Mexico - Mobilizing savings for growth (Vol. 2 of 2) : Detailed report

Report No. 16373-ME Mexico Mobilizing Savings for Growth (In Two Volumes) Volume ll: Detailed Report December 23, 1997 Mexico Country Management Unit Latin America and the Carribean Region Document of the World Bank  CURRENCY EQUIVALENTS Currency Unit = Mexico New Peso (MexNP) US$1.00 = MexNP8.072 (December 18, 1997) GOVERNMENT FISCAL YEAR January 1- December 31 ABBREVIATIONS AND ACRONYMS ABS Asset Backed Securities ADE Debtor Relief Program (Acuerdo de Apoyo Inmediato a los Deudores de la Banca) AFORES Retirement Fund Administrator (Administradores de Fondos de Ahorro para el Retiro) ASCRA Accumulating Savmgs and Credit Association CANAFO Mexican Development Savings Bank (Caja Mexicana de Fomento) CD Certificate of Deposit CEA Special Savings Account (Cuentas Especiales de Ahorro) CETES Mexican Treasury Bills CMCP Confederation of Mexican Cajas Populares CNBV National Banking and Securities Commission (Comisi6n Nacional Bancaria y de Valores) CONSAR National Commission of the Retirement Savings System (Comisi6n Nacional del Sistema de Ahorro para el Retiro) CPM Caja Popular Mexicana EU European Union FOBAPROA Bank Fund for Savings Protection (Fondo Bancario de Protecci6n al Ahorro) FOVI Bank Housing Finance Fund (Fondo de Operaci6n y Financiamiento Bancario a la Vivienda) FOVISSSTE Government Workers' Housing Fund (Fondo para la Vivienda de los Trabajadores del ISSSTE) GAPP Generally Accepted Accounting Principles GATT General Agreement on Trade and Tariffs GDP Gross Domestic Product GDS Gross Domestic Saving GNI Gross National Income GNP Gross National Product GNS Gross National Saving Vice President Shahid Javed Burki, LCR Country Director Olivier Lafourcade, LCC 1C Sector Manager Guillermo Perry, LCSPR Staff Member Zia Qureshi, LCCIA  MEXICO MOBILIZING SAVINGS FOR GROWTH VOLUME I: SUMMARY REPORT CONTENTS Preface ......................................... vii A. Overview ................................1.. . B. Mexico's Savings Performance .................................2 C. Why Savings Must Increase ...6.............................6 D. Policies for Mobilizing Savings .................................7 (i) Maintaining Macroeconomic Stability .................................8 (ii) Improving the Tax System ................................9 (iii) Extending Pension System Reform ................................ 11 (iv) Developing the Financial Markets ................................ 13 (v) Fostering Small Savings ................................ 17 Statistical Annex: Basic Macroeconomic Data ................................ 20 TEXT FIGURES 1 National Saving and Investment Balance .......3.....................................3 2 Public and Private Saving Rates ............................................3 3 Bank Liabilities and Credit Extended to the Private Sector .4 4Estimates of Private Sector Wealth.4 4 Estimaes of Pivate Sctor Welth .................................................................................................4 5 The Current Account and Portfolio Capital Inflows .5 l   VOLUME II: DETAILED REPORT CONTENTS Preface to Volume II ............................................ vii 1. The Decline in Mexican Saving: A Cost of Reform? .............................................1 I. Introduction .............................................1 II. Measurement of Saving ............................................3 National versus Domestic Saving ............................................3 Public-Sector Budget Accounts .............................................5 Private versus Public Saving .............................................7 111. Accounting for The Saving Decline ............................................ 10 Exchange Rate-Based Stabilization ............................................ 10 External Sector Liberalization ............................................ 12 Financial Sector Liberalization ............................................ 16 An Asset-Market Bubble ............................................ 18 Regression Analysis ............................................ 21 IV. Policy Implications ............................................ 23 2. Public Saving, Stabilization and Private Saving in Mexico .................... ........................ 29 I. Introduction ............................................ 29 II. Savings in Mexico and Around the World ............................................ 30 m. Savings Behavior in Mexico ............................................ 31 Measuring Savings ............................................ 31 Annual Data 1960-1994 ............................................ 34 Quarterly Data 1980-1995 ............................................ 35 Regression Analysis ............................................ 36 IV. A Model of Consumption and Saving in Mexico .............. .............................. 39 Effects of Shocks on Macroeconomic Aggregates ....................... ..................... 40 Predictions of the Model ............................................ 44 V. Conclusion ............................................ 46 3. Tax Policy, Domestic Savings and Investment in Mexico ................... ......................... 67 I. Introduction ............................................ 67 II. Current Tax Structure in Mexico ............................................ 68 iii  ,rv Contents III. Taxation of Investment in an Open Economy .................................................................... 68 IV. Capital Deepening versus Productivity Growth .................................................................... 74 V. Summary .................................................................... 77 4. Financial Markets and Savings Mobilization: Regulatory and Development Agenda .. 81 I. Overview .................................................................... 81 II. Financial Markets and Savings .................................................................... 83 III. Pension Funds .................................................................... 86 Recent Pension System Reform ............................. ....................................... 86 Policy Issues in Pension System Design .................................................................... 86 Recommendations .................................................................... 91 IV. Bankmg System . 92 Current Status of the Banldng System .92 Key Reform Issues .97 Recommendations .100 V. Stock Market .101 Recent Performance of the Mexican Stock Market .101 Causes of Small Corporate Universe, Low Market Capitalization and Illiquidity . 103 Recommendations .104 VI. Asset-Backed Securities Markets .105 Problems in Issuance of ABS in Mexico .106 Recommendations .108 VII. Mutual Funds .109 Recent Performance .109 Regulatory Issues .109 Recommendations .110 5. The Impact of Mexican Pension Reform on Savings ..113 I. Introduction .113 II. The Mexican Pension Reform ...................................................,,,,,,,,.. 115 The Existing System and Need for Reform.. 115 The New Pension System for Private Sector Workers . 117 Assessment, Risks and Uncertainties . 120 III. Methodology and Background Projections . 122 Impact of Pension Reform on Savings: Conceptual Background . 122 A Projections Model . 123 IV. Analysis of the Savings Impact . 128  Contents v Estimation ............................................................. 128 Assessment ............................................................. 131 V. Conclusion ............................................................. 132 6. Improving the Quality and Availability of Small Savings Instruments in Mexico . .................. 143 I. Introduction .............................................................. 143 II. Overview of Financial Savings Services for Low-Income Groups ....................................... 145 Formal Savings Services ............................................................. 145 Semi-Formal Savings Services ............................................................. 151 Informal Savings Services ............................................................. 153 III. Conclusions ............................................................. 155 IV. Policy recommendations ............................................................. 155 Reducing Transaction Costs ............................................................. 156 Fostering Institution-Building ............................................................. 157 TEXT TABLES 1.1 Saving and Investment Balance, 1970-94 .1 1.2 Relationship Between National Accounts Aggregates, 1970-94 .4 1.3 Alternative Measures of Public Sector Balance, 1970-94 .6 1.4 Alternative Measures of Public and Private Saving, Nominal Values .7 1.5 Alternative Measures of Public and Private Saving, Inflation-Adjusted .9 1.6 Interest and Exchange Rates, 1980-94 .12 1.7 The Mexican Current Account Deficit and Its Financing .14 1.8 Private Consumption and Its Components, 1980-94 .15 1.9 Net Credit and Liabilities of the Consolidated Banking System, 1981-94 .17 1.10 Price Indices for Mexican Equities and Land, 1980-94 .19 1.11 Private Sector Wealth and its Components, 1980-94 .20 1.12 Regression Estimates for the Private Saving Rate, 1971-94 .22 2.1 Growth Regressions with Savings Rates, 1960-1989 .47 2.2 Summary Statistics for Growth Rates .47 2.3 Summary Statistics for Growth Rates .48 2.4 Consumption Regressions ...................................... 49 2.5 Savings Regressions .50 4.1 Agenda for Financial and Capital Market Development .84 4.2 Size of Mexican Financial Institutions and Markets .85 4.3 Structure of Mexican Commercial Banks .93 4.4 Size Distribution of Commercial Bank Direct Funding .95 4.5 CNBV's Latest Estimate of the Fiscal Cost of Bank Restructuring Programs .97 4.6 Measures of Mexican Stock Exchange .102 4.7 Comparative Stock Market Indicators, 1994-1995 .................................................................... 102  _i Contents 5.1 hIflation and Minimum Wage Assumptions .............................................................. 133 5.2 Fiscal Costs of Reform .............................................................. 134 5.3 Net Accumulation in Individual Savings Accounts and Retirement Insurance Reserves ............. 135 5A4A Impact on National Savings under Alternative Financing Schemes Low Inflation Scenario ....... 136 5 5A Impact on National Savings under Alternative Financing Schemes High Inflation Scenario ...... 138 6 1 Bank Branches and Accounts .............................................................. 144 6.2 Agregate Savings and Financial Penetration .............................................................. 144 6.3 Small Savings Instruments in Mexico .............................................................. 145 TEXT FIGURES 1.1 National Saving and Investment Balance, 1970-94 ............................................................... 2 1.2 Alternative Measures of National Saving, 1970-94 ............................................................... 5 1.3 Public Sector Budget Balance, Nominal and Inflation-Adjusted ....................................................7 1.4 Nominal Values of Public and Private Saving ...............................................................8 1.5 Inflation-Adjusted Values of Public and Private Saving .............................................................. 12 1.6 The Current Account and Portfolio Capital Inflows .............................................................. 15 1.7 Bank Liabilities and Credit Extended to the Private Sector, 1980-94 ........................................... 18 1.8 Private Sector Wealth, Alternative Valuations .............................................................. 20 1.9 Actual and Predicted Private Saving Rate, 1971-94 .................................................................... 22 2.1 Level of Income and Savings Rate .. 51 2.2 National Savings Rate and Growth Rate of Real Per Capita GDP .. 51 2.3 Annual Time Series for Mexico, 1960-94 .. 52 2.4 Annual Time Series for Mexico, 1960-94 .. 52 2.5 Annual Time Series for Mexico, 1960-94 .. 53 2.6 Quarterly Time Series for Mexico, 1980Q1-1995Q2 .. 53 2.7 Quarterly Private and Public Sector Time Series .. 54 2.8 Response to TFP Shock in Tradables Sector .. 54 2.'9 Response to TFP Shock in Nontradables Sector .. 55 2.10 Response to a Decrease in Foreign Interest Rate .. 55 2.11 Response to an Increase in Foreign Iflation Rate ...:... 56 2.12 Response to a Decrease in Depreciation Rate of the Peso .. 56 2.13 Response to an Increase in Govenmment Consumption .. 57 2.14 Response to a Decrease in Tax Rate .. 57 2.15 Observable Shock Measures .. 58 2.16 Model Predictions .. 58 2.17 Model Predictions for Shares of GDP .. 59 4.1 Ten Most Active Stocks' Share of Total Value Traded ............................................................ 103 5.1 Fiscal Costs under Alternative Inflation Scenarios ............................................................ 126 5.2 Private Savings Accumulation under Alternative Inflation Scenarios .......................................... 128 5.3 Impact on National Savings under Alternative Financing and Inflation Scenarios ...................... 130 5.4 Ipact on National Savings under Alternative Assumptions about Substitution Effect .............. 130 5.5 Inpact on National Savings under Alternative Assumptions about INFONAVIT Reformn .......... 131  PREFACE TO VOLUME II The main findings and recommendations of this report (Mexico: Mobilizing Savings for Growth) are summarized in Volume I. This volume contains the detailed analysis underpinning these findings and recommendations, based on the background papers prepared for this report. As these papers were prepared separately, there is some overlap between them, but they are being presented here with relatively minor editing in order to retain the richness of their material as much as possible. There are six papers in all. The first paper 'The Decline in Mexican Saving: A Cost of Reform?" analyzes the causes of the substantial drop in Mexico's national savings rate between the mid-1980s and the early 1990s. An important focus of this paper is the role played in the savings rate decline by financial market developments-financial sector liberalization and the sharp rise in bank lending to the private sector, the surge in foreign capital inflows, and the boom in asset values. The second paper 'Public Saving, Stabilization and Private Saving in Mexico" contains a more detailed quantitative analysis of the determinants of private savings. Both an econometric analysis and a computable model are employed to explain the time-series behavior of private savings, especially the relationship between public and private savings and the role played by the stabilization program adopted by Mexico in the late 1980s (particularly through exchange rate and interest rate developments). TIhe third paper 'Tax Policy, Domestic Savings and Investment in Mexico" discusses the role of tax policy m promoting private savings and investment. Various options for tax reform to stimulate capital accumulation are examined. As a centerpiece of such reform, the discussion focuses on the expansion of the role of the value-added tax relative to the income tax in order to reduce the effective tax on capital. The fourth paper 'Financial Markets and Savings Mobilization: Regulatory and Development Agenda" discusses issues relating to making the financial and capital markets play a stronger role in mobilizing savings and allocating them efficiently. The discussion encompasses the banking system, pension funds, equity and debt securities markets, and mutual funds, focusing on the agenda for strengthening the regulatory and institutional framework for the development of these markets. The fifth paper 'The Impact of Mexican Pension Reform on Savings" focuses on the role of pension system reform in the mobilization of savings. It assesses the recent Mexican pension reform, estimates its likely impact on fiscal costs and private savings under different scenarios, and recommends follow-up reform actions to enhance the overall savings impact. Finally, the sixth paper 'Ymproving the Quality and Availability of Small Savings Instruments in Mexico" evaluates the small savings schemes currently available in Mexico, ranging from formal to non- formal. Based on this evaluation, a set of recommendations is made to improve the access of the poor and low-income savers to savings schemes that best respond to their needs and to ensure the soundness of those schemes. These papers were prepared by a team of contributors. The team included: Barry Bosworth, Brookings Institution, consultant (Paper I); Craig Burnside (Paper II); Roger Gordon, University of Michigan, consultant (Paper D); Hemant Shah, Mike Lubrano and Laura Mecagni (Paper IV); Ulpiano Ayala, University of the Andes, consultant (Paper V); and Catherine Mansell-Carstens, consultant (Paper VI). The team leader for the overall report was Zia Qureshi. vii   1. THE DECLINE IN MEXICAN SAVING: A COST OF REFORM? 1. INTRODUCTION 1.1 Since the mid-1980s, Mexico has been engaged in a vigorous program of stabilization and structural adjustment. Its efforts have been often cited in the international community as a model for the type of reforms that Latin America needs to adopt if it hopes to mimic the growth experience of East Asia. Great progress was made in the macroeconomic sphere with the elimination of the public-sector budget deficit and sharply reduced rates of inflation, and the Government enacted many of the structural reforms urged on the developing nations by international organizations. Mexico liberalized its trade regime, established both current and capital-account convertibility in its international transactions, privatized the national banks, and sharply reduced government regulation of the domestic financial system. As judged by the willingness of international investors to commit funds, the Mexican program appeared to be a huge success. By 1993-94, it was attracting large volumes of foreign direct investment, and portfolio capital poured in at record rates. 1.2 The actual gains in economic performance, however, have been rather disappointing. The economy did rebound from the 1986 recession, and the export sector has responded very strongly to the trade liberalization measures; but overall growth remained well below the rates achieved in the 1970s. And, when intemational investors, for a variety of reasons, soured on Mexico in 1994, it was hit by a severe external financing crisis and an economic collapse even worse than those of 1983 and 1986. 1.3 In retrospect, one of the most notable features of the Mexican experience was the failure of the reforms to move the economy in the direction of higher rates of domestic saving and capital accumulation. As shown in Table and Figure 1.1, domestic investment had by 1990 largely recovered from the depressed rates of the mid-1980s, but it did not reach the peak share of GNP achieved in the early 1980s, and it has remained far below the rates that have become common for East Asia. Even more surprising, the rate of national saving actually fell substantially, from an average 24 percent of GNP in 1981-85 to 16 percent by 1993. The decline in national saving is all the more remarkable given the sharply reduced public-sector deficit. With the deterioration of internal saving, Mexico became increasingly dependent upon net inflows of foreign resources, averaging in excess of 6 percent of GNP in 1992-94, to Table 1.1: Saving and Investment Balance, 1970-94 (Percent of GNP) Category 1970-79 1980-84 1985-87 1988-89 1990-91 1992-93 1994 1. Total Saving 22.8 24.7 20.6 21.7 22.7 23.2 24.1 External Saving 4.1 1.0 -1.3 1.9 3.7 6.9 7.3 National Saving 18.7 23.8 21.9 19.8 19.0 16.3 16.9 2. Domestic Investment 22.8 24.7 20.6 21.7 22.7 23.2 24.1 Public Sector 7.5 9.7 6.4 5.1 4.8 4.3 4.3 Private Sector 13.3 13.3 13.5 14.3 14.7 16.8 16.6 Stock Accumulation 1.9 1.8 0.7 2.3 3.2 2.1 3.2 Source: INEGI and author's calculations l  .2 Paper I Einance investment.! Several observers have pointed to the low rate of saving Figure 1.1: National Saving and Investment Balance, 197094 and the consequent excessive (Percent of GNP) dependency on external financing as 35 fundamental factors behind the peso Domestic crisis of 1994-95. Investic 25 1.4 The purpose of this 20- paper is to examine the behavior of the 15 - National Mexican saving rate in the years after Saving 1987 and to evaluate several 10 Foregin Financing hypotheses that have been put forth to 5 aLccount for its decline. The issue extends beyond Mexico because of 0 concerns that the fall in the domestic .5 saving rate may have been induced by 1970 1975 1980 1985 1990 the refonn program or by the foreign Source: INEGI and author's calculations capital inflows associated with liberalization of the capital account. If either of these factors were important, the Mexican experience would introduce a note of caution for other countries considering the enactment of a similar program. 1.5 The paper is divided into two major parts. The first focuses on issues of measuring saving and extends the Mexican national accounts to include a division of national saving between the public and private sectors. A substantial portion of the debate over the causes of the saving decline can be traced to differences in the measures of saving that various studies or commentators have used. In particular, there are important differences of view about whether the fall in saving was concentrated in the public or private sector. The second part of the paper focuses on a set of specific hypotheses that might account for reduced saving and develops data to evaluate their relevance to the Mexican situation. 1.6 The paper adopts national saving, as opposed to domestic saving, as the most appropriate measure of saving effort, and concludes that the decline began about 1987 and that it totaled about 5.5 percent of GNP between 1985-87 and l992-94.2 1.7 The allocation of the decline between public and private saving is dependent upon the rneasure of the public budget balance-whether it is defined to include or exclude the financial intermediation activities of government. It is also important to adjust for the effects of inflation on interest payments between the public and private sectors in order to derive a meaningful division of the aggregate saving. The Mexican data also appear to yield strong evidence of Ricardian equivalence-an inverse relationship between the public and private saving rates. However, the inverse relationship can be traced in part to mismeasurement of the real income component of interest payments on the public debt. During periods of high inflation, the Government made large interest payments to the private sector; but a large portion of that payment represented a repayment of principal and not income. Thus, fluctuations in inflation generate large variations in capital transfers between the Government and the private sector. If I While the size of the capital inflow is large by intemational standards, it represents a return to the pattem of the 1 970s. 2 The precise magnitude of the fall is sensitive to the choice of a base period because the saving rate was rising over the decade of 1975 to 1985. The decline is considerably smaller relative to the 1970s, when the saving rate averaged 18.7 percent, compared to a peak of 26 percent in 1983.  The Decline in Mexican Saving: A Cost of Reform? 3 those transfers are not excluded from the income measures, they result in the appearance of a strong inverse relationship between public and private saving. On the basis of our preferred definition of public saving, which excludes financial transactions and is adjusted for inflation, the fall in national saving is largely a private-sector phenomenon. 1.8 While it is difficult to firmly associate the saving decline with any single aspect of the Mexican reform program, the data developed in this paper suggest that much of the fall can be attributed to a boom in asset markets that generated large capital gains and a liberalization of financial regulation that permitted a large expansion of credit to the private sector. It also appears that a strong inflow of foreign portfolio capital, which was very large relative to the domestic equity market, contributed to the surge in capital gains. It is more difficult to attribute the fall in saving to the stabilization program, and the overall current account deficit emerges more as a consequence rather than a cause of the saving decline. The stabilization program and the liberalization of the external capital account, however, undoubtedly played an indirect role through their effects on expectations for the future course of the Mexican economy, and on developments in domestic financial markets. The Mexican experience stands out primarily as a warning about the potential dangers of rapid liberalization of financial markets without supportive institutional strengthening. HI. MEASUREMENT OF SAVING 1.9 In the opinion of several international organizations that have recently reviewed the Mexican national accounts, the available data are of relatively high quality. However, Mexico's national accounts present a balance of saving and investment only at the level of the nation as a whole, and a division between the public and private sectors must rely on fiscal accounts that reflect different concepts.3 Also, there were some uncertainties about the rates of saving reported for the early 1990s, given the magnitude of the structural changes that have taken place over the past 15 years. Data used in this report are derived from the 1980-based national accounts. Subsequent to the preparation of this report, the Mexican national accounts were rebased from 1980 to 1993 for the period after 1987. It is reassuring to note that while the revisions for some of the nominal components were large (mainly due to an increased number of economic activities and to the inflation accumulated between 1980 and 1993), the changes in the reported rates of saving and investment were small, confirming that the decline in Mexican saving was real. National versus Domestic Saving 1.10 At the aggregative level, the most significant measurement issues arise from the choice of the income concept used to define saving and the treatment of international payments. The international system of national accounts (SNA) incorporates a significant distinction between Gross Domestic Product (GDP), the total amount of income obtained from production within the geographical confines of a country, and Gross National Income (GNI), the total amount of income earned by residents of the country. The latter differs from the former by the amount of net factor income and transfer receipts from the rest of the world. For some countries the distinction can be quite important. While saving of industrial countries is nearly always defined as gross national saving (GNS), GNI less consumption, the development literature has often used an alternative concept of gross domestic saving (GDS), GDP less consumption. The second concept is a more straight-forward calculation from standard expenditure accounts, but it has a limited behavioral interpretation. GDP is not a meaningful measure of the income available to support domestic consumption because it includes the income earned by foreign factors of production, such as multinational corporations. 3 This will likely change in the near future as INEGI is expanding the accounts to include more sector information.  4 Paper I Table 1.2: Relationship Between National Accounts Aggregates, 1970-94 (Percent of GDP) Component 1970-79 1980-84 1985-87 1988-89 1990-91 1992-94 Gross Domestic Product 100.0 100.0 100.0 100.0 100.0 100.0 (+) net factor income -0.2 -4.7 -4.8 -3.6 -2.5 -2.7 Gross National Product 99.8 95.3 95.2 96.4 97.5 97.3 (-) capital consumption allowances 8.5 10.1 12.7 11.3 9.6 9.7 National Income 91.3 85.2 82.5 85.1 87.9 87.6 (+) transfers 0.2 0.2 1.0 1.1 1.1 0.8 National Disposable Income 91.5 85.4 83.5 86.2 89.0 88.4 (-) private consumption 72.5 63.0 66.3 69.9 71.3 71.2 (-) public Consumption 8.9 9.9 9.0 8.5 8.7 10.8 Net National Saving 10.2 12.5 8.2 7.8 8.9 6.4 (Gross National Saving 18.6 22.6 20.9 19.1 18.5 16.1 (Gross Domestic Saving 18.7 27.1 24.7 21.6 20.0 18.0 Source: INEGI. 1.11 At the aggregate level, total saving is singularly defined as equal to total capital inivestment, but the distinction between GDS and GNS alters its distribution between national and foreign sources. Under the GDS concept, foreign saving is defined as simply the negative of the net balance on trade in goods and non-factor services. Using GNS, it is defined more broadly to include the trade balance, net factor income and transfers-the negative of the current account balance. For net debtor countries, such as Mexico, the GNS concept attributes a larger portion of total saving to the foreign sector. 1.12 These measurement issues are of some significance in the Mexican case because of substantial changes in the country's balance on its transactions with other countries. The relationships among the national accounts aggregates are summarized in Table 1.2 for the period 1970-94.4 In the early 1970s, Mexico had only a small amount of international debt, and the offset of labor earnings from residents employed in the U.S. resulted in near equivalence between GNP and GDP. By the mid-1980s, however, net factor income payments had increased to about five percent of GDP, with the result that the income and saving of residents of Mexico were significantly less than implied by a focus on GDP. After 1983, a combination of lower international interest rates and debt relief resulted in a partial recovery in the raLtio of GNP to GDP.5 1.13 The implications for the various measures of the saving rate are highlighted in Figure 1.2. The GDS rate shows a very large increase between 1975 and 1983 followed by a precipitous decline over the remainder of the 1980s. In contrast, the rise in the GNS rate is only half as large and the post-1983 decline is considerably smaller. In the 1990s, the two series have declined in parallel; but the situation will ctLange again in the aftermath of the 1994-95 crisis, as Mexico is forced to curtail its reliance on foreign finiancing. For both saving concepts, it is evident that much of the recent decline in the Mexican saving rate might also be viewed as a return to normality, reversing the rise of 1975-83. The net national saving rate, 4 The national statistical institute (INEGI) supplied a set of data for the 1970s recomputed on the base of 1980 to maintain historical comparability. The statistics for the 1970s are, however, only partially comparable with the 1980s Data for individual years are reported in appendix Table 1. The focus of this paper is on the concept of gross national saving, and for consistency saving and the related concepts are scaled by GNP rather than GDP. The choice of the scalar has very little effect on changes in the reported ratios, however.  The Decline in Mexican Saving: A Cost of Reform? 5 which excludes capital consumption allowances, indicates a slightly larger Figure 1.2: Alternative Measures of National Saving, 1970-94 allowances, indicates a shghtly larger(Percent of GDP) decline because of a gradual rise in the ratio of capital to GNP. 35 Gross dometc Public-Sector Budget Accounts 30 sving 1.14 As mentioned 25 previously, the Mexican national 20 accounts do not include a public saVing sector. The national accounts do, 15 however, provide data on public- sector investment, defined to include 10 Nef ratl general govermment and the public !.ving enterprises. Thus, it is possible to 5 create a rough measure of public- o sector saving by adding the national 70 1975 19eo 1985 1990 accounts' measure of physical investment to an estimate of the net Source: Table 1.2 budgetary position of the government. In countries that do estimate a public sector account within their national accounts, the two measures of the overall government deficit are quite comparable. The two accounting systems often yield quite different measures of government capital formation, however; and it is not advisable to mix budgetary data on investment with that of the national accounts. 1.15 The Mexican budgetary accounts distinguish among: (1) the federal government, (2) organizations and enterprises subject to direct budgetary control, and (3) entities that are not controlled directly but whose income and expenditures are included in the budget. In addition, there is a residual category of difference between the detailed income-expenditure accounts and the government's financial accounts.6 Most measures of the public-sector budget include all four components. The category of federal government includes many of the standard functions of general government, but some organizations, such as social security, are included in the second category together with government-owned enterprises such as Pemex. The Federal District government is the most prominent entity that is not subject to direct budgetary control.7 Furthermore, the public-sector accounts do not include the budgets of state and municipal governments, but because they have very limited independent borrowing authority and rely heavily on transfers from the central government, their exclusion has a relatively small effect on the net public-sector budget balance. The basic budgetary data are summarized in Table 1.3 as a percent of GNP for the period 1970-94.' 6 The discrepancy is largely the result of timing differences in the reporting of income and outlays. 7 We did not attempt to reclassify the budgets of the organizations to construct a measure of the general government balance. Many of the enterprises are so thoroughly integrated into the budget that the distinction is of limited meaning. An examination of the operating balance of the major enterprises suggests that the trend in the balance of general government would be similar to that of the total. 8 Data for individual years are given in appendix Table 1.2.  6 Paper I Table 1.3: Alternative Measures of Public Sector Balance, 1970-94 (Percent of GNP) Component 1970-79 1980-84 1985-87 1988-89 1990-94 Public Sector Balance -5.0 -10.6 -13.4 -8.2 -0.3 Budgetary entities -4.6 -9.5 -12.2 -7.5 -0.2 Federal government -2.8 -7.8 -12.2 -7.6 -0.4 Org. and enterprises -1.8 -1.7 0.0 0.2 0.2 Extra-budgetary entities -0.8 -1.1 -0.8 0.0 0.1 Difference with financial accounts 0.4 0.0 -0.4 -0.8 -0.2 Financial intennediation 1.0 1.1 1.3 1.2 2.0 Financial balance -5.9 -11.7 -14.7 -9.4 -2.2 Economic balance -4.9 -10.6 -13.3 -8.2 -0.2 Prinarybalance -2.1 -1.3 4.3 8.6 5.3 Inflation adjusted: Economic balance -3.2 -4.0 -0.5 -2.7 2.0 Financial balance -4.2 -5.9 -2.7 -3.2 -1.1 Memo items: Debt renegotiation 0.0 0.0 0.0 0.0 0.6 Privatization revenues 0.0 0.0 0.0 0.0 1.5 Source: Director General de Planeacion Hacendaria 1.16 Finally, most budget presentations emphasize the concept of economic balance, which excludes govenmment lending activities. The economic balance also excludes the budgetary effects of the 1990 debt renegotiation and privatization revenues. Thefinancial balance was traditionally defined as the economic balance plus financial intermediation-the lending activities of governnent financial institutions. Beginning in 1993, however, the Mexican Govemrnment changed the concept of the financial balance to exclude financial intermediation, essentially eliminating it as a separate budget concept. They argued that the subsidy element of the lending programs had been largely removed, and that it was inappropriate to include purely financial transactions in the basic budget. The decision to ignore the loan programs in budgetary accounts has been criticized by some analysts who point to a generally poor record of loan recovery and continued emphasis on directed credit-part of which goes to sub-national govenmments. Thus, this study continues to use both concepts of the public sector balance. 1.17 The two budgetary measures are shown as a percent of GNP in Figure 1.3. Prior to the ] 990s, they moved very much in parallel. The financial balance was typically about one percent of GNP less than the economic balance. This pattern was broken in 1993-94 when financial intermediation expanded to over three percent of GNP, coincident with its exclusion from the official budget. Both measures show a dramatic improvement in the budgetary situation after 1987, when the economic balance went from a deficit equal to 16 percent of GNP to a 1992 surplus of 1.6 percent. The change may be a rather misleading indication of fiscal policy, however, because it is mainly a result of a large decline in inflation, and thus nominal interest payments on the public debt. The primary budget balance-excluding interest payments-shows that the major fiscal change, in terns of taxes and program outlays, occurred in tlle early to mid-1 980s, and there were relatively small changes in the non-interest balance in subsequent years. 1.18 The dominant role of interest payments suggests the usefiilness of a focus on a third oDncept, the operational budget balance, which excludes the inflation component of interest payments on the public debt. That is, in an inflationary environment, a portion of interest payments on debt represents a repayment of loan principal, amortzation, rather than constituting income to the recipient. The operational  The Decline in Mexican Saving: A Cost ofReform? 7 Figure 1.3: Public Sector Budget Balance, Nominal and Inflation-Adjusted (Percent of GNP) Nominal Values Inflabon-Adjusted 5 Primary >>X<~\ 5 Economic balane /balance 10 ,,,,,Fl 0 Fia: -5 0- -10 Financial balance -15 -15 1970 1975 1980 1985 1990 1970 1975 1980 1985 1990 Source: Director General de Planeacion Hacendaria. balance is computed on a monthly basis as the economic balance plus the inflation rate time the outstanding stock of peso-denominated debt.9 As shown in the second panel of Figure 1.3, the inflation adjustment makes a considerable difference, both because of the sharp variations in inflation and significant changes in the proportion of the domestic debt that is denominated in pesos. There are actually two measures of the operational budget balance, corresponding to the economic balance and the financial balance. They differ because the inclusion of the loan programs results in a much different measure of the net public debt. The most frequently used concept corresponds to the economic balance. The operational budget balance is in surplus in 1987, when the nominal deficit is still 16 percent; and it implies a much smaller post-1 987 shift in fiscal policy than the nominal data. Private versus Public Saving 1.19 The alternative estimates of public saving obtained by adding public-sector investment to the above measures of the budget balance are shown in Tables 1.4-1.5 and Figures 1.4-1.5. Looking first at the nominal measures of Figure 1.4, the most strildng feature is the strong inverse correlation between the measures of public-sector saving and the implied residual estimate of private saving. Between 1987 and 1992 the public saving rate, based on economic balance, improves by 16 percent of GNP, but any gain Table 1.4: Alternative Measures of Public and Private Saving, Nominal Values (Percent of GNP) Component 1970-79 1980-84 1985-87 1988-89 1990-91 1992-93 1994 Economic Balance: National Saving 18.7 23.8 21.9 19.8 19.0 16.3 16.9 Public Sector 2.6 -0.9 -6.9 -3.1 3.1 5.5 4.1 Private Sector 16.0 24.7 28.8 22.9 15.9 10.8 12.7 Financial Balance: National Saving 18.7 23.8 21.9 19.8 19.0 16.3 16.9 Public Sector 1.6 -2.0 -8.2 -4.3 2.1 3.5 0.6 Private Sector 17.1 25.8 30.1 24.0 17.0 12.9 16.3 Source: Author's calculations It is important to note that the adjustnent is limited to peso-denominated debt because a significant portion of the domestic debt is denominated in dollars.  l8 Paper I tD national saving is more than offset by a 23 percentage point drop in the private Figure(.4: NominalValues of PublicandPrivateSaving saving rate. These nominal data appear to (Percent of GNP) be strongly supportive of Ricardian 40 equivalence. Part of the indicated inverse 35 Economic Balance relationship between public and private 30 saving, however, is simply a product of the 25 treatment of nominal interest payments. In 20 Pnvate saving thde inflation-adjusted data of Figure 1.5, the 15 inverse relationship is much weaker: the \ 1987-92 improvement in the public-sector 10 saving rate is largely eliminated, and the 5 Pubic sectorSaving estimated decline in the private saving rate 0 is, dramatically reduced to about 6 -5 percentage points. _10 .- . . . . v 1970 1975 1980 1985 1990 1.20 The foregoing is confirmed by the results of a set of simple regressions 3 nrl Vales of the private on the public saving rate: 3Rrw BElarre 25 Privatesvirg r-sp = a + b rsg. 15 The first equation is based on the nominal s Picseorsi data and the coefficient on public saving implies a private-sector offset in excess of -s unity. The coefficient is highly significant, arLd this simple relationship explains 85 -15 percent of the variation in the private saving 1970 1975 1980 19B5 1990 rate. In contrast, the same regression using Source: Table 1.4 the inflation-adjusted data results in a much smaller offset, -0.52, and it is only weakly significant. It should be noted that even with the inflation- adjusted saving measures, the coefficient estimates may be biased due to measurement error. Since private saving is obtained as a residual, any error in the measure of public saving automatically gives rise to a negative correlation between the two saving rates. 1.21 Finally, any Equation Coefficient on t-statistic Dependent variable - conclusion about the allocation of the decline in national public savngNominald - Nominal saving between the public and private sectors is affected by 2 -0.52 1.8 Real- Real the choice of whether to include or exclude financial intermediation from the public-sector budget. If the financial transactions are excluded, all of the decline in the national saving rate since the mid-1980s can be traced to a decline in private saving. As shown in Table 1.5, this narrow measure of public-sector saving averaged 6.6 percent of GNP in 1992-94 compared to 5.7 percent in 1980-84. Thus, calculated as a residual, the private saving rate fell from 18.1 to 10.9 percent of GNP. Alternatively, if the financial transactions are included, the public-sector saving is lower in all years, and its sharp fall in 1993-94 implies that a strong recovery of the private saving rate was underway prior to the 1994-95 financial crisis.  The Decline in Mexican Saving: A Cost ofReform? 9 Table 1.5: Alternative Measures of Public and Private Saving, Inflation-Adjusted (Percent of GNP) Component 1970-79 1980-84 1985-87 1988-89 1990-91 1992-93 1994 Economic Balance: National Saving 18.7 23.8 21.9 19.8 19.0 16.3 16.9 Public Sector 4.3 5.7 5.9 2.4 7.2 6.6 5.0 Private Sector 14.4 18.1 16.0 17.4 11.8 9.7 11.9 Financial Balance: National Saving 18.7 23.8 21.9 19.8 19.0 16.3 16.9 Public Sector 3.3 3.8 3.7 1.9 5.2 3.2 0.0 Private Sector 15.4 20.0 18.2 17.9 13.8 13.1 16.9 Source: Author's calculations 1.22 Standard national accounting procedures would call for the exclusion of financial transactions, after accounting for any subsidy element, on the basis that, in competitive markets, public lending simply substitutes for loans from private banks. However, if financial markets are distorted by significant non-price rationing, public lending-even if it contains no subsidy element-can have real effects. In that case, an argument can be made for treating financial lending as comparable to transfer payments and for using the financial balance to compute public-sector saving. In the specific case of Mexico, it is notable that the surge of government lending in 1993-94 was not marked by any large increase in investment, even though investment was the declared purpose of most of the loans. Thus, the public- sector loans appear to have substituted for private lending, or they were used to support consumption. 1.23 Regardless, there is no consensus among the economists who follow the Mexican fiscal situation on whether to include or exclude the financial intermediation. In a recent article, Gil-Diaz and Carstens of the Mexican central bank adopt the view that government lending is equivalent to transfer payments.10 They use the inflation-adjusted financial balance to measure government saving, and they focus on the 1989-93 period, ignoring the large decline of national saving in prior years. With these assumptions, they can attribute the decline in saving to government, not the private sector; and they argue Figure 1.5: Inflation-Adjusted Values of Public and Private Saving (Percent of GNP) 40 40- Irfabtonusted Irflabor.A4usted 35 Econonmc Balance 35 FRanaal Balanoe 30 3D 25 Private sa1ng 25 Prvate saving 20 20 15 15 10 Publtcoseaair 10 Publicsector savgrrg 5- 5\ o 0 .5 .5 -10 . -10 1970 1975 19P 1985 1990 1970 1975 1985 1985 1_90 Source: Table 1.5 10 Gil-Diaz and Carstens (1995).  10 Paper I that there is no private saving puzzle to be explained. While their position may be extreme, it illustrates the importance of the differing concepts of public saving in any explanation of private saving behavior. III. ACCOUNTING FOR THE SAVING DECLINE 1.24 The greatest difficulty with an effort to account for the fall in Mexican saving is the surplus of possible explanations. So many aspects of the Mexican economy, with a potential impact on saving, were changing within a very short time span that it is difficult to discriminate among them. The rmajor issues, however, revolve around three hypotheses that trace the saving decline to: (1) an exchange rate-based stabilization program, which, operating through wage-price distortions or a lack of confidence that it would succeed, encouraged consumers to substitute current for future consumption; (2) liberalization of the extemal sector (particularly the large inflows of foreign capital); and (3) liberalization of the domestic financial system. All three of these explanations might also be combined as contributors to what emerges as a very large surge of asset-market prices in Mexico. 1.25 The analysis can be simplified by focusing only on those measures of the private-public division of saving that incorporate an inflation adjustment for interest payments. On this basis, there is no significant change in the Mexican fiscal situation until 1993-94, and the decline in national saving is largely a question of what happened to private saving. 1.26 It is convenient to use 1987 as the transitional Item 1985-87 1992-94 Change year, both because it marks a National Saving (% of GNP) 21.9 16.5 -5.4 relatively high level of the Version I Economic Budget Balance national saving rate and because Public 5.9 6.1 0.2 the Mexican stabilization Private 16.0 10.4 -5.6 pr-ogram was introduced in Version TI: Financial Budget Balance December of 1987. Given the Public 3.7 2.2 -1.6 potential for measurement error, Private 18.2 14.4 -3.8 however, it is best to average the data over a few years. On this basis, there was a 5.4 percentage point drop in the national saving rate between 1985-87 and 1992-94, heavily concentrated in the private sector. If the public-sector balance is defined to exclude financial transactions, all of the decline in the saving rate is in the private sector. Altematively, if government financial transactions are included, about 1.6 percentage points of the drop is within the public sector. However, the issue of how to treat government financial transactions is of major sigpificance only in 1993-94, well after much of the fall in the national saving rate. Exchange Rate-Based Stabilization 1.27 Exchange rate-based stabilization programs, similar to that of Mexico, have frequently been accompanied by a surge of private consumption.'" Several hypotheses have been advanced to account for an outcome that seems so counter to standard expectations that the reduction of inflation would require a period of austerity. In the early 1980s, Dornbusch and others argued that it was part of a more general phenomenon in which domestic demand rose in response to a sharp drop in real interest rates, appreciation of the real exchange rate, and higher real wages, all of which could be traced to sticky nominal wages and 11 For extensive documentation see Rebelo and V6gh (1995).  The Dedine in Mexican Saving: A Cost ofReform? 11 prices.'2 With capital mobility (interest-rate parity), a fixing of the exchange rate would lead to an immediate fall in the nominal interest rate, but the delayed response of prices would slow the adjustment on the real side. A related hypothesis by Calvo suggested that private agents expected the decline in inflation to be temporary.'3 Again, the lower nominal rate of interest would translate into an even larger decline in the expected real interest rate, the cost of current versus future consumption. 1.28 If the expansion of demand is concentrated in consumption, the above explanations are dependent upon a relatively high elasticity of substitution between current and future consumption, sonething that has been difficult to support empirically. It is troubling, moreover, to use an argument of '"emporariness" to sustain a decline in saving that extended over a period as long as that of Mexico. As shown in Table 1.6, nominal interest rates did fall promptly after the introduction of the stabilization program at the end of 1987, but the rate of inflation fell even faster. Thus, ex-post real rates of interest were higher in the years after 1987 than before. The highly repressed nature of Mexican financial markets prior to 1987 allowed the government to maintain a highly negative real return to savers during the prior years of high inflation, and the stabilization program would appear to have raised returns. If the problem was expectations, and if investors expected inflation to break out again, they should have been willing to pay a premium for indexed bonds (adjustabonos), which were introduced in 1989. Yet, the market rate on those bonds, column 3, appears to have been equal to or higher than the ex-post real rate on regular issues.'4 Furthermore, real wages actually fell in the first year of the program and rose at a relatively modest 2.5 percent annual rate over the 1987-94 period.15 The real wage did not regain its 1987 level until 1991. While the gains were modest, they still stand in sharp contrast to the 30 percent drop in the real wage between 1980 and 1987. Finally, as shown in a later section, there was no big post-1987 surge in the consumption of durables, as might be expected from an expectations-based explanation for the saving decline. 1.29 The behavior of the real exchange rate, which appreciated in the years after 1987 (columns 7 and 8), is more consistent with the basic argument that the stabilization plan provided an initial stimulus to demand. An appreciation provides a terms of trade or real income gain and promotes spending, followed much later by a loss of competitiveness and recession. It is argued, however, that the Mexican exchange rate began from a severely depreciated level in 1987, and that it never became overvalued relative to the longer-term average.16 Using national accounts measures of import and export prices, Mexico's terms of trade fell 40 percent between 1980 and 1987, recovered 13 percent by 1990, and remained constant in the first half of the 1990s. 1.30 The effects of the stabilization program on saving incentives were further complicated by significant changes in the Mexican tax system, which should have increased saving incentives. The current Mexican tax system is quite advanced in incorporating minimal penalties on saving. The fundamental changes were made in the early 1980s when Mexico began to move in the direction of an indexed tax system, but additional actions were initiated in 1987 to broaden the tax base and fiuther reduce marginal 12 Dombusch (1982). 13 Calvo (1986). 4 In additicon, the price of consumption relative to investment goods rose about 15 percent between 1985-87 and 1992-94. 15 This estimate of wage change is based on national accounts' measures of compensation per employee at the level of the national economy. Also, a more favorable picture is provided by the change in eamings within manufacturing, where growth averaged 5 percent annually. 16 Gil-Diaz and Carstens (1996).  12 Paper I Table 1.6: Interest and Exchange Rates, 1980-94 (Percent, annual average) Nominal Interest Real Interest Real Exchange Rate Deposit Cetes Indexed Deposit Cetes CPI Unit Labor Morgan- Year Rate 3-month 3-year Rate 3-month Change Costs Guaranty 1980 19.34 22.58 -8.04 -5.55 29.8 100.0 100.0 1981 25.58 30.85 -2.41 1.68 28.7 125.1 111.2 1982 38.53 45.73 -30.34 -26.72 98.9 84.4 80.9 1983 57.39 59.45 -12.93 -11.79 80.8 51.2 75.3 1984 48.84 49.65 -6.49 -5.98 59.2 52.3 87.9 1985 55.24 63.69 -5.20 -0.04 63.8 52.9 86.5 1986 75.91 88 71 -14.49 -8.27 105.7 34.5 61.7 1987 92.44 102.83 -25.75 -21.74 159.2 32.0 63.5 1988 52.70 63.98 0.69 8.13 51.7 38.9 76.8 1989 30.85 44.77 16.33 9.32 20.95 19.7 45.0 73.6 1990 27.88 35.03 12.57 -1.58 3.93 29.9 46.8 68.4 1991 16.57 19.82 5.81 -1.87 0.86 18.8 50.7 72.6 1992 14.48 15.89 3.20 2.27 3.53 11.9 55.0 73.7 1993 15.06 15.50 5.23 6.53 6.94 8.0 55.1 79.7 1994 13.32 14.68 5.92 5.86 7.13 7.1 50.7 76.7 Note: Interest rates are measured net of tax. The CPI change is expressed on a Dec.-Dec. basis. Both exchange rates are converted to 1980=100, and an increase in the index represents appreciation. The Morgan-Guaranty index is trade-weighted using relative wholesale prices. Source: Bank of Mexico tax rates. The top-bracket rate of the personal income tax was lowered from 55 to 35 percent in 1987. The corporate tax system is fully indexed, the top rate is 34 percent, and dividends are taxed only at their source. The government has also shifted toward greater reliance on the value-added tax. 1.31 In sumnary, while the outcome of the Mexican stabilization program is consistent with the hypothesized effect of a sharp drop in private saving, it is difficult to support the argument that it operated through a decline in the return to saving. The underlying details in terms of the relative price of current versus future consumption seem to have moved in the opposite direction. Furthermore, while there was some appreciation of the real exchange rate, it was most pronounced well after the initial decline in saving. As discussed later, it is possible that the stabilization program contributed to the saving decline through other more indirect mechanisms, such as playing a contributing role in the surge of financial asset prices. Excternal Sector Liberalization 1.32 Mexico does stand out in the magnitude and speed with which it liberalized its international economic relations in the 1980s. It began the decade as a highly protected economy, with a very extensive licensing system on imports and tariff rates that averaged near 25 percent. It also exerted very tight controls over inflows of financial capital. By 1990, it had established full capital account convertibility, most import licenses had been eliminated and the average tariff rate had been cut in half. It also experienced a tremendous increase in net capital inflows. Hypotheses have been advanced that relate the saving decline to these external account changes. 1.33 Capital Inflows. As pointed out in a recent review article by Obstfeld, there has long been a controversy in the development literature about the extent to which foreign resource inflows would lead to  The Decline in Mecican Saving: A Cost of Reform? 13 additional capital formation and growth versus a simple augmentation of current consumption."7 Foreign resource inflows could depress national saving by reducing domestic interest rates, relaxing credit constraints, or, most importantly, by moderating the pressures on public officials to make painful fiscal adjustments. On the other hand, the inflow could be an induced response to perceived improvements in the profitability of domestic investment. 1.34 Between 1985-87 and 1992-94, the Mexican current account balance moved from a surplus of 1.3 percent of GNP to a deficit of 7 percent (see Figure 1.1). When this is matched against the decline in the Mexican national saving rate, it would seem to provide compelling evidence for those who argue that foreign resource transfers are more likely to promote consumption than to add to domestic capital formation and growth. Over the period of 1970 to 1994 there is a strong negative relationship between the national saving rate and the current account deficit. A simple regression of the national saving rate on the current account balance yields a correlation coefficient of 0.7 and a regression coefficient that implies that 60 percent of any change in the net foreign inflow is absorbed by changes in consumption. As with the prior hypothesis, however, questions arise with the specification of the mechanism by which foreign resource inflows cause a decline in national saving. 1.35 Obstfeld argues that most of the existing empirical studies of the issue are severely flawed by the fact that saving, investment and the foreign inflow are all endogenous variables from which it is very difficult to infer causation. This problem is manifest in the Mexican case. Did the capital inflow play a role similar to a simple unrequited transfer in stimulating consumption, or was it a response to a fall-off in domestic saving in the presence of what foreign investors perceived to be good investment opportunities? 1.36 First, some doubts about the causal role of the capital inflows arise from noting that the inverse correlation between foreign and national saving is largely a product of the 1980s and 1990s. Up to the onset of the 1981 debt crisis Mexico had consistent current account deficits in the range of 4 percent of GNP and a rising national saving rate. After 1981, Mexico was excluded from international capital markets, and it had no choice but to generate a substantial trade surplus to meet its debt payments. The puzzle is why the experience with a net resource inflow in the 1990s differed from that of the 1970s. 1.37 Furthermore, the correlation is far less clear-cut if saving is separated into its private and public components. Prior to 1987, the foreign-sector balance was significantly correlated with only public- sector saving, and there was no obvious relationship with private sector behavior. In the post-1987 period, exactly the opposite situation emerged, when the public saving rate remained largely unchanged and private saving fell. If the foreign resource inflow was operating as an exogenously generated windfall, it should depress domestic market rates of interest or weaken public fiscal constraints. Neither of those events is evident in the Mexican case."8 As with the stabilization policy explanation, there are problems with the transmission linkage from a foreign resource inflow to a fall in saving. 1.38 What is different about the post-1987 episode is the extent to which the current account deficit was financed with an inflow of private portfolio capital, as opposed to reliance on official borrowing 17 Obstfeld (1995). Also, as he points out, the welfare benefits of foreign resource inflows have frequently been exaggerated by a focus on GDP instead of GNP, ignoring the fact that the owners of the foreign capital must be paid an amount roughly equivalent to their contribution to output. The domestic consequences are largely a redistribution of income from existing capital to labor. As discussed in the next section, the capital inflows did occur during a period of major financial reform and liberalization. Thus, the capital inflows may have eased credit constraints without being reflected in a decline of interest rates.  14 Paper I Table 1.7: The Mexican Current Account Deficit and Its Financing (Percent of GNP) Financing Current Reserve Direct Portfolio Other Account Asset Investment Capital Net Year Inflow Sales mnflow Iflow Inflow 1979 3.9 -0.2 1.0 -0.2 3.5 1980 5.5 -0.4 1.1 0.0 4.8 1981 6.6 -0.5 1.2 0.4 5.5 1982 3.1 6.2 1.0 0.3 -4.4 1983 -5.4 -0.6 2.0 -0.5 -6.3 1984 -2.9 -1.5 1.1 -0.3 -2.2 1985 -0.5 1.9 1.4 -0.4 -3.4 1986 1.1 0.4 1.7 -0.4 -0.5 1987 -3.1 -3.0 0.9 -0.7 -0.2 1988 1.4 6.1 1.2 0.6 -6.5 1989 3.0 0.1 1.4 0.2 1.3 1990 3.2 -0.9 1.1 1.4 1.6 1991 5.3 -2.8 1.7 4.5 1.9 1992 7.6 -0.5 1.4 5.6 1.2 1993 6.6 -2.1 1.2 8.2 -0.8 1994 7.9 4.8 2.2 2.2 -1.4 Source: International Financial Statistics. Other category comprises lending (including official) and errors and omissions. and bank lending, as in the earlier period (Table 1.7 and Figure 1.6). And, despite all the public attention devoted to direct investment inflows, they became only slightly more important, averaging 1.5 percent of G3NP in 1988-94 compared to 0.8 percent in the 1970s. In contrast, portfolio capital inflows surged from 0.6 percent of GNP in 1988 to a peak of 8.2 percent in 1993, before reversing course in 1994. Given the -epressed nature of Mexican financial markets, and controls on foreign inflows, inflows of portfolio capital were trivial prior to 1990. It is possible that these inflows impacted on private markets and decisions in ways that were not possible when the funds were channeled through public institutions and direct investment in the 1970s. A large portion the portfolio capital represented investment funds from the U.S. that went into a relatively thin Mexican equities market. In an international context, Mexico stands out among the developing countries in its reliance on private portfolio capital inflows.19 This issue is taken up atgain in a later section on financial market developments. 1.39 Trade Liberalization. Another suggestion about the possible influence of foreign sector reforms on private saving focuses on the reduction in trade barriers. Prior to the mid-1980s, Mexico imposed severe restrictions on imports, and the barriers were particularly high for consumer goods. Thus, the surge of consumption might be a response to an expansion of options for purchasing lower-ost, higher- quality imported goods. And, as shown in column (1) of Table 1.8, the combination of liberalization and exchange rate changes resulted in a dramatic reduction in the relative prices of imported consumption goods. The price index for imported consumption goods measured as a ratio to the price index for total consumption fell by one-half between 1987 and 1994. This is a far more dramatic change than suggested by the earlier examination of the real exchange rate. It clearly illustrates the importance of factoring the effects of trade liberalization into any consideration of the appropriateness of the exchange rate after 1987. 19 Brazil and Argentina also registered large inflows during the first half of the 1990s. Chile actively discouraged foreign inflows with reserve requirements and taxes. Within Asia, portfolio capital inflows have been substantial only for South Korea and then largely in the form of closed-end funds.  The Decline in Mexican Saving: A Cost ofReform? 15 1.40 Furthermore, there has been a Figure 1.6: The Current Account and Portfolio major increase in the proportion of Capital Inflows consumption represented by imported goods. (Percent of GNP) The import share, measured in constant prices, 10 rose from 1.2 percent of total consumption in Current 1987 to 6.8 percent in 1994. The greater role Account of imports, however, should have been largely 5 - reflective of a process of substituting foreign Po for domestically-produced goods in response to o Capital the change in relative prices. It is not clear why it should have translated into a much higher rate of overall consumption. -5 - 1.41 Some economists would cLassify the accumulation of consumer durables -1195 1990 as saving. Thus, if the increased expenditures 1980 Year on imports were largely durable goods, they might be included within a broad definition of Source: Table 1.7. saving. While the consumption and imports of durables increased relative to GDP after 1987, most of the rise in total consumption and the imported component relative to GDP was in the categories of nondurables and services (Table 1.8). There is no clear evidence of a rush to buy durable goods as a hedge against a resurgence of inflation.20 Table 1.8: Private Consumption and Its Components, 1980-94 (Percent of GNP) Consumption Outlays Relative Nominal Prices 1980 Prices Year Price of Total Durables Consumer hnports Total Durables Consumer hnports Imports Total Durables Total Durables 1980 100.0 67 0 8.8 2 S 0.8 67.0 8.8 2.5 0.8 1981 94.5 66.7 8.7 2.4 0.7 66.6 9.0 2.5 0.8 1982 109.8 65.3 7.4 1.5 0.4 66.8 7.7 1.4 0.3 1983 112.0 64.7 6.5 0.7 0.1 66.2 6.3 0.6 0.1 1984 110.6 66.7 6.7 1.0 0.1 65.7 6.4 0.9 0.1 1985 114.9 67.5 7.3 1.2 0.2 65.6 7 0 1 0 0.1 1986 142.5 72.3 7.5 1.4 0.3 66.8 6.5 0.9 0.1 1987 149.4 69.0 7.6 1.3 0.3 65.0 5.8 0.8 0.1 1988 123.8 72.1 8.0 1.9 0.6 64.8 6.2 1.3 0.4 1989 107.2 72.8 7.2 2.5 0.6 66.7 6.3 2.2 0.4 1990 98.9 73.0 7.0 3.1 0.8 67.4 6.7 2.9 0.5 1991 88.0 73.3 7.0 3.2 0.8 67.7 6.9 3 4 0.6 1992 80.9 74.1 7.1 3.6 1.0 68.7 7.2 4 2 0.8 1993 75.8 73.4 6.6 3.4 0.9 68.5 6 8 4 2 0.7 1994 75.4 72.0 6.2 3.7 1.0 68.7 6.7 4 7 0.9 Note: The pnce of nmported consumm ton goods is measured relative to the price deflator for total private cosumption on a base of 1980. Source: INEGI National Accounts 20 The relative price of imported durables, like that of all imported consumption items (Table 8), rose substantially between 1981 and 1987, and fell thereafter.  16 Paper ) 1.42 Finally, the appreciation of the peso and the elimination of trade restrictions may have increased competitive pressures on domestic Mexican enterprises, lowering the rate of profit and business saving. Thus, the decline in the overall private saving rate could have been concentrated in the business sector. The only information from the national accounts that bears on the issue is the reported rise in the share of total value added going to capital between 1987 and the early 1990s, from 50.4 percent in 1987 to 55.7 in 1990, before declining back to about 52 percent in 1993. That would seem inconsistent with the argument that business profits were greatly squeezed by the liberalization program. 1.43 Additional infonnation is available from the Mexican stock exchange for those companies that are listed on the exchange. Since 1989 data are available on the profits and sales of the listed companies. While there are significant variations in the number of reporting companies from year to year, the computed rate of profit on sales varies within a narrow range of 12.8 percent in 1989 to 14.2 percent in 1994, with no evidence of a downward trend Given that Mexican companies pay out a very small share of profits in dividends, the change in profits should be a reasonably good proxy for the change in retained earnings. Thus, the available evidence suggests that retained earnings could not have played a major role in the fall of the private saving rate, implying that it was concentrated in the non-corporate and household sectors. Financial Sector Liberalization 1.44 Since the work of McKinnon and Shaw, liberalization of the financial sector has been viewed as a way to raise the level of saving and the efficiency with which it is allocated.21 More recently, however, the experience with financial reforms suggests that they can lead to a short-term decline in saving because of the easing of prior quantity restrictions on credit.22 The elimination of credit restrictions allows individuals to increase their borrowing against existing assets, with the potential for some of the funds to spill over into consumption.23 While much of the recent literature focuses on saving and constraints on consumer borrowing, the effects of financial reforms appear broader in that increased access to credit also initiates a boom in housing and other asset markets, generating large capital gains as an additional stimulus to consumption. Recent examples in industrial economies are provided by the experiences of Japan and Sweden. 1.45 Much of this story also appears to apply to Mexico. It was certainly an example of a repressed financial system prior to 1987, with extensive credit controls and negative real interest rates. The structural reform program resulted in the elimination of interest rate controls, restrictions on private-sector lending, and reserve requirements. Reserve requirements were replaced by a focus on liquidity ratios where the required assets earned market interest rates. Most of the changes took place in late 1988 and early 1989.24 In addition, commercial banks were returned to private ownership in 1991-92, and granted permission to issue marketable securities. All this occurred in conjunction with a significant relaxation of 21 McKinnon (1973) and Shaw (1973). 22 McKinnon (1991) provides a useful discussion of some of the things that can go wrong One of the more dramatic recent examples was that of Sweden in the late 1980s when financial liberalization led to a sharp increase in housing credit, an asset market boom, and a negative household saving rate. 23 For a discussion and references see Deaton (1989). 24 Loser and Kalter (1992).  The Decline in Mexican Saving: A Cost of Reform? 17 the regulations on foreign financial transactions.25 Thus, the simultaneous liberalization of the external capital account and the internal financial system greatly increased the availability of credit to the private sector. 1.46 A summary of the changes in the balance sheet of the banking system is given in Table 1.9. There has been moderate growth in the ratio of total credit to GDP, as would be expected to accompany a decline in inflation and more competitive rates of return.26 The more dramatic change, however, is in the share of credit allocated to the private sector. Prior to 1987, the banking system was largely a conduit for channeling funds from the private to the public sector. With the shift to much smaller public budget deficits and greater reliance on direct-market bond issues, public sector lending has shrunk in importance and the banking system has emerged as a major source of financing for private firms and individuals. 1.47 The outstanding stock of private-sector loans, adjusted for inflation, rose at an annual rate of 30 percent between 1987 and 1994. That was far in excess of the growth in the domestic capital stock or private-sector bank deposits. Private-sector loans soared from 8.7 percent of GDP in 1987 to 17 percent in 1990, and 41 percent by the end of 1994. By 1992, the banking system was a net supplier of credit to the private sector (Figure 1.7). The situation changed again in 1994 when the banking system increased its dependence on foreign financing in support of a further expansion of its private sector lending. In addition, there is an obvious reversal of the pre-1988 pattern of financial dis-intermediation: the ratio of private- sector deposits to GNP rose from 16.8 percent in 1987 to 29.6 percent in 1994 (Table 1.9). Table 1.9: Net Credit and Liabilities of the Consolidated Banking System, 1981-94 (Percent of GDP) Net Liabilities Credit Extended Private Nonbank Private Sector Public Year Total Sector Finance Foreign Total Total Consumers Mortgage Sector Other 1981 37.5 26.7 2 1 8.7 37.5 15.9 0.8 0.9 21.7 0.0 1982 45.7 23.0 2.0 20.8 45.7 11.0 0.6 0.6 33.9 0.8 1983 40.1 21.1 2.4 16.6 40.1 9.4 0.4 0.7 29.3 1.4 1984 37.0 22.1 2 3 12.6 37.0 10.7 0.7 1.0 24.1 2.2 1985 38.4 19.3 2.8 16.3 38.4 10.1 0.8 1 0 27.2 1.1 1986 43.5 18.9 3.3 21.4 43.5 8.8 0.6 1.0 33.1 1.7 1987 35.7 16.8 3.1 15.8 35.7 8.5 0.6 0.8 26.4 0.8 1988 32.5 15.7 3.1 13.6 32.5 10.1 0.9 1.2 22 1 0.3 1989 34.3 18.2 3.3 12.8 34.3 14.5 1.3 1.3 20 8 -1.0 1990 31.3 19.4 3.0 9.0 31.3 16.9 2.0 1.4 16.1 -1.6 1991 32.9 23 2 2.7 7.1 32.9 21.4 2.2 1.4 12.8 -1.3 1992 34.2 25.3 2.6 6.3 34.2 27.6 3.1 4.1 8.2 -1.5 1993 36.7 27.1 3.6 6.0 36.7 31.9 3.0 5.6 4.6 0.3 1994 52.2 29.6 5.0 17.7 52.2 40.8 2.9 10.9 8.6 2 7 Note: Dataaredeflatedbytheax wnppiceindlexforDecmnberonabaseofl980=1 andexpressedas apecentofGDP in 1980prices Source: Bank of Mexico 25 The 1989 regulatory changes substantially reduced limits on foreign ownership and broadened the range of activities in which foreign investors could participate. 26 The sharp surge of total credit in 1994 was concentrated in the last half of the year and appears to be reflective of the growing problems of financing an outflow of private portfolio capital.  18 Paper 1 1.48 The credit Figure 1.7: Bank Liabilities and Credit Extended to the Private expansion is evident in a very Sector, 1980-94 sharp growth in consumer credit (Percent of GDP) after 1987; but the bulk of the loan expansion involved businesses and mortgages. Between 1987 and 1993, the 40 - stock of consumer credit rose from about one percent to three 30 the percent of GDP. Since the flow prvie edor of consumption spending should 2c be related to the change in the stock of debt, the growth in C e 10 Credk extended to consumer credit alone is not pdvate sector sufficient to account for a large portion of the drop in saving. The 0 more important effect would seem 1980 1985 1990 1995 to be in increasing the liquidity of Source: Table 1.9. private sector wealth through expanding the access of private business to credit. At least for corporations, there is no evidence that the loan funds were paid out as dividends. Thus, if the expansion of credit financed a consumption boom, it did it through non-corporate business or by raising asset market prices that generated capital gains for investors. The expansion in mortgage lending is very large in 1992-94, but it appears to occur after the bulk of the decline in private saving.27 Still, it added to the increased liquidity of private-sector assets. An Asset-Market Bubble ]L.49 One common effect of financiaL market liberalization is that it promotes large increases in asset prices. Investors are suddenly able to adopt far more leveraged portfolios as they use the expanded access to bank debt to purchase real estate or equities. Given the thinness of the newly developing asset markets and the inexperience of financial institution managers and regulators, this type of surge in asset prices can quickly become excessive. It is likely that the large inflows of foreign capital, made possible by the removal of capital controls, and a general atmosphere of excessive optimism also were important contributors. 1 .50 In the case of Mexico, the stock market averaged real gains of 40 percent annually after 1987, until the market turned down in 1994 (Table 1.10). The capitalized value of firms listed on the Mexican stock exchange rose from 6 percent of GDP in 1987 to 42 percent in 1993.28 Real estate prices collapsed in the early 1980s: an index of land prices compiled by the Bank of Mexico shows a 1987 real price of land at a fourth of its 1980 value. But, since 1987, the price collapse has been fully reversed, with the real price of land rising to above that of 1980. This surge in private-sector wealth stands out as the rmost dramatic change in the basic determinants of private saving. 27 The data on mortgage lending are drawn from the category "vivienda de interes social" in the Indicadores Economicos of the Banco de Mexico. 28 The capital value of listed finms rises faster than the index of stock prices because new firms have been added to the exchange listing. The number of new firms rose both as a result of privatization of public enterprises and because privately-held enterprises found it attractive to sell equities.  The Decline in Mexican Saving: A Cost of Reform? 19 Table 1.10: Price Indices for Mexican Equities and Land, 1980-94 Equity Market Capitalized Stock Land Value Price Prices Year (Percent of GDP) Index(1980=100) Index(1980=100) 1980 2.2 100.0 100.0 1981 1.0 51.6 136.3 1982 0.3 18.6 116.5 1983 0.9 37.0 88.0 1984 0.9 38.4 64.8 1985 1.8 64.9 48.0 1986 4.3 132.5 39.9 1987 5.8 115.3 27.1 1988 6.6 152.0 57.2 1989 10.9 251.0 83.5 1990 13.6 290.1 95.1 1991 28.6 555.9 99.4 1992 34.5 610.0 140.5 1993 45.7 835.7 130.3 1994 42.4 712.6 131.6 Note: Indices are measured on an end of year basis and adjusted for inflation using the consumer pice index for December of each year. Source: Mexican Stock Exchange and Bank of Mexico. 1.51 A rough estimate of private-sector wealth and the influence of asset price changes is shown in Table 1.11. Wealth is divided into three major components of net financial assets, residential housing, and non-residential capital. The residential and nonresidential capital stocks are constructed from the expenditure data of the national accounts. Initial levels of the capital stocks are based on a study by Goldsmith, and they are assumed to depreciate geometrically.29 In addition, the residential and non- residential capital stocks incorporate an allowance for land equal to 25 percent of the reproducible capital stock in 1980, adjusted in other years on the basis of the land price index of Table 1.10. Private-sector financial assets consist of the net asset position with the banking system plus nonfinancial holdings of public debt instruments.30 Missing from the wealth measure is an estimate of the net asset position of the private nonfinancial sector with the rest-of-world. 1.52 The measure of wealth shown in column 1 reports the wealth to GDP ratio on the basis of the cost of replacing reproducible capital. The second concept, market value, values the non-residential capital stock on the basis of the equity-market prices reported in Table 1.10.31 Normally, equity market prices are used only to value the stock of corporate capital less debt. However, since data are not available for the corporate sector in Mexico, the valuation adjustment is done at the level of private business.32 In 29 Raymond Goldsmith (1985). The rates of depreciation are 0.04 for residential and 0.07 for nonresidential capital. 30 The net financial position with the banking system is included as a rough means of incorporating a portion of the private sector's transactions with the govermment and foreign sectors. The net position is positive in the early period because private-sector deposits were loaned to the government. It turns negative in later years because the banks raise funds abroad to lend to the private nonfinancial sector. 31 Residential capital is valued at replacement cost with the assumption that any real capital gain is reflected in the land price index. 32 It is assumed that the equity market price index is an accurate reflection of the change in the value of all private enterprises.  20 Paper I Table 1.11: Private Sector Wealth and its Components, 1980-94 (End of year, ratio to GDP) Total Wealth Net Residential Nonresidential Capital Replacement Market Financial Housing Replacement Market Year Value Value Assets Value Value 1980 1.65 1.48 0.11 0.54 1.01 0.84 1981 1.74 1.16 0.12 0.56 1.06 0.48 1982 1.78 0.94 0.14 0.57 1.07 0.23 1983 1.80 1.06 0.13 0.59 1.08 0.34 1984 1 72 1.06 0.13 0.57 1.02 0.35 1985 1 67 1.19 0.11 0.57 0.99 0.50 1986 1 75 1.68 0.13 0.61 1.01 0.94 1987 1 72 1.55 0.13 0 61 0.98 0.81 1988 1.83 1.83 0.12 0.66 1.04 1.05 1989 1.89 2.48 0 12 0.69 1.07 1.66 1990 1.90 2.66 0.12 0.70 1.08 1.85 1991 1.89 4.11 0.09 0.71 1.10 3.32 1992 1.98 4.37 0 04 0.75 1 19 3.59 1993 2.03 5.63 0.04 0.76 1.23 4.83 1994 2.00 4.87 -0.02 0.77 1.25 4.12 Nlote: End of year financial stocks are deflated by the CPI of December and expressed as a ratio to GDP in 1980 prices. Nonresidential capital is measured net of credit from the financial sector. Source: Author's calculations as explained in text. constructing the estimate of market value, the q-ratio-the ratio of market value to replacement value-is assumed to be unity for all of the private sector in 1980. The version shown in Table 1.11 assumed that all private debt is business debt and multiplied the stock price index by the capital stock minus debt in the base year of 1980. An alternative assumption that all of the private debt is consumer debt implies a 1980 benchmark value of private equity equal to the capital stock. That yields a higher wealth to GDP ratio of 5.7 in 1994 (Figure 1.8). ]1.53 It is evident from the above calculations that the changes in Figure 1.8: Private Sector Wealth, Alternative Valuations asset market prices generated enormous (Ratio to GDP) capital gains to some investors in 6 6 Mexico. Given their magnitude, it is plausible to argue that the capital gains 5 Market Value 5 were an important factor behind the fall 4 4 in private sector saving. Using market values, private wealth rose from an 3 3 aLverage of 1.5 time GDP in 1985-87 to 2 4.9 in 1992-94. Even if only two a t percent of wealth is assumed to be I Value consumed m each year, the above rnagnitude of real capital gains would be 0 75 1980 1985 1990 1995 sufficient to lower the private saving rate by six percentage points over the Source: Author's calculations last decade. The timing of the change in aLsset prices also is close to that of the clecline in the private saving rate.  The Decline in Mexican Saving: A Cost of Reform? 21 1.54 The major objection to the emphasis on changes in asset prices as the source of the saving decline comes from the studies that have failed to find a link between equity market prices and saving in the U.S. The most recent study concluded that changes in equity market values had little or no effect on consumption.33 Furthermore, Mexico is noted for a high concentration of wealth, and it can be argued that changes in wealth among a relatively small proportion of the population and foreign investors could not have accounted for such a large surge in consumption. Other countries, such as Japan and the newly industrializing countries of Asia, have had large increases in equity m'arkets without any evidence of large effects on consumption and saving. However, the magnitude of the capital gain in Mexico was far in excess of anything experienced in the U.S.; and its interaction with a new ability to borrow from the banking system may have created a consumption opportunity that did not previously exist. True, this wealth effect benefited a relatively small proportion of the population, but one accounting for a large proportion of total consumption. Regression Analysis 1.55 The effort to use regression analysis to distinguish among the competing hypotheses is restricted by the lack of a large data sample, measurement problems in computing both the private saving rate and the various determinants, and a high degree of collinearity among the competing explanations. While the data on the private saving rate extend over a twenty-five year period, variation in the series is dominated by the post-1987 decline. It was possible to construct data series for several potential determinants of the change in the saving rate. But three of the major candidates, bank lending to the private sector, foreign private capital inflows, and private wealth are all very highly correlated with one another: all have their largest changes in the same post-1987 period. 1.56 Some basic results of correlating the change in the inflation-adjusted private saving rate with some of the measures discussed inthe preceding sections are reported in Table 1.12 (and Figure 1.9).34 Initially, the private saving rate was related to the level and rate of change in income per capita and the real rate of interest on deposit accounts. There was an absence of a significant positive correlation between changes in the real rate of interest and private saving. This result follows from the negative real rates of interest that prevailed during the late 1970s and early 1980s when private saving was high and the generally high real rates in the post-1987 period of a falling saving rate. Furthermore, unlike the results for many other countries, there was no consistent positive association between the saving rate and annual changes in income per capita. 1.57 Columns (1) - (3) of the table report the results of three alternative formulations that relate the private saving rate to private wealth, foreign portfolio capital inflows, and the stock of private-sector loans, all scaled by the level of GNP. Each of these formulations fits the saving data about equally well because they share a common break in trend after 1987, but efforts to combine them result in significant problems of multi-collinearity: the simple correlation coefficients among the three measures all exceed 0.75. Equation (1) implies that wealth holders consume about 2 percent of their wealth in each period. Equation (2) implies a one-to-one trade-off between portfolio capital inflows and private saving. hi the case of equation (3) the significant correlation is between the saving rate and the stock of private sector loans, 33 Poterba and Samwick (1995). 34 The private saving rate is derived from the economic balance concept of the budget. Similar regressions were estimated using the saving rate based on the financial balance, but they fit somewhat less well and did not imply significant differences.  22 Paper I Table 1.12: Regression Estimates for the Private Saving Rate, 1971-94 (Inflation-adjusted, percent of GNP) lndependent Regression Equation: Variables (1) (2) (3) (4) (5) Lagged GNP/capita 0.22 0.26 0.20 0.26 0.35 (2.8) (3.4) (2.3) (3.7) (6.4) Lagged wealth/GNP -1.87 -0.96 -0.98 (5.7) (2.2) (3.1) Portfolio capital -1.06 -0.66 -0.57 (6.2) (2.7) (3.3) Private sector loans -0.27 (5.0) Government saving -0.61 (4.4) R2 adjusted 0.59 0.63 0.52 0.69 0.87 SEE 1.80 1.70 1.94 1.56 1.12 ED.W. 1.39 1.95 1.30 1.73 2.2 Note: GNP/capita - 1980 prices. Lagged wealth/GNP - market value, 1980 prices, Table 1.11. Portfolio capital - capital inflows as a percent of GNP, Table 1.7. Private sector loans - percent of GNP, Table 1.9. Govermnent saving - inflation- adjusted. Figures in parentheses are t-statistics. rather that the change in the stock-flow-stock rather than flow-flow. That is consistent with an irnterpretation of the loan variable as a proxy measure of the liberalization of the private financial system. 1.58 The wealth and foreign portfolio capital variables are combined in equation (4). The result is a modest improvement in the overall fit of the equation (shown in Figure 1.9), but a sharp fall in the size of both coefficients. The results are quite unstable for small changes in specification, and it appears that tke portfolio capital term mamntamns its Figure 1.9: Actual and Predicted Private Saving Rate, significance primarily because it is the only 1971-94 variable that is correlated with the partial recovery of the private saving rate in 1993- 25 94.35 Other elements of the capital account Equation 4 balance have no significant correlation with the domestic saving rate. Thus, to the extent 15 that foreign resource inflows impact negatively on private saving, the effect is 10 evident only for portfolio capital. 5 159 Finally, a substantial 0 improvement in the fit of the equation can be 1970 1975 1980 1985 1990 1995 achieved by adding the public-sector saving I-Actual-Predict raLte, equation (5). It implies a private offset Actual-Predicted of about 60 percent of any change in government saving, and it reduces the Source: Tablel.12 35 The preference for the portfolio capital variable is even more pronounced in the regressions that are based on the alternative measure of the private saving rate derived from the government financial budget balance. As shown in Figure 5, the financial balance mplies a much stronger recovery of the private saving rate in 1993-94, when the rate of portfolio capital inflow slowed. For both saving rates, the portfolio capital variable loses its significance if the data period is restricted to 1971-92.  7he Decline in Mexican Saving: A Cost of Reform? 23 standard error to one percent of GNP with very little effect on the other coefficients. However, as discussed earlier (para. 20), this correlation may be somewhat exaggerated by measurement errors. Since the private-sector saving rate is computed as a residual, there is an automatic inverse correlation between any measurement errors in the public and private-sector saving. 1.60 In summary, the statistical correlations provide some support for the argument that the decline in the private saving rate was related significantly to changes in financial markets-particularly the sharp surge of asset market prices and the resulting capital gains. Domestic financial market deregulation and the large inflows of foreign capital are, in turn, plausible explanations for the rise in equity market prices that accounted for most of the surge in the wealth-income ratio. IV. POLICY IMPLICATIONS 1.61 The question of why the Mexican saving rate declined after 1987 is of considerable importance to other countries that are considering a similar comprehensive economic reform program. While it is difficult to sort out all of the changes that were occurring in Mexico after 1987 and their effect on saving, this analysis suggests that the Mexican experience should not simply be categorized as another case of a consumption boom initiated by a non-credible stabilization program. By most standards the Mexican stabilization program was successful in achieving a sustained reduction of inflation and the behavior of market interest rates was not consistent with an interpretation that the gains were widely expected to be temporary. Instead, the Mexican consumption boom appears to have been more closely related to financial reforms and the explosion of financial asset prices. The extent of the rise in asset prices may, in turn, be a reflection of the huge scope of the Mexican reforms and the speed with which they were implemented. 1.62 The Mexican experience suggests that reforming economies should be wary of large price increases in asset markets, induced by transitory effects of market liberalization and excessive optimism. Mexican authorities viewed the price increases in equity markets as a sign of the success of the reform program, but they also provided the seeds of the subsequent collapse. Furthermore, an open competitive financial market calls for a different type of financial regulation than that applied during the era of credit controls and allocations. Both the regulators and the market participants are likely to be quite inexperienced. It would seem better to proceed with financial liberalization gradually, underpinning it with a strengthening of the institutional framework for prudential regulation and supervision. This is particularly true if the domestic market liberalization is combined with external capital account convertibility. The inflow of foreign capital can easily overwhelm domestic markets that are still lacking in depth. Mexico could have forced a more conservative lending policy by requiring higher capitalization of domestic banks, and it could have used a variety of tax and reserve account measures, similar to those of Chile, to temper foreign portfolio capital inflows. 1.63 It is difficult to relate Mexico's experience directly to that of other reforming countries. Chile, the model for many of the Mexican reforms, began the process with a private saving rate near zero and did not approach the Mexican saving rate until the late 1980s. In contrast, the Mexican private saving rate of the mnid-1980s was among the highest in Latin America. Furthermore, various economies in Asia- Malaysia and Thailand in particular-had large current account deficits, but without notable declines in private saving. Instead, Mexico stands out more for the magnitude of change in domestic financial markets and the size of the inflows of foreign portfolio capital. The Asian countries have in general proceeded more slowly than Mexico with liberalization of both domestic financial markets and foreign portfolio capital inflows.  24 Paper 1 1.64 In retrospect, the decline in private saving was a principal cause of the financial crisis of 1994-95. It was the major force behind the expansion of the current account deficit, and the reliance on private portfolio capital proved to be as transitory and as dangerous as the emphasis on bank finance a decade earlier. While in concept the inward transfer of resources through a current account deficit could make a positive contribution to growth, Mexico is another case where the predominant share of the foreign resource flow went to consumption not investment. Perhaps, the combination of a high and rising domestic investment rate and a foreign resource inflow can be viewed in a favorable light, but the combination of a fhlling domestic saving rate and a resource inflow should be viewed as a threat to sustained growth. 1.65 Finally, the extent of the fall in private saving and the increased reliance on foreign capital inflows suggest that Mexico's fiscal policy was too expansionary. In the situation in which the fall in private saving could not be reversed, the need for greater national saving could have been addressed with a larger budget surplus. Additionally, fiscal restraint could have been complemented by some depreciation of the exchange rate.  The Decline in Mexican Saving: A Cost of Reform? 25 Appendix Table 1.1: National Accounts Aggregates (Billions of new pesos) Ass (+)Net Gross (-) Capial National Net Grss Gross Domestic Factor National Consumpfi National (+) Dis)osable (-)Piite (-)Public National National Domestic Year Product income Product Allowances Income Transfies Icome= Consumpin Cnumptin Saving Saving Saving 1970 0.465 0.004 0.469 0.039 0.430 0.001 0.431 0.352 0.031 0.047 0.086 0.081 1971 0.514 0.002 0.516 0.041 0.475 0.001 0.476 0.393 0.036 0.047 0.088 0.085 1972 0.590 0.005 0.595 0.048 0.546 0.001 0.547 0.445 0.047 0.055 0.104 0.098 1973 0.724 0.001 0.725 0.058 0.667 0.001 0.668 0.535 0.062 0.072 0.129 0.127 1974 0.939 0.000 0.939 0.076 0.864 0.002 0 865 0.689 0.080 0.096 0.171 0.169 1975 1.149 0.000 1.149 0.097 1.052 0.002 1.054 0.829 0.110 0.115 0.212 0210 1976 1.435 -0.007 1.428 0.123 1.305 0.002 1.307 1.024 0.146 0 137 0.260 0.265 1977 1.931 -0.018 1.913 0.173 1.740 0.004 1 744 1.344 0.193 0.207 0.380 0.394 1978 2.454 -0.035 2.419 0.221 2.199 0.004 2 203 1.693 0.247 0.263 0.484 0.514 1979 3.205 -0.053 3.152 0.289 2.863 0.005 2.868 2.167 0.324 0.377 0.666 0.714 1980 4.470 -0.129 4.341 0.384 3.958 0.006 3.964 2.909 0.449 0.607 0.990 1.112 1981 6.128 -0.217 5.911 0.527 5.383 0.007 5 390 3.945 0.660 0.785 1.313 1.523 1982 9.798 -0.558 9.240 0.956 8.284 0.017 8.300 6.036 1.026 1.238 2.195 2.736 1983 17.879 -1.055 16.824 2.176 14.648 0.045 14.693 10.881 1.574 2.238 4.414 5.424 1984 29.472 -1.618 27.854 3.359 24.495 0.076 24.571 18.590 2.722 3.259 6.618 8.160 1985 47.392 -2.122 45.270 5.331 39.939 0.329 40.268 30.575 4.374 5.319 10.650 12.443 1986 79.191 -4.255 74.937 10.871 64 066 0.871 64.936 54.209 7.208 3.519 14.390 17.774 1987 193.312 -8.849 184.462 25.284 159.178 2.332 161.509 127.268 16 996 17.246 42.530 49.049 1988 390.451 -14.588 375.863 46.763 329.101 4.319 333.419 270.998 33.741 28.680 75.443 85.712 1989 507.618 -17.511 490.107 53.637 436.470 5.145 441.616 356.900 42.915 41.801 95.438 107.803 1990 686.406 -20.369 666.037 66.239 599.798 9.922 609.720 486.354 57.799 65.568 131.806 142.254 1991 865.166 -18.165 847.001 82.703 764.298 6.598 770.896 621.208 77.971 71.717 154.419 165.987 1992 1,019.160 -26.302 992.854 98.237 894.617 9.346 903.962 735.865 102.751 65.346 163.583 180.544 1993 1,127.580 -30.654 1,096.930 112.881 984.047 8.482 992.529 805.684 121.951 64.894 177.775 199.945 1994 1,272.800 -34.499 1,238.300 121.000 1,117.300 9.000 1,126.300 891.199 147.313 87.788 208.788 234.288 Note: All data are expressed on a base of 1980 in current prices. Source: INEGI.  26 Paper 1 Appendix Table 1.2: Components of the Public Sector Budget Balance, 1970-94 (Millions of new pesos) Balane Gobiemo Ownims Extm- Balance Intunei Balae Balance DebReneg. y Year Pubioo Presupuestal F*edal yEmpresas Pesupuest IG-FF Primarj FnaL Fmcrieo Opadoal Pva 1970 -13.6 -9.9 -8.2 -1.7 -6.0 2.3 -5.8 -1.5 -15.1 -11.6 1971 -9.5 -9.2 -3.4 -5.8 -2.2 1.9 -2 -1.8 -11.3 -6.4 1972 -21.5 -19.6 -14.9 -4.7 -5.0 3.1 -12.4 -4.8 -26.4 -18.6 1973 -34.9 -32.9 -21.3 -11.6 -6.6 4.6 24.2 -8 6 -43 5 -17.3 1974 -48.7 -41.9 -27.1 -14.8 -11.8 4.9 -33.3 -11.6 -60.3 -27.9 1975 -88 9 -83.3 43.8 -39.5 -13.4 7.8 -66 -13.4 -102.3 -74.8 1976 -102.7 -98 2 -61.7 -36.5 -15.3 10.8 -63.1 -22.1 -124.8 -56.2 1977 -91 3 -97.4 -61.7 -35.7 -4.5 10.8 -40.7 -25.2 -116.6 -48.1 1978 -117.5 -121.4 -67 1 -54.3 -7.4 11.3 -51.4 -274 -144.9 -79.5 1979 -176.2 -167 1 -101.4 -65.7 -24.8 15.7 -82.8 -41.6 -217.8 -116.6 1980 -292.6 -249.9 -133.1 -116.8 -41.9 -0.8 -133.9 -43.3 -335.9 -160.9 1981 -797 1 -726.7 -399.8 -326.9 -70.4 -0.2 489.8 -69.1 -866.2 -612.8 1982 -1,531 8 -1,253.0 -1,170.3 -82.7 -190.3 -88.5 -246.6 -128.5 -1,660.3 -538.9 1983 -1,455 -1,489.0 -1,458.2 -30.8 -143.0 177.0 423.6 -85.6 -1,540.6 71.5 1984 -2,105.8 -1,875.1 -2,130.5 255.4 -218.2 -12.6 1,414.6 -399.0 -2,504.8 -88.4 1985 -3,808.9 -3,336 0 -3,581.4 245 4 -448.3 -28.6 1,880.6 -726.3 -4,635.2 -379.1 1986 -11,804.8 -10,463 0 -10,341.0 -1,22.0 -1,352.1 10.3 1,962.4 -880.8 -12,685.6 -1,908.9 1987 -29,060.6 -27,438.1 -27,466.5 28.4 -449.2 -1,173.3 11,034.4 -1,939.4 -31,000.0 3,470.4 1988 -42,479.3 -36,668.3 -37,844.6 1,176.3 -405.6 -5,405.4 32,741.2 -6,255.2 -48,734.5 -14,013.3 1989 -25,268.1 -26,146.8 -25,589.3 -557.5 591.8 287.1 41,938.3 -3,186.9 -28,455.0 -8,562.4 1990 -19,266.2 -16,515.1 -19,435.6 2,920.4 1,074.0 -3,825.0 55,131 -7,648.3 -26,914.4 12,220.6 20,063.1 1991 -4,724.5 -1,572.7 -701.1 -871.6 419.2 -3,571.0 47,489.3 -8,257.4 -12,981.9 24,664.3 29,913.8 1992 16,361.2 16,067.6 15,959.0 108.5 881.0 -587.4 60,415.7 -11,430.8 4,930.3 30,836.8 30,123.7 1993 8,242.8 8,546.1 4,156 1 4,390.0 -141 2 -162.1 43,989.5 -31,975.7 -23,732.9 17,135.8 0 1994 -1,734.9 -4,473.6 -9,927.6 5,454.0 575.9 2,162.8 31,118.5 -44,255.6 -45,990.5 8,260.4 5,081.2 Source: Director General de Planeacion Hacendaria  The Decline in Mexican Saving: A Cost of Reform? 27 REFERENCES Banco de Mexico. Various years. Informe Anual. Mexico, D.F. . Various years. The Mexican Economy. Mexico, D.F. Buira, Ariel. 1994. "The Main Determinants of Savings in Mexico." Paper presented at the Conference on the Role of Saving in Economic Growth. Federal Reserve Bank of Dallas, Dallas, Tex., March 18-19. Boone, Peter. 1994. "The Impact of Foreign Aid on Savings and Growth." London School of Economics, Department of Economics, London, U.K. Processed. Calvo, Guillermo A. 1986. "Temporary Stabilization: Predetermined Exchange Rates." Journal of Political Economy 94 (6): 1319-29. Calvo, Guillermo A., Leonardo Leiderman, and Carment M. Reinhart. 1993. "Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors." International Monetary Fund Staff Papers 40 (March): 108-51. Deaton, Angus. 1989. "Saving in Developing Countries: Theory and Review." Proceedings of the World Bank Annual Conference on Development Economics 1989. Washington, D.C. Dombusch, Rudgier. 1982. "Stabilization Policies in Developing Countries: What Have We Learned?" World Development 10 (9): 701-8. Dombusch, Rudgier, and Alejandro Werner. 1994. "Mexico: Stabilization, Reform and No Growth." Brookings Papers on Economic Activity 1: 253-315. Edwards, Sebasian. 1994. 'Vhy Are Latin America's Saving Rates So Low?" World Bank, Latin America and the Caribbean Region, Office of the Regional Vice President, Washington, D.C. Gil-Diaz, Francisco, and Agustin Carstens. 1996. "Some Hypotheses Related to the Mexican 1994-95 Crisis." Paper presented at the annual Allied Social Science Association meeting, San Francisco, Calif., January 5-7. Banco de Mexico, Mexico, D.F. Goldsmith, Raymond W. 1985. Comparative National Balance Sheets. Chicago, m.: University of Chicago Press. Loser, Claudio, and Eliot Kalter, eds. 1992. "Mexico: The Strategy to Achieve Sustained Economnic Growth." International Monetary Fund Occasional Paper No. 99. Washington, D.C. McKinnon, Ronald. 1973. Money and Capital in Economic Development. Washington, D.C.: Brookings Institution. . 1991. The Order of Economic Liberalization: Financial Control in the Transition to a Market Economy. Baltimore, Md.: Johns Hopkins University Press.  28 Paper I Obstfeld, Maurice. 1995. "Effects of Foreign Resource Inflows on Saving: A Methodological Overview." World Bank, Policy Research Department, Washington, D.C. Poterba, James M., and Andrew A. Samwick. 1995. "Stock Ownership Patterns, Stock Market Fluctuations, and Consumption." Brookings Papers on Economic Activity 2: 295-357. Rebelo, Sergio, and Carlos Vegh. 1995. "Real Effects of Exchange Rate-Based Stabilization: An Analysis of Competing Theories. " NBER Working Paper No. 5197. National Bureau of Economic Research, Cambridge, Mass. Shaw, Edward S. 1973. Financial Deepening in Economic Development. Oxford, U.K.: Oxford University Press. White, Howard. 1992. "The Macroeconomic hnpact of Development Aid: A Critical Survey." Journal of Development Studies 28 (2): 163-240.  2. PUBLIC SAVING, STABILIZATION AND PRIVATE SAVING IN MEXICO I. INTRODUCTION 2.1 This paper explores the behavior of savings in Mexico from 1960 to 1994. First, it compares Mexican economic performance with other countries over the past three decades. It asks whether growth and savings are linked in a systematic way. Second, the paper seeks to explain time series variation in savings in Mexico. This is done using annual data for 1960-1994 and quarterly data for 1980-1994. Third, the paper outlines a theoretical framework for gaining insight into changes in the Mexican savings rate, with a specific focus on events since 1980. 2.2 Consistent with other findings in the empirical growth literature, the results presented in this paper suggest a statistically significant positive correlation between savings and growth, though the link is weaker than the link between investment and growth. The results also suggest that over much of the past three decades, Mexico's savings rate has been close to typical for its income level, but it has declined significandy in recent years. Mexico's slower growth in the 1980s and 1990s appears to be partly but not wholly explained by its declining savings rate.' 2.3 The main findings of the empirical analysis can be summarized as follows: * Private savings appear to depend positively on the real interest rate, though the magnitude of the effect is not statistically significant. The effect on private consumption is larger and significantly negative, reflecting that higher real interest rates have also been associated with lower private income. . Private saving rises significantly (and private consumption falls significantly) when the peso depreciates more rapidly. Both this and the first finding are suggestive of the role of stabilization programs in determining savings rates in Mexico. The major stabilization of early 1988 brought down real interest rates and virtually ended the rapid depreciation of the peso of the previous five years. Both of these effects, according to the regressions, would have lowered private savings, as indeed occurred. * Higher rates of government consumption, holding the government deficit constant, appear to raise private consumption and lower private savings, but the magnitude of the effect is not significant in all regressions. * Government investment appears to have a negative impact on private savings and no significant impact on private consumption, again holding the government deficit constant, suggesting that government capital formation acts as a partial substitute for private capital formation. * Holding fixed the levels of government consumption and investment, government financing decisions appear to have no significant impact on private consumption. Furthermore, the budget deficit appears to be offset, though not wholly, by changes in private savings. This finding is suggestive that Mexican Mexico's nominal national savings fell by about 5.7 percent of nominal GNP between 1978-82 and 1990-94. The cross country regressions suggest that this would imply a 0.7 percentage point drop in the growth rate of per capita real GDP holding other factors constant. However, the growth rate in 1990-94 was 1.7 percentage points lower than the growth rate in 1978-82. 29  30 Paper 2 data are broadly consistent with the Ricardian equivalence hypothesis, since it is robust to alternative measures of the budget deficit.2 2.4 The theoretical model, motivated by these empirical results, assumes that the Ricardian equivalence holds and examines the other sources of shocks to the Mexican savings rate. Since government linancing (but not the level of spending) is assumed not to affect private consumption decisions, the model focuses on national rather than private savings. A calibrated version of the model is used in an attempt to explain movements in national savings in the 1980s and 1990s. The model successfully predicts that the rapid devaluations of 1982-83 and 1985-87 would have raised savings and depressed consumption. Furthermore, it predicts that the 1988-94 stabilization, which saw a decline in real interest rates and a steadying of the peso, would have been associated with a decline in savings. While the signs of these effects are correctly predicted, the magnitudes and persistence of the effects are not fully explained by the rnodel. The model is not successful in explaining well the role of government spending, nor does it capture the apparent association of declining savings in the 1990s with shocks to the financial sector, such as the sharp rise in credit to the private sector and in asset values. II. SAVINGS IN MEXICO AND AROUND THE WORLD 2.5 To understand the behavior of output and savings rates in Mexico, it is useful to put data on these variables in perspective by examining the stylized facts concerning savings and growth around the vvorld. 2.6 Figure 2.1 plots the gross national savings rate versus the logarithm of per capita GDP, measured in 1987 US$.3 The chart was constructed by taking decade averages of both variables for the periods 1960-69, 1970-79 and 1980-89 for a panel of countries in the World Bank data set. The chart indicates that the savings rate rises with the level of income, with eventual convergence of savings rates happening at high income levels. 2.7 Mexico was not an outlier in any of the three decades, as its average savings rate in each decade was rather typical for a country of its income level. Mexico's national savings rate was 14.9 percent inthe 1960s, 17.6 percent inthe 1970s and 21.4 percent inthe 1980s. This is as compared to 16.4 percent, 18.2 percent and 19.1 percent as predicted by a quadratic function fitted to the data. The high average for the 1980s is due mainly to the first half of the decade. The savings rate fell in the latter half of the decade. From 1990 to 1994 it averaged just 17.1 percent compared to a predicted rate of 19.0 percent. This is, perhaps, the most striking episode since 1960, at least in terms of the savings rate. 2.8 Interest in the savings rate often focuses on its relationship to the growth rate of income. Figure 2.2 illustrates the relationship between the national savings rate and growth in the same sample of countries and decades. The chart shows a modest positive relationship between savings and growth. A 10 2 At the very least, this finding means that the aggregate data behave consistently with Ricardian equivalence holding at the micro level. 3 The savings rate m Figure 1 is calculated as the ratio of gross national savings in current local currency to gross national product in current local currency. The World Bank measure of gross national savings that is available for the broadest set of countries and the longest time span excludes net foreign transfers, which accounts for differences between the savings rates derived from the World Bank's annual data, and those derived from the quarterly Mexican national accounts, which are presented in the next section. Transfers added more than 1 percent on average to Mexico's national savings rate in the 1980s.  Public Saving, Stabilization and Private Saving in Mexico 31 percentage point increase in the national savings rate is associated with a 0.5 percentage point increase in the growth rate of real per capita GDP. 2.9 The link between saving and growth, however, is stronger when one considers multiple factors that explain cross-country variation in growth, and when one averages over longer sample periods. Barro and Sala-i-Martin (1994) and Levine and Renelt (1992) describe growth regressions in which there is a positive relationship between the investment share of GDP and GDP growth. Table 2.1 presents similar regressions where the investment share has been replaced by the national or domestic savings rates. The regressions presented in the table use the average growth rate of real per capita GDP from 1960 to 1989 as the dependent variable. The explanatory variables are the initial level of real per capita GDP in 1960, the secondary school enrollment rate in 1960, the 1960-89 average growth rate of the population and the 1960- 89 average of the gross national (or domestic) savings rate.4 The regressions indicate that there is a large positive and significant relationship between savings and growth. The regressions cannot be directly interpreted in either case as evidence for or against causality since the savings rates and the growth rates are simultaneously determined. 2.10 Mexico's growth performance declined in the 1960-90 period. The growth rate of real per capita GDP averaged 3.4 percent in the 1960s, 3.0 percent in the 1970s and -0.8 percent in the 1980s. At the same time, Mexico's national savings rate was rising from 14.9 percent in the 1960s, to 17.6 percent in the 1970s and 21.4 percent in the 1980s. Only in the late 1980s and early 1990s did Mexico's savings rate fall, dropping to 17.1 percent in 1990-94 while growth averaged 1.1 percent. In light of the results presented in Table 2.1, if all other factors had been constant the Mexican growth rate in the 1990s should only have been marginally smaller (6 basis points) than it was in the 1970s. This suggests that other factors, including cycles of macroeconomic instability have clouded the underlying fundamental relationship between savings and growth in Mexico. lII. SAVINGS BEHAVIOR IN MEXICO Measuring Savings 2.11 From an accounting perspective domestic savings simply represent the portion of domestic product which is not consumed domestically. Similarly, national savings are simply the portion of national product plus net foreign transfers which is not consumed domestically. So domestic savings SD=GDP- C=I+NX, and national savings SN=GNP-C=I+NX+NFI, where C is total consumption, I is gross domestic investment, NX is net exports of goods and services and NP7 is net income from abroad, including transfers.5 2.12 From an economic perspective it is important to decompose saving into the components due to the private and public sectors. Most models, including the one to be described in Section IV, model private decisions as distinct from decisions made by the public sector. 2.13 Decomposition of Savings into Private and Public Components. Private saving may be computed from the identity, SP=SN-SG, where Sp is private saving and SG is public sector saving. Therefore, 4 This is analogous to Levine and Renelt's (1992) most robust regression, with the investment share replaced by the savings share. 5 For convenience of notation, GNP represents gross national product plus net foreign transfers.  32 Paper 2 the first step is to measure public sector saving. This is traditionally done by summing some measure of the public sector budget balance, A0, and a measure of public sector investment, IG. That is, SG=AG+ IG1 2.14 In the second step, private saving is computed residually from national and public sector saving as Sp=SN-SG. If desired this measure can then be used to compute a measure of private disposable income, Yp=Sp+Cp, where Cp is private consumption expenditure. :2.15 To implement this measurement strategy one must choose a particular concept of the public sector budget balance. For Mexico a number of choices are available: the financial balance of the consolidated public sector, which consists of the central government, its agencies, public sector enterprises and public sector financial intermediaries;6 the economic balance, which drops the financial intermediaries from the definition of the public sector; the primary balance, which is the economic balance minus net interest payments; and inflation-adjusted analogues to the financial and economic balances. 2.16 The benchmark public balance concept used here is the inflation-adjusted financial balance. The main reason for using the financial balance is that it is the most complete definition of the public sector. The reasons for the inflation adjustment are described in the next section. As it turns out, the results presented here are robust to using alternative measures of the budget balance, although the strength of the findings regarding Ricardian equivalence are affected.7 '2.17 Inflation Adjustments. The national accounts-derived measures of nominal private savings given above do not adjust interest flows for the effects of inflation and other shocks. In order to umderstand the problem it is useful to consider an ideal measure of private sector savings as suggested by Barro (1993). The discussion assumes a closed economy for simplicity. A discussion of open economies f-ollows. 2.18 Real savings should be measured as the change in the real wealth of the private sector. In a closed economy this wealth consists of two types of assets, physical capital and public sector debt. The ideal measure of real net private savings is S= K, - K,, + B -p1, wvhere Sp, is nominal net savings, Pt is the price level, Kt is real capital, and Bt is net holdings of public clebt. All variables are measured at the end of a period. Given this concept of real savings, the ideal niominal net private savings measure is Sp, = P(K,-K, )+B, - (I +t)Bt-I where or, = P, I P,_l -1 is the inflation rate. 6 As detailed data on the finances of state governments are not available for Mexico, they are excluded from the measures of public sector balance used here. Since these governments are fiscally small relative to the federal government, the impact of this is likely to be small. 7 In quarterly data, the point estimates are unaffected both by the inflation adjustment and by the budgetary concept used. The precision of the point estimate is somewhat affected. In annual data, point estimates are affected by the inflation adjustment (unadjusted data are more consistent with Ricardian equivalence) and, less so, by the budgetary concept.  Public Saving, Stabilization and Private Saving in Mexico 33 2.19 The national accounts-based measure described in the previous section corresponds to Spt = PtIpt + Bt - Bt-1, where Ipt is real private gross investment. The difference between the ideal and national accounts measures is that one is a measure of net savings while the other is a measure of gross savings. The difference is Spt -Sp, = P,FKlI + t,-, where a is the depreciation rate of capital. The intuition for the other component is straightforward: the inflation rate is the depreciation rate of nominal assets. 2.20 From this discussion it becomes clear that if interest income is a significant component of the private sector's income then adjustments for inflation may be important, especially in highly inflationary environments. Since this statement is true for Mexico over much of the period examined in this paper, this issue is important. 2.21 In the literature, the inflation adjustment is frequently applied to savings data, where possible, but a capital depreciation adjustment is not. Presumably this is because capital stocks and depreciation rates are perceived to move relatively smoothly over time, or because gross capital formation is viewed as the relevant variable. Therefore, the adjusted measure of private savings commonly found in the literature is Sp' = Spt - rtB, l. 2.22 A symmetric argument to the above applies to the public sector. The adjusted measures of the public sector balance and public savings are At= AGt + vtBt and SGt =Sot + OTtBt_-1. 2.23 In an open economy the adjustments required are more subtle, because the private sector holds foreign financial assets while foreigners hold part of the public sector debt. However, the reasons for the adjustments are the same. Adjusted private savings is SI = Sp, - ,tBptl + (dt - t)et IBpt , where Bp,1 is private sector holdings of domestic public debt, Bt,I is the net foreign financial assets of the private sector (measured in foreign currency), et is the nominal exchange rate, and dt is the rate of depreciation of the peso between t-I and t. This discussion assumes that all domestically held public debt is denominated in pesos, all foreign assets are denominated in foreign currency, and that no financial assets are indexed. A detailed discussion of the different inflation adjustments required for different types of assets is given in Appendix 2.1. The adjusted public balance and public savings are Gt AGt + itBp- l + (zt - dt)et,1B,l1 and SGt= SGt + 7tBpt,l + (t -d)e, BGt 1,  34 Paper 2 where BG1- is the net foreign debt of the public sector (measured in foreign currency). This implies an inflation-adjusted measure of national savings, S a, given by = Sta = St + (dt-_ Itt)Bt*l where BZ 1 is Mexico's aggregate net foreign assets. 2.24 As described above, the adjustments shown here are only valid under several assumptions about the nature of domestic and foreign debt. Contrary to these assumptions, Mexico has issued indexed debt, and foreigners have held some of its peso-denominated debt. Further complications would arise from the fact that foreign debt might be denominated in more than one foreign currency. Annual Data 1960-1994 2.25 Figures 2.3 and 2.4 plot annual time series for Mexico over the period 1960-94. A striking aspect of these figures is that since 1980, all variables have become more volatile, with investment and savings having become more volatile relative to GNP. This suggests that since 1980 cyclical shocks have been an important factor. Throughout the period savings have been less correlated with income than either consumption or investment, probably due to Mexico's access to world capital markets. Another interesting feature of the data is that consumption has been about as volatile as output, indicating that there may be barriers to consumption smoothing in Mexico.8 These facts are confirmed by Table 2.2 which presents summary statistics. 2.26 Splitting the sample in 1980 brings out another interesting aspect of the data. Although over the whole sample savings grew at roughly the same rate as GNP, both savings and investment grew imuch faster than GNP in the 1960-80 period. They also shrank much more rapidly in the 1980-94 period than GNP. While this finding is suggestive of the role of savings in Mexico's slow growth performance after 1980, it could still be that the drop in savings was caused by, rather than being the cause of, this slow growth. 2.27 Finally, Figure 2.5 plots the shares in nominal GNP of consumption, investment, savings and the current account balance. In the period up to 1982 consumption's share steadily declined until it iwas just below three quarters of GNP. This was accompanied by a steady rise in the savings and investment shares and a resulting fairly steady share to the current account balance. Through 1981, the savings share was consistently smaller than the investment share. This is consistent with the fact that ihroughout this period Mexico ran a steady current account deficit usually around 5 percent of GNP.9 In 1982 the investment share dropped dramatically as the economy plunged deeply into the recession associated with the debt crisis. The savings share followed in 1983 but stayed mainly above the investment share until 1988 in the wake of the reform program. During this period Mexico ran a current account surplus which averaged about 1 percent of GNP. In the wake of the December 1987 reforms there was a 8 Some caution in interpreting this fnding is required because the measure of consumption includes government consumption and the measure of income is GNP. However, the finding is robust to using private consumption and private income. 9 The current account deficit in Figure 5 includes the statistical discrepancy. Furthermore, the sign and magnitude of the current account deficit is quite sensitive to whether it is measured in real or nominal terms. The conventional base year for the Mexican national accounts is 1980. Measuring the current account in constant 1980 new pesos tends to shift up the entire time series presented in Figure 5, while preserving the general pattern.  Public Saving, Stabilization and Private Saving in Mexico 35 simultaneous boom in consumption and investment. As a result the situation was reversed dramatically. Until 1994 Mexico's savings share dropped steadily and the current account deficit widened to a large 6.9 percent share of nominal GNP in 1992-94. 2.28 For annual data, as suggested above, public savings may be measured as the sum of public investment plus the consolidated public sector balance (financial balance). Private savings are defined residually as national savings minus public savings. This decomposition is presented in Figure 2.5 which suggests that there are strong Ricardian effects at work. Government saving behavior seems to be offset by private behavior to a large degree. However, the effects of rapid inflation make Figure 2.5 potentially misleading. 2.29 As reported in several issues of the Bank of Mexico's Informe Anual, the Mexican authorities prepare inflation-adjusted estimates of the public sector financial balance which take into account the effects of inflation. These are referred to as the operational financial balance. The adjustments made are similar to, though more complex than, the ones described in the preceding section. The adjustments are applied only to the domestic debt rather than the foreign debt. One advantage of this approach is that savings flows are not distorted by high frequency fluctuations in the real exchange rate. Furthermore, the adjustments made by the authorities take into account the precise nature of the debt instruments. On the other hand, a disadvantage of this approach is that it ignores the effects of longer-term movements in the real exchange rate. Again, private savings are computed residually as the difference between national savings and public savings. One pitfall to this approach is that it implicitly assumes that what the authorities describe as the domestic debt is held domestically. However, a substantial and varying proportion of Mexican domestic debt is held by foreigners. When this is true, part of the inflation adjustment should be applied to the current account (and national savings) rather than to private savings. 2.30 The inflation-adjusted public and private savings series,10 based on official estimates of the operational financial balance, are presented in Figure 2.5. The negative correlation between public and private savings remains, but it is clearly weaker than for the unadjusted series. Quarterly Data 1980-1995 2.31 As for the annual data, the measure of public savings used for the quarterly data is the public sector financial balance plus public investment. Official inflation-adjusted data on the operational balance are not available on a quarterly basis. Therefore, for the quarterly data, inflation adjustments to the value of the financial balance are applied as described earlier. 2.32 Denoting the financial balance as AGt, the unadjusted measure of public savings used here is SGt=IGt+AGt. The inflation-adjusted measure of public savings is given by SGT = I GI + A GI + r ,B pt I X where IG corresponds to the measure of public investment in the quarterly Mexican national accounts provided by INEGI, it is the rate of inflation of the GDP deflator, and Bpt corresponds to the end-of-period peso value of the consolidated domestic public debt (economica amplia) published by the Bank of Mexico in the Informe Anual. This measure of the public debt considers the public sector to be the federal government, parastatal organizations and enterprises, plus official financial intermediaries, and thus corresponds to the definition of the public sector used to construct the financial balance. 10 These are the same series used by Bosworth (1996).  36 Paper 2 2.33 The unadjusted measure of private savings, Spt, is simply national savings, St, minus public savings. That is Spt=SrSGt. Similarly, the inflation-adjusted measure of private savings is just SPt = St - Sot. Since a measure of private consumption, Cpt, is available directly from the national accounts, two measures of private disposable income can be calculated as Ypt=Spt+Cpt and Y;, = S;, + Cp,. 2.34 The raw quarterly data are highly seasonal. As a result, a simple deseasonalizing filter wvas applied to all time series except rates of interest and inflation. The filter is described in Appendix 2.2. 2.35 Figure 2.6 plots quarterly time series for Mexico for 1980Q1-1995Q2. The charts indicate that, as in the annual data, savings were less correlated with income than either investment or consumption. Consumption was also about as volatile as GNP indicating that there might be barriers to consumption smoothing. Investment was badly hurt by the debt crisis of 1982 and fell quite steadily through 1986. Consumption followed a similar, if less dramatic, path. In the wake of the reforms of 1987 both investment and consumption boomed, with total investment per capita rising about 24 percent in real terms between 1987Q4 and 1994Q4, and private consumption per capita rising 14 percent in real terms over the same period. These changes came during a period in which GNP per capita rose by a mere 6 percent in real terms. 2.36 Figure 2.7 looks at more disaggregated data which tell a more detailed picture. The unadjusted public and private savings figures suggest strong Ricardian effects. This remains broadly true for the inflation-adjusted series, although the patterns in these series are quite different. The large public deficits of the 1982-88 period almost become surpluses. This arises because there are enormous inflation adjustments to be made in the period 1982-1989, when both nominal interest rates and inflation were high. 2.37 Table 2.3 presents summary statistics for per-capita growth rates over the same period. The table highlights Mexico's poor growth performance over the past 15 years, with per capita aggregates actually shrinking over the sample period. Worst of all, investment fell more rapidly than any of the other aggregates, at almost 1 percent per annum. Consumption is more volatile than GNP, but it is considerably simoother than private income, so there is evidence of some degree of consumption smoothing in the data. Investment, as in the annual data, is much more volatile than output, by a factor of about 4.5. Similarly, savings is also more volatile than output. Not surprisingly, quarter-to-quarter changes in the variables are less correlated with GNP than in annual data. Regression Analysis 2.38 This subsection presents results of regressions which seek to explain the variation in private savings and consumption. These regressions should be interpreted cautiously. They do not necessarily identify causal links, since some of the right hand side variables are likely to be simultaneously determined. They are meant to be informative about what the sources of variation in the savings rate might be, and about the comovements between private savings and other macroeconomic variables. 2.39 The dependent variable in each of the benchmark regressions is either the real per capita level of private consumption (CP) or the real per capita level of inflation-adjusted private savings (SP). The right hand side variables in the benchmark regressions presented here were chosen from among:  Public Saving, Stabilization and Private Saving in Mexico 37 * a proxy for permanent income (YPERM) given by the trend component of real per capita GNP as defined by the Hodrick and Prescott (1980) filter,1' * a proxy for transitory income (YTRAN) given by the cyclical component of real per capita GNP as defined by the Hodrick and Prescott filter, * the ex-ante real interest rate on Mexican treasury bills (R), * the depreciation rate of the peso (D), * real per capita government consumption (CG), * real per capita government investment (IG), * a proxy for the marginal tax rate, given by taxes as a fraction of GNP (TX), * the real, per capita, inflation-adjusted financial balance of the public sector (AG), * the relative price of tradable goods (RPTR), and * the real per capita level of private domestic credit (PDC). 2.40 Ideally the specification would be based on a theoretical relationship expressing private consumption as some function of the relevant state variables. In a small open economy model, these would typically include the level of physical capital, the level of net foreign assets, the real rate of interest, the depreciation rate of the currency (or the growth rate of the money supply if this were the relevant monetary policy variable), government consumption and investment expenditure, and, possibly, various variables related to government budget financing. 2.41 Since not all of these variables are available for Mexico, certain compromises must be made. The proxy for permanent income, YPERM, is included to attempt to capture the effects of the wealth variables. In case the other right hand side variables do not adequately capture transitory factors affecting consumption, transitory income, YTRAN, is also included. As might be expected, including YTRAN does affect the point estimates associated with other variables, but it does not affect the qualitative findings. It does improve the fit of the estimated consumption functions. The income measures are based on GNP, because under the null hypothesis of Ricardian equivalence this is the appropriate measure of income. 2.42 The motivation for the real interest rate variable, R, is simply that savings or borrowing decisions are likely to be a function of the real interest rate. In a small open economy model, investment, consumption, the current account and subsequently savings, are all affected by the real interest rate. The ex-ante real interest rate was measured by the difference between the nominal interest rate on Mexican treasury bills and the expected rate of inflation computed as the fitted values from a regression of the current inflation rate on past inflation, the past nominal interest rate and the past depreciation rate of the peso. 2.43 The depreciation rate of the peso, D, is also included in the regressions. Fluctuations in the rate at which the peso depreciates can have effects because they affect the real return to holding peso balances. More rapid depreciation acts as a higher tax on peso balances. To the extent that these balances are useful in reducing transactions costs, this higher tax will have real consequences for consumption, investment, savings and output. 2.44 Public sector consumption, CG, is included in the regressions as it measures one of the components of physical resources extracted by the government from real GNP. These are resources The Hodrick and Prescott filter defnes the trend component of a series as a smooth centered moving average of the data. The smoothness of the trend depends on a smoothing parameter which is commonly set to different values for data of varying frequency: 1600 for quarterly data and 100 for annual data.  38 Paper 2 unavailable for private consumption. Furthermore, a test of whether government financing decisions matter requires government consumption to be entered as an explanatory variable. 2.45 Public sector investment, IG, is included in the regressions for similar reasons as public consumption. However, its effect is expected to be different because public investment produces public capital, which may be either a substitute for or a complement of private capital in production. 2.46 The public sector budget balance, AG, and investment, IG, are entered separately, rather than including their sum (public saving). The budget balance does not affect private consumption under the null hypothesis of Ricardian equivalence, but this hypothesis makes no statement about public investment. 2.47 In the regressions using quarterly data, taxes as a fraction of GNP, TX, is also included. Ideally, marginal tax rates would be used instead, as these would be the relevant variables for the consumption function. Since a time series of marginal tax rates was not available, the average tax rate was used. (The tax variable was not used in the annual regressions as the requisite data were not available over the full annual sample). 2.48 There are other variables which sometimes provide additional explanatory power. First, real per capita private domestic credit, PDC, is included in some of the regressions. It has been argued that the credit expansion of the late 1980s and early 1990s was associated with the drop in savings during that period. Second, for quarterly data, the relative price of tradable goods, RPTR, is available. This variable imoves sharply during periods of stabilization, and, as a result, may be highly correlated with consumption. Macroeconomic models, such as the one described in Section IV, suggest a strong correlation between the relative price of tadables and shocks which have wealth effects. As a result, RPTR may pick up some aspect of these shocks that are not captured by the other regressors used here. :2.49 The results of the regressions using private consumption as the dependent variable are presented in Table 2.4. Results are presented for both annual and quarterly data. The results indicate that a considerable portion of the variation in private consumption in Mexico can be explained by the variables included in the benchmark regressions. Both permanent and transitory income are positively correlated with private consumption. 2.50 Higher interest rates and more rapid depreciation of the peso are both associated with lower consumption, although the relatonship is not always significant. The direction, but not the magnitude, of the effect is robust to excluding transitory income from the regression. 2.51 Interestingly, the coefficient on government consumption is positive, though insignificant, in the annual data, when transitory income is excluded from the regression. However, when transitory income is included the coefficient is close to zero. In the quarterly data the coefficient is positive and significant when transitory income is excluded, but becomes much smaller when it is included. This suggests that government consumption is highly correlated with transitory income. 2.52 The results from the annual data suggest that government investment has a negative impact on consumption, although the size of the effect is only large when transitory income is included in the regression. For quarterly data a role for government investment is only found when transitory income is included in the regression, and the magnitude of the effect is small. If the coefficient on govermnent investment were -1, holding permanent and transitory income fixed, this would suggest that government investment acts like a lump-sum tax and that it produces nothing of value to private consumers. A coefficient of 0 would indicate that government investment was a perfect substitute for private investment,  Public Saving, Stabilization and Private Saving in Mexico 39 with the impact being felt on private investment not consumption. A coefficient in between suggests that public investment is an imperfect substitute for private investment. 2.53 The tax rate appears to have a negative impact on consumption in quarterly data. Once transitory income is included the tax rate is no longer significant. This does not mean taxes are unimportant: they may be an important determinant of transitory income. 2.54 The inflation-adjusted government budget balance has very little impact on private consumption. While there is some evidence from the annual regressions that budget deficits raise private consumption, the magnitude of the effect is modest and is statistically insignificant. The bulk of the evidence appears to support the Ricardian proposition. 2.55 In the annual data, private domestic credit provides no additional explanatory power. It is significant in the quarterly regressions, and enters with a positive sign, but it is driven out of the regression if the relative price of tradables is included. This suggests that these series were driven by common shocks (such as financial sector shocks generating wealth effects) not captured by other variables in the regressions, at least in the post-1987 stabilization period. 2.56 The results of the regressions using private savings as the dependent variable are presented in Table 2.5. Since private savings is a linear combination of a subset of the variables in the consumption regression,12 there is a trivial relationship between the coefficients in the savings regressions and the ones in the consumption regressions. 2.57 Both permanent and transitory income are positively correlated with private savings, although, in some cases, the correlations are not significant. Higher interest rates and more rapid depreciation of the peso are both associated with higher savings, although the relationship is not always significant. Interestingly, the coefficient on government consumption is negative, though insignificant, in the annual data. 2.58 The results from the annual data suggest that government investment has an insignificant but negative impact on savings. For quarterly data the effect is significant and negative. In the baseline regression the coefficient is -0.7, which suggests that public investment is not a perfect substitute for private investment or savings. As a result, when the government spends less on investment, the private sector does not replace this investment peso for peso. The results for private domestic credit and the relative price of tradables are consistent with the results from the consumption regressions, suggesting a role for financial sector related shocks and wealth effects. 2.59 The government inflation-adjusted budget balance is offset substantially by private savings. The regressions using annual data put the point estimate at -0.73, and for quarterly data the point estimate is very close to -1. Again, the results provide supporting evidence of Ricardian effects. The precise estimate of the offset coefficient, however, appears sensitive to possible measurement errors. IV. A MODEL OF CONSUMPTION AND SAVING IN MEXICO 2.60 The previous section presented regression evidence that suggested the following: (a) private consumption is largely unaffected by the government budget balance, holding the levels of public sector consumption and investment fixed, while private saving to a substantial extent offsets movements in 12 The identity relating the variables is SP=YPERM+YTRAN-C-CG-IG-AG.  40 Paper 2 the budget balance; (b) private consumption is positively correlated with public sector consumption, while private saving is negatively correlated with it; and (c) private consumption tends to fall and private saving tends to rise during periods of more rapid peso devaluation, and in periods of high real interest rates. This section describes a model which assumes that the first finding holds (the government budget balance does not influence private consumption) and attempts to explain the other findings theoretically. 2.61 More broadly, the model is used to understand how different shocks might have affected the Mexican economy in the 1980s and 1990s, focusing on the effects on private consumption and savings. Ihe shocks examined here include shocks to productivity, changes in the foreign nominal interest rate and foreign inflation rate (these have effects through purchasing power parity), changes in the rate of depreciation of the peso, fiscal policy shocks and changes in tax rates. 2.62 The model in this section is derived from Rebelo and Vegh (1995). The model describes an economy in which there are forward-looking consumers who have preferences over two types of goods: tradables and nontradables. Consumers allocate consumption so that the marginal rate of substitution between the two goods is equal to their relative price. 2.63 Consumers in the model also suffer disutility from work, which takes place either in the tradables or nontradables sector. In the version of model presented here, the labor market clears, so that the real wage is set equal to the marginal rate of substitution between consumption and leisure. 2.64 Production takes place in two sectors: a tradables sector where output is produced using capital and labor, and a nontradables sector where output is produced using a fixed factor, say land, and labor. Labor is freely mobile between the two sectors, so that the nominal wage is equated across them. Capital is accumulated using a standard investment technology which allows for adjustment costs. 2.65 The model treats Mexico as a small open economy, so that both the private and public sectors can borrow from the foreign sector at the world nominal interest rate measured in foreign currency. Unfortunately, this does not allow the model to capture those aspects of international borrowing and lending relating to risk. 2.66 The exchange rate is treated as an exogenous policy variable, whose rate of depreciation is selected by the government according to a rule. This makes the supply of money an endogenous variable, since money must be allowed to adjustto sustain any particular value ofthe exchange rate.l3 Money enters the model through a standard transactions technology. Purchases of consumption and investment have associated transactions costs which are smaller the greater are money balances. 2.67 A detailed description of the model is provided in Appendix 2.3. The appendix also describes how the parameters of the model were selected to replicate key features of the Mexican economy. Effects of Shocks on Macroeconomic Aggregates 2.68 This section describes the effects of different shocks on various macroeconomic aggregates in the model. The importance of any shock depends on its typical size and persistence. This means that a discussion of the effects of a particular shock should be based on an empirical assessment of these factors. In what follows, when an empirical measure of a shock is available, the law of motion of the shock is 13 The opposite case where the money supply follows a rule and the exchange rate is endogenously determined can also be examined.  Public Saving, Stabilization and Private Saving in Mexico 41 calibrated to match the data. When no measure of the shock is available, it is assumed to follow a random walk with drift; i.e., the shock is assumed to be persistent with unpredictable innovations. 2.69 Shocks to Productivity. Each of the goods (tradables and nontradables) is assumed to be produced using production technologies which shift over time. The model allows for exogenous shocks to total factor productivity in each sector. Positive shocks to productivity in either sector raise the level of wealth and permanent income as long as they are persistent. As a result, they are accompanied by a permanent rise in the level of private consumption. 2.70 As illustrated in Figure 2.8, a persistent rise in the level of productivity in the tradables sector leads to a permanent increase in the desired level of capital and the level of investment. In the short run, the rise in consumption and investment demand outstrips the increase in tradables output due to the productivity increase. As a result, that demand is met by a worsening of the trade balance. This effect outweighs the increase in investment in the very short run (about 2 years), so that domestic and national savings actually fall. However, the long-run increase in output is larger due to capital accumulation. This eventually eliminates the trade balance effect, with the result that in the medium- to long-run, domestic and national savings go up. 2.71 A rise in productivity in the nontradables sector does not lead directly to an increase in investment demand, since capital is not used in the nontradables sector. However, there is an indirect effect because the relative price of tadables goes up. This leads to a modest increase in the desired level of capital in that sector, and thus to an increase in investment. As a result, the impacts of the shock on the trade balance and savings are similar to the previous case, though of smaller magnitude. 2.72 Shocks to the Nominal Interest Rate on Foreign Debt. The model is a small open economy model, with Mexico having access to foreign capital at an exogenous nominal rate of interest. The model assumes certainty equivalence which means that the ex-ante real rates of return on all assets must be equated in equilibrium. There are three assets in the model: (i) foreign assets, (ii) peso money balances and (iii) capital. Peso interest-bearing securities play no role in the model because they are held in zero net supply. 2.73 The model can most easily be understood by considering the components of the real retums on these three assets. The real return on foreign assets is simply the foreign nominal interest rate minus the foreign inflation rate r -nT, both of which are determined by factors abroad. The real return on peso balances (which bear no interest) is the marginal savings in transaction costs that they induce, denoted s, minus the domestic inflation rate s-ir. 2.74 Agents in the model hold money balances because s is positive, or put differently, because money balances reduce the transactions costs associated with consumption and investment purchases. The return to accumulating the third type of asset consists of three components, the marginal product of capital net of depreciation, MPK, minus an adjustment cost, a, which is greater the more rapidly the stock of capital is changed, minus a transactions cost, t, associated with investment purchases. So the return to capital is MPK-a-t.  42 Paper 2 :2.75 A decrease in the foreign nominal interest rate, with no change in the foreign rate of inflation, lowers the real interest rate on foreign debt (or assets). Basic intuition suggests that this will lower the demand for foreign assets, and raise the demand for pesos and capital with a subsequent fall in the real rates of return to these assets. 2.76 The model assumes purchasing power parity so the domestic inflation rate, z, is equal to 1he foreign inflation rate, ir, which we assumed to be unchanged, plus the rate of depreciation of the peso, di, which is assumed to be an exogenous policy instrument. Thus, Yr is unchanged. Therefore, s must fall to decrease the return to holding pesos. The model assumes that s is diminishing in the ratio of real balances to real expenditure so a decrease in the return is effected by an increase in this ratio.14 2.77 As real balances become more abundant t falls, raising the return to capital. Therefore, an overall decrease in the return to capital is effected through a decline in MPK-a. In the short run this occurs dlue to a shift in labor away from tradables and into nontradables, which lowers MPK. In the medium run, the level of capital rises, which lowers MPK and raises a. In the very long run, only APK is lower. 2.78 Figure 2.10 illustrates the effects on consumption and savings of a 50 basis point decrease in the nominal interest rate, which is assumed to be transitory but long-lived. Because the decrease in the interest rate is persistent, the long-run increase in the level of capital will have a significant positive impact on permanent income. Private consumption will rise persistently as a result. With both investment and consumption demand going up, and tradables output actually falling in the short-run (due to the effect on labor), this demand is met through a significant worsening of the trade balance. As a result there is a long- lived negative effect on domestic savings which is only reversed by the eventual increase in tradables output. The impact on national savings is not as severe, because the fall in interest rates reduces the maagnitude of interest flows associated with existing foreign debt. 2.79 Shocks to the Foreign Inflation Rate. The effects of shocks to the foreign inflation rate are similar to those of the nominal interest rate since they both affect the real return on foreign assets. An inrcrease in the foreign inflation rate, with no change in the foreign nominal interest rate, decreases the real return on foreign assets. Furthermore, via purchasing power parity, if the depreciation rate of the domestic currency remains unchanged, the domestic inflation rate will rise by as much as the foreign inflation rate, directly decreasing the real return on peso balances. In equilibrium, the return to capital must also fall, and as intuition would suggest, this occurs through capital accumulation. 2.80 Capital accumulation, as usual, raises the level of permanent income, so there is an increase in private consumption, illustrated in Figure 2.11. The effect on savings is similar to that found for the interest rate: in the short- to medium-run the impact is significant and negative. In the long-run savings rise due to the increase in investment and income and the decline in the cost of servicing existing debt. 2.81 While inflation in the United States was falling throughout the early part of the 1980s, it is important to keep in mind that nominal interest rates were falling as well. These simultaneous changes would have tended to have offsetting effects. Real rates in the U.S. were very low in 1980 but rose sharply thereafter and then declined steadily throughout the remainder of the decade until 1992 when they began to rise again. These changes likely had important effects on the Mexican economy. 14 This ratio is the mverse of the velocity of money.  Public Saving, Stabilization and Private Saving in Mexico 43 2.82 Shocks to the Rate of Devaluation. The real return on foreign assets is unaffected by changes in the rate of peso devaluation. As a result, policy changes regarding this rate of devaluation do not change the equilibrium rates of return on peso balances or capital. The return to holding peso cash balances is s-a, but ,r=i+d, through purchasing power parity. A persistent decline in the rate of peso devaluation, d, leads to a persistent decline in the domestic inflation rate, ir, holding X fixed. But this tends to raise the return to holding pesos. This must be offset by an increase in the ratio of real balances to expenditure that reduces s. 2.83 An accumulation of real balances relative to expenditure lowers t and raises the return to holding capital. This must be offset by other factors that reduce the return. In the short-run these are increased adjustment costs, a, associated with rapid capital accumulation, which in the long-run leads to a reduction in MPK. 2.84 As with the interest rate and inflation rate shocks, the accumulation of capital has a significant impact on permanent income leading to an increase in private consumption. Investment demand and consumption demand are met in the short-run through a worsening of the trade balance which offsets the increase in investment, leading to a short-run decrease in domestic savings. Eventually, output increases and the trade balance improves, leading to a long-run increase in domestic savings. The response of national savings is similar because the shock has no impact on the cost of foreign debt service. 2.85 Consideration of Figure 2.12 makes clear the possibility that fluctuations in the rate at which the peso has depreciated have played an important role in economic fluctuations in Mexico. It illustrates the effect of a permanent and unanticipated 50 percent decrease in the rate of devaluation of the peso. The law of motion of the rate of depreciation is perhaps the most difficult to model, as it has tended to follow a regime switching process rather than a simple autoregressive process. Here it is assumed that each of these regime switches was unanticipated, but was believed to be permanent. 2.86 Shocks to Government Consumption Expenditure. As Christiano and Eichenbaum (1992) discuss, shocks to government consumption have no effect on aggregate variables (such as output and hours worked) if government consumption is a perfect substitute for private consumption. To illustrate the maximum possible effect of government consumption shocks it is useful to assume that private consumers derive no utility from government consumption. While this is unrealistic, government spending shocks have even less impact if government consumption directly provides utility. 2.87 Since the data used here do not distinguish between types of government consumption, we consider the effects of equiproportional increases in government consumption of the two types of goods. This increase in government consumption acts like a lump-sum tax. The data suggest that fluctuations in government consumption have been highly transitory, which would imply that each fluctuation has little impact on permanent income. As Figure 2.13 illustrates, this means that a 1 percent rise in government consumption, because it is transitory, has a very small negative impact on private consumption. 2.88 There are some minor distributional effects. Since government consumption is a larger fraction of nontradables output than it is of tradables output, equiproportional increases in government consumption of both types of good lead to a relative scarcity of nontradables for private use. This leads to a rise in the relative price of nontradables. As a result there is a slight transitory shift in labor and output to nontradables but a slight shift of consumption towards tradables. This extra tradables consumption must, therefore, come from a short-run worsening of the trade balance. Since the shock has virtually no impact on investment, the net effect on savings is small and negative.  44 Paper 2 2.89 The model does not explain the empirical findings of Section HI. There, increases in government consumption appeared to result in increases in private consumption. The model predicts modest effects in the opposite direction. One explanation would be that the model incorrectly assumes that government consumption provides no utility to consumers. A different prediction would arise from a model in which government and private consumption act like complements rather than substitutes. 2.90 Shocks to the Tax Rate. The model incorporates marginal income taxation in a rather crude form as a flat tax on value-added. The steady state tax rate was set at 15.5 percent which roughly matches the average tax burden as a fraction of GDP for Mexico from 1980Q1 through 1987Q4. Figure 2.14 illustrates the effects of a transitory 10 percent decrease in the tax rate. That is, the tax rate falls to 13.9 percent, but then rises back to 15.5 percent. 2.91 Since a tax on output acts indirectly as a tax on both capital and labor, a decrease in the tax rate increases the desired levels of capital and labor supply. However, since the tax cut is temporary, tihe response of investment is fairly weak, as most of the income generated by the investment is taxed later, alt the higher rate. There is a transitory effect on labor supply because the tax cut period is a relatively good time to work. Because little of the additional output generated by the tax cut goes to investment, there is a larger consumption and export response. The small increase in investment plus the improvement of the trade balance lead to a small increase in savings. 2.92 There has been little change in taxation in Mexico as a fraction of GDP over the period l980Q1-1995Q2, so these sorts of shocks cannot be expected to explain much of the time series variation in macroeconomic aggregates observed over this period. A caveat to this statement would be that the mLodel does not capture the subtleties of changes in the structure of taxation. Furthermore, this does not mLean that changes in tax policy would not have a significant impact on savings in Mexico. Predictions of the Model 2.93 The data for Mexico allow for direct measurement of the foreign nominal interest rate, the foreign inflation rate, the rate of depreciation of the peso, government consumption and taxes as a fraction oi GDP. This section explores the predictions of the model assuming that these shocks, alone, drove fluactuations in the Mexican economy in the period 1980Q1-1995Q2. Data on these measurable shocks are plotted in Figure 2.15. These clearly illustrate that shocks to government consumption and taxes were, for the most part, transitory. The model, being highly stylized, predicts strong responses to small fluctuations in the real interest rate and the rate of depreciation of the peso. As a result, if the shocks illustrated in Figure 2.15 were fed directly to the model they would imply excessively volatile fluctuations in quantity vazriables. As a result, smoothed paths for the shocks, illustrated in Figure 2.15 by dashed lines, were fed through the model to generate predictions for aggregate quantities. 2.94 The results of these experiments are illustrated in Figures 2.16 and 2.17 which illustrate the predictions for aggregate quantities and shares of GDP, respectively. These figures indicate that the model does quite well at capturing the general direction of the fluctuations actually observed. The rise in real rates of interest in the early 1980s imply a sharp decline in output, consumption and investment, although the decline in the data was deeper and longer lasting than the model predicts. The sharp decline in the rate of depreciation of the peso in 1988 (the peso depreciated at an average annual rate of about 2 percent in the last three quarters of 1988 after depreciating at an average annual rate of almost 95 percent dutring the period 1985Q3-1988Q1), together with the continued low real rates of interest of the late 1980s, imTplies that consumption, investment and output should all have risen sharply and the trade balance should have worsened in that year. Furthermore, domestic savings should have fallen initially and risen later as  Public Saving, Stabilization and Private Saving in Mexico 45 the investment effect outweighed the trade balance effect. The actual expansion seen in the data lags behind that predicted by the model. Furthermore, the model predicts an even sharper rise in investment than was observed in the data. This finding is symmetric to the fact that the model predicts a greater widening of the trade deficit than was observed in the actual data. 2.95 The model does a fairly good job of explaining the variation in domestic savings rates over the entire period. With the sharp rise in real rates, and the rapid depreciation of the peso in the early 1980s, the model predicts that savings should have risen during this period. These effects were reversed in the late 1980s, so savings should have fallen during this period. Both of these effects are found in the data. The model does not do a good job, however, of explaining the further drop in savings in the 1990s. 2.96 The model underpredicts the response of the relative price of nontradable goods to the stabilization of the late 1980s. This is illustrative of a basic failing of many dynamic optimizing models; they often are unable to simultaneously rationalize fluctuations in quantities and in prices. Were the model calibrated to match the relative price movements seen in the data, it would predict too much variation in the quantity variables. 2.97 What do these findings say about the consumption boom in Mexico in the late 1980s and early 1990s? Per capita private consumption (measured in 1980 new pesos) rose 13.8 percent between 1987Q4 and 1994Q4. The model predicts that consumption should have risen by 11.1 percent. So the model has little difficulty in predicting a strong consumption boom.15 However, in this same period investment rose by 24.0 percent while the model predicts that it should have risen by 34.8 percent. This suggests that the main puzzle lies more with investment than with consumption; specifically, why did investment not respond more to the decline in real interest rates, the stabilization of the peso and strong inflows of capital? 2.98 Part of the puzzle appears to be that in the early part of the stabilization program the response to the improved economic environment was sluggish. This may have been due to stickiness in the labor market. Rebelo and Vegh (1995) show that wage stickiness tends to dampen and propagate the effects of shocks in this sort of model. 2.99 Another explanation is that in this early period confidence in the stabilization program may have been low. Mexican borrowers no doubt paid risk premia above the U.S. real rate of interest. If this risk premium varied over time, the pattern of real interest rates faced by Mexican borrowers may have been quite different from the one in U.S. rates. 2.100 Major changes in financial markets not captured by the model may also have played an important role. Private credit rose dramatically relative to GDP in this period. But, since the main puzzle lies with investment, not private consumption, a credit based explanation can go only part of the way. 2.101 There is still a savings puzzle. While the model predicts that domestic savings should have fallen in 1988 in response to the stabilization of the peso at the end of 1987, it should have stabilized near that lower level and risen thereafter because the further declines in world interest rates should have been offset by the dynamic response of the economy to the peso stabilization. The actual decline in savings in Mexico was more persistent and deeper. Even assuming that shocks are slow to propagate it is difficult to explain the drop in savings that was observed in the period 1991-93 using the model and shocks examined 15 The model predicts consumption better than savings because it overpredicts the output response to the shocks that occurred.  46 Paper 2 here. The importance of the relative price of nontradables variable in the regressions suggests that perhaps some important shocks, especially in the financial sector and with important wealth effects (such as the sharp rise in private credit and the asset market boom), may have been omitted from the analysis that are being proxied by this variable and that provide part of the explanation for the continued decline in the savings rate in the early 1990s.'6 2.102 One problem with the discussion here is that durable consumption goods are treated in the same way as nondurables and services. So, part of the boom in consumption could be considered to be a lboom in investment and savings. This appears to be a partial explanation of the investment and savings puzzles. Between 1987 and 1994, real per capita consumption of durables in Mexico rose 21 percent while the real consumption of nondurables and services rose by 8 percent. However, there was no particular lurables boom in 1991-93 to explain the further worsening of the domestic savings rate at that time. V. CONCLUSION 2.103 This paper explored the behavior of savings in Mexico from 1960 to 1994. Cross-country evidence presented here indicates a statistically significant positive correlation between savings and growth. 'The results also suggest that while over the past three decades Mexico's national savings rate has been close to typical for its income level, in the late 1980s and early 1990s it has dropped below the average rate predicted for its income level. 2.104 Regression evidence leads to three main findings. First, private saving appears to depend positively on real interest rates and the rate of depreciation of the peso. Second, private saving depends negatively on government consumption and investment. Third, the government's budget deficit is substantially offset by private saving behavior. 2.105 The theoretical model suggests two major sources of shocks to the Mexican savings rate in the 1980s and 1990s: fluctuations in world real interest rates and dramatic fluctuations in the nominal value of the peso. Falling world real interest rates and a stable peso in the post-1987 reform period were likely mnajor factors in reducing national savings, as they would have induced substantial positive wealth effects in Mexico. 16 See Bosworth (1996) for a discussion of the possible role of such financial sector shocks.  Public Saving, Stabilization and Private Saving in Mexico 47 TABLES Table 2.1: Growth Regressions with Savings Rates, 1960-1989 GDP60 -0.95 -0.74 (0.21) (0.23) ENROLL 2.51 2.50 (1.48) (1.64) GRPOP -0.71 -0.88 (0.24) (0.27) GNSRATE 12.12 (2.43) GDSRATE 4.33 (1.89) R2 0.41 0.28 Observations 85 85 Note: The dependent variable is the average growth rate (in percent) of per capita real GDP measured in 1987 US$ from 1960 to 1989. The set of right-hand side variables includes GDP60 (the logarithm of the level of per capita GDP in 1960 in 1987 US$), ENROLL (the secondary school enrollment rate in 1960 measured as a fraction), GRPOP (the average annual growth rate of the population from 1960 to 1990 measured in percent), GNSRATE (the average share of gross national savings in gross national product from 1960 to 1989 measured as a fraction) and GDSRATE (the average share of gross domestic savings in gross domestic product from 1960 to 1989 measured as a fraction). The ratio of nominal variables was used to define the savings rates. Standard errors are indicated by parentheses. Source: GDP, GNS and GDS from the World Bank National Accounts; population and ENROLL from World Bank Social hndicators. Table 2.2: Summary Statistics for Growth Rates (Annual Data) Variable 1960-1994 1960-1980 1981-1994 Mean GNP 2.0% 3.4% -0.0% Consumption 1.8% 3.0% 0.1% Investment 3.3% 6.1% -0.6% National Savings 2.9% 5.2% -0.4% Standard Deviation GNP 3.1% 1.9% 3.5% Standard Deviation Relative to the Standard Deviation of GNP Consumtion 0.95 1.06 0.97 Investment 3.57 4.12 4.03 National Savings 2.09 2.83 1.95 Correlation with GNP Consumption 0.91 0.82 0.91 Investment 0.86 0.77 0.92 National Savings 0.69 0.48 0.71 Note: Each of the series is measured as the growth rate of the original series measured in per capita constant 1980 new pesos. National savings excludes net transfers. Source: Real measures of GNP, consumption, investment and national savings were obtained directly from the World Bank National Accounts; population from the World Bank Social Indicators.  48 Paper 2 Table 2.3: Summary Statistics for Growth Rates (Quarterly data 1980Q2 - 1995Q2) Variable Mean Std. Dev. (aCx) CX/COGNP Correlation with GNP GNP -0.2% 2.0% 1.00 1.00 Consumption -0.2% 2.2% 1.10 0.67 (Private) -0.2% 2.4% 1.23 0.69 (Public) 0.1% 3.4% 1.74 0.07 Investment -0.8% 8.9% 4.53 0.75 National Saving -0.1% 6.5% 3.33 0.52 Private Saving 1.0% 18.0% 9.16 0.23 Private Savingt 1.6% 22.9% 11.6 0.28 Private Income -0.2% 5.0% 2.55 0.46 Private Incomet -0.2% 5.4% 2.74 0.50 * Indicates unadjusted data. Indicates data adjusted for inflation as described in the text. Note: Each of the series is expressed in terms of its growth rate, in seasonally adjusted per capita constant 1980 new pesos terms. Real series were created by deflating nominal series with their corresponding deflators or were constructed as described below. The real series were then divided by the population and were seasonally adjusted. All variables are expressed in ecquivalent annual rates of change. Source: GDP, consumption (total, private and public), investment, exports and imports of goods and services, and their corresponding price deflators were obtained from the Mexican national accounts (NEGI). Net foreign income and transfers (WI) was obtained from the International Financial Statistics of the 1MF. Real NFI was created by dividing NFI by the GDP deflator. Real national savings was computed as real investment plus real current account deficit. Real GNP was computed as real GDP plus real NFI. The nominal fmancial balance was obtained from data provided by the Mexican authorities. The inflation-adjusted version of this was constructed as described in the text, where the inflation rate was computed using the GDP deflator, and the domestic debt was obtained from data provided by the Mexican authorities. Real budget balances were os)tained by dividing nominal balances by the GDP deflator. Real private savings measures were obtained as the difference between real national savings and the corresponding real budget balances. Real private income was obtained as the sum of real private savings and real private consumption. Population was obtained from the World Bank Social Indicators database, and was interpolated to the quarterly frequency by assuming constant within year growth rates. This is the optimal interpolation scheme if the logarithm of population follows a random walk with drift.  Public Saving, Stabilization and Private Saving in Mexico 49 Table 2.4: Consumption Regressions Annual Data Quarterly Data YPERM 0.37 0.56 0.53 0.93 1.05 0.80 0.62 (0.22) (0.21) (0.22) (0.21) (0.11) (0.13) (0.15) YTRAN 0.68 0.60 0.73 0.69 0.63 (0.27) (0.31) (0.06) (0.06) (0.06) R -0.023 -0.012 -0.010 -0.023 -0.012 -0.009 -0.010 (0.024) (0.022) (0.022) (0.02) (0.009) (0.008) (0.008) D -0.029 -0.018 -0.017 -0.013 -0.005 -0,007 -0.006 (0.008) (0.008) (0.009) (0.004) (0.002) (0.002) (0.002) CG 0.87 -0.11 -0.41 2.06 0.55 0.23 0.61 (0.61) (0.65) (0.86) (0.70) (0.38) (0.37) (0.40) IG -0.37 -0.90 -0.65 -0.13 -0.31 -0.06 0.08 (0.26) (0.31) (0.56) (0.22) (0.11) (0.13) (0.14) TX -0.59 -0.16 -0.13 -0.10 (0.15) (0.09) (0.08) (0.08) AG -0.07 -0.28 -0.27 0.02 -0.01 0.00 0.02 (0.23) (0.21) (0.22) (0.08) (0.04) (0.04) (0.04) PDC 0.06 0.11 0.01 (0.11) (0.04) (0.06) RPTR -2.76 (1.30) R2 0.76 0.83 0.83 0.76 0.94 0.95 0.95 Note: The dependent variable is CP The sample for the annual data is 1971-1994, and for the quarterly data it is 1980Q2- 1995Q2. Standard errors are in parentheses. Unit of Measurement. All real quantity variables are measured in constant 1980 new pesos at annual rates, and are expressed as levels. CP is real per capita private consumption. YPERM is real per capita permanent income while YTRAN is real per capita transitory mcome (construction of these series is described in the text). R is the real interest rate in annual percent. D is the depreciation rate of the peso in annual percent. CG and IG are real per capita public consumption and investment, respectively. TX is taxes (obtained from budget data provided by the Mexican authorities) as a percentage of GNP. AG is the real, inflation-adjusted, per capita public sector financial balance. PDC is real per capita private domestic credit, defined as nominal private domestic credit (obtained from the International Financial Statistics) divided by the GDP deflator. RPTR is the relative price of tradables nonnalized to I in 1980, derived from the national accounts (INEGI).  so Paper 2 Table 2.5: Savings Regressions Annual Data Quarterly Data YPERM 0.44 0.47 -0.05 0.20 0.38 (0.21) (0.22) (0.11) (0.13) (0.15) YTRAN 0.32 0.40 0.27 0.31 0.37 (0.27) (0.31) (0.06) (0.06) (0.06) R 0.012 0.010 0.012 0.009 0.010 (0.022) (0.022) (0.009) (0.009) (0.008) D 0.018 0.017 0.005 0.007 0.006 (0.008) (0.009) (0.002) (0.002) (0.002) CG -0.89 -0.59 -1.55 -1.23 -1.61 (0.65) (0.86) (0.38) (0.37) (0.40) IG -0.10 -0.35 -0.69 -0.94 -1.08 (0.31) (0.56) (0.11) (0.13) (0.14) Tx 0.16 0.12 0.10 (0.09) (0.08) (0.08) AG -0.72 -0.73 -0.99 -1.00 -1.02 (0.21) (0.22) (0.04) (0.04) (0.04) PDC -0.06 -0.11 -0.01 (0.11) (0.04) (0.06) RPTR 2.76 _ (1.30) 'R 0.82 0.83 0.95 0.96 0.96 Note: The dependent variable is SP. The sample for the annual data is 1971-1994, and for the quarterly data it is 1980Q2- 1995Q2. Standard errors are in parentheses. Unit of Measurement. All real quantity variables are measured in constant 1980 new pesos at annual rates. SP is real, inflation-adjusted per capita private savings constructed as described in Table 2.3. See Table 2.4 for the other definitions.  Public Saving, Stabilization and Private Saving in Mexico 51 FIGURES Figure 2.1: Level of Income and Savings Rate 0.4 0.35 + + + + + i 0.35 + ++ + + 0.25 + + &'+ + 5 6 7 8 9 10 Log of GDP per Capita Note: The chart plots decade averages (1960s, 1970s, 1980s) of the gross national savings rate (defined in the text) against the logarithm of the level of real per capita GDP measured in constant 1987 US$. Observations for Mexico, including the average for 1990-94, are plotted as filled circles rather than crosses, and are linked, in chronological order by the thick line. The thin line is a quadratic funaction fitted to the scatterdiagram. Some observations used to fit the curve have been omitted for readability. Figure 2.2: National Savings Rate and Growth Rate of Real Per Capita GDP 0. 1 . X 8 +++ + + 04 + ~+ 1 t:i~~~~Lgo DPprCpta 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Gross National Savings iRate Note: The chart plots decade averages (1 960s, 1 970s, 1980s) of the growth rate of real per capita GDP measured in constant 1987 US$ against the gross nationlal savings rate. The savings rate is calculated as nominal national savings over nominal GDP, both measured in current local currency. Observations for Mexico, including the average for 1990-94, are plotted as filled circles rather than crosses, and are linked, in chronological order by the thick line. The thin line is a linear function fitted to the scatterdiagram. Some observations used to fit the curve have been omitted for readability.  52 Paper 2 Figure 2.3: Annual Time Series for Mexico, 1960-94 (Per Capita Constant 1980 New Pesos) GNP Consumption 70 6 * 0 30 20 60 70 80 90 60 70 80 90 Year Year Investment NationaI Savings 20L X 20 15 ~~~~~~~~~~~15 0 0 5 5 0 ~~~~~~~~~~~~~0 60 70 80 90 60 70 80 90 Year Year Source: See Table 2.3 for sources of the data Figure 2.4: Annual Time Series for Mexico, 1960-94 (Growth Rates of Per Capita Constant 1980 New Pesos) GNP Consumption 10 IC 5 5 0q n / e 0 v _Nv -5 .5 -10 -1o 60 70 80 90 60 70 80 90 Year Year Investment National Savings 20 20 -20 - 20 - -40 -40 60 70 80 90 60 70 80 90 Year Year Source: See Table 2.3 for sources of the data.  Public Saving, Stabilization and Private Saving in Mexico 53 Figure 2.5: Annual Time Series for Mexico, 1960-94 (Shares of GNP) Consumpton Investm ent 0 8 0 2 075 60 70 80 90 60 70 80 90 Year Year National Savtngs CurrentAccount 02> ' O 1 -O I 60 70 80 90 60 70 80 90 Year Year Public Savings Pnmate Savings 0 ~~~~~~043 60 70 80 90 60 70 80 90 Year Year Inflation Adjusted Public Savings Inflation Adjusted Prtivate Savings n0 2 - 0 2{~ ~ 60 70 80 90 60 70 80 90 Year Year Note: Each of the shares is expressed as a decimal fraction of GNP. The shares are constructed as the series in current new pesos relative to GNP in current new pesos. Source: See Table 2.3 for sources of the data. Figure 2.6: Quarterly Time Series for Mexico, 1980Q1-1995Q2 (Per Capita Constant 1980 New Pesos) G NP Consumption 70 ~~~~~~~~~~60 65 55 o 0 2 60 CL 50 * 0 55 - ~~~~~~~~~45 50 40 80 85 90 95 80 85 90 95 Year Year 25 ~~Investment 20National Savings 20 18 -~~~~~~~~2 o 0 :16 1 4 5 ~~~~~~~~~~~10. 580 85 90 95 80 85 90 95 Year Year Note: One 1980 new peso = 43.6 1980 US$. Source: See Table 2.3 for the sources of the data.  54 Paper 2 Figure 2.7: Quarterly Private and Public Sector Time Series (Per Capita Constant 1980 New Pesos) Pnvate Cornsunpbon Pibbc Consrumpon 50 ,,.8 , ,,- 40 35 4 30 L 80 85 90 95 80 85 90 95 Year Year Pnvate Savirgs ktlabon Adjusted Private Savings 25 , , 25 20 20 15 15 10 10 80 85 90 95 80 85 90 95 Year Year PLbiIc Savings hation Adjusted Public Savings 10 10X Dr 5 5 0 C -5 -5- -Ic -lo, 80 85 90 95 80 85 90 95 Year Year Note: One 1980 new peso = 43.6 1980 US$. Source: See Table 2.3 for the sources of the data. Figure 2.8: Response to TFP Shock in Tradables Sector Tradable Productivity Shock Output Consumption 1.5 ~~~~~~~~~1.5 1.5 1 aH 1 1 0.5 0.5 0.5 20 40 20 40 0 20 40 time time time Investment Domestic Saving National Saving 3 12 ,1 0 20 40 0 20 40 0 20 40 time time time Trade Balance Relative Price ofNontradables Real Balances 0 3---3- -