasticities to make cyclical adjustments to these components of the budget.7 At this stage an important set of assumptions must be made: one must decide which revenue and expenditure components fall into the automatic category and which fall into the discretionary category. Because the assumption is that the business cycle causes those that fall into the automatic category, while those in the discretionary category potentially cause the cycle, only those components that fall into the automatic category should be adjusted. 7. Details of how the elasticities are estimated and used to make cyclical adjustments are provided in the technical appendix. 17 Chapter I 1.47 For the purposes of this chapter, the following revenue and expenditure categories in Mexican data were considered for adjustment: * Income tax revenue, Rlt * Taxes on domestic goods and services, excluding gasoline, R2t * Taxes on intemational trade, excluding taxes on PEMEX imports, R3t * Other tax revenue, R4t * Governnent transfers net of transfers to the public sector, X, . Estimates of the elasticities of these revenue and expenditure categories with respect to the cyclical component of output are found in table 1.4. As the elasticity for transfer payments was not statistically significant, no adjustments to this item were made. Adjustments to the four revenue categories were made. 1.48 It should be reemphasized that the decision to adjust some revenue/expenditure categories and not others is based on strong a priori assumptions about causality rather than on a statistical test. The notion is that tax revenues behave cyclically largely because most tax systems rely on statutory tax rates on various types of economic activity-this naturally leads to cyclical movements in tax revenue. Similarly, in many countries transfer programs are structured to respond automatically to business cycle movements. As a result, it seems reasonable, from a theoretical perspective, to treat the cyclical movements of tax revenues and transfers as being driven by the various factors that drive the business cycle, rather than being the causes, themselves, of the business cycle. Expenditure categories such as wages and salaries and capital expenditure are highly procyclical, but they are typically not adjusted for the cycle-the implicit argument against adjustment is that these categories of expenditure are fundamentally more discretionary. Of course, if all revenue and expenditure categories were adjusted the adjusted surplus would be uncorrelated, by construction, with the business cycle. 1.49 As described in more detail in the technical appendix, the adjusted surplus measure is given by (1) At = At - Rj, [exp (eR>y )-1 i=, where A, is the standard budget surplus measure, ej1 is the elasticity of Rj, with respect to output and y, is the cyclical component of output, as measured using the HP filter. 1.50 As table 1.3 indicates, oil revenue represents about a third of the Mexican public sector's budgeted revenue. There is some question as to whether the cyclically adjusted budget measures should also reflect this fact. On the one hand, if one wants to make adjustments that purely reflect the business cycle's effect on the budget, one would make no adjustments to oil revenue. On the other hand, if the purpose of estimating a cyclically adjusted fiscal surplus is to isolate those components of the budget that are not driven by exogenous forces, correcting for oil prices 18 Chapter I is important. Two alternative measures of the cyclically adjusted budget surplus are presented below: one that makes no adjustment for oil prices, !A, and one that does, namely: (2) A, = A - RO[exp(eROpt )-1]. Here Rot represents petroleum revenue, eRG is the elasticity of the cyclical component of petroleum revenue with respect to the cyclical component of the world oil price, and po is the cyclical component of the world oil price. Table 1.4 provides an estimate of eRO. It is positive and statistically significant, and indicates that a 1 percent rise in the world oil price is matched by a 0.4 percent rise in petroleum revenue. Table 1.4. Estimates of Revenue and Expenditure Elasticities Revenue source Elasticitya Standard error t-statistic Income tax revenue 2.33 0.36 6.56 Taxes on domestic goods and services 0.51 0.33 1.55 Taxes on trade 5.25 0.84 6.25 Other tax revenue 1.47 0.81 1.82 Transfer paymnents 0.65 0.89 0.73 Petroleum revenue 0.41 0.08 6.82 Notes: The estimates were computed using ordinary least squares. a. Elasticity refers to output elasticity in all cases except petroleum revenue, where it refers to oil price elasticity. Source: Author's calculations. Budget Surplus Estimates for Mexico 1.51 Figure 1.8(a) presents data on the public sector's primary surplus measured as a percentage of GDP.8 Note that in the early 1980s the public sector was in a large primary deficit position that it was forced to reverse as of 1982 with the onset of the debt crisis. Throughout the rest of the 1980s and into the 1990s the government remained in a strong primary surplus position, usually at more than 5 percent of GDP. As inflation was stabilized, the government no longer needed to run such a large primary surplus and it was scaled back to less than 5 percent for most of the later 1990s. 1.52 The economic balance, which includes nominal interest payments, is illustrated in figure 1.8(b). It paints a different picture, but as has been argued, the economic balance is deceptive, because during periods of high inflation interest flows largely reflect compensation for inflation rather than income to the recipient. Whether the economic balance should be the focus of the analysis is not obvious, because small-scale macroeconomic models usually emphasize government purchases of goods and services, as well as taxes and noninterest transfer payments. Of course, large interest payments can impinge upon the rest of fiscal policy. 8. These figures are quarterly surpluses relative to quarterly GDP, where both have been seasonally adjusted. 19 Chapter 1 1.53 The cyclically adjusted primary balance is illustrated in figure 1.8(c). At first glance, the cyclically adjusted balance and the standard measure of the primary balance appear to be quite similar. The difference between the two measures is plotted in figure 1.8(e), and at times it is substantial. Whenever it is negative, it indicates that fiscal policy was more contractionary than indicated by the standard primary surplus. Whenever it is positive, fiscal policy was more expansionary than indicated by the standard primary surplus. Note, for example, that fiscal policy looks more expansionary during the 1990-94 with the adjusted budget figures. 1.54 The adjustments for petroleum revenue can also be substantial. Petroleum revenue moves closely with the world oil price, which is highly volatile. Figure 1.8(f) shows that adjustments of revenue for exogenous movements in oil prices have amounted to as much as 2 percent of GDP. For example, oil revenue shot up in 1991 because of the Gulf War. Removing this effect from the data requires an adjustment in the amount of 2 percent of GDP; however, figure 1.8(d) shows that the overall picture is little changed by taking movements in oil revenue due to oil prices into account. Other factors dominate movement in the budget surplus. 1.55 Using the primary balance as a summary measure, Mexico's fiscal policy is procyclical in the following sense. Countercyclical policy, or leaning against the wind, is usually described as running deficits during recessions and surpluses during expansions. In other words, policy is countercyclical when the budget surplus is procyclical. However, the correlation between Mexico's primary surplus as a percentage of GDP and the GDP's cyclical component, measured in percent, is -0.35. This indicates that Mexico has tended to run higher surpluses during hard times and smaller surpluses or deficits during good times. 1.56 The cyclical adjustment of the primary surplus only makes this fact starker. The correlation between the cyclically adjusted primary surplus and GDP is -0.44.9 This suggests that discretionary policy, rather than leaning against the wind, seems to lean quite strongly with the wind. 1.57 The finding here is consistent with the discussion in Gavin and others (1996). They study a number of Latin American countries and find that in the typical country, fiscal policy is much more procyclical than in the OECD. Both revenue and expenditure are typically much more sensitive to the business cycle than in the OECD, but the expenditure effect is stronger. 1.58 This finding may reflect the fact that during recessions, Mexico's public sector, like that in many Latin American countries, faces a hard budget constraint. Perhaps discretionary policy cannot be expansionary in the traditional sense because the government is liquidity constrained. But this begs the question: how did the government become liquidity constrained in the first place? Was procyclical fiscal policy itself the culprit? Whatever the reason, policymakers should note that policy moves with the business cycle rather than against it. 9. This is to be expected, because the cyclical component of the budget surplus is highly correlated with the cyclical component of GDP. The cycle- and oil-adjusted budget surplus has a correlation with GDP of -0.40. 20 Chapter I Figure 1.8. The Budget Surplus and the Fiscal Impulse, 1980-98 (a) Primary Balance (b) Economic Balance 15 5- -15 ^20 s ! 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (c) Cyclically-Adjusted Primary Balance (d) Cycle- and Oil-Adjusted Primnary Balance 15 - . ..............15! 10- C 101S........ 1.5 2.0 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (c) Cy(ca) Cyciedbased PriAdjustm ent (BaOilh) Cycal OR-Adjust me nt s 1.5- 2.0 10- ~~~~~~~~~~~10- o E 05 _ 0 \ .3 \ F 94\t ~~~~ 0. --0.5- 10- ~~~~~~~~~~~~~~-10. -1.0 -1.0 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (g)o Fiscale: se b Cyc lAdjustment (h m ls Adjustment s 2.5- 21 o ~~~~~~~~~~~~~~~~~10- 0.5- 5-1 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~05 - -2 1. -2 80 8 2 84 86 9 2 94 9 88 84 86 88 90 92 94 96 98 j S gour ical Ipublcseco bacdounts.eAjsmnt(Fsa mplebsdo Bt dutet 1~~~~~~~~~~~~2 Chapter 1 1.59 How are the cyclically adjusted budget figures useful to policymakers? Presumably they are useful because they isolate the component of fiscal policy that is assumed to be exogenous with respect to the business cycle from the part that is determined by the business cycle. One could argue that it is this component of policy that is discretionary, and policymakers should be given some sense of the effects of these discretionary policies on economic activity. Underpinning the various approaches discussed is the assumption that a simple Keynesian model can be used to think about the economy. It is this model that allows the effects on output to be identified. To summarize fiscal policy with one budget number one must assume that the same Keynesian multiplier applies to government purchases of goods and services, taxes, and transfer payments. Because even the simplest Keynesian models are inconsistent with such an assumption, the possible limitations of the cyclically adjusted budget balance as an indicator of fiscal policy are immediately obvious. 1.60 An altemative fiscal indicator that the IMF uses is the fiscal impulse (the discussion here is loosely based on Chand 1993). This indicator compares the stance of fiscal policy in two successive budget years, but it continues to treat governmnent purchases, taxes, and transfers virtually symmetrically. The fiscal impulse measure is based on the so-called cyclical effect of the budget, which is defined as the difference between the actual budget surplus and the budget surplus that would have been achieved in the absence of discretionary policy. In its simplest form, this method treats all movements in government expenditure that are not proportional to trend output as discretionary. It further treats all changes in revenue because of changes in the average rate at which revenue is raised as discretionary. Thus the discretionary component of the budget surplus is AD = R, - rY, - (X, - xY,*) =/, r-t ) where F is the ratio of revenue to output in a typical year, x is the ratio of expenditure to output in a typical year, and the other variables are defined as before. The fiscal impulse is simply defined as the negative of the change in the discretionary budget surplus, expressed as a fraction of GDP."' 1.61 The discretionary primary budget surplus for Mexico was calculated, along these lines, by setting r and x equal to their sample averages for 1980-98. The difference between the primary budget balance and the discretionary primary budget balance is its nondiscretionary component. While the IMF method makes bigger adjustments to the budget surplus for nondiscretionary fiscal policy than the methods described earlier, the overall picture remains unchanged. The discretionary budget surplus remains countercyclical, and its correlation with the cyclical component of GDP is -0.46. It is also highly correlated with the cyclically adjusted budget surplus calculated previously. The correlation coefficient between the two series is 0.998. Hence this section continues to use l . 10. The fiscal impulse is the negative of the change in the discretionary budget surplus, because budget deficits are assumed to provide a positive imnpulse to economic activity. 22 Chapter 1 1.62 The fiscal impulse, FI/A = _(YA _), in some sense measures the change in policy stance. Whenever it is positive, policy is moving toward a more expansionary position. Figure 1.8(g) shows the fiscal impulse calculated using the cyclically adjusted budget surplus, while figure 1.8(h) shows the fiscal impulse calculated using the cycle- and oil-adjusted budget surplus: FIB = -( _J - AB1). Both series are expressed as backward-looking, four-quarter, moving averages, because the quarterly observations are extremely volatile. The two measures are almost perfectly correlated. They both indicate that in every case where Mexico has gone into recession, fiscal policy has moved toward a more contractionary stance. 1.63 As Chand (1993) acknowledges, the IMF measure of discretionary fiscal policy, like the cyclically adjusted budget surplus measures, is somewhat flawed in that it treats government purchases, transfers, and taxes symmetrically, at least with regard to the effects of their discretionary components on output. Furthermore, the cyclically adjusted budget surplus and the discretionary budget surplus are useful tools of policy analysis only to the extent that they are important factors in the determination of output. This is usually taken on faith, but whether discretionary policy matters for economic activity, even if it is identified correctly, is not obvious. It will matter only if the economy behaves as if it were a simple Keynesian model. The next section addresses this issue for Mexico. How Fiscal Policy and Output Affect Each Other in Mexico 1.64 Does discretionary fiscal policy actually affect output? Does it affect output in ways that Keynesian macroeconomic models would predict? That is, is so-called expansionary fiscal policy actually expansionary? This section explores these questions in a more complex framework than the previous section by explicitly modeling the feedback between fiscal policy and output. Using a vector autoregressive (VAR) approach, the model is enhanced by explicit consideration of oil prices (the main determinant of Mexico's terms of trade), the real exchange rate (a possible indicator of monetary policy, wealth effects, or public confidence), the U.S. Federal Funds rate (an indicator of U.S. monetary policy) and GDP in the United States (an indicator of the demand for Mexican exports). 1.65 The approach in the previous section is clearly somewhat limiting. Under the identifying assumption that some revenue and expenditure components are endogenous to the business cycle while others are not, it allows the calculation of cyclically adjusted budget figures, but it does not permit hypotheses about the dynamic effects of the cyclically adjusted budget to be tested. Furthermore, it does not take into account the full effects of such factors as the terms of trade, external demand, and world interest rates. 1.66 This section addresses the following questions using the VAR approach: * How does discretionary policy affect economic activity? * What were the fiscal impulses to output in historical episodes? * Is there significant reverse causality from output to the budget not captured with the methodology used in the previous section? 23 Chapter I * Is there enough persistence in the budget to allow for short- or medium-term forecasting of financing? The next subsection describes the time series included in the VAR. The second subsection answers the first two questions about effects on output. The third subsection deals with the last two questions concerning feedback effects on the budget and the potential for forecasting. A Small VAR Model of the Mexican Economy 1.67 A modified VAR is specified for a 6 x 1 vector of time series, zt, where z, consists of * The logarithm of the world price of oil (expressed in constant 1993 pesos per barrel) * The logarithm of real GDP in the United States (measured in constant 1992 chained dollars) * The U.S. federal funds rate (measured in percent per year) * The Mexican fiscal surplus (measured in percentage of GDP), * The logarithm of real GDP in Mexico (measured in constant 1993 pesos) * The logarithm of the real Mexico-U.S. exchange rate. 1.68 The logarithm of the oil price, pot, is included not only because oil prices affect the public sector budget balance, but also because they largely determine Mexico's terms of trade. They may therefore have an effect on economic activity as well as the real exchange rate. The logarithm of real GDP in the United States, Yut, is included because it can be used as an exogenous indicator of the demand for Mexican exports. The U.S. Federal Funds rate, rut, is used as an indicator of monetary policy in the United States. Because Mexico's ability to borrow funds might well depend on conditions in the world financial market, American monetary policy is likely to play some role in business cycle fluctuations in Mexico. 1.69 For the Mexican fiscal surplus the cycle- and oil-adjusted primary budget surplus as a fraction of GDP, Xt = A, / Y,t, is used in the benchmark VAR. The unadjusted primary surplus is also used to verify the robustness of the findings. The logarithm of Mexican real GDP, YMt, is included to examine the feedback between the fiscal surplus and output. Finally, the logarithm of the real exchange rate, s,, is included, because it is a useful variable that may reflect other shocks to wealth in Mexico, as well as short- and medium-term changes in Mexican monetary policy. The logarithm of the real exchange rate is measured as s* = ln(StP, / PF), where St is the nominal exchange rate in pesos per dollar, P, is the U.S. GDP deflator, and P, is the Mexican GDP deflator. 1.70 So z, = (pot yut rut X YMt s7)t. The technical appendix describes how a structural VAR model for zt can be identified, estimated, and used to answer the questions posed. 24 Chapter I How Does the Fiscal Surplus Affect Output? 1.71 Table 1.5 presents the impulse response function of output with respect to a fiscal shock. It shows how output would respond to an improvement in the fiscal surplus of 1 percent of GDP. The table shows that within two to three quarters, output would fall by -0.6 percent. This is a substantial decline. By the sixth quarter after the shock, the decline in output reverses and output expands modestly. Note also that the fiscal surplus moves almost symmetrically and in the opposite direction of output. So unanticipated fiscal contractions cause output to decline initially, and to continue to fluctuate in the opposite direction to the fiscal surplus. This is consistent with basic Keynesian theories. Table 1.5. Impulse Response Functions from the VAR Response of output to Response of adjusted fiscal surplus to Number shofk quarte Output shock Fiscal shock Output shock Fiscal shock (percent) (percentage of GDP) (percent) (percentage of GDP) a i.oo -0.33 0.00 1.00 1 0.73 -0.44 0.04 0.60 2 0.65 -0.60 -0.08 0.66 3 0.26 -0.62 0.33 0.33 4 0.17 -0.30 0.13 0.45 6 -0.11 0.20 0.07 0.21 8 -0.08 0.32 0.12 0.05 12 0.05 0.21 0.09 -0.19 16 0.07 0.01 0.00 -0.18 Note: The output shock is an unanticipated I percent increase in Mexico's output. The fiscal shock is an unanticipated I percent of GDP improvement in the cyclically adjusted fiscal surplus. Source: Author's calculations. 1.72 In tenrs of policy design, not only does the response of output to unanticipated fiscal shocks matter, but how output's response to other shocks is affected by the behavior of the fiscal surplus also matters. For example, does the fiscal surplus insulate output from the effects of other shocks? There is no simple way to address this issue, because it pertains to experiments on the feedback rule for fiscal policy. The VAR identifies one feedback rule prevailing during one sample period, but it cannot, strictly speaking, be used to answer questions about the possible impact of alternative feedback rules. 1.73 To expand on the non-insulating properties of fiscal policy suppose one computes the response of the fiscal surplus at period t + k to a shock at period t, and compares this with the response of output to the same shock. If the response of output is positive, then the shock is expansionary. If the response of the fiscal surplus is also positive, one could think about this as policy leaning against the wind. The VAR model includes six shocks, the responses to which can be computed out to any arbitrary k. For k = 24, this implies 150 responses of output (at 25 different dates to 6 different shocks) and 150 responses of the fiscal surplus. In only 49 of these cases do output and the fiscal surplus move together. In 98 of the cases they move in opposite directions. This suggests a strong tendency of fiscal policy to magnify, not dampen, the effect of 25 Chapter I shocks on output, given the presumption that improvements in the fiscal surplus are contractionary. 1.74 Another way to judge the importance of different shocks is the variance decomposition. The variance decomposition breaks forecast errors for each time series into their various sources. For examnple, if one were forecasting output four periods ahead, the forecast error would depend on unanticipated shocks to each of the variables in z,. The relative importance of the six shocks will vary depending on the forecast horizon. If a shock is relatively important at short horizons, then it is a shock that has relatively short-lived, but important, effects on output; however, if a shock is relatively important at long horizons, then it has more of a long-run impact. Table 1.6 illustrates the variance decomposition for Mexican GDP at different forecasting horizons. Table 1.6 Variance Decomposition of Output (percentage of the variance of the forecast error) Forecast Fraction of variance due to shocks to horizon (quarters) Oil prices U.S. GDP Federal funds Fiscal surplus GDP Real exchange rate 0 0.0 2.6 4.9 15.1 77.4 0.0 1 19.4 1.2 2.6 18.4 52.2 6.3 2 20.9 0.8 1.9 25.2 41.4 9.8 3 24.4 0.9 1.8 29.1 31.5 12.4 4 27.2 1.4 4.8 26.6 26.9 13.1 6 24.4 3.1 15.7 22.2 21.8 12.9 8 20.2 4.0 26.1 20.9 18.2 10.8 12 18.5 3.6 36.4 19.1 13.5 9.0 16 22.9 3.3 37.9 16.1 11.2 8.6 Source: Author's calculations. 1.75 The table indicates that fiscal shocks are in important source of fluctuations in GDP, especially at horizons of about two to six quarters, where they account for almost 25 percent of the variance in GDP. At the one-year horizon, fiscal shocks are as important as any other source of business cycle fluctuations. II 1.76 Another way to examine the importance of the budget for output is to compute a historical decomposition of fluctuations in GDP into their sources. This can be done quite easily as a by-product of the VAR estimation. The decomposition divides all deviations of GDP from trend into seven components: a part due to conditions at the beginning of the sample period, and six parts each due to the six different shocks in the VAR. Figure 1.9 plots the portion of GDP caused by fiscal shocks. Note that the fiscal shocks have typically contributed a substantial 11. Notice that shocks to the real exchange rate have no effect on output at the 0 horizon because the real exchange rate is placed last in the VAR ordering. This placement is based on the notion that all shocks to the economy, both real and nominal, are likely to be quickly reflected in both prices and the nominal exchange rate. It is also arguable, however, that one source of shocks to the fiscal surplus is the real exchange rate. If the real exchange rate is placed before the fiscal surplus in the ordering then the impulse response of output with respect to a fiscal shock is in the same direction as indicated in Table 1.5, though a little less strong. On the other hand, the variance decomposition does change significantly: for example at a horizon of 8 quarters real exchange rate shocks and fiscal shocks account for 23 and 13 percent of the variation in output, respectively. 26 Chapter 1 portion of the variation in output, with the possible exception of the expansion of the early 1990s which appears to have had little to do with fiscal expansion. The recession of 1995, also appears not to have been caused by contemporaneous or lagged fiscal shocks to any significant degree. 1.77 From this analysis it seems fair to conclude that the reforms of the late 1980s brought with them an improvement in budgetary institutions-fiscal policy has played a less directly destabilizing role. This does not mean that the feedback rule for fiscal policy cannot play an important stabilizing role in the future, through further institutional changes. Figure 1.9. Cyclical Fluctuation in Output Caused by Fiscal Shocks 4- -4- -8 80 82 84 86 88 90 92 94 96 98 - Fiscal Shock Component + Cyclical Component of GDP Source: Author's calculations. Dynamic Behavior of the Fiscal Surplus 1.78 Table 1.5 also shows that the adjusted fiscal surplus has little response to an output shock. The lack of a contemporaneous response is based on the identifying assumptions made in estimating the VAR. The subsequent dynamic response is small and tends to be positive, indicating that increases in output tend to improve the fiscal situation.12 12. This finding is robust to using the unadjusted primary fiscal surplus in the VAR with an altemative ordering where GDP comes before the fiscal surplus. With this ordering, strictly exogenous shocks to GDP appear to have little imnpact on the fiscal surplus over any horizon. 27 Chapter 1 1.79 The final question is whether this simple VAR model can help predict the financing needs of the public sector. Undoubtedly, it can. Ultimately, the public sector's need for financing is driven by the primary surplus, because financing needs are derived from past and current values of the primary surplus. Financing needs will be predictable as the consequence of any predictability in the budget surplus, and the primary surplus turns out to be highly predictable. In fact, a look at the reduced form of the simple VAR given by (A14), shows that more than 90 percent of the one-step ahead variation in the primary fiscal surplus is predictable. Lagged values of the primary surplus, lagged values of the real exchange rate, and lagged values of the world oil price are all important predictors of the future primary surplus. 1.80 To measure the degree of predictability in the primary surplus it is useful to compute its variance decomposition. Table 1.7 presents the variance decomposition for a VAR in which the unadjusted primary fiscal surplus is used in place of the adjusted fiscal surplus. The second column shows the standard deviation of the forecast error at different forecasting horizons. At a horizon of one quarter the standard deviation of the forecast error is 1.3 percent of GDP, which can be compared with the standard deviation of the fiscal surplus itself, which is 3.7 percent of GDP. Even at a horizon of two years, the standard deviation of the forecast error is only 2.5 percent of GDP, indicating that about a third of the variation in the fiscal surplus, is predictable at this horizon. Table 1.7. Variance Decomposition of the Unadjusted Primary Fiscal Surplus Forecast Standard Fraction of variance due to shocks to horizon deviation of (percentage of variance of forecast error) (quarters) forecast error Oil prices U.S. GDP Federal funds GDP Fiscal surplus Real exchange rate 0 1.3 13.9 0.9 0.0 4.8 80.4 0.0 1 1.7 30.0 3.1 2.3 3.1 61.3 0.3 2 2.0 33.2 2.3 2.4 4.5 54.8 2.8 3 2.1 33.7 2.3 4.4 4.9 51.0 3.8 4 2.2 31.0 3.6 4.6 4.5 50.6 5.8 6 2.4 27.6 3.9 5.8 4.2 48.9 9.6 8 2.5 25.9 3.7 6.6 3.8 46.2 13.9 12 2.6 24.2 5.3 8.2 4.2 41.8 16.3 16 2.8 23.5 8.1 10.6 4.6 38.7 14.6 Source: Author's calculations. 1.81 Most of the variation in the forecast error at all horizons is due to fiscal policy shocks, although oil prices also play an important role. This suggests that the process of fiscal policymaking itself induces a great deal of unpredictable volatility in the economy. 28 Chapter I Policy Conclusions 1.82 The following subsections highlight the main findings of this chapter and explain why they are important to policymakers. Fiscal Policy Is Procyclical 1.83 Mexico's fiscal policy tends to be procyclical. In this way, Mexico resembles many countries in Latin America. Procyclical policy appears to exacerbate business cycle fluctuations. This begs the question as to how and why fiscal policy behaves in this way. Automatic Stabilizers Are Weak 1.84 In most of the OECD, taxes and social programs act as natural stabilizers of the business cycle. Tax revenue, and, in some cases, even marginal tax rates, tend to accelerate during cyclical upturns. Spending on social programs such as unemployment insurance and welfare increases during cyclical downturns. These factors tend to make the fiscal surplus move with the business cycle, so that fiscal policy has a natural, and automatic, tendency to dampen business cycle fluctuations. However, in Mexico these automatic stabilizers are weak. While income tax revenue, taxes on imports, and VAT revenue all move procyclically, they represent only a fraction of the public sector's revenue. Much of this revenue-ultimately about a third of public sector revenue-comes from petroleum, and this revenue is, if anything, countercyclical. 1.85 Transfer payments appear to be neutral with respect to the business cycle. This is probably due to the small fraction of transfer payments that is dedicated to social assistance type spending. Discretionary Policy Is Strongly Procyclical 1.86 Not only are automatic stabilizers weak in Mexico, but the rest of the budget has tended to be strongly procyclical. Cyclical upturns are an opportunity to expand public sector investment and other forms of discretionary spending. Cyclical downturns bring austerity. However, these effects only exacerbate the business cycle. This chapter has presented evidence that the discretionary component of policy may have become less destabilizing since the early 1990s. Is Procyclical Policy Really a Problem? 1.87 Procyclical policy makes business cycles more severe. In Latin Arnerican countries it imposes high financing costs as periods of expansion end and recessions begin. As Gavin and others (1996) point out, international credit dries up when investors recognize that an industrializing economy is entering a recession. Whether this is exogenous behavior by investors or not is irrelevant. The fact remains that a recession makes financing a given level of debt more difficult, so entering a recession with a relatively high level of debt and then having to pay a higher price for that debt is extremely detrimental to the economy. This is what drives the fiscal corrections that must be made during recessions. Furthermore, it is what limits the government's ability to shelter the economy's most vulnerable members during bad times. 29 Chapter I What Can Be Done? 1.88 Automatic stabilizers can be improved. Part of the problem in Mexico is that a third of all revenue comes from petroleum in one formn or another, and oil prices have tended to be countercyclical. A decreased reliance on oil revenue would be helpful, as would more broadly targeted taxation of economic activity. Transfer programs could be more narrowly targeted to soften the blow of cyclical fluctuations. 1.89 Mexico could also improve its discretionary fiscal policy. Discretionary spending, especially public investment, is highly volatile and highly sensitive to the business cycle. This suggests that the planning of projects and financing of public expenditure could be better managed so as to smooth it more evenly over time. A commitment to discretionary policies that lead to larger surpluses during recessions would help the government smooth its financing needs. 1.90 The government could try to smooth its financing needs by smoothing its use of oil revenues. Interestingly, during the period studied in this chapter (1980-98), oil revenue has been, if anything, countercyclical, so it has been a relatively stabilizing feature of the budget. 1.91 Finally, the government can project its financing needs. Projections of financing over the medium term, along with economic forecasts, should allow the government to manage fiscal policy in such a way that a fiscal crunch does not always coincide with a cyclical downturn. 30 Chapter I References Beveridge, Stephen, and Charles R. Nelson. 1981. "A New Approach to the Decomposition of Economic Time Series into Permanent and Transitory Components with Particular Attention to Measurement of the 'Business Cycle'." Journal of Monetary Economics 7(March): 151-74. Blanchard, Olivier. 1990. "Suggestions for a New Set of Fiscal Indicators." Working Paper no. 79. OECD Department of Economics and Statistics. Paris. Brown, E. Cary. 1956. "Fiscal Policy in the Thirties: A Reappraisal." American Economic Review 46 (December): 857-79. Buiter, Willem H. 1993. "Measurement of the Public Sector Deficit and Its Implications for Policy Evaluation and Design." in Mario I. Blejer and Adrienne Cheasty, eds., How to Measure the Fiscal Deficit. Washington, D.C.: International Monetary Fund. Chand, Sheetal K. 1993. "Fiscal Impulse Measures and Their Fiscal Impact." in Mario I. Blejer and Adrienne Cheasty, eds., How to Measure the Fiscal Deficit, Washington, D.C.: International Monetary Fund. Chouraqui, J. C., R. Hagemann, and N. Sartor. 1990. "Indicators of Fiscal Policy: A Reexamination." Working Paper no. 78. OECD Department of Economics and Statistics. Paris. Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 1998. "Monetary Policy Shocks: What Have We Learned and to What End?" Working Paper no. 6400. National Bureau for Economic Research, Cambridge, Mass. de Leeuw, Frank, and Thomas M. Holloway. 1982. "The High-Employment Budget: Revised and Automatic Inflation Effects." Survey of Current Business 62 (April): 21-33. . 1983. "Cyclical Adjustment of the Federal Budget and Federal Debt." Survey of Current Business 63 (December): 25-40. de Leeuw, Frank, Thomas M. Holloway, Darwin G. Johnson, David S. McClain, and Charles A. Waite. 1980. "The High Employment Budget: New Estimates, 1955-80." Survey of Current Business 60 (November): 13-43. EC (European Community), Directorate-General for Economic and Financial Affairs. 1995. "Technical Note: The Commission Services' Method for the Cyclical Adjustment of Government Budget Balances." European Economy (60): 35-88. Fellner, William. 1982. "The High-Employment Budget and Potential Output-A Critique." Survey of Current Business 62 (November): 25-33. 31 Chapter I Gavin, Michael, Ricardo Hausmann, Roberto Perotti, and Emesto Talvi. 1996. "Managing Fiscal Policy in Latin America and the Caribbean: Volatility, Procyclicality, and Limited Creditworthiness." Working Paper no. 326. Inter-American Development Bank, Office of the Chief Economist, Washington, D.C. Giomo, Claude, Pete Richardson, Deborah Roseveare, and Paul van den Noord. 1995. "Estimating Potential Output, Output Gaps, and Structural Budget Balances." Working Paper No. 152. OECD Department of Economics, Paris. Hodrick, Robert J., and Edward C. Prescott. 1997. "Postwar U.S. Business Cycles: An Empirical Investigation." Journal of Money, Credit and Banking 29: 1-16. Holloway, Thomas M. 1984. "Cyclical Adjustment of the Federal Budget and Federal Debt: Detailed Methodology and Estimates." Bureau of Economic Analysis Staff Paper 40. U.S. Department of Commerce, Washington, D.C. IIMF (International Monetary Fund). 1993. "Structural Budget Indicators for the Major Industrial Countries." World Economic Outlook (October): 99-103. Newey, Whitney K., and Kenneth D. West. 1987. "A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix." Econometrica 55: 703-708. Price, W. R., and P. Muller. 1994. "Structural Budget Indicators and the Interpretation of Fiscal Policy Stance in OECD Economies." OECD Economic Studies (Autumn). 32 Appendix Alternative Definitions of the Trend in GDP AL.1 POTENTIAL OUTPUT. The level of potential output is typically defined as the level of output that could be produced if the economy was at full employment or was at the natural rate of employment. The IMF and OECD measures of the cyclically adjusted budget surplus are ultimately based on some measure of potential output. Potential output is usually constructed with reference to some production function that determines GDP as a function of the levels of capital and labor in the economy. Suppose output, Y, , is written as a function Y, = f (K,, N, A,) of the levels of capital, Kt, labor, Nt, and technology, A,. Then potential output is given by Y = f (KI, N, A,), where N1 is the level of full or natural employment and A, is the trend level of technology. Making the concept of potential output operational is difficult because it requires a measure of capital. An annual series on the capital stock is available for Mexico, but ideally, some estimate of capital services that took variable utilization into account would be used. Furthermore, the parameters of the production function, f () , must be estimated. Because technology is unobservable, these estimates must be used to decompose fluctuations in output according to their sources: fluctuations in capital, labor, and technology. Finally, the level of full or natural employment and the trend level of technology must be estimated. Generally speaking, practitioners cannot agree on how to define full or natural employment. For these reasons, the concept of potential output is not used here. A1.2 PIECEWISE LINEAR TREND. Figure l.AI(a) illustrates a piecewise linear trend fitted to real GDP, with a breakpoint at 1988Q1 (first quarter of 1988), the date at which the government's stabilization program was implemented. As table l.Al shows, the break in the trend is significant at approximately the 5 percent significance level. 33 Figure 1.A1. Trends in Real GDP (a) Piecewise Linear (b) Hodrick-Prescon 1500 - 1500 ° 1400 - 1400- , 1300 - 1300 - -1200 -1200- 900 900 80 82 84 86 8 8 90 92 94 96 98 80 82 84 86 8 8 90 92 94 96 98 (c) Beveridge-Nelson (d) Peak-to Peak 1500 - 1500 -.I °1400 - PK °1400 - c1300 -N 1300 -5a/ 1100 - 1100- 1000 1000 900 900 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 Sources: Author's calculations. Table 1A c Estimates of a Piecewise Linear Trend in the Logarithm of Seasonally Adjusted Real GDP Coefficient Standard error t-statistic Constant 13.8 0.051 269 Post-1988 dummy -0.155 0.083 -1.87 Trend 0.0017 0.0023 0.75 Trend x post-1988 dummy 0.0052 0.0027 1.95 Nvote: The estimates were computed using ordinary least squares. The standard errors and t-statistics are robust to heteroskedasticity and serial correlation. A Newey and West (1987) estimator with five lags was used to compute the standard errors. Source: Author's calculations. A1.3 The deviations fom trend implied by the piecewise linear trend are illustrated in figure l.A2(a). GDP was below trend during the troughs of each of the recessions, as well as at the verCobeginnings ofthe 1980 and 1989-94 expansions. 34 Figure 1.A2. Cyclical Components of Real GDP (a) Piecewise Linear (b) Hodrick-Prescott 8 -8 6 - 6 4 -4 C)(c Beeig-elo d ea-oPa 8 0 . -2 , . -2 -4 -4 -6 -6 -8 ~~~~~~~~~~~~~~~~~~~~-8 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 (c) Beveridae-Nelson (d) Peak-to-Peak 0 6 --2 - 4 --4 - - 2 --6- C) 0 ~~~~~~~~~~~~~~~~~~ ~~-8 -2 - -10 d -4 --12- -6 --14- -8 -16 80 82 84 86 88 90 92 94 96 98 80 82 84 86 88 90 92 94 96 98 Source: Author's calculations. trend will simply equal the original series for all t, but if 2 is very large, changes in the slope of the trend are avoided, and in the limit, the trend will simply be a straight line. The conventional value of 2 for quarterly data is 1,600, and this is used here. The trend obtained is similar to the one obtained using the piecewise linear trend. Consequently, the deviations from trend, illustrated in figure .A2(b), are highly correlated with those obtained using the piecewise linear trend. A1.5 BEVERIDGE-NELSON DECOMPOSITION. Another popular trend concept is the permanent component of a time series as defined by the Beveridge and Nelson (1981) decomposition. This procedure involves fitting an ARIMA model to the logarithm of output, y, = In Y, . Consider the following model: 35 (A2) Ayt - u = a, (yt-I - p) + a2 (YI-2 -i) + A + ap (y, p - ,u) + , where u is the mean of Ayt. The permanent component of y., yt , is given by its current value plus any predicted stochastic growth in the series: (A3) Yt = Yt + Et (yt+l - u) + Et (Yt+2 -u) + A. A1.6 An estimate of the perrnanent component can be obtained by estimating equation (A2) and using it to compute the expectations on the right-hand side of equation (A3). Here, equation (A2) was estimated using maximum likelihood, and the order of the autoregression, p, was chosen according to the Schwartz criterion, which selected p = 2. The resulting trend estimates are plotted in figure l.Al(c). The deviations from trend are plotted in figure l.A2(c). The most notable aspect of the trend is that it closely tracks the original series. The deviations from trend are much smaller, and they behave differently than those identified using the piecewise linear trend and the HP filter. They are typically most negative during the early part of cyclical upturns and most positive at the end of expansions. Thus they appear to be more useful as indicators of cyclical turning points than as indicators of the cycle itself. For this reason, the Beveridge- Nelson procedure is not used here. A1.7 PEAK-TO-PEAK TREND. Finally, an ad hoc procedure that is sometimes used is to draw a peak-to-peak trend line so that observed output is never above the trend. In this case, a linear trend in the logarithm of Y, was used to connect the peaks in 1981Q4 and 1998Q2 as illustrated in figure .AlA(d). With the trend specified in this way, output never lies above the trend. The deviations from trend are plotted in figure l.A2(d). They are highly correlated with the deviations from trend defined by the HP filter and the piecewise linear trend, with correlation coefficients of 0.78 and 0.73, respectively. Given the high correlation with the other measures, and the ad hoc nature of the peak-to-peak trend, the cycle defined in this way is not used here. Alternative Definitions of the Cyclical Component of the Budget A1.8 The baseline method of adjustment used in this chapter follows the methodology of the EC fairly closely. First, a limited number of expenditure and revenue categories are selected for adjustment. To illustrate the method of adjustment, take as an example personal income tax revenue, 7Y, one of the revenue categories that is usually adjusted. Its elasticity with respect to output, eTY is given by (A4) eTy = aln( ) A1.9 The elasticity might be estimated using a purely statistical model of the relationship between income tax revenue and GDP. It could also be obtained with reference to statutory tax rates, and a statistical model of the relationship between personal income and GDP, as in the method employed by the Bureau of Economic Analysis of the U.S Department of Commerce, discussed below. 36 Al.10 In this chapter estimates of the elasticities of various revenue and expenditure categories with respect to the output gap were obtained using the following statistical model, illustrated in the case of income taxes: (A5) t = eTYy + E,t where rTy, and y' represent the cyclical components of income taxes and output, respectively, as measured using the HP filter. Al.11 Given an estimate of the elasticity, the EC method adjusts income tax revenue by the amount (A6) - g [eXp(eTYyt )- 1] where y' is the cyclical component of output in logarithmic percentage terms. If the cyclical component is zero, clearly no adjustment to tax revenue is made. If the cyclical component is positive and the elasticity is positive, then the adjustment will be negative. This makes intuitive sense: during a cyclical upturn tax revenues rise simply because the economy is expanding. To adjust for this effect, tax revenue should be adjusted downward. A1.12 In general, with a method such as this the adjusted budget surplus is easy to compute. Any standard budget surplus measure, A, is defined as the difference between revenue, Rt, and expenditure, X,. To adjust the budget surplus for the business cycle and create a new budget surplus measure denoted A, one uses data on the cyclical component of output, y', along with estimates of the revenue and expenditure elasticities. Suppose there are N revenue categories, {Rl......,RJ} and M expenditure categories, {X,,..., Xm }, to be adjusted. Suppose the elasticity of R it with respect to output is given by eR1, while the elasticity of Xj, with respect to output is given by exi I The adjusted surplus measure is given by At = A, + adjustment (A7) = (Rt X, ) - R[exp(eRjy' ) i]-E Xjj [exp(exyu )-l j=1 j=, A1.13 The Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce has a concept of the budget surplus, which was originally described as a high employment budget surplus. It is described in numerous papers, and attempts to compute the budget surplus that would prevail were the economy at full employment and discretionary policies were unchanged. (See, for example, de Leeuw and Holloway 1982, 1983; de Leeuw and others 1980; and Holloway 1984). Later variants make adjustments for the effects of inflation on the budget surplus, via its effects on interest expenditure and indexed transfer programs. Al.14 The BEA's approach is relatively complicated and involves going through the budget component by component making individual adjustments. For example, to compute adjusted 37 personal income taxes, the adjustment procedure first asks what personal income would be at full employment. It denotes this level of income as yp and the actual level of personal income as Yp . To compute Yp the method proposed involves estimating the elasticity of changes in Yp with respect to changes in the output gap, which is the difference between actual output and potential output. Roughly speaking, the estimated Yp adds to Y. that elasticity times the measured output gap. The adjustment process also recognizes that personal taxes are not unit elastic with respect to personal income. In other words, when personal income rises by 1 percent, personal taxes may rise by some different amount, say e percent. So adjusted personal tax receipts, Tp , will be given by (A8) where Tp represents actual personal tax receipts. Al .15 The BEA method presents a number of difficulties in the Mexican context. First, rather than directly relating each revenue and expenditure category to the output gap, it relates them indirectly. In the example, personal taxes are related to personal income, which is then related to the output gap. For Mexico, measuring the relevant income concepts would add a difficult layer of complexity. In addition, the output gap concept requires the assessment of potential output, which is a difficult task even for the United States. Fellner (1982) has argued that the potential output concept is not useful, because true potential output depends on a number of unobservables that are not involved in its estimation. The BEA method sets potential output equal to what is referred to as middle-expansion trend gross national production, or gross national product at the natural rate of unemployment (see de Leeuw and Holloway 1983 for the methods used to compute these concepts of potential output). For these reasons a simpler method is adopted here. A1.16 The IMF and OECD methods resemble each other and are described in some detail in IMF (1993) and Giorno and others (1995). Like the BEA method, they both require obtaining an estimate of potential output. Suppose that output is given by a function of capital, labor, and technology, and suppose further that this function takes the Cobb-Douglas form: (A9) Y, = AtK aNI-a where the notation is defined as before. Then potential output is given by (A1O) Y' = A;K, (N; a where N, is the natural level of employment and A, is the trend level of technology. Generally speaking, the IMF and OECD measures of potential output are derived by first estimating equation (A5) and obtaining estimates of the level of technology, A,. The HP filter-based trend 38 of the series A, is generally used to define A; . The OECD computes the natural level of employment using a model of unemployment rates consistent with nonaccelerating inflation, while the IMF method uses unemployment rates defined by the HP trend in observed unemployment to define natural employment. Both methods use estimates of the actual capital stock in estimating potential output. A1.17 Once the estimate of potential output is obtained, the method for estimating the cyclical adjustments is similar to the one described in the main text. In particular, on the revenue side the EC and OECD make adjustments to corporate taxes, personal income taxes, social security taxes, and indirect taxes. On the expenditure side, the EC and the OECD make an adjustment that is more complicated, and only adjusts for the effects of the business cycle on unemployment benefits. They use a model linking the output gap to the unemployment rate, and hence to the level of unemployment benefits. All other expenditure categories are assumed to be discretionary. Details of the VAR Model Al .18 The VAR is defined in terms of z, = (pO, Yu ru, . YMt St) s. A structural VAR model for z1 that permits contemporaneous feedback between the variables is given by (Al1) Bzt = A(L)z,-l + st where B is a nonsingular square matrix, A(L) is a kth-ordered polynomial in the lag operator, and s, is a vector of mutually orthogonal serially uncorrelated shocks. Premultiplying (Al 1) by B-1 one obtains (A12) z, = C(L)zt, + ut where C(L) = B-'A(L) is a kth-ordered polynomial in the lag operator and u, = B-K1s is a vector of potentially correlated error terms. The standard procedure for estimating VARs is to choose k, and then simply run ordinary least squares regressions for each equation implicit in (A12). The covariance matrix of the residuals from these ordinary least squares regressions, along with a set of identifying restrictions on B, is used to estimate B. Al.19 The procedure used here is nonstandard in that it imposes zero restrictions on various parts of A(L) as well as of B. Specifically, because one could argue that the world oil price is determined strictly exogenously with respect to the other variables in zt, it is assumed that the first equation in (Al 1) is given by (A13) Pot = All]Po1l + A12po0-2 +***+ A1kIP0-k +61t 39 A1.20 One could further argue that U.S. real GDP and the U.S. Federal Funds rate are determined exogenously with respect to the last three variables in zr.. Furthermore, Christiano, Eichenbaum, and Evans (1998) assume that the U.S. Federal Reserve observes enough information on the U.S. economy that the Federal Funds rate should feed back on contemporaneous output, but not enough for the reverse to occur. Hence it is assumed here that the next two equations in (Al1) can be represented as (A14) Yu. =-B2p + (X2, A2 A(3A)zA A + A + (A'1 Ak2 Ak3)zl,k + and (A15) rut =B3 p- -B32y +(A31 A[2 A13)z,,,+A +(A 1 A 2 Ak )zltk + 3, where zlt = (ot YUt rut) mutually orthogonal shocks to the six variables. A1.21 The rest of the VAR is standard. The identifying assumptions imposed on the remaining equations are as follows. The cyclically-adjusted fiscal surplus is assumed to feed back contemporaneously on the external variables, that is z,,; the level of GDP is assumed to feed back contemporaneously on the external variables and the fiscal surplus; and finally, the real exchange rate is assumed to be determined by contemporaneous feedback from all the other variables. This implies that B has the usual lower triangular form with ones on the diagonal. Al.22 With these identifying assumptions the model can be estimated equation by equation using ordinary least squares. The analysis in the main text was conducted using data that had been HP filtered. This is inappropriate in the context of a VAR model, because a simple autoregression cannot, by construction, whiten the data.' For this reason, the VAR analysis in this section is conducted under the assumption that each of the time series in the VAR has a linear deterministic trend (or no trend at all). This allows the VAR to be estimated in levels, with a trend included on the right-hand side. 1. The HP filter is an approximately asymmetric double-sided filter. Thus each observation consists of a double sided moving average of whatever the true serially uncorrelated innovations in the data are. Given this, an autoregression defined over HP filtered data can never have a serially uncorrelated error as equation (A7) assumes. 40 INFRASTRUCTURE, EXTERNAL SHOCKS, AND MEXICO's FIscAL AccOUNTS 2.1 This chapter develops a dynamic general equilibrium model to analyze issues of stability in the Mexican economy. It focuses on whether increased provision of infrastructure can reduce the impact of exogenous shocks on the real economy. That is, would the impact of a shock be less if higher levels of infrastructure spending were in place at the time of the shock? 2.2 The answer to this question is a qualified yes. All else being equal, higher stocks of infrastructure will tend to reduce the declines in real income caused by certain types of shocks. We reach this conclusion by carrying out a series of numerical exercises based on a model that incorporates various types of estimated Mexican data. The model incorporates four types of infrastructure in the production process: electricity, telecommunications, transportation, and education. In general, the estimates indicate that in Mexico, increased provision of infrastructure, either by the public or private sector, tends to be cost reducing. 2.3 Using a variety of Mexican data sources, we calibrate our model to the years 1995-97 as part of a six year simulation for the years 1995-2000. We subject the model to two types of shocks. The first is a shock to the interest elasticity of money demand that causes the absolute value of the elasticity to decline. Such a shock might be caused by a sudden loss of confidence in the banking system, and tends to increase holdings of money and reduce bank deposits. Over time, this shock brings about an increase in the real interest rate, a deflation, and a reduction in real gross domestic product (GDP) amounting an annual average of about half a percentage points over the six years of the simulation. 2.4 We then suppose that prior to the shock infrastructure spending was higher, in real terms, on each of the four types of infrastructure. The increased provision of infrastructure reduces private sector costs and, as a result, real GDP rises to a somewhat higher level than in the initial preshock simulation. We thus conclude that higher levels of infrastructure stocks can indeed insulate the Mexican economy from certain types of shocks. 2.5 The second shock is an external shock: stagnation of the world's real income for the six years of the simulation. This lowers the demand for Mexican exports and, accordingly, the rate of growth of real Mexican income. In this case, as before, a higher level of expenditure on infrastructure before the shock and throughout the period of the simulation tends to neutralize the impact of the shock on real Mexican income. We therefore conclude that the positive implications Chapter 2 of increased infrastructure outweigh the negative impact on the budget deficit. Enhanced infrastructure prior to shocks seems to offer a way to avoid the ex post remedies that have been tried so often, frequently with little success. 2.6 In the past 25 years Mexico's economyhas been subjected to a variety of shocks, both internal and external, including sudden increases and declines in world oil prices, changes in U.S. interest rates, economic collapses in Russia and Asia, and bank panic in Mexico. Our aim is to determine whether changing the provision of certain types of infrastructure can mitigate the effects of such shocks. If this is indeed the case, then choosing alternative infrastructure paths could help stabilize the Mexican economy. Table 2.1 indicates the degree of fluctuations in Mexican real GDP during 1980-97. Table 2.1 Growth rate of real GDP, 1980-97. Year 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 Reai GDP 8.4 8.8 -0.6 -0.4 3.6 2.6 -3.8 1.5 1.1 3.2 5.1 4.3 3.6 1.9 4.4 -6.2 5.2 7.0 Source: Cuarto Informe de Gobierno, Annex. 2.7 Negative growth rates in real GDP correspond to easily identifiable shocks. For example, the decline in 1982 followed the collapse of oil prices, while the 1995 drop came after the financial panic and bank failures of 1994. The governmnent has used a wide variety of policies to try to alleviate the effects of these shocks. Many of these policies have involved either conditional foreign borrowing, such as programs of the International Monetary Fund, or have revolved around domestic fiscal austerity. Thus the remedies for the shocks have, in general, been primarily financial. In addition, the austerity programs have usually involved reducing capital expenditures, thereby reducing the rate of growth of public infrastructure. 2.8 The adjustments that have been undertaken in response to these shocks have generally been after the fact. Thus, for example, fiscal austerity measures have been imposed after the budget deficit has increased. Capital controls have been utilized after there have been losses in reserves. Foreign loans have been incurred after Mexican enterprises have been unable to service their outstanding debts. hiterest rates have been increased and the exchange rate has been fixed after their has been a loss in confidence in the banking system. 2.9 Many claim that adjustments must take place after the shock because by definition, the shock could not be anticipated. Here we consider an alternative approach that recognizes that shocks are diverse and are difficult to foresee. Casual observation might lead to the view that countries with high levels of certain types of infrastructure are more resistant to shocks than countries with inadequate infrastructure. For example, the Republic of Korea, which has developed a high level of educational infrastructure, seems to be recovering from the recent shocks in Asia better than with less developed education systems. Russia, which has little infrastructure of any type, appears to have been hit especially hard by similar types of shocks. 2.10 Although relatively little work has been done on the interaction between infrastructure and a country's ability to withstand shocks, investigators have carried out considerable research on a 42 Chapter 2 related topic, that is the relationship infrastructure provision and a country's rate of real growth. Aschauer (1997) carries out an analysis of a broad range of countries. Among a variety of policies, he considers the impact on economic growth of an increase in public capital, financed by a reduction in government consumption. He finds that the impact of a shift of government expenditure equal to one standard deviation of the sample distribution for public investmnent is, on average, 2.15. In the case of Mexico, the relevant elasticity is 2.20. Given the structure of the estimates, this would mean that an increase in public investment by 63 percent of actual output would lead to an increase in the average annual growth rate of about 0.6 percent. These magnitudes are quite close to those generated by our simulations, even though ours are generated by a general equilibrium framework while these come from a partial equilibrium analysis. Note also that Mexico lies at approximately the middle of the group of 46 sample countries. Uruguay had the highest estimated elasticity, 5.04, while Algeria had the lowest, 0.38. 2.11 One might suppose that not all types of infrastructure are equal. Thus, for example, having an extensive network of roads might have less of an impact on a country's ability to sustain shocks than having a well-educated populace. At the same time, infrastructure might have different impacts on different sectors of the economy. For instance, what might protect manufacturing from, for example, an oil price shock, might do little for agriculture. 2.12 We therefore construct a dynamic model that incorporates an arbitrary number of types of infrastructure. These different types of infrastructure will have varying impacts on the different sectors of the economy. The next section discusses the Mexican background to our analysis. The following section develops the analytical model. Next we discuss data sources and simulation results obtained from the model using Mexican data, before summarizing our findings and reaching a conclusion. Background 2.13 A variety of reasons exist for reducing infrastructure expenditures. Governments often determine that the easiest way to reduce public expenditure is to cut capital spending. They feel they do not need to cut public payrolls drastically, as would be the case if they reduced civil service employment or wages. The costs, if any, of reduc ed formation of infrastructure are usually not felt for some time. Mexico is an extreme case of government reduction in public investment. Public investment fell from 12.1 percent of GDP in 1981 to 4.2 percent of GDP in 1989. At the same time economic growth slowed down markedly. Real GDP, which grew by 8.8 percent in 1981, grew by only 1.1 percent in 1988, while private investment remained approximately constant as a percentage of GDP. 2.14 Drawing conclusions from a category as broad as public capital formation is, of course, difficult. Table 2.2 shows the stock of capital in three key types of infrastructure: electricity and gas, transportation, and communications. The rates of growth of all three changed drastically during 1970-90. For example, from 1970 to 1980 the stock of capital in electricity and gas grew by 12.8 percent per year. From 1980 to 1990 the corresponding rate of growth was only 0.3 percent per year, and the stock of electricity and gas infrastructure declined from 1981 to 1990. The stock of capital in 43 Chapter 2 transportation grew by 8.7 percent from 1970 to 1980, but only by 3.5 percent from 1980 to 1990. Similarly, communications capital grew by 49.8 percent per year from 1970 to 1980 and by 4.1 percent from 1980 to 1990. Table 2.2. Stocks of Infrastructure, 1970-90 (millions of pesos at constant prices) Year Electricity and Gas Transportation Communications 1970 19013.1 9062.5 147.8 1971 22285.4 9562.9 153.6 1972 26091.1 9863.7 180.7 1973 30141.8 10066.8 218.2 1974 33986.5 10045.8 248.3 1975 37846.9 12244.8 1398.7 1976 44798.6 17055.9 3658.4 1977 46669.3 18495.8 5524.5 1978 51148.1 19536.6 7458.9 1979 55761.8 20368.6 9137.7 1980 63419.0 20921.4 8436.3 1981 70175.4 22809.5 10342.3 1982 73962.6 23875.1 11124.0 1983 73587.9 24511.5 11276.7 1984 71613.8 24630.4 11818.7 1985 69216.9 25213.3 12471.3 1986 66555.8 25096.6 11689.6 1987 64195.5 26583.0 11975.3 1988 65630.1 31746.4 13409.7 1989 65340.5 30763.4 12769.1 1990 65110.6 29372.1 12564.5 Source: La Encuesta de Acervos, Depreciaci6n y Formaci6n de Capital del Banco de Mexico. 2.15 Feltenstein and Ha (1998) develop a two-period model that analyzes the extent to which inadequate levels of public infrastructure have contributed to slow economic growth in Mexico. Feltenstein and Ha (1995) use a partial equilibrium model to estimate the income shortfall due to the decline in the stock of Mexican infrastructure. Infrastructure may be either publicly or privately provided. If the infrastructure is publicly provided, as was the case in Mexico until 1990, then changes in its rate of provision will effect the government's budget deficit, with corresponding changes in the domestic interest rate. For example, an increase in the public provision of infrastructure may, initially, increase the productivity of private capital, leading to increases in output. The rise in interest rates caused by larger budget deficits may, however, dampen private investrnent, leading to a less than expected eventual increase in private output. 2.16 Here we develop a multi period model that allows us to examine not only the effect of changes in both publicly and privately provided infrastructure on GDP, but also upon the stability of the Mexican economy in response to shocks. Relatively little analysis of the relationship between 44 Chapter 2 infrastructure provision and a country's ability to withstand shocks has been carried out for other countries. Nonetheless, we can make certain comparisons. (Our examples are taken from World Bank 1998.) 2.17 In Mexico, as in many other countries in Latin America, the decline in real growth in the period 1984-94 is closely associated with a decline in total factor productivity. Indeed, about 2/3 of the decline in the rate of growth of Mexico's per capita real income may be explained by the decline in total factor productivity. The remaining portion of the decline comes from the slow down in capital accumulation. Although physical capital accumulation slowed during this period in Mexico, educational achievement continued to rise, in contrast to most of the other countries in Latin America. 2.18 Evidence concerning Mexico's ability to respond to shocks is mixed. Several studies cited in World Bank (1998) note that economic reforms took place essentially after the fact, in much the same pattern as in most of Latin America. Thus, reforms would be instituted after the shock occurred, although in the Mexican case, the improvements after the reforms came more slowly than in many other countries. Chile and Mexico offer a useful comparison. In the early 1980s, Chile suffered an output collapse similar to that of Mexico. Unlike Mexico, however, Chile had already completed its trade and financial liberalization programs by the time the shocks occurred. As a result, Chile's initial collapse was followed by a sustained output recovery. In Mexico, however, measures such as trade liberalization followed after the collapse. Also, Mexico nationalized its banking system, thereby reducing savings, and hence private investment, while Chile liberalized its financial system, thereby increasing private savings and, accordingly, the rate of capital formation. 2.19 Along with the failure to carry out financial reformns in advance of the shocks, did Mexico's failure to provide adequate rates of growth of infrastructure contribute to the severity of shocks? Using estimated parameters for the entire Mexican economy, in particular, for the elasticities of output with respect to infrastructure, we investigate this issue later. Let us now turn to the model. Model Structure 2.20 This section analyzes the concepts discussed with the help of a dynamic model designed to permit simulation. The model is based on n discrete time periods. During each period all agents optimize over a two-period time horizon. That is, in period t agents optimize given prices for periods t and price expectations for period t + 1, and the future thereafter. When period t + 2 arrives, agents reoptimize for periods t + 2 and t + 3, based on new information about period t + 2. In particular, after events that take place during the first two periods, shocks may occur. For example, certain sectors in the economy may have become insolvent, leading to bank defaults. Because these defaults were not anticipated in period t, they cause agents to recalculate the value of their assets as they optimize over the next two periods. Alternatively, some external shock could have occurred such as an oil price change, or a foreign interest rate change, that may have implications for the domestic real economy. 45 Chapter 2 2.21 The following subsections describe specific details of the model. (See Feltenstein 1992 and Feltenstein and Morris 1990 for the basic structure of production and demand in a perfect foresight context.) Production 2.22 We assume that, in addition to conventional factors, the process of production indirectly requires of liquidity, that is, monetary assets, to finance investment. These assets are therefore implicitly incorporated as factors in the production function. We assume that domestic production takes place in various sectors that use inputs of capital, labor, and land. In addition, infrastructure enters as inputs into private production, possibly as cost-reducing elements. Therefore the productive structure of the model includes at least three factors of production and three categories of financial assets: domestic currency, bank deposits, and foreign currency. Also, there maybe multiple types of infrastructure, some of which are costless to the private producer and some of which require payments, for example, publicly provided education maybe free, while privatelyprovided electricity requires payment. Each of these factors of production, as well as the financial assets, are replicated in each period and, accordingly, have a price in each period. Domestic currency in period 1 is taken as the numeraire. 2.23 We use an input-output matrix, A,, to determine intermediate and final production in period t. Corresponding to each sector in the input-output matrix, value added is produced using capital, labor, land, and infrastructure' We may now specify the following problem for the firim. Let yKI , YLi be the inputs of capital and labor to thejth sector in period i (we will ignore land to simplify notation). Let YGi be a vector representing the outstanding stocks of infrastructure of various types in period i. The production of value added in sectorj in period i is then given by Va1i = vaii (yK{ , YLi YGI) Sectorj pays value added taxes on inputs of capital and labor, given by tKij, tLij, respectively, in period i.2 Hence the effective price for labor and capital paid by sectorj is PLY ( + tLY)PLY, PKj (l + tKY)PK# Thus if P = (IPKi, ILY) are the prices of capital and labor in period i, then the prices of 1. The use of neoclassical value added functions "sitting above" an input-output matrix is conmmon (see Shoven and Whalley 1984 for this approach). An application and a detailed description of functional formrs are given in Feltenstein (1986). Thus, for example, agriculture would use land and labor while manufacturing would use capital and labor. 2. The interpretation of these taxes is, thus, as a profit tax and a personal income tax that is withheld at the source. 46 Chapter 2 goods charged by enterprises, pi, are given by (pi) = va (P, YGi) (I + t)(I - A )-' where va(P, YGd) is the vector of cost-minimizing, value added per unit of output. 2.24 We now suppose that each of the one or more capital types is produced via a sector-specific investment technology that uses inputs of capital and labor to produce new capital. Investment is carried out by the private sector and is entirely financed by domestic borrowing.3 Producer's/investor's may receive an investment tax credit as well as a depreciation allowance, and pay a profit tax on the returns to their investments. The appendix provides the specification of the investor's problem. 2.25 Investor's take out a loans from the banking system to cover their costs. The operational assumption is now made that when feasible new investment, as a percentage of the existing sectoral capital stock, falls below a certain minimum threshold, firns are unable to pay the debt obligations 4 that they incurred to finance their capital formation. Accordingly, banks holding these assets now hold corresponding bad debts. This situation might occur if, unexpectedly during the period, interest rates in the economy rose sufficiently so as to reduce firms' profitable investment below the predetermined threshold. This assumption implies that each firm has a lower feasibility bound for its operations, reflected by its level of investment, below which it cannot operate. This threshold, expressed as a percentage of the existing capital stock, is taken here as exogenously determined, for instance, by pre-existing technology. We also assume, for simplicity, that the threshold is uniform across sectors. 3. We assume that all foreign borrowing for investment is carried out by the govemment, so that, implicitly, the governiment is borrowing for the private investor, but the debt thereby incurred is publicly guaranteed. 4. It is thus claimed that, as a proxy, a firm whose investments fall below some predetermined rate is, in practice, bankrupt. 47 Chapter 2 Banking 2.26 Without loosening generality, here we simplify the structure of the banking sector. Each broad sector of the economy has one bank. To be specific, we will suppose that the economy has five such sectors. Each bank lends primarily, but not exclusively, to a certain sector; therefore, banks are potentially, but not fully, specialized. To streamline the simulations, we assume that the assets of each bank include 50 percent of the outstanding debt of its particular sector. It then holds 12.5 percent of the debt of each of the remaining four sectors.5 This assumption about the diversification of assets avoids the ease with which the insolvency of a particular sector leads to the automatic insolvency of its related bank. A solvency requirement is then imposed on individual banks: if 8 percent of a bank's assets are in default, caused by a corresponding insolvency of its borrowers, then the bank is declared insolvent and is seized by the government. Depositors in the seized bank find their assets frozen. Of course, a bank declared insolvent cannot continue to lend.6 2.27 Thus, the bank's supply of loans, and hence its assets determines the demand for loans from the productive sectors of the economy. Of course, given the existence of a maximum lending to capital ratio, that bank's capital restricts its supply of loans. The bank's liabilities (deposits) are determined by consumers' savings behavior, which in turn is derived from the intertemporal optimization of consumption. Consumption 2.28 We permit an arbitrary number of consumers, each of whom has Cobb-Douglas demand functions. The consumers may also differ in their initial allocations of scarce resources and financial assets. The consumers maximize utility functions subject to intertemporal budget constraints-that have as arguments the levels of consumption and leisure in each of the two periods. The model is similar to that of a standard cash in advance structure. Consumers save by holding money, bank deposits, government bonds, and foreign currency. They require money for transactions purposes, but their demand for money is sensitive to changes in the interest rate. In addition, the consumers' demand for bank deposits is sensitive to their perception of the solvency of the banking system, reflected in the level of non performing assets in the system. In particular, as banks increasingly incur bad loans, the consumers' interest elasticity of money declines, leading them to reduce their bank deposits, which reflects the notion that consumers worry about the safety of their deposits as they perceive banks becoming progressively more insolvent. 5. Clearly these percentages are arbitrary, and should serve only for sinplification and illustrative purposes. Any initial pattem of distribution of bank assets across different sectors will provide results consistent with the hypotheses of the chapter. 6. An 8 percent loss of assets would be tantamount to a total liquidation of capital. While other values could be used for the purpose of the simulation, 8 percent corresponds to intemational standard practices. 48 ChapRter 2 The Government 2.29 The government collects income, profit, and value-added taxes, and import duties. It pays for the production of infrastructure and public goods, as well as for subsidies. In addition, the government must cover both domestic and foreign interest obligations on public debt. The deficit of the central government in period 1, DI, (as before, 1 denotes period i and 2 denotes period i + 1) is then given by Di = G1 + Si + ri Bo + rFJ ei BFO - Ti where SI represents subsidies given in period 1. GI is spending on goods, infrastructure, and services; while the next two terms reflect domestic and foreign interest obligations of the government based on its initial stocks of debt. TI represents tax revenues. 2.30 The resulting deficit is financed by a combination of monetary expansion, and domestic and foreign borrowing. If AyBGI represents the face value of domestic bonds the government sells in period 1 and CFI represents the dollar value of its foreign borrowing, then its budget deficit inperiod 2 is given by D2= G2 + S2 + r2(A YBGI + Bo) +e2rF2(CFl + BFo) -T2 where r2(AyBGI+Bo) represents the interest obligations on its initial domestic debt plus borrowing from period 1, and e2rF2(CFL+BFO) is the interest payment on the initial stock of foreign debt plus period 1 foreign borrowing. The Foreign Sector and Exchange Rate Deternination 2.31 The foreign sector is represented by a simple export equation in which aggregate demand for exports is determined by domestic and foreign price indexes and world income. We also permit exogenously determined exports. In the case of Mexico, this would be represented by oil exports, which are not market determined. 2.32 The government also chooses an exchange rate regime. The model permits essentially any regime, from fixed to floating. In our simulations we will use a modified floating rate. We have taken foreign lending to be exogenous.7 Thus gross capital inflows are exogenous, but the change in reserves is endogenous. The supply of foreign reserves in period i, yFGi, is given by YFGi YFG(i-) + Xi Mi + XF(i1) XFi + CFi - 7. Capital inflows in Mexico are, of course, endogenously determined by a variety of factors. We are mnaking a simplifying assumption here, because the determrination of capital flows is not the primary focus of our analysis. 49 Chapter 2 Here xFi represents the demand for foreign assets by citizens of the home country, so XF(i-4) - XFi represents private capital flows. The govermnent has a demand for foreign assets that is determined by an exchange rate rule. Let yFi represent whatever the government feels to be the critical level of foreign reserves in period i. The government wishes to peg the exchange rate in period i, ei, at its level of the previous period, ei ,,. It will, however, adjust the exchange rate if its stock of reserves, YFGi, deviates from its target, YFi. Money Supply 2.33 Changes in the money supply in period i, AMs8, are now given by A Msi = A ym* + A OMOI + ei YFGi ei- , YFG(i-I) where Aymlfi is deternined the government financing of its budget deficit, and A OMOi represents money created by the central bank via open market operations. The remainder of the right-hand side represents the domestic currency value of the external sector balance. Data Sources, Calibration, and Simulations 2.34 Because our model does not permit an analytical solution, we will use a numerical solution method to derive certain qualitative conclusions about the effect of alternative paths of infrastructure growth on stability in the presence of shocks.8 We derive a fixed point that corresponds to an intertemporal equilibrium. This equilibrium thus represents a set of prices in each period at which all factor and financial markets clear. 2.35 To simulate our model we have used production, demand, export, and migration parameter estimates reported in Feltenstein (1992), Feltenstein and Ha (1995, 1998), Feltenstein and Shah (1995), and Murphy and Feltenstein (1999). We have taken initial allocations to be the stocks at the end of 1994. We carry out a simulation for a six year period representing 1995-2000. All exogenous parameters for the first three years take on values equal to their Mexican values over the period 1995-97. We then assume that these parameters maintain the same values for the final 3 years ofthe simulation. 2.36 We wish to use this benchmark case as a reference point for alternative policy simulations. We use a variety of data sources, apart from our estimated parameters. These are as follows: * There are nine sectors that correspond to Mexican national income accounts, namely, 1. agriculture; 8. The solution to the model depends on finding a fixed point for the underlying Arrow-Debreu economy. This fixed point does not permit an analytical solution, but depends on an iterative numerical methodology. We use a variant of Merrill's algorithm to solve the problem. 50 Chapter 2 2. mining and resource extraction; 3. manufacturing; 4. construction; 5. electricity, gas, and water; 6. commerce, restaurants, and hotels; 7. transport and communications; 8. financial and housing services; 9. social services. * Sectoral value added (in real terms) is taken from Estados Unidos Mexicanos (1998, p. 27), for 1997. Real value added per unit of final output is fixed, but the shares of factor inputs vary. 2.37 Production functions for value added, incorporating infrastructure are derived from Murphy and Feltenstein (1999). These incorporate estimated elasticities for four types of infrastructure: electricity, transportation, communications, and education. These estimates are based on seemingly unrelated regression estimates of a two-equation system: sectoral cost and the labor share of total costs. Capital and labor shares are computed using the "NIPA" data. Both estimates impose constant returns to scale and include dummies for - Sector - Sector * factor prices - Sector * output - Sector * infrastructure. 2.38 The elasticities are calculated using the sectoral average of the appropriate wage and output measures for the estimation subsample (as opposed to calculating the elasticities for each year using that year's sector wage and output and then averaging the elasticities over the estimation period). The standard errors are calculated using the estimated coefficient variance-covariance matrix and assuming the average wage and output figures are constants. That is, if b = column vector of infrastructure coefficients A row vector of "weights" (for example, average wage) V = variance-covariance matrix then the elasticities (e) are e=A*b and V(e) = A * V(b) * A'. 51 Chapter 2 2.39 Interestingly, almost all the elasticities are relatively small, that is on order of magnitude of l 0.25 l or so. The electric ones are generally negative, transport and communications have mixed signs, and the education elasticities generally have positive signs. Two possible problems arise relative to the education estimates. First the education measure monotonically increases through time, therefore, a positive relationship between education and cost is not surprising. Geographically less aggregated data might let us obtain better estimates. Second, some kind of simultaneity may be at work in that shifts in the Mexican economy toward higher cost (and higher quality) production processes are associated with greater availability (and reliance on) educated labor. 2.40 We used annual data to estimate the cost equations for 1974-93. We used 14 sectors, which are slightly different than the 9 sectors, based on national income accounts, of the general equilibrium model.9 The 14 sectors are: mining, food products, textiles, wood products, paper, chemicals and petroleum, non-metallic minerals, basic metals, machinery, other manufacturing, construction, commerce and hotels, financial services, and medicine. Table 2.3 shows the sectoral elasticities of the different types of infrastructure. 9. Our capital stock data, on which the production functions are based, have a slightly different format than the national income accounts. The capital stock and infrastructure data are obtained from Bank of Mexico (1998). 52 Chapter 2 Table 2.3 Cost Elasticities by Sector and Infrastructure Type Sector Electricity Transportation Communications Education Mining -0.103 -0.389 0.028 2.127 (0.180) (0.195) (0.075) (0.507) Food Products -0.741 0.129 -0.001 0.307 (0.143) (0.212) (0.061) (0.617) Textiles -0.129 -0.155 0.006 0.801 (0.133) (0.207) (0.050) (0.490) Wood products -0.585 0.112 -0.019 0.494 (0.139) (0.203) (0.060) (0.436) Paper -0.678 0.025 0.048 0.521 (0.134) (0.206) (0.062) (0.531) Chemicals and -0.288 0.036 -0.083 0.243 Petroleum (0.154) (0.207) (0.087) (0.676) Nonmetallic rminerals -0.456 (0.209) -0.027 0.603 (0.137) (0.209) (0.061) (0.583) Basic metals -0.077 0.196 -0.054 -1.176 (0.125) (0.207) (0.057) (0.557) Machinery -0.422 -0.259 0.059 0.028 (0.142) (0.206) (0.061) (0.608) Othermanufacturing -0.592 0.250 -0.015 2.101 (0.131) (0.207) (0.050) (0.430) Construction -0.383 -0.052 0.029 1.915 (0.184) (0.213) (0.060) (0.431) Comnmerce and hotels -0.440 0.055 -0.024 1.517 (0.239) (0.214) (0.076) (0.548) Financial services -0.342 -0.112 -0.005 0.706 (1.134) (0.197) (0.069) (0.553) Medicine -0.298 -0.037 0.018 0.369 (0.196) (0.202) (0.077) (0.513) Source: Authors' calculations. * Intermediate and final production structure is given by the Mexican input-output matrix for 1985. This is the most recent input output-matrix available and is taken from Mexico's national accounts (INEGI,1998). This is a 72 x 72 matrix, but we have written an aggregation program that adds rows and columns to reduce the dimensionality. For our simulations we used a 9 x 9 version of the matrix, which corresponds to our national income account sectors. * The structure of government production is taken from Murphy and Feltenstein (1999). * Average effective indirect tax rates are derived from INEGI (1998) on a sectoral basis for 1993, the most recent year available. * The average effective import tariff rate is taken from Estados Unidos Mexicanos (1998), p. 24. * The factor inputs corresponding to the different sectors are 53 Chapter 2 1. Agriculture: land, rural labor 2. Mining and resource extraction: capital 1, urban labor 3. Manufacturing: capital 1, urban labor 4. Construction: capital 2, urban labor 5. Electricity, gas, water: capital 3, urban labor 6. Commerce, restaurants: capital 4, urban labor 7. Transport: capital 4, urban labor 8. Financial services: capital 5, urban labor 9. Commerce: capital 5, urban labor * We derive shares of different financial instruments in financing the budget deficit of the public sector from the same computer package. This gives us financing from domestic borrowing, money creation, and foreign borrowing. * Foreign borrowing by the government, in US dollars and foreign borrowing by the private sector are taken from Estados Unidos Mexicanos, EUM, (1998) pp. 122 and 108 for 1997. * Initial allocations of factors, capital, land and labor are taken from CD Cuentas Nacionales for 1993. We define a unit of the factor as the amount that earned 1 mxp in the base year. * Initial allocations of money, Ml, for 1997 and while bonds, derived as the 1997 internal debt of the federal government are taken from EUM, (1998), pp.92. Allocations of foreign U.S. dollar assets are given by central bank reserves for 1997. EUM, (1998), pp.82. * Utility weights, assumed to be the same for both urban and rural consumers, are taken as expenditure shares. These in turn are derived from the aggregation of the 1985 input-output matrix. 54 Chapter 2 Simulations The Benchmark Case 2.41 We now run a six period simulation, that should serve as our benchmark exercise. The goal is to demonstrate that the parameterized model generates results that are consistent with historical outcomes in Mexico. Accordingly, we use 1994 as the base year and run a simulation for the following six years. We have historical data for 1995-97, so we can compare actual and simulated outcomes for those years. We assume that exogenous policy parameters, such as tax rates, takes on their historical values. Table 2.4 shows the resulting macroeconomic outcomes. Table 2.4 A Benchmark Simulation, 1995-2000 (numbers in parenthesis are historical values). 1995 1996 1997 1998 1999 2000 Nominal CTDP 100.0 (100.0) 134.9 (136.3) 186.6 (173.3) 249.5 302.6 419.6 Price level 100.0 (100.0) 129.0 (129.6) 170.1 (153.9) 230.8 269.5 378.6 Inflation ratea na 29.0 (29.6) 31.9 (18.8) 35.7 16.8 40.5 Real GDP 100.0 (100.0) 104.6 (105.2) 107.9 (112.6) 108.0 112.3 110.8 Interest rateb 5.0 (37.8) 14.3 (20.7) 16.1 (12.4) 29.0 21.6 35.3 Trade Balance (% GDP) 1.4 (2.7) 4.1 (2.1) 2.6 (0.0) 3.8 2.1 2.9 Exchange rateb 100.0 (100.0) 122.5 (118.4) 198.5 (123.3) 238.2 325.2 397.4 a. The GE model generates a price level in each year. It is normalized to 1 00 in the first year, hence 1995. Thus you can calculate the rate of inflation in each year after the first year as P(t)/P(t-l)-l. There is no price level for 1994, since that is before the model begins, hence, you cannot calculate a simulated rate of inflation for 1995. Of course there is a real world rate of inflation available for 1995, but it is pointless to compare it with a non-existent model number for that year. b. Exchange rates are normalized to the 1995 rate. We use period averages for the historical exchange rates. Source: Author's calculations except as indicated. 2.42 We can make certain observations. Our model does reasonably well in replicating growth rates in both nominal and real income for the three years for which we can make comparisons. Real GDP stagnates after year five, primarily because of the rising interest rate. The simulated trade balance shows a small surplus throughout the period of the exercise, which is consistent with the assumed floating exchange regime, as well as with historical reality. At the same time, our exchange rate regime generates a small, but steady, real devaluation. A Shock to Confidence in the Banking System 2.43 It now seems reasonable to use our model to generate counterfactual simulations. As a first example let us consider an internal shock. In particular, let us suppose that the economy suffers an unexpected exogenous shock that causes increased anxiety among consumers regarding their bank deposits. One could argue that such an event is currently happening in Mexico in reaction to the crisis in Brazil. That is, a financial crisis in one country causes a loss of confidence in the home country, even though no reason for such a loss of confidence is apparent. This anxiety is reflected by a fall in the interest elasticity, as demand for money becomes less interest-sensitive because the public is suspicious of the banking system. 55 Chapter 2 2.44 To give some numerical value to our example, we will suppose that the elasticity of money demand declines from its estimatedvalue of 0.269 to 0.219.1O This reductioninthe elasticitymeans that, in response to an interest rate increase, consumers are less likely to shift their portfolio structure from money into bank deposits. Such a reduction might be caused by, for example, a loss of confidence in the banking system's stability. Of course, the actual numerical value of the reduction we impose is arbitrary. All other exogenous parameter values are assumed to stay the same, and no policy response to the shock occurs. The results are given in Table 2.5 Table 2.5 Reduction in the Interest Elasticity of Money Demand, 1995-2000 1995 1996 1997 1998 1999 2000 Nominal GDP 90.6 118.7 158.5 209.9 252.0 343.7 Price level 90.8 114.0 147.2 195.4 225.4 311.8 Inflation rate na 25.6 29.1 32.7 15.4 38.3 Real GDP 99.8 104.2 107.7 107.4 111.8 110.2 Interest rate 4.5 14.2 16.3 30.9 23.1 38.3 Trade balance (% GDP) 1.4 4.2 3.2 4.1 2.8 3.4 Exchange rate 91.9 108.8 174.6 203.3 277.2 332.5 Source: Author's calculations. 2.45 The relatively small decrease in the interest elasticity has had measurable effects on the equilibrium outcomes. As might be expected, the price level has declined and the real interest rate has increased. As a result, real GDP has declined by about 0.6 percentage points, on average, as real investment has declined. The trade balance has remained relatively unchanged, given the model's floating exchange rate. 2.46 Now suppose that infrastructure had been growing at a faster rate from the time the shock took hold. In particular, consider a 60 percent increase in real spending on infrastructure in each year of the simulation, divided between the four types of infrastructure in proportion to their shares in the total stock of infrastructure. Thus the higher rates of infrastructure growth are already in place when the interest elasticity shock occurs. Table 2.6 presents the results of this exercise. 10. The numerical value of this reduction is, of course, arbitrary and is used solely to give a sense of the sort of magnitudes that we mnight consider. 56 Chapter 2 Table 2.6 Interest Elasticity Decline Combined with an Infrastructure Increase 1995-2000. 1995 1996 1997 1998 1999 2000 Nominal GDP 97.5 128.6 175.2 235.2 285.6 396.2 Price level 96.2 121.5 157.2 211.8 242.5 341.2 Inflation rate na 26.3 29.4 34.7 14.5 40.7 Real GDP 101.4 105.8 111.4 111.1 117.8 116.1 Interest rate 5.4 15.2 18.3 32.2 25.2 39.2 Trade balance (% of GDP) 1.4 4.1 2.7 3.8 2.3 3.0 Exchange rate 96.3 115.6 187.9 223.1 304.4 383.0 Source: Author's calculations. 2.47 The increased expenditure on infrastructure has raised the price level. Such an outcome might be expected because of the rise in the budget deficit. At the same time, however, the real interest rate has fallen, primarily because of the increased productivity of capital. This, in turn, is a result of the higher provision of public infrastructure from the beginning of the simulation. As a result, real GDP has risen compared with table 2.5, and is now actually higher than in table 2.4, where no elasticity shock occurred. All other macro economic aggregates remain approximately unchanged. Thus providing higher rates of infrastructure spending before a shock may, indeed, negate the effects of the shock. 2.48 As noted in the introduction, Aschauer's (1997) partial equilibrium approach indicates that a 63 percent real increase in government capital expenditures would lead to about a 0.6 average increase in the rate of real GDP growth. A comparison of tables 2.4 and 2.5 shows that a 60 percent real increase in the corresponding capital expenditure causes the overall growth rate of real GDP to rise by 0.7 percent, consistant with Aschauer's results. Trade Shock 2.49 Let us now consider another type of shock, a stagnation in world income. In particular, real world income will remain constant for the entire six years of the simulation. This will have a direct impact on Mexico's economy, because real world income affects aggregate demand for export's equation. Indeed, the estimated elasticity of export demand with respect to world income is 2.33, and in the benchmark simulation world income was assumed to increase. All other parameters will remain the same as in the benchmark case. Table 2.7 gives the results of this exercise. 57 Chapter 2 Table 2.7 Trade Shock: Real World Income Stagnates, 1995-2000 1995 1996 1997 1998 1999 2000 Nominal CUTP 99.2 133.8 183.9 244.6 303.2 40W.9 Price level 99.1 128.5 170.4 227.9 270.2 371.5 Inflation rate na 29.6 32.6 33.8 18.6 37.5 Real GDP 100.0 104.1 108.0 107.3 112.2 110.1 Interest rate 4.7 13.8 15.1 28.1 20.4 33.3 Trade balance (%GDP) 1.6 3.7 3.5 3.1 3.3 1.8 Exchange rate 100.0 141.1 205.8 272.0 340.6 447.4 Source: Author's calculation 2.50 A comparison ofthis case with the benchmark (table 2.4) shows that real GDP is consistently about 0.5 percent lower, the nominal exchange rate depreciates more rapidly, and the Mexican currency undergoes real depreciation. The directions and magnitudes of these changes seem plausible. Let us now suppose that the rate of spending on infrastructure was 60 percent higher, in real terms, than in the benchmark case, and that this was in place before the shock occurred. Table 2.8 gives the outcome of this exercise. Table 2.8 Trade Shock Combined with an Infrastructure Increase, 1995-2000 1995 1996 1997 1998 1999 2000 Nominal GDP 107.6 144.4 203.7 274.4 345.0 469.4 Price level 105.7 136.7 182.5 247.5 291.7 405.8 Inflation rate na 29.4 33.5 35.6 17.9 39.1 Real GDP 101.8 105.6 111.6 110.8 118.3 115.7 Interest rate 5.8 14.6 16.9 29.2 21.4 34.0 Trade balance (% of GDP) 2.1 3.5 3.1 2.7 2.7 1.4 Exchange rate 107.2 149.7 221.3 269.8 373.4 496.7 Source: Author's calculations. 2.51 As in the case ofthe internal shock and infrastructure increase, the higherrate of infrastructure spending has brought about a significant increase in real GDP compared with table 2.7. Indeed, the increased provision of infrastructure has lowered the real interest rate prior to the imposition of the shock, even in comparison with the benchmark case. The increased budget deficit has brought about a slightly reduced trade surplus in most years, while the domestic price level has risen and the exchange rate has depreciated, compared with table 2.7. Thus once again the results indicate that having a higher rate of infrastructure spending in place at the time of a shock tends to alleviate the real effects of the shock. 2.52 One might ask whether our results stem from the importance of infrastructure in resisting shocks. Could they, on the other hand, be simply the outcome of a public sector spending increase; that is, an increase in aggregate demand? We have therefore carried out a pair of simulations in 58 Chapter 2 which all private cost elasticities with respect to all types of infrastructure are assumed to be 0. Thus infrastructure is completely ineffective and increases in stocks will not lead to any private sector cost efficiencies. We report two simulations. The first is the outcome of the trade shock, assuming 0 values for all elasticities. The second, also assuming 0 elasticities, imposes the same 60 percent real increase in infrastructure spending as in the previous examples. We will only report the real GDP outcomes, which are given in Table 2.9. Table 2.9: Infrastructure Elasticities = 0 1995 1996 1997 1998 1999 2000 Shock 97.7 101.7 104.7 104.2 107.8 105.9 Shock plus infrastructure 97.9 101.8 104.8 104.3 107.8 105.8 increase Note: The numbers in the table are real GDP. We observe that increasing real public expenditure on worthless infrastructure has no effect upon real GDP in the long run, although there is a slight improvement in the initial year. Thus a Keynesian shift in the IS curve leads to no increase in real income, and hence no insulation against the shock, unless the expenditure is on useful infrastructure. 59 Chapter 2 Conclusion 2.53 We constructed a dynamic general equilibrium model to analyze issues of stability in the Mexican economy and focused on whether an increase in the provision of infrastructure can reduce the impact of exogenous shocks on the real economy. That is, would the impact of a shock be less if higher levels of infrastructure spending were in place at the time of the shock? 2.54 Our model incorporates four types of infrastructure in the production process. These are electricity, telecommunications, transportation, and education. As part of another study (Murphy and Feltenstein 1999), we have estimated their impact on sectoral production functions. In general, increased provision of infrastructure, either by the public or private sector, tends to be cost reducing. We have then incorporated the estimated infrastructure elasticities in the fully parametarized general equilibrium model. 2.55 Using avarietyofMexican data sources, we calibrated themodel to the years 1995-97 aspart of a six-year simulation for the years 1995-2000. The model generates reasonably accurate approximations of the macroeconomic outcomes of the actual Mexican economy, therebyjustifying its use for counterfactual simulations. We then generate a shock to the interest elasticity of money demand, causing the absolute value of the elasticity to decline. Such a shock might be caused by a sudden loss of confidence in the banking system, and tends to increase holdings of money and reduce bank deposits. Over time, this shock brings about an increase in the real interest rate, a deflation, and a reduction in real GDP amounting to about 0.5 percentage points, on average, over the six years of the simulations. 2.56 We then suppose that prior to the shock, infrastructure spendingwas 60 percent higher, inreal terms, on each of the four types of infrastructure. The increased provision of infrastructure reduces the private sector's costs, and as a result, real GDP rises to a somewhat higher level than in the initial, pre shock simulation. We thus conclude that it is, indeed, possible for higher levels of infrastructure stocks to insulate the Mexican economy from certain types of shocks. 2.57 The final set of simulations involves the imposition of an external shock: stagnation of the world's real income for the six years of the simulation. This lowers the demand for Mexican exports, and accordingly, the rate of growth of real Mexican income. We then suppose that expenditure on all types of infrastructure was uniformly 60 percent higher before the shock occurred and throughout the period of the simulation. Again, this reverses the decline in real Mexican income and largely neutralizes the effects of the shock. 2.58 Thus, the positive implications of increased infrastructure provision seem to outweigh the negative impact on the budget deficit. Enhanced infrastructure, prior to shocks, seems to offer a useful way to avoid the ex post remedies that have been tried so often, and frequently with little success. 60 Chapter 2 References Alberro, J. 1989a. "In Search of a Stable Demand for Money in Mexico During the 1969-1987 Period," unpublished. . 1989b. "Capital Outflows in Mexico During the 1970-1987 Period," unpublished. Aschauer, David. 1997. '?ublic Infrastructure, Capital, and Economic Growth: Some Results Pertaining to the Mexican Economy" unpublished discussion paper. Ball, S. and A. Feltenstein. 1997. "Basic Macroeconomic Options for Bangladesh: A Numerical Analysis" forthcoming, Journal of Asian Economics. Banco de Mexico. 1998. La Encuesta de Acervos, Depreciacion y Formacion de Capital del Banco de Mexico, Computer disks. Estados Unidos Mexicanos. 1998. Cuarto Informe de Gobierno, Anexo, Poder Ejecutivo Federal: Presiencia de la Repuiblica, Mexico. 1NEGI. 1998. Cuentas Nacionales de Mexico, CD-ROM. Feltenstein, A. 1986. "An Intertemporal General Equilibrium Analysis of Financial Crowding Out: A Policy Model and an Application to Australia," Journal of Public Economics, (November), 1986, pp. 79-104. Feltenstein, A. 1992. "Oil Prices and Rural Migration: The Dutch Disease Goes South,"Journal of International Money and Finance, n. 11, pp. 273-291. Feltenstein, A., and J. Ha. 1995. "The Role of Infrastructure in Mexican Economic Reform," World Bank Economic Review, v. 9, n.2, pp. 287-304. 1998. "An Analysis of the Optimal Provision of Public Infrastructure: A Computational Model Using Mexican Data," forthcomning, Journal ofDevelopment Economics. Feltenstein, A., and S. Morris. 1990. "Fiscal Stabilization and Exchange Rate Instability: A Theoretical Approach and Some Policy Conclusions using Mexican Data," Journal of Public Economics, (August), 42, pp. 329-356. Feltenstein, A., and A. Shah. 1995. "General Equilibrium Effects of Investment Incentives in Mexico," Journal of Development Economics, 46, pp. 253-269. International Monetary Fund. Mexico - Recent Economic Developments (various issues),. Jarque, Carlos. 1988. "An Empirical Study of the Determinants of Production in Mexico" 61 Chapter 2 unpublished manuscript. Jung, W.S. 1988. "Asset Demands in Mexico," unpublished University of Kansas discussion paper. Murphy, Russell and Andrew Feltenstein. 1999. "Private Costs and Public Infrastructure: The Mexican Case", unpublished, Virginia Tech. Paquete de Finanzas Publicas y Deuda Publica (Base de Datos). 1998. 3 disk computer package, provided by the Ministry of Finance of Mexico, that provides disaggregated fiscal time series data. Shoven, J.B., and J. Whalley. 1984. "Applied General Equilibrium Models of Taxation and International Trade," Journal ofEconomnic Literature, vol.22, (September), pp. 1007-105 1. World Bank. 1998. Mexico: Enhancing Factor Productivity Growth, Country Economic Memorandum, August. Zedillo, Emesto. 1986. "Capital Flight: Some Observations on the Mexican Case," Paper presented at the Conference on Capital Flight and Third World Debt. Institute for International Economics, Washington, D.C. Appendix Investment Let us adopt the following notation: ki = investment tax credit in period i (percent) di = depreciation allowance in period i (percent) tki profit tax rate (percent) CHi = the cost of producing the quantity Hi of capital in period i ri = the interest rate in period i PKi the return to capital in period i PMi the price of money in period i Suppose that the rental price of capital in period i +l is PK(+,) . If CHi is the cost of producing the quantity of capital, Hi, then future debt obligations must equal the return on new capital. Hence 62 Chapter 2 CHIO - k, - d,) = (l- tK2)PK2Hl (Al) where r, the interest rate in period i, given by ri = 1/PBi (A2) where PBi is the price of a bond in period i. Consumption 2.59 Here, and in what follows, x denotes a demand variable and y denotes a supply variable. To avoid unreadable subscripts, 1 will refer to period i and 2 will refer to period i + 1. The consumer's maximization problem is thus Max U(x), x = (xl,xLUl,xLrl,X2,XLu21XLT2) (A3) Such that (I +ti)PiXi + PLriXLri + PMiXmi +PBiXBi +eiPBflxBF= Ci (A4) PKIKO + PAIAO + PLUILUI + PL,IL,I + PMIMo + roBo + PBI Bo + e,PBFIBFO + TR, = N1 PK2 (1 -)KO + PLu2Lu2 + PLr2Lr2 + PM2XM B + r1XB1 + PB2XBI + e2PBF2XBFI + TR2 N2 C, = N, logP,ixmi =a +blog(1 +tj)8xj--c1ogrj; c=c(DEFIASSET) (A5) log PBi XBi-log eiPBFi xBFi=atZ+ p log rj-log-e+IrFi )(A6) PB2XB2 =S(l+t2 )P2X2 (A7) 63 Chapter 2 where Pi= price vector of consumption goods in period i xi= vector of consumption in period i C =j value of aggregate consumption in period i (including purchases of financial assets) Ni = aggregate income in period i (including potential income from the sale of real and financial assets) ti = vector of sales tax rates in period i PL,u =price of urban labor in period i L,i = allocation of total labor to urban labor in period i XYL,,i= demand for urban leisure in period i PLri= price of rural labor in period i L,i= allocation of total labor to rural labor in period i PKi= price of capital in period i Ko = initial holding of capital PAi = price of land in period i Ao= initial holding of land 6 rate of depreciation of capital PMi price of money in period i (money in period 1 is the numeraire, and hence has a price of 1; a decline in the relative price of money from one period to the next represents inflation) xmi = holdings of money in period i PBi = discount price of a certificate of deposit in period i ri= domestic interest rate in period i xBi = quantity of bank deposits, that is, certificates of deposit in period i ei = the exchange rate in terms of units of domestic currency per unit of foreign currency in period i XBFI =quantity of foreign currency held in period i TRi = transfer payments from the government in period i a, b, a, B estimated constants DEF = the total value of non-perforning assets in the banking system ASSET = total assets of the banking system C = a functional form that depends negatively upon the ratio of non performing assets to total assets in the banking system 2.60 The left-hand side of equation (A4) represents the value of consumption of goods and leisure, as well as of financial assets. The next two equations contain the value of the consumer's holdings of capital and labor, as well as the principal and interest that the consumer receives from the domestic and foreign financial assets that he held at the end of the previous period. The equation C1 = Ni then imposes a budget constraint in each period. Equation (A5) is a standard money demand equation in which the demand for cash balances depends upon the domestic interest rate and the value of intended consumption. There is, however, one modification. The interest elasticity, c, 64 Chapter 2 depends upon the share of nonperforming bank assets in total assets. If there are no bad assets, then c takes its econometrically estimated value. As nonperforming assets rise, c declines. 2.61 Equation (A6) says that the proportion of savings made up of domestic and foreign interest bearing assets depends on relative domestic and foreign interest rates, deflated by the change in the exchange rate. In period 2 (i + 1) we impose an exogenous savings rate on the consumers, as in equation (A7). Thus savings rates are endogenously determined by intertemporal maximization in period i, but are fixed in period i + 1. When period i + 1 begins, the consumer's holdings of financial assets may be different than those incorporated in the above problem, because defaults may have occurred. The consumer then optimizes again for periods i + 1, i + 2, based on his or her new, unexpected holdings of financial assets at the beginning of period i + 2. Foreign Sector and Exchange Rate Determination 2.62 The specific form of the export equation is: AXno =oa[ er ]±+2Ay\Ywi (A8) /Aei7rFi where the left-hand side of the equation represents the change in the dollar value of exports in period i, 7ri is inflation in the domestic price index, Aei is the percentage change in the exchange rate, and 7rFi is the foreign rate of inflation. Also, AyWi represents the percentage change in world income denominated in dollars. Finally, a, and a2 are corresponding elasticities. 65 INFRASTRUCTURE, PRIVATE COSTS, AND PAYOFFS FROM ADDITIONS TO INFRASTRUCTURE 3.1 This chapter provides estimates of the potential (partial equilibrium) payoffs from increased investment in public infrastructure and calculates (in a static context) the optimal infrastructure stocks implied by the elasticity estimates. The chapter also considers the role that public infrastructure plays in improving the efficiency of the private sector. In particular, it focuses on the short-run, static gains that accrue to private firms because of government investments in electric, transportation, and communications infrastructure. We also consider the potential role of education in reducing private sector costs, but ultimately conclude that the education estimates are not particularly informative, primarily because of the nature of the data used. 3.2 The role public spending plays in enhancing economic productivity has long been a concern for policymakers. In recent decades, public expenditure has primarily been evaluated in terms of two roles: enhancing macroeconomic stability and mitigating market failure. An equally important role concerns the ability of public investments in infrastructure capital to reduce the costs of private firms. Particularly for developing economies, this role may be critical because it may allow the private sector to become more resilient to external shocks. A major concern related to the recent fiscal adjustment in Mexico is that it was carried out partly by depleting public infrastructure stocks. This depletion could significantly retard future growth by imposing an additional drag on private sector costs and output. 3.3 This chapter shows that public infrastructure in Mexico has generally small, but significant, negative effects on private sector costs. The base case estimates of the elasticity of private sector costs with respect to infrastructure suggest a mean value of -0.106 across 14 sectors of the economy (with a range of -0.563 to 0.355). In general, electric infrastructure appears to have the most beneficial effects on private sector costs (mean base case elasticity of -0.171), and transportation infrastructure has the next most beneficial effects (mean base case elasticity of -0.165). Communications infrastructure appears to have little effect on private sector costs (the mean base case elasticity is 0.019), but all the base case t-ratios are less than 2.0, indicating that the estimates are not statistically significant.' Education infrastructure elasticity 1. Only 2 of the 14 are larger than 1.0. Chapter 3 estimates are mixed and difficult to interpret, and most are also statistically insignificant.2 3.4 The communications and education elasticity estimates are surprising. Why are these estimates so different from our priors? Why are the education estimates at such variance with different, but potentially related figures such as estimates of the social returns to human capital of roughly 10 to 20 percent. (See Psacharopoulos 1994, who reports social returns of 10 to 14 percent for Mexico in the mid 1980s and 12 to 18 percent for Latin America as a whole). For education, these results are likely the result of the combination of two facets of the analysis rather than indications of the true productive (or unproductive) nature of investments in education. The analysis is based on aggregate data for several sectors of the Mexican economy during the period 1960-93. We measure technical progress by time. Unfortunately (for the analysis), Mexican investment in education has been relatively steady in recent decades which means that there is little opportunity econometrically to separate the effects of time (technical progress) and education. 3.5 The most sensible interpretation of the education elasticities is that our aggregate data do not allow us to infer much about the productive spillovers of public investment into human capital. An alternative approach, based on observations of individual workers and firms, would be more appropriate for understanding the role of education, because we would be able to observe more variation of educational levels and private sector costs across firms. Investigators typically use a micro data approach to estimate social returns to human capital investments. 3.6 The small, positive, and insignificant estimates of the elasticity of private sector costs with respect to investments in communications infrastructure are similar to the estimates of the productivity-enhancing effects of high technology investments in the United States (Stiroh 1998). Some studies (Canning 1998) have found that communications infrastructure (measured in terms of main lines per capita) is productive, but in general the productivity-enhancing effects of high technology are not immediately apparent in aggregate data. Several explanations could account for this odd result. One possibility is that reduced technology costs induce firms to factor substitute, but do not lead to significantly lower costs. In the United States there are indications, even from the micro data, that while technology use and lower costs are associated, firms that are productive in other ways (good performers) tend to be the ones that use technology, rather than technology use leading to lower costs (McGuckin, Streitweiser, and Doms 1996). This is similar to Hulten's (1996) argument that, in the words of the paper's title, how well you use it may be more important than how much you have. 3.7 Using the electricity, transportation, and communications estimates, rough calculations based on these elasticities suggest that a 1 percent increase in public infrastructure stocks would cost approximately mxp 6.6 billion and provide annual benefits across 14 sectors of the economy of mxp 12.4 billion (both in terms of real 1980 pesos). These are static, partial equilibrium estimates, but they suggest that at least in the short run, additional investment in public infrastructure stocks could be welfare improving. Given sensible depreciation rates, the present value of these benefits over future years might be roughly nine times the single year gross benefits. If the base case elasticity estimates are correct, static calculations of the optimal size of 2. Only 4 of the 14 have t-ratios larger han 2.0 in absolute value. 67 Chapter 3 infrastructure stocks suggest that electric and transportation stocks should have been 2.5 and 4 times as large as they actually were in 1993. 3.8 However, one of the striking features of the elasticity estimates is that they are generally small (the effects of public infrastructure on private sector costs do not appear to be large) and somewhat noisy (the estimated effects vary, sometimes significantly, across sectors). This suggests that a degree of caution is warranted in interpreting the estimates. If large potential benefits were associated with public infrastructure, the effects should have been clearer in the data. The estimates do suggest a role for public sector infrastructure capital in reducing private sector costs, but a modest role. 3.9 The results presented in this chapter represent improvements over the existing literature in several respects as follows: * We consider the role of education as a potential type of public infrastructure. * We explicitly evaluate the role of different assumptions about cost function characteristics such as returns to scale. - We calculate the precision of estimates of the elasticity of private sector costs with respect to public infrastructure, rather than just point estimates. - We consider the potential effects on the estimates of cross-sector heteroskedasticity and cross-time autoregression. * We calculate the optimal infrastructure levels implied by the elasticity estimates. 3.10 The next section of the chapter briefly discusses the relevant policy environment. It then introduces the basic model and the related conceptual issues, briefly considers the data used for analysis, and discusses empirical implementation. This is followed by the empirical results and a concluding section. Several annexes discuss details of the production cost estimation model, econometric specification issues, construction of the dataset, and detailed regression results. Background 3.11 Our dataset is confined to the period 1960-93, so the overview focuses on this period as well, recognizing that unfortunately we neglect the most recent years. This is necessitated by our need to construct a dataset from several sources, not all of which are currently available for the same periods in the form required (see appendix C). 3.12 Gross domestic product grew strongly during 1960-93. Average growth in real terms was 0.049 per year during this time, and only 1982, 1983, and 1986 saw real declines in aggregate output (INEGI 1997). Furthermore, this growth was experienced across the economy (table 3.1). If anything, infrastructure growth was even more impressive; during this period. Electricity and communications stocks (total net capital stock, in real 1980 pesos, see appendix 68 Chapter 3 C) grew by 0.137 and 0.255 annually (compound annual rates, see table 3.2). Table 3.1: Compound Annual Growth Rates 1960-93 (real 1980 pesos) Major Sector GDP Sector Sector Growth Mining 2 0.048 Food and tobacco 3 1 0.043 Textiles 3 2 0.029 Wood products 3 3 0.034 Paper 3 4 0.051 Chemicals 3 5 0.070 Nonmetallic minerals 3 6 0.055 Basic metals 3 7 0.052 Machinery 3 8 0.068 Other manufacturing 3 9 0.028 Construction 4 0.048 Conmmerce, hotels 6 0.051 Financial services 8 0.047 Medicine 9 0.048 Table 3.2: Ihfrastructure Compound annual Growth rates, 1960-93 and 1983-93 (real 1980 pesos) Sector 1960-1993 1983-1993 Electricity 0.137 -0.003 Transport 0.063 0.029 Communic 0.255 0.075 ations Education 0.021 0.021 3.13 Physical infrastructure stocks have increased at rates comparable to, or perhaps a little higher than, rates in several other Latin American countries (Table 3.3 2). The table presents 2 Based on data from Canning (1998), which also include measures of telephones, unpaved roads, and rail lines. The mnain lines data (communications) and paved roads data (transportation) are of most interest here since they present the best opportunities for spillover effects which would appear in the national accounts data. 69 Chapter 3 averages of annual growth rates for several periods; the first column is for the whole period; the last column is for the most recent ten years. Growth has been, on average, somewhat lower in Mexico in recent years. Table 3.3: Physical infrastructure (average annual growth rates) 1950-1995 1950-1970 1970-1990 1980-1995 1985-1995 Electric Mexico 0.120 0.176 0.071 0.068 0.065 Argentina 0.057 0.072 0.049 0.034 0.021 Brazil 0.081 0.096 0.082 0.038 0.030 Chile 0.048 0.053 0.036 0.033 0.060 Colombia 0.084 0.119 0.064 0.056 0.070 Venezuela 0.098 0.121 0.096 0.057 0.053 Telephone main lines Mexico 0.087 0.074 0.096 0.082 0.091 Argentina 0.053 0.035 0.046 0.077 0.076 Brazil 0.093 NA 0.099 0.068 0.064 Chile 0.087 0.081 0.068 0.120 0.138 Colombia 0.079 0.081 0.073 0.090 0.086 Venezuela 0.092 0.094 0.088 0.079 0.074 Paved roads Mexico 0.020 0.058 -0.045 0.019 0.019 Argentina 0.045 0.086 0.031 0.014 0.012 Brazil 0.123 0.212 0.067 0.051 0.039 Chile 0.041 0.091 0.021 0.018 0.038 Colombia 0.052 0.077 0.038 0.021 0.024 Venezuela 0.042 0.086 0.018 0.027 0.033 70 Chapter 3 3.14 Our financial measures are more convenient for analysis, and are reasonable approximations to the measures of physical infrastructure reported by Canning (1998). Figure 1 shows changes in (log3) electric, transport, and communications infrastructures over past several decades. The physical and financial measures of electric infrastructure, with the exception of the early 1960s, moved pretty much in tandem over the past few decades. The transportation measures also move similarly, although there is a substantial jump in (financial) value in the late 1970s. Prior to, and after, this two year jump, the measures move in similar fashions. The value of communications infrastructure also makes a substantial jump in the late 1970s; an increase which we do not see in the telephone main lines data. This, like the jump in transport infrastructure, may be due to quality changes which the physical measures do not capture. Figure 3.1 Changes in electric, transport, and communications infrastructure (log3) 1000000 - 1000000 l ,.. - ---------- 100000 - -- .o ~~~~~~~~~~~~~100000 .' 10000~~~~~~~~~~~~~~100 1000 10000 1960 1965 1970 1975 1980 1905 1990 1995 1960 1965 1970 1975 1980 1985 1990 1995 - Electric generating capacity - - - - Electic inc atucture PavedToads ------- TranspoTtation infrastnrctire 100000 - _ _ 10000 1OO'. _. ___ loo -- 10I 1960 1965 1970 1975 1980 1985 1990 1995 - Tel Main limes ------ Comm. infrastructure 3.15 Table 3.4 shows correlations between the physical and financial measures; the 3 The data are reported in logs to facilitate comparisons of growth rates, which are more relevant in this context than magnitudes. 71 Chapter 3 correlations between the logs of financial and physical measures are all over 90 percent; the correlations between growth rates are lower, but with the exception of the transport measures, even the growth rates of financial and physical measures are positively correlated. Table 3.4 Correlations: physical and financial infrastructure measures In,frastructure ln(Levels) Growth rates Electric (generating capacity) 0.9152 0.4766 Transport (paved roads) 0.9855 0.0638 Comimunications (main lines) 0.9774 0.3594 3.16 Clearly, however, investment has slowed significantly in recent years. This recent decline is even more noticeable in the post-sample period 1994-98. It prompts the concern that while necessary for short-term, pressing budgetary reasons, these reductions in public infrastructure will ultimately prove to increase costs in the private sector, perhaps even to the extent of raising private sector costs by more than the short-term public sector savings. The Model 3.17 The model of private sector costs in the Mexican economy is based on a standard translog cost model: an approximation of a general private sector cost function from a first-order Taylor series expansion in logs of a transcendental function (see Berndt 1991, chapter 9 and the references cited therein; Takayama 1990). In the following presentation, the data dimensions of interest (and their associated indices) are private inputs (indexed by i (- {L,K}) government infrastructure (indexed by m), sectors (indexed byj), and time (indexed by t). For brevity, the time index is ordinarily suppressed. 3.18 In each sector], the basic specification relates production costs (c) to input prices (w), 72 Chapter 3 technical progress (t), and output levels (y): 3 Cj = a,j + aLjwLj + aKjwKj + bljwLj2 + bjwLJwKj + bKKjWKj (1) + wijtbiTj + wijyjbyij + yjtbTr +nay; + byj2 + byjYW +bywKJ + Uj 3.19 Given the limited data available, we exploit assumed conmmonalties between the different sectors of the economy by estimating a single empirical model that treats each sector- year combination as a separate observation of a single production cost function. As our primary interest here is not the simple private cost function, but the effects of government infrastructure investments on private sector costs, we augment equation (1) with stocks of government infrastructure (see Feltenstein and Ha 1995; Nadiri and Mamuneas 1994). +KaLWL +aKW +bLLWL +bLKWLWK +bKWK + ±witbiT + ,wiybyi + ytb77 (2) i E +ayy+by2 +±bYywL +byKYWK + Pmg. +Epjimgmwi + pymg.y+u m m i m 3.20 However, the sectors will not be similar in all respects. We allow for differences in intercepts, in the marginal effects of input prices (w), and in output levels (y) and government infrastructure (g). 3.21 A sensible cost function will satisfy conditions that are not inherent in equation (2), such as homogeneity of degree one in input prices. We impose homogeneity of degree one in private factor input prices a priori. We do not impose inequality constraints such as ai > 0 orpm < 0. We also consider two further sets of restrictions on the estimated model: constant degree of homogeneity in output and constant returns to scale in private inputs. (Unlike, for instance, Nadiri and Mamunoas 1994 who impose CRS restrictions a priori). Appendix A discusses the specific restrictions and the implications of the restrictions for estimation in more detail. In general, we restrict our attention to the constant degree of homogeneity in the output case. Our estimates suggest that the Mexican economy has not been characterized by constant returns to scale, so the constant degree of homogeneity model is more appropriate than the CRS one. 3.22 One of the convenient aspects of working with a translog specification such as equation (2) is that the elasticities of private sector costs with respect to public infrastructure stocks are easily obtained. In sectorj, the elasticity with respect to infrastructure m is 3Notation is standard: vectors are denoted by bold type and logs (natural) by lower case. 73 Chgater 3 (3) elm = Pm + PimWi + PymY 3.23 Because the estimated elasticities are simply linear combinations of estimated coefficients, their variances are simple to calculate as well. Let A = [1 WL WK y] and b=[P. PL. PK. Pym] . Then ejm= Ab and V1(ejim )=AV(b)A'. In the elasticity calculations below, we construct values for A from the average (over the relevant sample horizon) values of w and y; we treat those values as constants. The Data 3.24 We briefly discuss the data used in this analysis here. See appendix A for complete, detailed documentation. Sources 3.25 All data derive from official Mexican government sources. The two primary sources are INEGI (1997) and Banco de Mexico (1995). For 1980-93, the data were taken from the government sources with little adjustment. For earlier periods (1970-80 and 1960-70), data series were often constructed from multiple sources. Appendix C describes these procedures in detail. 3.26 The sectors we use are determined by the Sistema de Cuentas Nacionales de Mexico (SCNM). Our basic delineation of the Mexican economy includes 17 sectors. One, agriculture, is dropped because of a lack of capital stock data. We used two others, electricity (sector 61, major sector 5) and transport and communications, (Sectors 64 and 65, or major sector 7) are used to construct our estimates of public infrastructure. This left us with 14 sectors for analysis (see table 3.6 for a list). The variable definitions are as follows: pb Gross sectoral output in real 1980 pesos pib Sectoral value added in real 1980 pesos labor Workers employed in the sector (millions) gwage Mean annual sectoral wages in real 1980 pesos lmpts Sectoral payments to labor in real 1980 pesos cap Total sectoral net capital stock in real 1980 pesos tbill Treasury bill rate (Mexico) cmpts Sectoral payments to capital in real 1980 pesos ielec Electricity infrastructure (total net capital stock of Sector 61 of the SCNM) itran Transport infrastructure (Sector of the SCNM) icomm Communications infrastructure (Sector 65 of the SCNM) primed Index of adult population with at least primary school 74 Chapter 3 Table 3.5 presents the variables of interest and their sample means and standard deviations. Table 3.5 Primary Data: Means and Standard Deviations (millions of real 1980 pesos) Variable 1960-93 1960-69 1970-79 1980-93 Pb 333,788 156,940 298,529 485,294 (349,484) (131,115) (250,968) (440,396) Pib 209,708 98,471 186,757 305,555 (283,186) (116,221) (221,609) (363,613) Labor 0.777 0.414 0.735 1.067 (1.347) (0.645) (1.160) (1.727) Gwage 114.0 85.0 134.4 120.2 (52.3) (35.5) (47.7) (56.5) Lpmts 71,277 37,780 74,411 92,964 (110,401) (52,010) (103,922) (137,032) Cap 38,963 17,308 39,159 54,292 (34,646) (15,939) (29,598) (39,312) Tbill 0.150 0.039 0.070 0.288 (0.136) (0.012) (0.025) (0.108) Cpmts 117,082 40,467 104,317 180,925 (164,152) (47,853) (134,742) (206,483) Ielec 195,837 47,624 178,566 314,041 (117,121) (26,934) (58,108) (21,461) Itran 114,124 35,558 90,504 187,115 (69,499) (4,729) (30,611) (29,910) Icomm 30,784 163 10,817 66,917 (33,718) (106) (12,535) (19,640) Primed 0.163 0.134 0.165 0.182 (0.024) (0.016) (0.007) (0.014) Source: Author's calculations. Sample horizon 3.27 Our dataset covers 1960-93. Because of the significant changes in the Mexican economy during this time, we estimated models for several subsamples of this period. In particular, we consider models of the periods 1960-79, 1974-1993, and 1960-93. Our base case was the 1974- 93 subsamples because this period should most closely reflect the modem Mexican economy. Data Limitations 3.28 This analysis is subject to several potential limitations. First, the data include only the 75 Chapter 3 period through 1993 thus to the extent that the experience of the past five years is particularly informative about the effects of public sector infrastructure on private sector costs, the analysis will be deficient. By restricting the base case estimates to 1974-93, we hope that this problem will be partially mitigated as the estimates are based on relatively recent history. 3.29 , This chapter takes an initial step toward evaluating the role public education has played as a component of the Mexican economy's infrastructure. Nonetheless, this step is not entirely satisfying, in part because of data limitations (see appendix C). The private sector cost elasticities that are most compelling are those that ignore education. This is a limitation of the data. The analysis does not provide compelling evidence either that education has no cost saving role or that it has a significant cost saving role. Instead, despite our attempts to date, we are unable to provide evidence one way or the other. 3.30 The Mexican economy exhibits considerable variation across different geographic areas and within the same industrial sector. This variation is masked in the aggregate data with which we work. One can consider this as a form of measurement error. Hence our coefficient estimates are likely to be biased (in absolute value) toward zero. Estimation of the Model Methods 3.31 The translog model allows for convenient simplification of the cost equation (2) into a set of share equations (see appendix A). Each represents the share of total costs for a private factor. In our case, with only two factors (labor and capital), we only need to estimate one share equation because the shares sum to one. We arbitrarily choose to use the labor share equation (given our estimation method, the coefficient estimates should be invariant to this choice). Therefore, let the translog model encompass a cost equation and a labor share equation. This allows us to estimate (in a seemingly unrelated regression framework) the two equations simultaneously and gain efficiency. Infrastructure 3.32 The data include measures of public infrastructure in four areas: electricity, transportation, communications, and education. In the cases of electricity, transportation, and communications, we consider total net capital (the sum of net capital in buildings and construction projects, machinery, transportation equipment, and office equipment) as our measure of infrastructure. Typically, these stocks are measured in real 1980 pesos. In the case of education, we use an index representing the proportion of the adult population with at least a primary school education. 3.33 The education measure is suspect for several reasons. First, while the development and labor literatures speculates that primary school education is particularly important for economic development, this claim is by no means certain. For this reason, alternative measures of educational infrastructure deserve close attention in the future. Second, even if primary school 76 Chapter 3 education is the appropriate infrastructure benchmark for Mexico, our measure is likely to be prone to measurement error. One form of measurement error arises because the penetration of primary school education is likely to vary across different parts of the Mexican economy: by geographic area, by workers' ages, and by sector of the economy. A second form of measurement error arises from the lengths to which we had to go to construct our index (see appendix C). Finally, because of the significant and relatively sustained growth in educational attainment in the Mexican work force during the past 30 years, our measure exhibits little variation that is, growth is concentrated somewhat in the early and later years, but is monotonic (see figure 3.2). More than 85 percent of the variation in the education index can be explained by a time trend; however, our model already captures technical progress by a time trend. For this reason, the education measure is unlikely to provide informative estimates. Figure 3.2 Education infrastructure index Education Infrastructure Index 0.25 ' 0.20 0. 90 0.15 0 0 -u ., '0.10 0 E 0 0. 0.00. 1960 1965 1970 1975 1980 1985 1990 1995 For all these reasons, we consider estimates both with and without the education infrastructure variable. On the whole, the without education estimates are more plausible. Estimation limitations 3.34 The formulation of the problem in equation (2) is not entirely satisfactory. There are two estimation issues related to the role of input prices. The first is that for the economy as a whole, presumably the prices of capital, r, and labor, w, are endogenous. If so, OLS and similar estimators will be inconsistent. Ideally, we would instrument for factor prices, however, in the absence of good instruments, we argue that factor prices in Mexico are determined in a world market. For our purposes, private factor prices are not endogenous. 3.35 A second objection to the structure above is that instead of government infrastructure 77 Chapter 3 stocks, one would preferably use prices (much as we use private factor prices). If infrastructure is acquired and utilized in the same fashion as other inputs, this objection is appropriate. Here, the use of stocks may not be problematic. Recall that our interest is in the spillover effects of publicly provided infrastructure; we estimate those spillovers through the translog model. The translog specification is appropriate not because our priors (based on, e.g., engineering data) suggest that costs vary as in equation (1), but because equation (1) is a good approximation to an arbitrary cost function. We allow the data to tell us how costs are affected by the inputs. Use of private inputs such as capital and labor, from standard cost theory, are determined by factor prices; therefore r and w are crucial regressors. By extension, this is why one would want to include infrastructure usage costs in equation (2). However, in the case of infrastructure, utilization, or congestion, would be more desirable than prices; firms do not demand infrastructure services, they use the services that are available. Presumably private sector costs vary with utilization, z: C zGu or in logs: c z+g+u 3.36 The problem is an omitted variable one; we estimate c g + u rather than c z + g + u . The effects of utilization on costs will be, in our formulation, attributed to the disturbance term u; if cov(z,g) 0, the estimates of the effect of government infrastructure, g, on private costs c will be inconsistent. However, if utilization is uncorrelated with infrastructure levels, the estimates of the cost reducing effects of infrastructure will not be adversely affected by the omission of utilization from the estimated model. A sufficient, but not necessary, condition for this (omission of utilization not biasing the estimated effects) is that government infrastructure is utilized at the same rate each period. But, a weaker condition, that utilization is not correlated with infrastructure levels, is also sufficient. This assumption is more plausible and is the one that we make. 3.37 The Mexican economy is apt to exhibit considerable variation across different geographic areas and within the same industrial sector. This variation is masked in the aggregate data with which we work. One can consider this as a form of measurement error; hence, our coefficient estimates are likely to be biased (in absolute value) toward zero. Results 3.38 Our primary interest is in the effects of public infrastructure on private sector costs. Our summary measures of these effects are the elasticities of private sector costs with respect to the various infrastructure measures. The base case consists of estimates based on * The period 1974-93 * The stocks of electricity, transport, and communications infrastructure (but not education) * A constant degree of homogeneity in output. 78 Chapter 3 Overview 3.39 Tables 3.6-3.9 present the estimated elasticities and their standard errors for four different specifications of the basic translog model. The first, which we call our "base case," includes the three main (measurable) infrastructure stocks: electricity, transport, and communications. Recall that we exclude education, not because it is unimportant, but because we cannot econometrically separate the effects of education from the effects of our measure of technological progress. The base case estimates are calculated imposing the constraint that costs are homogeneous of degree one in output (doubling output doubles private sector costs). 3.40 The "no restrictions" specification considers how the estimated effects of infrastructure change if we do not impose as many constraints on the basic translog model. The only constraints imposed on the basic translog specification are those necessary for homogeneity of degree one in input prices. The estimates, overall, are not too different from the base case estimates. The no restrictions estimates of the effects of electric infrastructure are slightly more negative, while the estimated effects of transport infrastructure are less so. The estimated communications effects are quite similar. 3.41 The constant returns to scale model is similar to the base case, but an additional constraint is imposed: the scale of operations does not change marginal costs. Imposing the constant returns to scale constraint generates larger (more negative) estimated effects of electric infrastructure, but smaller (more positive) estimated effects of transport and communications infrastructure. A test of rejects the null hypothesis that costs are characterized by CRS. 3.42 The cost elasticity estimates are generally plausible: as table 3.6 shows, the magnitudes are not overly large and most elasticities have appropriate signs (see appendix A for more details about the estimated models). Of the 14 base case elasticity estimates for electricity, 10 are negative with values of roughly -0.20. The estimated standard errors are in the range of 0.10- 0.16. The smaller (in absolute value) elasticity estimates are, by conventional standards, insignificantly different from zero. The larger values (for instance, for sectors 2, 4, and 5), however, do tend to be several times the size of their standard errors. The transportation estimates are fairly similar. Of the 14, 12 are negative and many take on values around -0.15 or -0.35. Again, the standard errors are roughly 0.15. 3.43 The communications elasticities, however, are generally quite small (roughly 0.05), the wrong sign (9 of the 14 are positive), and imprecisely estimated (standard errors of 11 of the 14 are larger than their corresponding coefficients). 3.44 The alternative cases generally present a mixed picture. The model with education generates coefficient estimates that are mostly negative (9, 10, 12, and 10 of the 14 sectors for each of the electricity, transport, communications, and education infrastructures, respectively). However, only 13 of the 56 are at least twice their corresponding standard errors (and one of these is a positive elasticity of 1.135 for education in sector 1). 79 Chapter 3 Table 3.6: Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Base Case Rector Fle7tric Trqnqnnrtntinn Commiminstincn Mining 0.003 -0.086 -0.022 (0.018) (-0.532) (-0.360) Food and tobacco -0.419 -0.179 0.043 (-3.416) (-1.183) (0.911) Textiles -0.095 -0.495 0.063 (-0.776) (-3.449) ( 1.578) Wood products -0.563 -0.433 0.071 (-4.445) (-3.282) ( 1.503) Paper -0.382 -0.140 0.064 (-3.148) (-0.840) ( 1.291) Chemicals 0.001 -0.083 -0.061 (0.004) (-0.488) (-0.916) Non-metallic minerals -0.200 -0.157 0.009 (-1.610) (-1.092) ( 0.190) Basic metals 0.225 -0.394 0.035 (1.895) (-2.894) (0.799) Machinery -0.021 -0.338 0.046 (-0.169) (-2.347) ( 0.932) Other manufacturing -0.396 0.355 -0.030 (-3.283) (2.482) (-0.724) Construction -0.283 0.019 0.020 (-1.707) (0.142) (0.430) Commerce, hotels -0.272 -0.003 -0.008 (-1.525) (-0.017) (-0.124) Financial services -0.085 -0.136 -0.009 (-0.680) (-0.817) (-0.162) Medical services 0.098 -0.245 0.041 ( 0.595) (-1.661) (0.676) Note: Figures in parenthesis are t-ratios Source: Author's calculations. 80 Chapter 3 Table 3.7: Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks Education Sector Electric Transportation Communications Education Mining 0.374 -0.353 -0.051 1.135 (2.382) (-2.179) (-0.797) (2.690) Food and tobacco -0.429 -0.033 -0.020 -0.743 (-3.488) (-0.188) (-0.392) (-1.434) Textiles -0.145 -0.236 0.012 -1.119 (-1.306) (-1.372) (0.293) (-2.645) Wood products -0.558 0.105 -0.059 -1.394 (-4.810) (0.625) (-1.180) (-3.602) Paper -0.399 -0.03 -0.007 -0.741 (-3.531) (-0.175) (-0.135) (-1.647) Chemicals 0.019 -0.003 -0.155 -0.450 (0.143) (-0.017) (-2.095) (-0.796) Nonmetallic minerals -0.183 -0.058 -0.05 -0.692 (-1.578) (-0.333) (-0.980) (-1.407) Basic metals 0.188 0.022 -0.076 -1.586 (1.725) (0.127) (-1.583) (-3.448) Machinery 0.007 -0.380 0.011 -0.180 (0.056) (-2.209) (0.216) (-0.357) Other manufacturing -0.356 0.092 -0.002 0.033 (-3.207) (0.535) (-0.049) (0.088) Construction -0.182 -0.136 -0.003 0.050 (-1.167) (-0.768) (-0.059) (0.132) Commerce, hotels -0.224 0.043 -0.082 -0.254 (-1.098) (0.242) (-1.281) (-0.540) Financial services -0.001 -0.145 -0.066 0.175 (-0.009) (-0.890) (-1.138) (0.381) Medical services 0.073 -0.086 -0.049 -0.534 (0.427) (-0.512) (-0.754) (-1.251) Note: Figures in parenthesis are t-ratios. 81 Chapter 3 Table 3.8: Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks No Restrictions Sector Electric Transportation Communications Mining -0.126 -0.028 -0.041 (-0.818) (-0.181) (-0.624) Food and tobacco -0.412 -0.109 0.038 (-3.421) (-0.749) (0.759) Textiles -0.187 -0.510 0.106 (-1.538) (-3.581) (2.420) Wood products -0.625 -0.272 0.057 (-5.046) (-2.046) ( 1.062) Paper -0.452 -0.038 0.050 (-3.757) (-0.244) ( 0.929) Chemicals -0.076 0.007 -0.131 (-0.542) (0.041) (-1.996) Non-metallic minerals -0.274 -0.074 0.008 (-2.269) (-0.544) ( 0.164) Basic metals 0.172 -0.257 0.016 (1.469) (-1.923) ( 0.333) Machinery -0.100 -0.294 0.047 (-0.815) (-2.053) ( 0.928) Other manufacturing -0.391 0.468 0.004 (-3.069) (3.358) (0.074) Construction -0.364 -0.050 0.051 (-2.249) (-0.357) (0.962) Commerce, hotels -0.240 -0.023 -0.025 (-1.308) (-0.153) (-0.403) Financial services -0.162 -0.091 -0.026 (-1.290) (-0.580) (-0.461) Medical services 0.095 -0.205 0.010 (0.566) (-1.424) (0.163) Note: Figures in parenthesis are t-ratios. 82 Chapter 3 Table 3.9: Estimated Private Sector Cost Elasticities with Respect to Public Infrastructure Stocks CRS Sector Electric Transportation Communications Mining _n Q -01 79 n (R (-3.002) (-0.901) ( 1.164) Food and tobacco -0.807 0.044 0.063 (-5.802) (0.251) ( 1.133) Textiles -0.276 -0.153 0.027 (-1.943) (-0.931) ( 0.570) Wood products -0.740 -0.074 0.065 (-4.972) (-0.494) ( 1.176) Paper -0.776 0.048 0.102 (-5.646) (0.248) ( 1.745) Chemicals -0.375 0.019 0.027 (-2.525) ( 0.093) (0.344) Non-metallic minerals -0.571 0.063 0.028 (-4.028) ( 0.380) (0.502) Basic metals -0.212 -0.460 0.113 (-1.624) (-2.902) (2.230) Machinery -0.446 -0.405 0.122 (-3.124) (-2.408) (2.120) Other manufacturing -0.692 0.897 -0.101 (-4.962) ( 5.634) (-2.102) Construction -0.669 0.499 -0.012 (-3.447) ( 3.287) (-0.210) Commerce, hotels -0.771 0.290 0.025 (-3.724) ( 1.668) (0.342) Financial services -0.470 -0.089 0.058 (-3.326) (-0.456) ( 0.878) Medical services -0.342 -0.121 0.105 (-1.808) (-0.700) ( 1.450) Note: Figures in parenthesis are t-ratios. 3.45 Detailed coefficient estimates are shown in appendix D. Appendix C considers additional specification issues, such as heteroskedasticity and autoregression. In general, these heteroskedasticity and autoregressive consistent estimates are larger (not in absolute value) than 83 Chapter 3 the base case estimates (table 3. 10). Table 3.10. Mean Elasticities Category Electric Transportation Communications Base case -0.171 -0.165 0.019 OLS, robust -0.065 0.031 0.072 OLS, random effects -0.036 -0.012 0.060 FGLS, hetero, AR(1) -0.019 -0.005 0.069 Source: Author's calculations. 3.46 This lends further support to the conclusion that the role of public infrastructure in reducing private sector costs in Mexico is modest. A notable feature of table 3.10 is that the estimated infrastructure effects are generally largest for electric infrastructure and smallest for communications, with the effects of transportation capital falling somewhere in the middle. The effects of communications infrastructure appear to be consistently positive, implying that additional public sector communications infrastructure increases private sector costs. Payoffs from Additions to Infrastructure 3.47 There are several ways to try to evaluate the importance of these estimates for policymakers. Estimates of the optimal stocks of public infrastructure would be useful. Unfortunately, in a static context, whether the optimum is knowable is not clear. Here we estimate static elasticities, essentially the slope of the line relating changes in private sector costs to changes in infrastructure stocks. If the elasticities are negative, then naively, the estimates suggest increasing infrastructure infinitely. An alternative approach is to consider the estimated costs and benefits that would be associated with different increases in infrastructure stocks. Note that these are static calculations that ignore the shadow cost of public funds; ignore firms' reactions to infrastructure increases; and ignore financing costs, such as increases in the real interest rate. 3.48 We consider three alternatives. The first is an illustrative elasticity calculation: what is the result of a 1 percent increase in public infrastructure stocks? Second, what would be the effect of an increase in infrastructure stocks of the same size as current gross investment by 60 percent (to 5 percent of GDP)? All this additional investment is assumed to add to infrastructure stocks, that is, we assume that depreciation is covered separately. 84 Chapter 3 Static Costs and Benefits of Increased Infrastructure 3.49 The basic calculation is a simple comparison of the value of the estimated reductions in private sector costs following from an increase in public sector infrastructure stocks. The cost of a d percentage increase in infrastructure stocks is (4) Edgm while the estimated benefits across the sectors 1 < j < 14 are: (5) E E dej,,cj . m 3.50 Using our base case estimates and the 1993 values for infrastructure stocks and sectoral costs, the estimated cost of a 1 percent increase in infrastructure would be rnxp 6.62 billions (real 1980 pesos). The total benefits in private sector cost reduction would be mxp 12.35 billions (real 1980 pesos). If these benefits were accrued over several years, the total would be larger. At depreciation and discount rates of 10 percent, the present value would be approximately mxp 55.6 billion. The two scenarios based on larger expansions of public infrastructure stocks would be associated with similarly larger multiperiod payoffs. Table 3.11 presents estimated costs and benefits for the three scenarios. Table 3.11: Static Costs and Benefits of Increased Infrastructure (real 1980 pesos, billions) Private sector cost Investment costreuto reduction I % increase in infrastructure stocks 6.6 12.4 Increase by current gross spending amount (3% GDP) 186.9 348.7 Increase current amount by 60% (5% GDP) 311.5 581.1 3.51 These are, of course, static calculations which ignore general equilibrium effects of additional infrastructure investments through channels such as wage and interest rates. See Feltenstein and Murphy (1999) for a discussion of added investments in infrastructure stocks in the context of a general equilibrium model of the Mexican economy. 3.52 A second approach to evaluating appropriate levels of infrastructure stocks is to recognize that we estimate constant elasticities, but not constant peso changes. Therefore, our expression of infrastructure stocks in currency rather than physical terms allows a simple 85 Chapter 3 calculation which yields an estimate of "optimal" infrastructure stocks. For infrastructure stock m, by definition the elasticity of private sector costs with respect to stock m is: dCl (6) em C dG /Gm A static notion of optimality would require that the marginal benefits of increases in infrastructure stock (- dCG ) be equal to the marginal costs of those increases. Infrastructure stocks are measured in pesos and we ignore potential infrastructure production nonlinearities. Therefore the marginal cost of infrastructure increases is -1 (it costs 1 peso to increase stocks by 1 peso). Then, at the optimum level of infrastructure Gm : (7) -dC-1 =0 dGm A slight rearrangement of (7) using (6) gives: (8) Gm * e=C This calculation of the optimal stock implicitly assumes: 1. Total private sector cost remains constant (C is fixed) 2. Infrastructure production has constant returns to scale (marginal cost of producing infrastructure is constant) 3. Increasing infrastructure stocks does not affect the shadow cost of government funds (which is implicitly assumed to be zero: the cost of 1 peso is 1 peso) 4. Only current period benefits matter in calculating optimal levels. Optimal Infrastructure Stocks 3.53 The calculated optimal infrastructure stocks are shown in Table 3.12. The calculations are based on the mean base case elasticities for each infrastructure type with three different elasticity assumptions for communications: -0.019 (negative of base case), 0.00, and -0.010. The table also includes calculations using panel elasticity estimates (feasible GLS, heteroskedastic consistent, 1 period autoregressive structure; see Table 3.10) with the communications elasticity set to 0.00. 86 Chapter 3 Table 3.12: Optimal Infrastructure Stocks (real 1980 pesos, billions) em Actual Gm Optimal Gm* Ratio Gm/Gm* Base case Electric -0.171 335.0 826.8 0A41 Transportation -0.165 217.4 797.8 0.27 Communications -0.019 110.0 91.9 1.20 Communications 0.000 110.0 0.0 Communications -0.010 110.0 48.4 2.27 Panel Electric -0.019 335.0 91.9 3.65 Transportation -0.005 217.4 24.2 8.98 Comnmunications 0.000 110.0 0.0 3.54 Unfortunately, because the elasticity estimates are not particularly precise, neither are the calculated levels of optimal infrastructure stocks. If the base case estimates are accurate then current levels of electric and transportation infrastructure are too low; the optimal levels are 2-4 times larger. However, the panel estimates suggest much lower elasticities and hence optimal infrastructure levels. Consideration of the multi-period flow of benefits from infrastructure capital would suggest higher optimal stocks. Conclusions 3.55 This chapter provides estimates of the effects of public sector infrastructure stocks on private sector production costs. The estimates, although problematic in some respects, are generally quite sensible. The estimates are reasonably consistent across different empirical models and, in the base case, estimated precisely. On average, the elasticity of private sector costs with respect to public infrastructure costs is approximately -0.10 to -0.15. Given the size of public infrastructure stocks and private sector costs, these elasticity estimates suggest that moderate increases in public sector infrastructure stocks would be welfare improving. The imprecise nature of the estimates suggests that caution is appropriate, however, and the communications and education estimates in particular should be regarded as comments on the data rather than definitive statements about productivity. 87 Chapter 3 References Banco de Mexico. 1995. "La Encuesta de Acervos, Depreciacion y Formacion de Capital del Banco de Mexico". Computer files. Bemdt, E.R. 1991. The Practice of Econometrics: Classic and Contemporary. Reading, Mass.: Addison-Wesley. Canning, D. 1998. "A Database of World Infrastructure Stocks, 1950-1995", Unpublished working paper. Estados Unidos Mexicanos. 1998. Cuarto Informe de Gobierno, Annexo. Poder Ejecutivo Federal: Presidencia de la Republica, Mexico. Feltenstein, A., and Ha. 1995 "The Role of Infrastructure in Mexican Economic Reform", World Bank Economic Review, 9(2): 287-304. Feltenstein, A., and R. Muiphy. 1999. "Can Infrastructure Protect Against Shocks?: An Analysis of the Situation of Mexico". Unpublished working paper. Greene, W. H. (2000). Econometric Analysis (Fourth ed.). Upper Saddle River, N.J.: Prentice Hall. Hulten, C. R. 1996. "Infrastructure Capital and Economic Growth: How Well You Use It May Be More Inportant Than How Much You Have." Working paper 5847, National Bureau of Economic Research, Cambridge, Mass. IMNF. 1998. International Financial Statistics CD-ROM. INEGI.1997. Cuentas Nacionales de Mexico. CD-ROM. McGuckin, R. H., M. L. Streitsweiser, and M. E. Doms. 1996. 'The Effect of Technology Use on Productivity Growth". Working paper CES 96-2. Center for Economic Studies, U.S. Census Bureau, Washington, D.C. Nadiri, M. I., and T. P. Mamuneas. 1994. "The Effects of Public Infrastructure and R&D Capital on the Cost Structure and Performance of U.S. Manufacturing Industries". The Review of Economics and Statistics 76 (1): 22-37. Psacharopoulos, G. 1994. "Returns to Investment in Education: A Global Update". World Development 22(9): 1325-43. Stiroh, K. J. 1998. "Computers, Productivity, and Input Substitution", Economic Inquiry, 36(2): 175-91. 88 Chapter 3 Takayama, A. 1990. Mathematical Economics, (2nd ed.), Cambridge University Press, Cambridge, U.K. 89 Chapter 3 Appendix A. Estimation Model Details Let the augmented translog model from equation (2) be given by 2 2 c = ao + aLwL + aKwK + bLLwL + bLKWLWK + bKKWK + witbiT + E wiybyi + ytb, (9) 2 + a(y + bryy + byLywL + byKywK + I Pm*g + IIpjmgm wi + Ipymg.Y + u m m i m where the *d coefficient vectors include sector-specific effects. We gain some additional efficiency in our estimates by noting that from Shepard's Lemma, the above specification implies, that the share of total cost attributable to each private factor input is (10) SL =aL +bLKWK +bLLWL +bLYY+jPL.9 +UL (11) S