78194 v2 Report No. 78194-PY A Public Expenditure Review for Paraguay Supplementary volume with selected background papers November 25, 2013 Argentina, Paraguay and Uruguay Country Management Unit Poverty Reduction and Economic Management Latin America and the Caribbean Region Document of the World Bank Preface This supplementary volume of the Paraguay’s Public Expenditure Review provides a number of background papers and material that was prepared as part of this PER. The topics are closely linked with the overarching storyling presented in the first volume of this Public Expenditure Review. The preparation of the PER also included a medium term debt management strategy (MTDS) mission to Paraguay and the preparation of the corresponding MTDS report. Given that the audiences of the present report and the MTDS report may differ, it is not included into this volume. 1) Paraguay: Estimation of the Structural Fiscal Balance and Fiscal Rule Proposal, Guillermo Le Fort V., Gonzalo Escobar and David Contreras 2) Paraguay: Agriculture Commodity Prices and Tax Revenue Collection, Edgardo Favaro, Friederike (Fritzi) Koehler-Geib, Nathalie Picarelli, Agustin Inaci 3) Evolution and Composition of Tax Revenue in Paraguay. Effects of the Tax Reform of 2004, Osvaldo Schenone 4) Assessing the Poverty and Social Impact of Fiscal Policies and External Shocks in Paraguay, Carolina Diaz-Bonilla, Martin Cicowiez 5) Social Spending, Taxes and Income Redistribution in Paraguay, Sean Higgins, Nora Lustig, Julio Ramirez, Billy Swanson 6) Equality of Opportunities and Public Spending in Paraguay, Jose Cuesta and Pablo Suarez Becerra 7) Paraguay’s BOOST database, a description, Massimo Mastruzzi, Eduardo Andrés Estrada, Renato Busquets, and Francisco Vazquez Ahued Table of Contents Chapter 1. Paraguay: Estimation of the Structural Fiscal Balance and Fiscal Rule Proposal, by Guillermo Le Fort V., Gonzalo Escobar and David Contreras. ..................................................................................... 1 Executive Summary ............................................................................................................................... 1 Introduction ............................................................................................................................................ 4 1. Descriptive Statistics of the Paraguayan Economy ........................................................................... 5 2. Estimation and Projection of the Structural Fiscal balance............................................................. 18 3. A Structural Fiscal Rule for Paraguay ............................................................................................ 35 4. Concluding Remarks ....................................................................................................................... 47 5. Technical Appendix ........................................................................................................................ 48 6. Paraguay’s Data Base...................................................................................................................... 60 7. Bibliographical References ............................................................................................................. 60 Chapter 2. Paraguay: Agriculture Commodity Prices and Tax Revenue Collection, by Edgardo Favaro, Friederike (Fritzi) Koehler-Geib, Nathalie Picarelli, Agustin Inaci............................................................ 63 Executive Summary ............................................................................................................................. 63 Introduction .......................................................................................................................................... 63 1. The Agricultural Sector in Paraguay with its Soy and Beef Subsectors ......................................... 64 2. Agriculture supply elasticities (soybean and beef) ......................................................................... 69 3. Tax Revenue collection, agriculture and economic activity ........................................................... 74 4. Conclusions ..................................................................................................................................... 81 5. Bibliography ................................................................................................................................... 82 6. Appendix ......................................................................................................................................... 82 Chapter 3. Evolution and Composition of Tax Revenue in Paraguay, Effects of the Tax Reform of 2004, by Osvaldo Schenone .................................................................................................................................. 97 Executive Summary ............................................................................................................................. 97 Recommendations .............................................................................................................................. 100 Introduction ........................................................................................................................................ 101 1. Situation Prior to Tax Reform Law 2421/04................................................................................. 102 2. Tax reform of law 2421/04 ........................................................................................................... 106 3. Conclusions and Recommendations ............................................................................................. 126 4. Bibliography ................................................................................................................................. 141 Chapter 4. Assessing the Poverty and Social Impact of Fiscal Policies and External Shocks in Paraguay, by Carolina Diaz-Bonilla and Martín Cicowiez........................................................................................ 142 Introduction ........................................................................................................................................ 142 1. Country Context: Growth, Poverty, and Inequality Trends .......................................................... 144 2. Policy Scenarios ............................................................................................................................ 146 3. Methodology and Data .................................................................................................................. 147 4. Simulations and Results ................................................................................................................ 149 5. Conclusions ................................................................................................................................... 160 Chapter 5. Social spending, taxes and income reditribution in Paraguay Sean Higgins (Tulane), Nora Lustig (Tulane), Julio Ramirez (Cadep), Billy Swanson (UC Davis) ....................................................... 161 Overall context ................................................................................................................................... 161 1. Paraguay’s tax and government benefits system .......................................................................... 164 2. The social protection system ......................................................................................................... 166 3. Social promotion and action programs ......................................................................................... 167 4. Data and Methodology .................................................................................................................. 169 5. Results ........................................................................................................................................... 173 6. Impact of taxes and social spending on inequality and poverty .................................................... 173 7. Conclusions ................................................................................................................................... 188 8. References ..................................................................................................................................... 188 9. Appendix ....................................................................................................................................... 190 Chapter 6. Equality of Opportunities and Public Spending in Paraguay, by Jose Cuesta and Pablo Suárez Becerra ....................................................................................................................................... 198 Abstract .............................................................................................................................................. 198 Introduction ........................................................................................................................................ 198 1. The Human Opportunity Index: Concepts and Measurement ....................................................... 200 2. Choice of Opportunities and Circumstances for Paraguay ........................................................... 204 3. Description of Results ................................................................................................................... 210 4. Benefit Incidence Analysis of Opportunities ................................................................................ 218 5. Education ...................................................................................................................................... 218 6. Health care .................................................................................................................................... 223 7. Policy implications: targeting additional spending ....................................................................... 226 8. Summary and Conclusion ............................................................................................................. 229 9. References ..................................................................................................................................... 230 10. Appendix ....................................................................................................................................... 231 Chapter 7. Boost Database for Paraguay, Massimo Mastruzzi, Eduardo Andrés Estrada, Renato Busquets, and Francisco Vazquez Ahued ................................................................................................. 241 Introduction ........................................................................................................................................ 241 1. Structure of Paraguay’s National Budget...................................................................................... 242 2. Data Sources ................................................................................................................................. 249 3. Particularities of the Data and Database Organization ................................................................. 250 4. How to Use the BOOST Database for Paraguay .......................................................................... 253 List of Figures-Chapter 1 Figure 1.1: Fiscal Revenue Composition ........................................................................................ 5 Figure 1.2: Fiscal Revenue percent of GDP ................................................................................. 6 Figure 1.3:Composition of Tax Revenue ........................................................................................ 6 Figure 1.4: Composition of indirect taxes....................................................................................... 7 Figure 1.5: Indirect Taxes as percent of GDP................................................................................ 7 Figure 1.6: Fiscal Spending as percent of GDP ............................................................................ 8 Figure 1. 7: Primary Fiscal Expenditure as percent of GDP......................................................... 8 Figure 1. 8: Public Sector Balance ( percent of GDP) .................................................................. 9 Figure 1. 9: Public Debt (as percent of GDP) ............................................................................. 10 Figure 1.10: Changes in Debt and Primary Balance ..................................................................... 10 Figure 1.11: Real GDP ................................................................................................................. 11 Figure 1.12: Real GDP - percent changes ................................................................................... 12 Figure 1. 13: Composition of GDP by Sector of Origin ............................................................... 12 Figure 1.14: Volatility of GDP by sector ...................................................................................... 13 Figure 1.15:Volatility of GDP and of Export Products ................................................................ 14 Figure 1.16: Tax Revenue Volatility ............................................................................................ 15 Figure 1.17: Direct Taxes Volatility ............................................................................................. 15 Figure 1.18: Indirect Taxes Volatility........................................................................................... 16 Figure 1.19: Volatility of revenues, public expenditure and GDP ............................................... 16 Figure 1.20: Volatility of government spending ........................................................................... 17 Figure 1.21: Primary spending volatility ...................................................................................... 17 Figure 1.22: Volatility of the GDP components ........................................................................... 18 Figure 1.23: Participation of top 10 export products .................................................................... 19 Figure 1.24: GDP Growth ............................................................................................................. 21 Figure 1.25: Labor force Growth .................................................................................................. 21 Figure 1.26: Unemployment and Solow Resource Use Index ...................................................... 23 Figure 1.27: Growth of investment and of the capital stock ......................................................... 24 Figure 1.28: Growth of total factor productivity .......................................................................... 25 Figure 1.29: GDP Gap .................................................................................................................. 26 Figure 1.30: Unemployment rate ................................................................................................. 26 Figure 1.31: GDP GAP and Estimated Structural Balance........................................................... 35 Figure 1.32: Historical Fiscal Revenues ....................................................................................... 36 Figure 1.33: Projected Fiscal Revenues ........................................................................................ 36 Figure 1.34: Public debt and interest spending as percent of GDP with a fiscal target of 1.23 percent and 20 years’ time ............................................................................................................ 42 Figure 1.35: Public debt and interest spending as percent of GDP with a fiscal target of 2.2 percent and 10 years’ time ............................................................................................................ 42 Figure 1.36: Public debt and interest spending as percent of GDP with a fiscal target of 1.62 percent and 20 years’ time ............................................................................................................ 43 Figure 1.37: Fiscal Balance Evolution ( percent of GDP) ........................................................... 44 Figure 1.38: Tax revenue as percent of GDP in 2010.................................................................. 60 List of Tables-Chapter 1 Table 1.1: Ordinary Least Squares Estimation for Total Fiscal Revenues ....................................28 Table 1.2: Ordinary Least Squares Estimation ..............................................................................29 Table 1.3: Ordinary Least Squares Estimation Log of Total Fiscal Expenditure .........................30 Table 1.4: Ordinary Least Squares Estimation Log of Current Fiscal Expenditure ......................31 Table 1.5: OLS Estimation Logarithm Capital Expenditure, 1990-2010 (T = 21) .......................32 Table 1.6: OLS Estimation Total Revenues Logarithm.................................................................32 Table 1.7: OLS Estimation Tax Revenues Logarithm ...................................................................33 Table 1.8: Targets for the Primary fiscal balance as a function of the Long Term Goal for Fiscal Net Worth and Time Horizon to achieve them. .............................................................................41 Table 1.9: OLS estimation Log of Total Fiscal Revenue ..............................................................48 Table 1.10: OLS Estimate First Difference logarithm Total Revenue ..........................................49 Table 1.11: OLS Estimation First Difference Total Revenue logarithm .......................................49 Table 12: OLS Estimate First Difference Total Revenue logarithm .............................................50 Table 1.13: OLS Estimate Total Revenue logarithm .....................................................................51 Table 1.14: OLS Estimation Total Revenue logarithm .................................................................51 Table 1.15: OLS Estimation Total Revenue logarithm, 1990-2010 (T = 21) ................................52 Table 1.16: OLS Estimation logarithm of Taxes on Goods and Services .....................................53 Table 1.17: Estimation logarithm of First Difference of Taxes on Goods and Services ...............54 Table 1.18: OLS Estimation logarithm of First Differences of Taxes on Goods and Services .....54 Table 1.19: OLS Estimation logarithm of First Differences of Taxes on Goods and Services .....55 Table 1.20: OLS Estimation logarithm of First Differences of Taxes on Goods and Services .....55 Table 1.21: OLS Estimation logarithm International Trade Taxes................................................56 Table 1.22: OLS Estimation logarithm Taxes on Corporations and Companies ...........................56 Table 1.23: OLS Estimation logarithm First Differences Taxes on Corporations and Companies57 Table 1.24: OLS Estimation logarithm First Differences Taxes on Corporations and Companies57 Table 1.25: OLS Estimation logarithm First Differences Taxes on Corporations and Companies58 Table 1.26: OLS Estimation logarithm First Differences Taxes on Corporations and Companies58 Table 1.27: Augmented DickeyFuller Test Total Revenue and GDP ...........................................59 Table 1.28: Dickey Fuller Augmented Test Estimated Errors .......................................................59 List of Figures-Chapter 2 Figure 2.1: Annual real GDP growth versus agricultural value growth ....................................... 65 Figure 2.2: International beef price (USD per ton Australian beef) and Inclan Tiao (1994) volatility break points ................................................................................................................... 68 Figure 2.3: Cattle herd size in million .......................................................................................... 68 Figure 2.4: Soybean and Beef Exports (in tons per month) .......................................................... 69 Figure 2.5: Soybean and Beef Export Trends ............................................................................... 69 Figure 2.6: Trends in Price of Soybeans and Beef........................................................................ 70 Figure 2.7: Short- and Long-run Soybean Price Elasticities......................................................... 73 Figure 2.8: Short-and Long-run Beef Price Elasticities ................................................................ 73 Figure 2.9: Tax Revenue Collection and Level of Economic ....................................................... 74 Figure 2.10: Summary Elasticity of response: TAXREAL-IMAEP ............................................ 75 Figure 2.11: Johansen Method. Agriculture (XSOY, XBEEF) and Level of Economic Activity (IMAEP) ....................................................................................................................................... 75 Figure 2.12: Vector Error Correction Estimates ........................................................................... 78 List of Tables-Chapter 2 Table 2.1: Sectoral shares in total value added ............................................................................. 65 Table 2.2: Product Contribution to Total Export Growth ............................................................. 66 Table 2.3: Main International Export Markets for Paraguayan Beef (2010) ................................ 68 List of Figures-Chapter 3 Figure 3.1: Paraguay: Real IMAGRO Revenue ......................................................................... 112 Figure 3.2: Tax Contribution of Agricultural Sector Compared to Share of GDP ..................... 114 Figure 3.3: Tax Contribution of Industry and Commerce Sectors Compared to Shares of GDP 115 List of Tables -Chapter 3 Table 3.1: Tax Revenues in Latin America ( percent of GDP) and Per Capita GDP (in US$) . 101 Table 3.2: Paraguay: Tax Collection 1990-2003 ( percent of GDP) ......................................... 102 Table 3.3: Paraguay: IMAGRO Collection 1996-2003 ( percent of GDP) ............................... 103 Table 3.4: Paraguay: IRACIS Collection 1994-2003 ( percent of GDP) .................................. 104 Table 3.5: Paraguay: Tax Revenue 2004-2010 ( percent of GDP) ............................................ 106 Table 3.6: Small and Large Taxpayers, 2000-2011 .................................................................... 107 Table 3.7: Estimated PIT Revenue Collection (billions of 2011 G. and percent of GDP) ...... 109 Table 3.8: Number of Single Tax and IRPC Contributors, 2003-11 (thousands)...................... 110 Table 3.9: IMAGRO Payment Regimes ..................................................................................... 111 Table 3.10: Number of IMAGRO Taxpayers, 2003-2011 (thousands) ...................................... 112 Table 3.11: Tax Revenue by Economic Activity, 2007-11* ( percent of total SET revenue) ... 113 Table 3.12: Share of Economic Sectors in GDP (2007-11) ........................................................ 114 Table 3.13: Effective IRACIS Rates After 2006 and Law 2421/04 ........................................... 116 Table 3.14: Number of IRACIS Taxpayers 2003-11 (thousands) .............................................. 117 Table 3.15: Maquila Regime (US$ millons) ............................................................................... 118 Table 3.16: Number of VAT Taxpayers, 2003-11 (thousands) .................................................. 119 Table 3.17: Tax expenditure Estimate for Bank Interest Exemption, 2008-11 (billions G and percent of GDP) .......................................................................................................................... 120 Table 3.18: Estimate of Tax expenditure of VAT Exemption for the Agricultural Sector, 2004, 2006, 2008 and 2010 (billions G. and percent of GDP) ........................................................... 121 Table 3.19: Tax expenditure of VAT Exemptions to Petroleum-based Fuels, 2006-11 (billions G. and percent of GDP) ................................................................................................................. 122 Table 3.20: Tax expenditure of VAT Exemptions and Reduced Rates, 2007-09 Average ....... 123 Table 3.21: Number of ISC Taxpayers, 2003-11 (thousands) .................................................... 123 Table 3.22: ISC Rates ................................................................................................................. 124 Table 3.23: Customs Duty Exemptions for the Import of Raw Materials, 2003-11 .................. 125 Table 3.24: Composition of Tax Revenue .................................................................................. 126 List of Figures-Chapter 4 Figure 4.1: Poverty and Extreme Poverty Trends, Paraguay ...................................................... 144 Figure 4.2: Gini Coefficient Trend, Paraguay (2003-2011) ....................................................... 144 Figure 4.3: GDP at Factor Cost ( percent growth) ..................................................................... 150 Figure 4.4: Education outcomes for the baseline and the “Exemption” simulations.................. 153 Figure 4.5: Under-5 mortality rates for the baseline and the “Exemption” simulations............. 154 Figure 4.6: Rates of access to clean water for the baseline and “Exemption” simulations ........ 154 Figure 4.7: Change in Moderate Poverty by Simulation relative to Base .................................. 154 Figure 4.8: Change in Gini for Household Per Capita Income by Simulation relative to Base . 154 List of Tables-Chapter 4 Table 4.1: Real macro indicators by simulation ........................................................................ 150 Table 4.2: Government receipts and spending in first report year and by simulation in final report year .............................................................................................................................................. 150 Table 4.3: Real government consumption -- annual growth from first to final report year by simulation ................................................................................................................................... 151 Table 4.4: MDG Indicators ........................................................................................................ 152 Table 4.5: Real macro indicators by simulation ....................................................................... 155 Table 4.6: Real government consumption -- annual growth from first to final report year by simulation.................................................................................................................................... 155 Table 4.7: Government receipts and spending in first report year and by simulation in final report year .............................................................................................................................................. 156 Table 4.8: Employment by factor -- annual growth by simulation from first to final report year ..................................................................................................................................................... 157 Table 4.9: Unemployment rate by labor type, simulation, and year ( percent) ......................... 157 List of Figures-Chapter 5 Figure 5. 1: The Evolutioon of Overall and Extreme Poverty in Paraguay, 2006-2010............. 162 Figure 5. 2: The Evolution of Inequality in Paraguay, 2003-2010 ............................................. 162 Figure 5. 3: Paraguay Social Spending as a percent of GDP, 2002-2010 ................................ 163 Figure 5. 4: Gini Coefficient for Each Income Concept in Argentina, Bolivia, Brazil, Guatemala, Mexico, Paraguay, Peru and Uruguay ........................................................................................ 174 Figure 5. 5: Inequality in Paraguay, 2010 ................................................................................... 175 Figure 5. 6: Reduction in Gini and Redistributive Effectiveness in Argentina, Bolivia, Brazil, Guatemala, Mexico, Paraguay, Peru and Uruguay. .................................................................... 176 Figure 5. 7: Decline in Headcount Index and Poverty Reduction Effectiveness in Argentina, Bolivia, Brazil, Guatemala, Mexico, Paraguay, Peru and Uruguay. .......................................... 177 Figure 5. 8: Concentration Shares of Taxes and Transfers in Paraguay, 2010. .......................... 182 Figure 5. 9: Concentration Coefficients by Spending Category in Paraguay, 2010. .................. 184 Figure 5. 10: percent of Direct Transfer Benefits Going to the Poor in Argentina, Bolivia, Brazil, Guatemala, Mexico, Paraguay, Peru and Uruguay. .................................................................... 185 Figure 5. 11:. percent of Direct Transfer Beneficiaries who are Poor in Argentina, Bolivia, Brazil, Guatemala, Mexico, Paraguay, Peru and Uruguay. ........................................................ 185 Figure 5. 12: percent of Poor Receiving At Least One Direct Transfer .................................... 186 List of tables-Chapter 5 Table 5. 1: Central Government Revenues in Paraguay, 2005, 2008 and 2010. ........................ 164 Table 5. 2: Social Spending as a percentage og GDP in Paraguay, 2005, 2008 and 2010. ....... 165 Table 5. 3: Taxes, Transfers, Inequality and Poverty in Paraguay, 2010. .................................. 174 Table 5. 4: Incidence of Taxes and Transfers in Paraguay, 2010. .............................................. 178 Table 5. 5: Average Benefits per Member of a Beneficiary Household in Paraguay, 2010. ...... 187 List of Figures Chapter 6 Figure 6. 1: Graphical Interpretation of HOI .............................................................................. 201 Figure 6. 2: Coverage and HOI in Paraguay, 2010 ..................................................................... 210 Figure 6. 3: Contribution of Circumstances to Overall Inequality (Shapley Decomposition), 2010 ..................................................................................................................................................... 211 Figure 6. 4: Vulnerability Profile for Educational Opportunities, 2010 ..................................... 212 Figure 6. 5: Change in the HOI across Educational Opportunities, 2003–10............................. 215 Figure 6. 6: Change in the Coverage and the HOI Gap over Time, 2003–10 ............................ 217 Figure 6. 7: Share of Aggregate Public Expenditure on Education (age 5-17), 2009 ................ 220 Figure 6. 8: Share of Public Expenditure on Secondary Education (age 16-17), 2009 .............. 221 Figure 6. 9: Distribution of Unitary Public Expenditures on Education Net of Private Household Contributions (age 5-17), 2009 ................................................................................................... 221 Figure 6. 10: Distribution of Unitary Public Expenditures on Elemental Education Net of Private Household Contributions (age 5-17), 2009 ................................................................................. 222 Figure 6. 11: Distribution of Unitary Public Expenditures on Secondary Education Net of Private Household Contributions (age 5-17), by Quintiles of Households, 2009 ................................... 222 Figure 6. 12: Share of Public Expenditures on Health Care (age 0-17), 2009 ........................... 224 Figure 6. 13: Share of Public Expenditures on Health Center Care (age 0-17), 2009 ................ 224 Figure 6. 14: Share of Public Spending on Hospital Care (0-17), 2009 ..................................... 224 Figure 6. 15: Distribution of Unitary Public Expenditures on Health Care Net of Private Household Contributions (age 0-17), 2009 ................................................................................. 225 Figure 6. 16: Distribution of Unitary Public Expenditures on Health Center Care Net of Private Household Contributions (0-17), 2009 ....................................................................................... 226 Figure 6. 17: Distribution of Unitary Public Spending on Hospital Care Net of Private Household Contributions (age 0-17), 2009 ................................................................................................... 226 Figure 6. 18: Share of Public Spending on Secondary Education by Circumstance Group (age 15- 17), 2009 ..................................................................................................................................... 228 Figure 6. 19: Share of Public Expenditure on Hospital Health Care by Circumstance Group ... 229 List of Tables, Chapter 6 Table 6.1: Opportunities in Paraguay ......................................................................................... 207 Table 6.2: Summary Statistics of Circumstances Used in Analysis ........................................... 209 Table 6.3: Circumstance Groups and Their Probabilities to Attend School and Access to Health Care Services (when sick) ........................................................................................................... 227 List of Figures, Chapter 7: Figure 7. 1: Structure of the Administrative Classification ...................................................................... 243 Figure 7. 2: Structure of Central Administration ...................................................................................... 243 Figure 7. 3: Structure of Decentralized Entities ........................................................................................ 244 Figure 7. 4: Hierarchy of the Variables of the Functional Classification ................................................. 244 Figure 7. 5: Hierarchy of the Variables of the Classification by Object of Expenditure .......................... 247 Figure 7. 6: Hierarchy of the Variables of the Classification by Source of Funding ................................ 249 Figure 7. 7: Interface of the BOOST Database for Paraguay ................................................................... 255 Figure 7. 8: Variable Settings for Central Administration and Decentralized Entities ............................. 256 Figure 7. 9: Variable Settings for Central Administration and Decentralized Entities ............................. 257 Figure 7. 10: Example of Filter for Executive Branch by Nivel (Level).................................................. 258 Figure 7. 11: Example of Time Series Analysis: Initial Budget by Entities in the Executive Branch (2003- 2012) ......................................................................................................................................................... 259 Figure 7. 12: Example of Execution Report for the Municipality of Asunción (2006-2010) ................... 260 Figure 7. 13: Example of Execution Report for the Judicial Branch (2009-2012) ................................... 261 List of Tables, Chapter 7 Table 7. 1: List of Variables for Central Administration and Decentralized Entities................. 252 Table 7. 2: List of Variables for Municipalities .......................................................................... 253 Chapter 1. Paraguay: Estimation of the Structural Fiscal Balance and Fiscal Rule Proposal, by Guillermo Le Fort V., Gonzalo Escobar and David Contreras. Executive Summary This article presents a proposal for the development of a structural fiscal policy for Paraguay, including an assessment of structural revenues and the derivation of a primary balance goal on the basis of a target for public net worth. The structural fiscal framework is based on a set of macroeconomic variables that are used for the estimation of the structural fiscal revenue and balance, and also for the derivation of a target for the structural primary balance. These include the long-term real interest rate relevant for the public debt and the GDP growth trend. The GDP growth trend was estimated using a Cobb-Douglas aggregate production function, resulting in a 4.5 percent to 4.8 percent annual growth estimate, while the long-term interest rate was estimated on the basis of international references and Paraguayan risk. The GDP gap, defined as the logarithmic difference between GDP and GDP trend, is a key variable for the estimation of structural fiscal revenues. In the estimation of structural fiscal revenues, cyclical variations of actual fiscal revenues were linked only with GDP since we did not find a price of a commodity export or weather related variable presenting a statistically significant relationship with Treasury revenues since fiscal revenues are mainly composed of indirect taxes. In the estimation of structural fiscal revenues, actual fiscal revenues were linked only with GDP since we did not find a price of a commodity export presenting a statistically significant relationship with Treasury revenues. A co-integration model was used to estimate the elasticity of total fiscal revenues with respect to GDP, which resulted in a value of 1.8 implying that economic growth allows for more than proportional increases of fiscal revenue, that is, revenues, as a share of GDP, increase with economic growth. The large incidence of indirect taxes stands out from historical data; however, this should be modified with the strengthening of direct taxes that intends to be addressed by the introduction of an income tax. Total public revenues exhibit a strong pro-cyclical behavior so that when the GDP gap is positive there are temporary positive revenues that expand total income over structural revenues; the opposite happens when the gap is negative. Public expenditure does not show a similar behavior; its evolution is neutral to the GDP cycle, confirming that the estimation of a structural balance only requires a correction of revenues and not of expenditures. The equation for public debt, expressed as a percentage of GDP, links the future level of public debt with its initial level, with the structural primary fiscal balance and with the key macroeconomic variables: real long-term GDP growth and the real interest rate relevant to the Paraguayan public debt. The structural primary fiscal balance is the policy target to be chosen so that public debt reaches a certain goal in a certain number of years; it is calculated free of the effects of the economic cycle, which, in the medium and long term average zero; but, should be checked annually by computing the actual primary structural balance on the basis of the estimates that connect the GDP gap with fiscal revenues. Structural fiscal revenue for one particular year represent the level fiscal revenue would have reached had the GDP gap been zero in that year. 1 The assumptions used for the development of the structural fiscal policy proposal include a long- term real GDP growth equivalent to 4.5 percent per annum and a relevant real interest rate of 6.5 percent per annum. Given that Paraguay does not have a developed public debt market, the relevant interest rate estimated for Uruguay, 5.75 percent, was used as a reference; 75 b.p. were added due to the difference of three grades between the risk rating of the two economies. The assumption of a constant interest rate during the entire period under consideration could be reviewed to the extent that Paraguay improves its public debt indicators and, as a consequence, risk for investors decreases; this would lead to a lower real interest rate as its risk rating improves. . The issue of Paraguayan public debt in the international financial markets could facilitate disseminate information on the use of a fiscal rule in Paraguay and confirm a lower risk perception. If the relevant interest rate decreases, the target for the primary fiscal balance could be less demanding, or, the goal for the fiscal net worth could be achieved in a shorter period. For the derivation of the structural primary fiscal balance target different alternatives are considered regarding the long-term goal for public net worth and for the time horizon to achieve it. As proposal for the net worth objective it was considered a range between the current level of net debt, a negative net worth equivalent to 20 percent of GDP, and one in which the net debt becomes negative, reaching a positive public net worth equivalent to 10 percent of GDP. A range between 5 to 25 years is considered as possible horizons to achieve the long-term goal; if the public debt goal is to maintain the current level of indebtedness, 5 years are more than enough, while 25 years seem appropriate in order to reach the most ambitious goal of a negative net public debt. A structural fiscal primary surplus of 0.43 percent of GDP is the target consistent with the less demanding long-term fiscal goal of maintaining the current level of debt. That small primary surplus allows for the payment of interests and no additional debt would be required, but, does not reduce current levels of indebtedness, for which a higher primary surplus would be needed. A structural primary fiscal surplus of 1.2 percent of GDP is required in order to achieve the complete elimination of public debt in a period of 20 years. Of course, there are still more ambitious alternatives for the fiscal policy goal. However, an objective of zero debt in the medium term appears to be very achievable and provides significant guarantees of stability since it generates a protective cushion of savings against crisis and the sustainability of fiscal policy appears to be significantly secured with the objective of zero fiscal net worth. With an indebtedness of 20 percent of GDP, similar to the current level, it is more likely that negative shocks may cause the debt to increase up to unsustainable levels. Economic disturbances affecting GDP and other variables as a result of domestic or external shocks can produce significant impact on the fiscal results and the level of public indebtedness. Once the fiscal net worth objective is reached it will be necessary to define another target for the primary fiscal balance, this time for the indefinite future. If for example, the initial long-term goal considered was a zero fiscal net worth, after it has been achieved it will be necessary to compute, using the same parameters, a new target for the primary structural fiscal surplus, this time, for the indefinite future. In the case of the example considered, it would be a primary surplus of 0 percent of GDP for the long term, rather than a 1 or 2 percent of GDP surplus. That is, a country with zero net public debt does not need to generate a primary surplus in order to pay 2 interest or to reduce its public debt; it is enough to keep a permanent zero balance to avoid incurring on positive net indebtedness. A structural fiscal policy such as the proposal is a long term State policy that transcends and sets a framework for the specific fiscal policies implemented by the different Governments during their term. In that sense, the technical parameters defining such policy should be determined outside the political contingency by technical groups that provide guarantees of impartiality. The estimates of the macroeconomic variables could be carried out by a Committee of independent experts. For example, highly regarded economists that represent different points of view related to the Paraguayan fiscal policy. This Commission should calculate the macro variables and the structural primary balance goal associated with the fiscal wealth target and the period of time to achieve it. These include estimates of the relevant real interest rate and the GDP long-term growth rate; it should also review the estimations of structural revenue and the annual expenditure budget consistent with it and, therefore, with the long-term fiscal goal. The decision of a long-term fiscal goal should be democratically adopted and maintained during the selected period. The technical group should review the value of the relevant parameters from time to time, for example, every five years. A technical report must be submitted by the authority to Parliament and the public explaining the methodology and the results obtained in the estimation of structural revenue and structural balance. Fiscal transparency, as for example, the availability of data and information, is an additional requirement for a proper operation of the fiscal rule. Such a fiscal policy must be supported by the timely availability of high quality data on the public sector: fiscal operations, earnings data, accrued expenses and others related to national accounts, as well as information on assets and liabilities, properly endorsed by informative sources. The information should be detailed and public through accessible media (web page). In addition, a better knowledge of the contingent liabilities of the public sector is required. In addition, the development of a public debt market, both domestic as well as international, is advisable for Paraguay. An international Paraguayan public debt market would enable the information regarding the fiscal policies under execution to be disseminated, and, in case of adopting a structural fiscal policy, would allow further benefits as country risk decreases. This would be of benefit not only for the public sector as its financing cost decreases, but also for the Paraguayan private sector that could see an increased availability of external financing. The development of a domestic public debt market might have a very positive effect on the diversification of the treasury’s financial risk; for example, developing instruments denominated in local currency. It would also be beneficial for the private sector since that is the basis for the development of a local capital market. 3 Introduction In establishing a structural fiscal policy the goal is to contribute to macroeconomic stability generating predictability of the fiscal policy and strengthening the solvency of the public debt. The purpose of this work is to recommend policy measures that facilitate the development of a structural fiscal policy in the Paraguayan economy, which generate predictability and credibility that may contribute to the stabilization of the cycle and may decrease the likelihood of insolvency. For this purpose, this report presents a brief review of the Paraguayan fiscal situation, as well as estimates to determine a structural fiscal balance, and, proposes elements for a structural fiscal rule. This report on a structural fiscal policy for Paraguay is divided into three sections. In the first one, the Paraguayan macroeconomic data as well as the fiscal accounts statistics that were used are described and interpreted; the second section offers estimates and projections for the structural fiscal balance, proposing that the origin of the deviations is in the gap between actual GDP and trend GDP. The third section provides a design and simulation of a structural fiscal policy for Paraguay, including a goal for the structural balance that seems most appropriate for the country as well as its effects on the level of public debt in the medium and long term; in addition, some recommendations for the implementation of a fiscal rule that contributes to the improvement of the country's macroeconomic stability are presented. Time series were constructed based on three different sources of information; despite not being necessarily and fully compatible, all of them are considered to be reliable. Data from different sources was compiled seeking to ensure its compatibility; but, despite our best efforts, there are limitations that could affect the results. The database was assembled with information from reliable sources such as ECLAC, the International Monetary Fund and the World Bank; however, there is no guarantee that these series are fully compatible. In addition, due to the lack of data in the series, it was necessary to estimate and interpolate certain variables and parameters that may be inaccurate, particularly, in the case of the labor market. The second part of this report presents the development of the analytical framework used to estimate the trend GDP, the GDP gap and the fiscal variables used to determine the structural fiscal rule. This second section also describes the estimates and projections of the structural fiscal balance and fiscal revenue of Paraguay, which were estimated based on the trend GDP which in turn was derived from a Cobb-Douglas type of production function and from time series that can be traced back to the 1960s for the stock of capital, the labor force, unemployment and total factor productivity. Considering that the main source of volatility and deviations from the trend of fiscal revenue in Paraguay corresponds to the GDP cycle, this second part presents an analytical framework to define and estimate the trend GDP and the corresponding structural fiscal variables, building on structural fiscal policy experiences in other countries such as Chile. The third part of this report presents a proposal for a structural fiscal target that would enable the public debt sustainability in the medium and long term by reducing the public sector vulnerability to shocks affecting its solvency. That is, a proposal, design and simulation of a structural fiscal policy is described, including a measure for the structural balance that seems most appropriate for Paraguay. Since the objective of establishing a structural fiscal target is to 4 generate predictability that ensures the solvency or sustainability of the financial position of the public sector—or at least to reduce its historical vulnerability—a proposal for a long-term goal for the public sector net worth is also presented. Finally, certain policy measures are proposed for the implementation of the rule in order to improve fiscal transparency and thus contribute to macroeconomic stability. The report presents some recommendations on policy reforms that contribute to transparency and are necessary for the proper implementation of a stabilizer structural fiscal policy. 1. Descriptive Statistics of the Paraguayan Economy The objective of this chapter is to provide a simplified review of the evolution experienced by the Paraguayan economy in the last two decades. The performance of the productive activity as well as the evolution of the fiscal condition during this period is shown; this will be used as the basis for the development of the proposal for a structural fiscal policy explained later. Paraguay’s Public Sector1 Fiscal Revenue and Spending In the past 20 years tax collection has represented, on average, 71 percent of the public sector revenue being this the main source of fiscal income; but, this collection is equivalent scarcely to 13 percent of GDP on average. The remaining third is explained by non-tax income as is the case of royalties from hydroelectric plants in the basin of the Paraná River, capital revenue, and, in lower amount, external grants. (Figure 1.1 and 1.2) Figure 1.1: Fiscal Revenue Composition Source: LE&F based on data from ECLAC and the World Bank 1 All the information of the public sector considers the Central Government. 5 Figure 1.2: Fiscal Revenue (percent of GDP) Source: LE&F based on data from ECLAC and the World Bank Almost ¾ of tax revenues correspond to indirect taxes. When breaking down tax revenues, it is possible to determine that since 1992 up to date the indirect taxes component is the one that generates the highest contribution to total fiscal revenue with an average of 71 percent of total tax collections in the last two decades. (Figure 1.3) Figure 1.3:Composition of Tax Revenue Source: LE&F based on data from ECLAC and the World Bank Indirect tax revenues depend mainly on taxes associated with consumption. Within excise a large proportion is originated in the collection of levies applied to the consumption of goods and services; these taxes represent 4.7 percent of GDP. The high weight of these taxes is the result of 6 a tax reform carried out in 1991, the effect of which is reflected starting 1992, reaching in 2010 80 percent participation within total tax revenues, a figure that was equivalent, that same year, to 12.3 percent of GDP. This level of participation clearly shows a huge dependence of tax revenues on taxes associated with domestic private spending. This situation can be modified in the short term with the entry into force of a recent tax reform which introduced beginning in 2013 an income tax. (Figure 1.4 and 1.5) Figure 1.4: Composition of indirect taxes Source: LE&F based on data from ECLAC and the World Bank Figure 1.5: Indirect Taxes as percent of GDP Source: LE&F based on data from ECLAC and the World Bank 7 Primary spending accounts for near 13 percent of GDP while interest expenditure has fallen from 1.5 percent to 0.5 percent of GDP. Primary fiscal expenditure represents on average, from 1990 to 2012, around 13 percent of GDP, reaching its maximum level in the years 2000 and 2009 with shares above 16 percent of GDP. Meanwhile, interest spending on these two decades has represented, on average, around 1percent of GDP, exhibiting its maximum level in 2002 with a 1.5 percent of GDP; however, since then interest expenditure has fallen reaching only 0.5 percent in the 2010 fiscal year. This favorable evolution is explained both by the reduction of the interest rate applicable to the Paraguayan debt as well as by the reduction of the public debt as a percentage of GDP. (Figure 1.6) Figure 1.6: Fiscal Spending as percent of GDP Source: LE&F based on data from ECLAC and the World Bank Consumption of goods and services, together with wage payments, represent the bulk of public expenditure. The breakdown of government spending shows that the consumables item, including disbursements for goods, services and wages, constitute the bulk of the central government expenditure: spending on consumables is up to 9.3 percent of GDP, followed by expenditure on social benefits with 3.9 percent of GDP. (Figure 1.7) Figure 1. 7: Primary Fiscal Expenditure as percent of GDP Source: LE&F based on data from ECLAC and the World Bank 8 Fiscal Results Since 2003 to date the public sector of the Paraguayan economy has consistently presented primary surpluses. Between the years 1996 to 2002 Paraguay exhibited, continuously throughout the period, primary deficits; in 2000, as a result of a deep banking crisis that had started in 1995, the country experienced its highest primary deficit: the equivalent of 3.5 percent of GDP. Since 2004, public finances have shown surplus balances, both at the primary and global level, even during the period of 2008 to 2009 when the country faced a severe drought; despite this shock, public finances resisted and maintained a positive balance, while interest payments remained stable throughout the period. (Figure 1.8) Figure 1. 8: Public Sector Balance (percent of GDP) Source: LE&F based on data from ECLAC and the World Bank Public Debt Paraguay’s public debt has declined from a 45 percent of GDP in 2003 to less than 20 percent in 2010. During the last two decades public debt has fluctuated between 15 percent and 45 percent of GDP. The crisis that took place in the years 1995 and 2002 should be considered when analyzing the level of public indebtedness in Paraguay; in both cases, the impact on the banking and financial system caused the public debt to raise considerably, from levels under 20 percent of GDP in 1994 to over 45 percent of GDP in 2002. Faced with this scenario the tax reform in 2003/04 had a positive impact on the fiscal result; the public sector improved its solvency, and the level of public indebtedness fell reaching in 2010 a level close to 15 percent of GDP. (Figure 1. 9) 9 Figure 1. 9: Public Debt (as percent of GDP) Source: LE&F based on data from ECLAC and the World Bank Increases in the level of indebtedness come along with primary budget deficits. When reviewing the evolution experienced by the public sector borrowing it is clear that the biggest fluctuations were generated between the years 1997 to 2002, period that matches the financial crisis experienced by Paraguay and that were also reflected in the primary deficit observed in the same period. (Figure 1.10) Figure 1.10: Changes in Debt and Primary Balance 15% 10% 5% 0% -5% -10% -15% Debt changes Primary balance Residue -20% Source: LE&F based on data from ECLAC and the World Bank 10 Paraguay’s Economic Growth2 Fiscal Real GDP and its Components3 At the beginning of the new millennium the Asian crisis and the crisis in Argentina generated direct contagion effects on the Paraguayan economy, either through international trade or through the financial system. In particular, the crisis experienced by the Argentinian banking and financial system had a strong impact in Paraguay’s financial system. During the period from 1998 to 2002 the Paraguayan economy presented an average GDP growth rate of -0.4 percent per annum, well below the 3 percent averaged between the years 1990 to 2010. (Figure 1.11) Figure 1.11: Real GDP* (USD ‘000 of 1994) *Until 2012, GDP figures do not include the binational power plants at Itaipu and Yacyreta. Source: LE&F based on data from ECLAC and the World Bank Since 2000 the Paraguayan economy has shown a positive growth rate, except for the year 2009 as a result of a severe drought. In the eight years following the Argentinian crisis and until 2010, the economy has displayed sustained growth at average rates of 4.9 percent per year, with the exception of 2009 when, as a result of a severe drought, the economy exhibited a contraction close to 4 percent. (Figure 1.12) 2 All the information of the public sector considers the Central Government. 3 The information used in this study considers ECLAC as its main source of information. Accordingly, the bi- national hydroelectric plants are included in the estimate for Paraguay’s GDP. Source: http://www.lanacion.com.py/articulo/42901-bcp-corrige-a-la-baja-el-crecimiento-del-2010-al-incluir-a- binacionales.html 11 Figure 1.12: Real GDP - percent changes Source: LE&F based on data from ECLAC and the World Bank The service sector contributes with 60 percent of Paraguay’s GDP. When the Paraguayan GDP is broken-down a bias toward the service or tertiary sector can be observed; it represents near 60 percent of GDP. The secondary sector, mainly constituted by the industrial segment, is responsible for a 22.3 percent of total activity in Paraguay; the remaining 18.2 percent corresponds to the primary sector, particularly related to agriculture. (Figure 1.13) Figure 1. 13: Composition of GDP by Sector of Origin Source: LE&F based on data from ECLAC and the World Bank Real GDP Volatility The volatility of the service sector is slightly lower than the one displayed by total GDP. The coefficient of variation4, for total GDP and for the main sectors of origin, is used in order to measure the GDP variability; the procedure was applied to the original series, as well as to the 4 Ratio between the standard deviation of a variable and its average; allows to compare volatility. 12 series depicting the cycle; the latter were obtained by filtering in the original series so as to obtain trends and then calculate gaps around the trend movements5. The primary sector presents the highest volatility with a coefficient close to 0.25, over the 0.15 of total GDP; the primary sector concentrates the main export products. The secondary sector, the least volatile segment, presents a 0.06 indicator, while the coefficient associated with the service sector shows a value of 0.14, similar to total GDP. (Figure 1.14) Figure 1.14: Volatility of GDP by sector 0,3 ORIGINAL SERIES HP FILTER SERIES 0,25 CF FILTER SERIES BW FILTER SERIES 0,2 0,15 0,1 0,05 0 GDP AGRICULTURE INDUSTRY SEDRVICES Source: LE&F based on data from ECLAC and the World Bank The volatility of the agricultural sector depends directly on the international prices of export products. The volatility of the agricultural activity can be explained by its dependence on the cycles of international prices, such as those of soybean and meat; the prices of these exports display a higher volatility than GDP in its original series, while the cyclical series corrected by the filters exhibit a lower variability than the 0.15 corresponding to GDP. (Figure 1.15) In addition to international prices, climatic factors also contribute to this volatility. 5 The following filters were used: Hodrick and Prescott, Christiano and Fitzgerald filters, and the Butterworth filter. The cyclical series, or gaps, were obtained by removing from the original series the trends calculated with the application of these filters. 13 Figure 1.15:Volatility of GDP and of Export Products 0,3 ORIGINAL SERIES 0,25 HP FILTER SERIES CF FILTER SERIES 0,2 BW FILTER SERIES 0,15 0,1 0,05 0 SOY MEAT GDP Source: LE&F based on data from ECLAC and the World Bank Volatility of fiscal revenues During the period under analysis fiscal revenues appear to be slightly more volatile than fiscal spending. Nevertheless, they present less volatility than the majority of individual tax revenues, except for the case of taxes associated with foreign trade. The components of indirect taxes, such as general taxes on goods and services, specific taxes and even the taxes associated with international trade, present a volatility above that of total GDP. This probably reflects that the volatility of domestic demand, and, in particular of consumption, is greater than the GDP volatility, so, changes in the level of economic activity would cause more pronounced variations in the collection of these taxes in comparison to other sources of fiscal revenue. (Figure 1.16, 1.17 and 1.18) 14 Figure 1.16: Tax Revenue Volatility 1,4 1,2 ORIGINAL SERIES 1 HP FILTER SERIES CF FILTER SERIES BW FILTER SERIES 0,8 0,6 0,4 0,2 0 DIRECT TAX REVENUES INDIRECT TAX REVENUES OTHER TAX REVENUE SOCIAL SECURITY Source: LE&F based on data from ECLAC and the World Bank Figure 1.17: Direct Taxes Volatility 2 1,8 1,6 ORIGINAL SERIES 1,4 HP FILTER SERIES CF FILTER SERIES 1,2 BW FILTER SERIES 1 0,8 0,6 0,4 0,2 0 REVENUE, INCOME AND CAPITAL CORPORATIONS AND OTHER PROPERTY GAINS BUSINESS Source: LE&F based on data from ECLAC and the World Bank 15 Figure 1.18: Indirect Taxes Volatility 1,4 1,2 ORIGINAL SERIES 1 HP FILTER SERIES CF FILTER SERIES BW FILTER SERIES 0,8 0,6 0,4 0,2 0 GENERAL GOODS AND SPECIFIC ON GOODS AND TRADE AND INTERNATIONAL OTHER TAXES SERVICES SERVICES TRANSACTIONS Source: LE&F based on data from ECLAC and the World Bank Fiscal spending volatility The item that presents the greatest volatility within public expenditure corresponds to interest spending. Interest expenditure shows more volatility than primary spending. Considering the components of primary expenditure, spending on goods, services and wages as well as capital spending, all exhibit a similar volatility to social benefits, item that in addition bears a low incidence within the budget of the country. The volatility of interest expenses could be reflecting the fluctuations of the exchange rate and of international interest rates to which the Paraguayan debt is subject. On the other hand, capital expenditure volatility is residual, i.e. it may reflect the efforts to adjust the fiscal balance to available resources. (Figure 1.19, 1.20 and 1.21) Figure 1.19: Volatility of revenues, public expenditure and GDP 0,45 ORIGINAL SERIES 0,4 HP FILTER SERIES CF FILTER SERIES 0,35 BW FILTER SERIES 0,3 0,25 0,2 0,15 0,1 0,05 0 TOTAL EXPEDITURE TOTAL TAX REVENUE GDP Source: LE&F based on data from ECLAC and the World Bank 16 Figure 1.20: Volatility of government spending 0,40 ORIGINAL SERIES HP FILTER SERIES 0,35 CF FILTER SERIES BW FILTER SERIES 0,30 0,25 0,20 0,15 0,10 0,05 0,00 PRIMARY EXPEDITURE PAYMENT OF INTEREST Source: LE&F based on data from ECLAC and the World Bank Figure 1.21: Primary spending volatility 0,50 ORIGINAL SERIES HP FILTER SERIES CF FILTER SERIES BW FILTER SERIES 0,45 0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 PRIMARY EXPEDITURE PAYMENT OF INTEREST GOODS, SERVICES AND SOCIAL SECURITY WAGES Source: LE&F based on data from ECLAC and the World Bank Fiscal spending volatility The volatility of the components of domestic demand is greater than the volatility of GDP. The different components of domestic demand are several times more volatile than total GDP. The volatility of personal and government consumption stands out since both exceed the volatility of capital formation and amount to around six times the GDP volatility (Figure 1.22) 17 Figure 1.22: Volatility of the GDP components 0,80 ORIGINAL SERIES HP FILTER SERIES CF FILTER SERIES BW FILTER SERIES 0,70 0,60 0,50 0,40 0,30 0,20 0,10 0,00 GDP GOVERMENT CONSUMPTION PRIVATE CONSUMPTION INVESTMENT EXPORTS IMPORTS Source: LE&F based on data from ECLAC and the World Bank 2. Estimation and Projection of the Structural Fiscal balance This chapter covers the estimations and projections that will be used to derive the structural fiscal rule for Paraguay. The definition of a structural balance requires an estimate of structural or trend fiscal revenue and of the underlying factors to that trend. Possibly among them are the gross domestic products, the price of some export commodities, and weather related variables such as rainfall. The relation of fiscal revenue with the price of one or more export commodities was considered since the Paraguayan economy, like other Latin-American countries, shows a high level of specialization in commodity exports. Actually, Paraguay exports a variety of primary agricultural and livestock commodities, including, among others, soy beans, cereals and meat, and, its degree of specialization is relatively high when the top ten individual export products are considered. Indeed, comparing Paraguay’s degree of export specialization with other Latin-American countries, it is only surpassed by Ecuador, Bolivia, and Venezuela (though this country was not included in the chart). In spite of that, Paraguay does not show any specific or individual export product playing a significant role in revenue generation, such as is the case of oil in Venezuela, Ecuador and Trinidad and Tobago, soy beans and other grains in Argentina, or copper in Chile and Peru. (Figure 1.23)6. The characteristics of the Paraguayan tax system and its heavy reliance on indirect taxes, rather than export diversification is the underlying cause for 6 ECLAC, the source of information, does not include data on Venezuela 18 the non-existence of a direct link between export commodity prices and tax revenue. There could be an indirect link that operates through GDP, but that cannot be peaked up using the single equation cointegration methodology of this work. That is, higher commodity prices increase GDP and domestic private spending and through that, more indirect taxes are collected. Figure 1.23: Participation of top 10 export products 90% 80% 70% 60% 50% 40% 30% Argentina Bolivia Brasil Chile Colombia Ecuador Perú Uruguay PROMEDIO Paraguay Source: LE&F based on data from ECLAC and the World Bank An estimation of a stable and balanced relationship between both, fiscal revenue and its components, with the trend GDP and its gap, is a requisite for the estimation of a structural balance. Therefore, in order to identify the relevance of each component the elasticities of total and aggregate fiscal revenue and expenditure with respect to the economy’s actual and trend GDP were estimated; in particular, the elasticity of fiscal revenue with respect to GDP plays an essential role in the determination of the value of structural income and of the structural balance itself7. The elasticity of fiscal revenue to GDP was estimated through co-integration and error- correction models.8 For the estimation of fiscal revenues a function of GDP, a cointegration relationship was estimated under the hypothesis of the existence of a long-term relationship between revenues and GDP; in addition, some dummy variables were included in order to capture changes in the country’s tax regulation as well as other variables that may affect the revenue cycle, including, among them, prices of export products. In addition, estimates were made for the relationship between total and disaggregated fiscal revenue and expenditure with gross domestic product identifying the relevance of each component. These estimates were made on the basis of cointegration models, however, error correction models of the first differences were also estimated so as to illustrate on the short-term relationship of these variables. The 7 At a disaggregated level the relations between tax revenue and the GDP gap did not show statistical significance 8 The current estimations do not take into account the potential structural break caused by the 2004 tax reform. While the tax-to-GDP ratio increased moderately after the 2004 tax reform this increase occurred towards the end of the sample and was small in size. As a result, it was not possible to recognize this change in the estimations and it is possible the increase introduces some upward bias in the estimated elasticity. However, the authors believe that the bias would be small and would not significantly change the elasticity estimated for the entire sample period. 19 results indicate a significant relationship between the aggregate tax revenue with GDP, but not at the disaggregated level of each of the fiscal revenue items with GDP.9 Not only the price of export products but also the hydrography of the basins of the rivers would be relevant in the determination of tax revenue cycles. According to some point of views, the prices of the main export commodities, like soy beans and beef meat, could be considered to be relevant variables in the determination of cyclical and trend tax revenue and hence for the definition of a structural fiscal policy for Paraguay. In addition, fiscal revenue should also be function of the country’s rainfall because of its impact on the volume of electricity generation (in particular in the Parana basin) and on agricultural exports; both have direct effects on the amount of royalties and taxes collected by the State. However, prices of major exports did not show a statistically significant relationship with fiscal revenue, and, a confirmation and quantification of its relationship with rainfall was not possible due to data limitations. Estimation of the trend or structural The trend GDP was estimated on the basis of a Cobb Douglas aggregate production function. Real GDP (y) can be analytically split up in a permanent or trend component (Yp), estimated through the production function under a “normal” productivity and degree of use of resources, and a transitory component (Ytran) that represents productivity deviations and above or below normal use of resources. The logarithmic difference of GDP and its trend, also called the GDP gap, represents the economic cycle; it has an average value equal to zero, is positive in periods of expansion or high activity and negative during downturns and recessions. A Cobb-Douglas aggregate production function, with constant returns to scale and under the assumption of a 0.35 value for α, was used for the estimation of trend GDP. In the aggregate production function A is the total factor productivity, K is the capital and N the actual workforce, while α is a parameter that represents the share of capital within total income. The value of α was obtained from a figure estimated by de IMF (0.35)10 since its value could not be derived from the national accounts. The superscript P is used to represent the permanent trend or permanent component of the variables: total factor productivity, the stock of capital and employment. (Figure 1.24) ( ) ( ) ( ) ( ) ( ) 9 Only tax originated revenues were included; tax revenues represent 71 percent of the central administration income 10 International Monetary Fund (2011): Paraguay consultation - Article IV corresponding to 2011. IMF Country Report No. 11/239. 20 Figure 1.24: GDP Growth 15% Efective GDP Growth GDP Growth (HP Filter) GDP Growth PF GDP Growth (CF Fiter) GDP Growth (BW Filter) 10% 5% 0% -5% Source: LE&F based on data from ECLAC and the World Bank Actual employment, defined as a quality adjusted labor index, is calculated as the product of the number of workers (N #), the number of hours worked on average by each worker (h), and an index of labor qualification represented by the average years of schooling. The series on the number of workers N # was obtained from ECLAC for the period 1970-2011. Unfortunately, the series for the average number of hours worked and the average years of schooling of the labor force are not available for Paraguay; therefore, we have assumed a value of 1 for all observations of both variables waiting to eventually obtain information in this regard. (Figure 1.25) Figure 1.25: Labor force Growth 8% Labor Force Growth (HP Filter) Labor Force Growth 7% 6% 5% 4% 3% 2% 1% 0% -1% -2% -3% Source: LE&F based on data from ECLAC and the World Bank 21 The unemployment rate, effective and natural, is used to represent the degree of utilization of resources. The natural unemployment rate is represented as the trend of the original unemployment rate series. The unemployment rate (U) is defined as the difference between the labor force (L), all those willing and ready to participate in the labor market, and the number of people actually employed N #, expressed as a percentage of labor force. On the other hand, the trend or natural rate of unemployment is obtained through the application of an HP filter on the unemployment rate series11. ( ) The growth rate of employment in Paraguay is high by international standards, with a trend rate of over 3.5 percent per year. The value of the trend of actual employment is obtained by applying a Hodrick-Prescott filter (HP filter) to the labor force series, and multiplying the result by one minus the natural rate of unemployment. In recent years, with the recovery of the debt crisis, the growth rate of trend employment has increased peaking slightly above 3.5 percent per year. Given demographic projections, the trend growth rate of employment should decrease gradually, falling below 3 percent by the end of the next decade. ( ) ( ) The Solow labor utilization index is obtained from the relation between the actual rate of unemployment and the natural rate of unemployment. The Solow index, which measures the intensity of use of the labor factor, is defined on the basis of the relationship between the actual unemployment rate and the natural unemployment rate, so that the value of the index is 1 when actual and natural unemployment rates are equal, greater than one when the actual unemployment rate is below the natural rate and less than one when the actual unemployment rate is above the natural rate. By subtracting 1 to the Solow utilization index an alternative measure of the output gap can be obtained. (Figure 1. 26) ( ) ( ) 11 In order to obtain the trend of the unemployment rate based on annual data the OECD recommends a value of lambda = 100, where lambda is the parameter of the HP filter 22 Figure 1.26: Unemployment and Solow Resource Use Index 16% Natural Unemployment Rate 6% Unemployment Rate "Solow index - 1" 14% 4% 12% 2% 10% 0% 8% -2% 6% -4% 4% -6% 2% 0% -8% Source: LE&F based on data from ECLAC and the World Bank The growth rate of gross fixed capital formation is estimated at 3.2 percent for the coming years on the basis of an average medium-term investment rate and the GDP growth rate projected by the IMF-WEO. The estimation of the stock of capital was based on the capital accumulation equation, where K corresponds to the stock of capital, I is investment defined as gross fixed capital formation, and δ is the depreciation rate. The gross capital formation series (investment in the national accounts) was obtained from ECLAC, and the depreciation rate, estimated at 8 percent per year, was obtained from previous studies12. According to the initial value of the stock of capital (1970), the capital per worker ratio amounted to 6.3 in 1970. Investment was projected using the last five years average of the investment to GDP ratio and the IMF-WEO projections for Paraguayan GDP through 2020. Furthermore, on that basis the estimated trend growth rate for gross fixed capital formation is of 3.2 percent, that is, below the growth rate of employment, thus, a continuous slight decline for the capital-labor ratio in the coming years can be inferred. (Figure 1.27) 12 Fernández Valdovinos, C. and A. Monge Naranjo, 2004, "Economic Growth in Paraguay", Economic and Social Study Series, (Washington: Inter-American Development Bank) 23 Figure 1.27: Growth of investment and of the capital stock 50% 40% Growth of Capital Gross Fixed Capital Formation Growth 30% 20% 10% 0% -10% -20% -30% -40% Source: LE&F based on data from ECLAC and the World Bank The estimated stock of capital actually used (K) was obtained by correcting the trend stock of capital Kp with the Solow’s utilization index (S), assuming equal degree of utilization of labor and capital. The actual capital stock was used to calculate the residual value corresponding to total factor productivity, which has a trend growth rate of 1.7 percent per year. The total factor productivity (PTF = A) is obtained as the residual of the production function, subtracting from actual GDP the contribution of capital and of actual labor, duly adjusted by their respective intensity of use. In order to obtain the total factor productivity trend the Hodrick and Prescott filter is applied to the A series. Consequently, the resulting trend PTF shows an annual growth rate around 1.7 percent. In the projection this rate is expected to remain at the historical rate (Figure 1. 28) ( ) ( ) ( ) where: A: total factor productivity 24 Ap: trend total factor productivity K: stock of capital N: actual workforce δ: depreciation rate y: real GDP α: share of capital in total income13 Figure 1.28: Growth of total factor productivity 15,0% 10,0% TFP Growth Potential TFP Growth 5,0% 0,0% -5,0% -10,0% Source: LE&F based on data from ECLAC and the World Bank The trend of GDP growth for the years 2015-2020 is estimated between 4.5 percent and 4.8 percent, slightly above Paraguay´s historical average (4.3 percent). The results indicate that in recent years the Paraguayan trend GDP growth has accelerated, reaching a rate of near 4.8 percent per year in the period 2008-2011. As other experiences indicate, potential growth should decline to more "normal" rates after the post-crisis period as the recovery process is fully completed. However, given the IMF_WEO forecast and the assumptions used in this article, our forecast indicates a convergence of GDP trend growth to rates of 4.5 percent to 4.8 percent for 2015-2020, well over the average or trend GDP growth of late last century. When Paraguay decides to really implement a structural balance policy the assumptions used for the projections of growth, labor force and gross capital investment should be subject to extensive discussion; the final design of such a policy should be the result of the opinions and agreements between a diversity of experts on the Paraguayan economy and should be re-examined annually or at least every five years. Several filters were used to estimate series for trend GDP alternatives to that of the aggregate production function presented above; however, the results were quite similar, 13 Value estimated by the IMF, where α is equal to 0.35. International Monetary Fund (2011). Paraguay -Article IV Consultation - Corresponding to 2011. IMF Country Report No. 11/239 25 only much smoother. For comparison purposes trend GDP was estimated by directly applying the HP filter to the actual GDP series and then comparing it with the one obtained with the production function method explained in previous paragraphs. In addition, the Butterworth Christiano-Fiztgerald filters were used also applying them directly to the GDP series. In terms of levels and trajectories of the trend GDP, the use of different filters shows similar results than those obtained with the production function. In any case, when analyzing the first differences of the calculations of the trend GDP it can be noticed that the HP filter applied directly results in a much smoother series than that obtained from the production function. In fact, far more profound cycles and with much more pronounced annual variations can be distinguished when using the production function model. (Figures 1.29 and 1.30) Figure 1.29: GDP Gap 25% GDP Gap PF GDP Gap (HP Filter) 20% GDP gap (CF Filter) GDP gap (BW Filter) 15% 10% 5% 0% -5% -10% -15% Source: LE&F based on data from ECLAC and the World Bank Figure 1.30: Unemployment rate 16% 14% Unemployment rate Natural unemployment rate (HP) 12% 10% 8% 6% 4% 2% 0% Source: LE&F based on data from ECLAC and the World Bank 26 Elasticities with respect to GDP The relationship between total government revenue and GDP was estimated. An estimation of a stable relationship between both, fiscal revenue and its components, with the trend GDP and its gap is a requisite for the estimation of a structural fiscal balance. The elasticity of fiscal revenue with respect to GDP will be pivotal in the assessment of the value of structural revenues and of the structural balance. That is, transitory revenue can be estimated as the output gap multiplied by the β1 elasticity. Then, the estimated regression is: ( ) Several models were estimated and were dismissed since they lacked significance, including a breakdown by type of public revenue and the inclusion of other explanatory variables. The model in which total revenues are a function of GDP was chosen because the alternative econometric models for the estimation of total revenue did not result in statistically significant elasticities. Among them, a model in which total revenues were a function of GDP along with prices of meat and soy —two major export products in Paraguay— was considered; it was not possible to find a statistically significant response of fiscal revenue to the prices of major exports, perhaps due to the limited available sample, or because the effect of commodity prices on fiscal revenue act indirectly through GDP. No data are available in order to intend a correlation of fiscal revenue with rainfall in the Paraná Basin. In addition, we attempted to estimate revenue through disaggregated taxes (taxes on goods and services, on businesses, foreign trade, etc.), but all these models were dismissed since their parameters were found to have no statistical significance. So, perhaps due to limitations of the available data estimates of the elasticities of fiscal revenue with respect to real GDP, its trend and its gap were the result of the regression of the aggregates and not of the elasticities of individual taxes or individual income sources. Cointegration and Error Correction Model The order of integration of the relevant variables was verified, finding out that all of them are integrated of order one, while the existence of cointegration between total revenues and GDP could not be rejected by the data. A statistically significant relationship between total fiscal revenue and the GDP of the Paraguayan economy was established, with both variables expressed in logarithmic terms. In addition, as evidenced by the results, both variables appear to be integrated of order 114. The elasticity of total revenue with respect to GDP is quite high (1.9), reflecting in part the growing trend of government revenue as a percentage of GDP, and the high incidence of indirect taxes, more associated to domestic spending than to income. As it is well known, the cyclical volatility of domestic demand is much higher than that of GDP. Following this methodology, the first regression estimated reflects the long term behavior of the variables: Where: 14 The results of unit root tests can be found in Appendix 6.1.6 27 ln( I _ Tot )t = Natural logarithm of Total Revenue in period t. ln( PIB) t = Natural logarithm of GDP in period t. ln( I _ Tot )t  0  1 ln( PIB)t  t ln( I _ Tot ) t  15.6452  1.844291ln( PIB) t Table 1.1: Ordinary Least Squares Estimation for Total Fiscal Revenues In Logs, 1990-2010 (T = 21) HAC Standard Deviations with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -15.64522 1.466162 -10.67087 0 Ln(PIB) 1.844291 0.088857 20.75583 0 Media de la vble. dep. 14.78506 D.T. de la vble. dep. 0.274289 Suma de cuad. residuos 0.063559 Suma de cuad. regresión 0.057838 R-cuadrado 0.957759 R-cuadrado corregido 0.955536 Estadístico F 430.8046 Prob (Estadístico F) 0 Log-verosimilitud 31.10554 Criterio de Akaike -2.771956 Criterio de Schwarz -2.672478 Crit. de Hannan-Quinn -2.750367 Source: LE&F based on data from ECLAC and the World Bank Then, given the cointegration between total revenue and GDP, an Error Correction Model was estimated in order to capture the short-term relationship between the variables. For such purpose, the methodology proposed by Engle - Granger (1987) 15 was followed. Cointegration can be verified given the stationary nature of the errors estimated in the regression above. Since the error residuals are found to be stationary, that is, the null hypothesis of unit root for the residuals is rejected at a 95 percent confidence, a cointegration relationship was established. The existence of a long-term relationship between total revenues and trend GDP can be inferred since under the null hypothesis there is cointegration between these two variables. However, in the short term the existence of certain imbalances that are captured by the estimated errors, may divert this relationship.16. ˆ t 1 t  ln( I _ Tot ) t   0  1 ln( PIB) t   15 Engle and Granger, 1987, Cointegration and Error Correction: Representation, Estimation and Testing, Econometrics, 55 251-276 16 Results in Annex 6.1.7 28 Table 1.2: Ordinary Least Squares Estimation First Difference Logs of Total Fiscal Revenues 1991-2010 (T = 20) HAC Standard Deviations with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -15.64522 1.466162 -10.67087 0 D Ln(PIB) 1.034366 0.298947 3.460031 0.003 U_hat (-1) -0.604082 0.214789 -2.81245 0.012 Media de la vble. dep. 0.04979 D.T. de la vble. dep. 0.06468 Suma de cuad. residuos 0.040638 Suma de cuad. regresión 0.048893 R-cuadrado 0.488743 R-cuadrado corregido 0.428595 Estadístico F 8.125692 Prob (Estadístico F) 0.003338 Log-verosimilitud 33.60903 Criterio de Akaike -3.060903 Criterio de Schwarz -2.911543 Crit. de Hannan-Quinn -3.031747 Source: LE&F based on data from ECLAC and the World Bank According to the error correction model, the short-term elasticity of the public sector revenue with respect to GDP is 1.03, while in the long-term its value is close to 1.8. The effect of the growth over time of the size of the government relative to GDP is included in the long-term elasticity. This elasticity will be used to estimate the public sector structural revenues as a function of the actual public revenues and the GDP gap. The short-term elasticity captured in the error correction model is much lower.  ln( I _ Tot) t  -15.64522  1.034366  ln( PIB) t - 0.604082  ˆ t 1 Analysis of Cycles in Public Revenue and Expenditure We sought to establish the behavior of Paraguay’s fiscal expenditure and revenue with respect to its GDP cycle. To determine the behavior of fiscal revenue and expenditure in this regard both variables were correlated with trend GDP and the GDP gap; the results indicate that fiscal revenues generally show a pro-cyclical behavior, while fiscal expenditures do not present a significant relationship with the GDP cycle. In order to analyze the behavior of Total Public Expenditure, the model to be estimated corresponds to: ln(Gasto) t   0  1 ln( BrechaPIB ) t   2 ln( PIB _ tend ) t  t where: ln(Gasto) t = Natural logarithm of Public Expenditure in period t. ln( BrechaPIB )t = Natural logarithm of GDP gap in period t. ln( PIB _ tend )t = Natural logarithm of Trend GDP in period t. 29 From the results obtained the model was established as: ln(Gasto) t  -14.2791 - 0.86946 ln( BrechaPIB ) t  1.761336 ln( PIB _ tend ) t Table 1.3: Ordinary Least Squares Estimation Log of Total Fiscal Expenditure 1990-2010 (T = 21) HAC Standard Deviations, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -14.2791 2.914757 -4.898898 0.0001 BrechaPIB -0.86946 0.787665 -1.103845 0.2842 Ln(PIB_tend) 1.761336 0.176639 9.971361 0 Media de la vble. dep. 14.79179 D.T. de la vble. dep. 0.287709 Suma de cuad. Residuos 0.237275 Suma de cuad. regresión 0.114813 R-cuadrado 0.856677 R-cuadrado corregido 0.840753 Estadístico F 53.79539 Prob (Estadístico F) 0 Log-verosimilitud 17.27439 Criterio de Akaike -1.359465 Criterio de Schwarz -1.210248 Crit. de Hannan-Quinn -1.327081 Source: LE&F based on data from ECLAC The results indicate that government spending has a cyclically-neutral behavior. It can be concluded that the constant coefficient and that of the trend GDP are significant at a 99 percent confidence. The GDP gap coefficient appears to be not significantly different from zero; thus, it can be inferred a non-statistically significant reaction of public spending to the GDP cycle. Fiscal current expenditure does not seem to present a statistically significant relationship with the GDP cycle, while the constant and trend GDP are significant at a 99 percent confidence. That is, current expenditure shows a cyclically-neutral behavior. In the case of Total Current Expenditure, the following model was estimated and its results are listed in Table 1.4. ln(Gasto _ Corriente)t  0  1 ln( BrechaPIB )t   2 ln( PIB _ tend )t  t where: ln(Gasto _ Corriente) t = Natural logarithm of current fiscal expenditure in period t. ln( Brecha _ PIB)t = Natural logarithm of GDP gap in period t. ln( PIB _ tend )t = Natural logarithm of Trend GDP in period t. 30 Table 1.4: Ordinary Least Squares Estimation Log of Current Fiscal Expenditure 1990-2010 (T = 21) HAC Standard Deviations, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -17.50443 2.238915 -7.818262 0 BrechaPIB -0.784506 0.60503 -1.29664 0.2111 Ln(PIB_tend) 1.663352 0.135682 12.25917 0 Media de la vble. dep. 9.949093 D.T. de la vble. dep. 0.264719 Suma de cuad. residuos 0.139998 Suma de cuad. regresión 0.088191 R-cuadrado 0.90011 R-cuadrado corregido 0.889011 Estadístico F 81.09893 Prob (Estadístico F) 0 Log-verosimilitud 22.81408 Criterio de Akaike -1.887055 Criterio de Schwarz -1.737838 Crit. de Hannan-Quinn -1.854671 Source: LE&F based on data from ECLAC and the World Bank Similar results are observed in the case of the central government capital expenditure which also presents a cyclically-neutral behavior. As in the previous case, the model that estimates capital expenditure indicates that the constant and the trend GDP coefficients are significant at a 99 percent confidence level, while the output gap appears to be non-significant. The model estimated for capital expenditure is presented below, and the detailed results can be found in Table 1.5. ln(Gasto _ Capital )t  0  1 ln( BrechaPIB )t   2 ln( PIB _ tend )t  t where: ln(Gasto _ Capital ) t = Natural logarithm of Capital expenditure in period t. ln( Brecha _ PIB)t = Natural logarithm of GDP Gap in period t. ln( PIB _ tend )t = Natural logarithm of Trend GDP in period t. By contrast, in the case of the public sector revenue clear traces of significant pro-cyclical behavior were found. Total revenues have a pro-cyclical behavior with a positive and significant effect of the GDP gap on them. The results indicate that the constant and the trend GDP coefficients are significantly different from zero at a 99 percent confidence. The GDP gap coefficient is also significantly different from zero at 95 percent confidence and the value of the parameter is positive. Then, the output gap has a positive effect on total government revenue, which shows a pro-cyclical behavior. In the case of Total Revenue, the following relationship was estimated and its results are listed in Table 1.6. ln( ITOT ) t   0  1 ln( Brecha _ PIB) t   2 ln( PIB _ tend ) t  t where: ln( ITOT )t = Natural logarithm of total revenues in period t. 31 ln( Brecha _ PIB)t = Natural logarithm of GDP gap in period t. ln( PIB _ tend )t = Natural logarithm of Trend GDP in period t. Table 1.5: OLS Estimation Logarithm Capital Expenditure, 1990-2010 (T = 21) HAC Standard Deviations, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -28.87245 6.922551 -4.170782 0.0006 BrechaPIB -1.174331 1.870705 -0.627748 0.5381 Ln(PIB_tend) 2.271418 0.419519 5.414339 0 Media de la vble. dep. 8.617532 D.T. de la vble. dep. 0.430334 Suma de cuad. residuos 1.338382 Suma de cuad. regresión 0.27268 R-cuadrado 0.638641 R-cuadrado corregido 0.59849 Estadístico F 15.90596 Prob (Estadístico F) 0.000105 Log-verosimilitud -0.890565 Criterio de Akaike 0.37053 Criterio de Schwarz 0.519748 Crit. de Hannan-Quinn 0.402914 Source: LE&F based on data from ECLAC Table 1.6: OLS Estimation Total Revenues Logarithm 1990-2010 (T = 21) HAC Standard Deviations, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -15.85658 1.198032 -13.23553 0 BrechaPIB 0.819889 0.323748 2.532487 0.0208 Ln(PIB_tend) 1.856878 0.072603 25.57583 0 Media de la vble. dep. 14.78506 D.T. de la vble. dep. 0.274289 Suma de cuad. residuos 0.040085 Suma de cuad. Regresión 0.047191 R-cuadrado 0.97336 R-cuadrado corregido 0.9704 Estadístico F 328.8356 Prob (Estadístico F) 0 Log-verosimilitud 35.94563 Criterio de Akaike -3.137679 Criterio de Schwarz -2.988461 Crit. de Hannan-Quinn -3.105295 Source: LE&F based on data from ECLAC Another of the variables analyzed corresponds to Total Tax Revenue which also shows a pro-cyclical behavior; that is, all estimated coefficients were significantly different from zero at a 99 percent confidence. Then, the output gap has a positive effect on total tax revenues that, therefore, exhibit a pro-cyclical behavior. The model used is presented below, and its results are depicted in Table 1.7. 32 ln( IT )t  0  1 ln( BrechaPIB )t  2 ln( PIB _ tend )t  t where: ln( IT ) t = Natural logarithm of Tax Revenue in period t. ln( Brecha _ PIB)t = Natural logarithm of GDP gap in period t. ln( PIB _ tend )t = Natural logarithm of Trend GDP in period t. The model provides the following results: Table 1.7: OLS Estimation Tax Revenues Logarithm 1990-2010 (T = 21) HAC Standard Deviations, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -17.32299 1.527129 -11.3435 0 BrechaPIB 2.093586 0.412681 5.073128 0.0001 Ln(PIB_tend) 1.92555 0.092547 20.80624 0 Media de la vble. dep. 14.44743 D.T. de la vble. dep. 0.286356 Suma de cuad. residuos 0.065133 Suma de cuad. Regresión 0.060154 R-cuadrado 0.960285 R-cuadrado corregido 0.955872 Estadístico F 217.614 Prob (Estadístico F) 0 Log-verosimilitud 30.84872 Criterio de Akaike -2.652259 Criterio de Schwarz -2.503042 Crit. de Hannan-Quinn -2.619875 Source: LE&F based on data from ECLAC Estimation of a Structural Fiscal Balance The structural fiscal balance is estimated as the difference between total structural fiscal revenues —the ones that would have existed if GDP had remained at its trend level — with actual expenditures. Consequently, the structural fiscal revenue is the fiscal revenue that would be obtained under a zero GDP gap. Then, the structural fiscal balance will result from the difference between total structural fiscal revenues and total fiscal expenditures. The estimated total structural revenues can be obtained using the estimated regression for total revenue and assuming an output gap equal to zero. Thus, the structural fiscal balance is: Fiscal revenues (T) have a permanent component (Tp) and a transitory one (Ttran); both can be expressed as a percentage of GDP and in that case they are denoted in lowercase letters. On the other hand, the structural fiscal balance as a percentage of GDP (sfb) can be obtained by subtracting total expenditures (both primary (gprim) and interest spending (gint)) from structural or permanent revenue. Finally, the primary structural fiscal balance (psfb) can be obtained by subtracting interest expenses from the structural fiscal balance: 33 The actual fiscal balance, expressed as a percentage of GDP (fb), is the structural balance plus the public transitory income which can be positive or negative. In order to achieve a stable structural fiscal balance potential positive transitory revenue should be saved while possible negative transitory income is “un-saved” through a lower temporary fiscal balance or higher temporary deficit. The structural fiscal revenue and the transitory fiscal revenue are derived from the relationship obtained between total fiscal revenue and GDP. In the case of Paraguay, permanent and transitory fiscal revenue derive from the economic cycle which leads to fluctuations in the total income of the public sector (IT). In the total revenue equation βo corresponds to the position constant, while β1 is the elasticity with respect to GDP and εt represents disturbances or factors not included in the equation that also affect total government revenues. Using this relationship, it is possible to correct fiscal revenue by eliminating the effect of the output gap on them. That is, the cyclical effect of GDP was eliminated and the trend GDP was multiplied by the previously estimated elasticity to determine the total effect. [ ] The representation of revenues as a percentage of GDP is given by: [ ] The worst fiscal result since 1990 corresponds to a structural fiscal deficit of 4.6 percent in 2000. Our estimates indicate that Paraguay’s total structural fiscal balance showed, between 1990 and 1998, a continuous surplus that reached a maximum of 4.5 percent of GDP in 1995. Afterwards the situation was reversed moving to a sustained structural deficit between 1999 and 2002, possibly due to the effects of the Asian crisis. The maximum structural fiscal deficit, around 4.6 percent of GDP, is reached in 2000. Thereafter the structural balance started to 34 improve and in 2003 turned into a surplus, condition that has continued since then despite the effects of the debt crisis in Argentina, one of Paraguay’s major trading partners. (Figure 1.31) Figure 1.31: GDP GAP and Estimated Structural Balance Source: LE&F based on data from ECLAC and the World Bank Since 1992 and given its fiscal results, Paraguay has not required new credit facilities or financing from multilateral agencies. Since 2002, public finances show progress mainly because debt repayments have exceeded the debt disbursements received during the period. This is coupled with higher economic and revenue growth which have eliminated the need for new fiscal financing. In addition, Paraguay has showed expenditure containment, which has allowed to improve the fiscal balance so as to reach structural fiscal surplus from the year 2007 to date; this, despite the drought that hit the country during the years 2008 and 2009. 3. A Structural Fiscal Rule for Paraguay This chapter presents a proposal for a structural fiscal rule that was developed on the basis of the estimates for structural fiscal balance presented in previous chapters. Several alternatives for a long-term fiscal objective and for the horizon to reach them are presented; these alternatives long term goals are the basis for the calculation of the annual fiscal target on the structural primary balance. Fiscal Projections Future fiscal revenues were projected on the basis of the GDP projections and of the error correction model that associates both variables; then, structural revenues were obtained using the estimated GDP gap and the long term elasticity of fiscal revenue to GDP. The error correction model, developed in the previous section, was used for the projection of fiscal revenue; this model reflects the short-term relationship between fiscal revenue and GDP; these 35 revenues are denominated Actual Total Revenues. Then, total projected structural revenue was obtained by correcting these results with the output gap weighted by the long term elasticity. So, when GDP is below its trend level Actual Revenues reflect this gap by standing below Structural Revenues which are not affected by the gap. Historical and projected total public revenues, both structural and actual, are shown in Charts 32 and 33. In the projected values the differences between structural and actual revenue are minimal since the projected GDP gap is very close to zero, and, the shocks affecting total revenues are projected at their expected total value, that is, zero. Figure 1.32: Historical Fiscal Revenues-(in logarithms) Source: LE&F based on data from ECLAC and the World Bank Figure 1.33: Projected Fiscal Revenues (in logarithms) Source: LE&F based on data from ECLAC and the World Bank A production function allows for the projection of the trend GDP between 2012 and 2020. Future values of trend GDP have been projected on the basis of a production function, the projected trend productivity (Ap), the trend of the stock of capital (Kp) and of employment (Np). 36 Projections for GDP were obtained from the World Economic Outlook of the International Monetary Fund (IMF-WEO). The projections for investment (I) were obtained by assuming that the rate of investment to GDP remains constant at the average of the last five years; this is the input for the projection of the stock of capital. Employment was projected assuming that the elasticity of employment to GDP remains constant at 0.55, while the labor force was projected using demographics, assuming a constant participation rate equal to the last five years average and using the official population projections. Both sources provide the projected unemployment rate. The capital accumulation equation allows for the projection of the stock of capital on the basis of the projections of gross fixed capital formation (I). The projected trend capital equals the previous period capital, minus depreciation, plus the investment of the period. ( ) The projection of trend employment requires the projection of the trend labor force and of the natural unemployment rate. The trend labor force is obtained by applying an HP filter to the historical and projected (up to 2020) labor force figures. Similarly, the trend unemployment rate is obtained with historical and projected figures. by applying an HP filter to the historical and projected unemployment rate series. ( ) ( ) Projected total factor productivity is originated on the projections of actual K, of trend capital adjusted by level of use, of actual employment and of actual GDP. Trend total factor productivity was estimated throughout 2020 on the basis of the projected total productivity series. Trend productivity can be obtained by applying an HP filter to the actual and projected productivity series. ( ) ( ) Trend GDP projections are obtained by applying to the production function the trend projections of capital, of employment and of total factor productivity. The GDP gap is projected as the logarithmic difference between effective and trend GDP. ( ) Having the elasticity with respect to trend GDP and to the gap, total projected fiscal revenue may be corrected in order to obtain structural fiscal revenues which would represent fiscal revenue under a zero gap. A scenario with an active fiscal policy, where the structural balance target is achieved, is assumed when projecting each fiscal year spending. In the following equations government primary spending (as a percentage of GDP, pg) is determined by a budget based on permanent or structural income (tP) and by a target for the structural primary balance (ps*). Interest expense (ig) is determined by the debt itself and the 37 assumed real interest rate17; total government spending (g) is obtained by adding interest expense to primary spending. Structural Fiscal Target and Long Term Fiscal Goal The structural fiscal target aims to ensure the sustainability of public debt and, consequently, of plans for fiscal spending over the medium and long run. The primary structural balance target aims to provide fiscal sustainability by meeting certain goals for the public sector net debt or its negative the net financial wealth or net worth (B). The target for the structural fiscal balance is directed at generating debt sustainability over the long-term, which can be any level of debt. In our approach we define a goal for net fiscal worth in order to go beyond debt sustainability and reduce the vulnerability to shocks by targeting a reduced level of net debt. Under this view, the target for the primary structural balance must be derived from the long-term goal regarding the public sector net debt or its negative the net worth. The actual final balance represents the change in net financial wealth (B) or minus the change in net debt (-B), we will use the former. where: i is the nominal interest rate, B is net worth, T is fiscal revenue, PG primary spending and PS the primary fiscal surplus. Thus, an equation of wealth or public net worth (B) consistent with a structural primary balance can be obtained through: π inflation, λ real GDP growth rate and r the real interest rate; where public net worth depends on its value in the previous period, the discount rate and the primary structural fiscal balance (psb). ( ) ( ) ( ) [ ] ( )( ) ( ) [ ] ( ) ( ) 17 It should be noted that only those interest corresponding to the actual paid rate are included in the interest expenses item. Nominal interest in excess of real interest is a form of prepayment of the real value of the debt and is not considered as an expense in this exercise. 38 Where b represents the public net worth as a percentage of GDP; psb the primary fiscal surplus, also in percent of GDP; r the real interest rate for public sector assets and liabilities; and lambda, the long term real growth rate of GDP. The structural surplus must be permanent. In order to define a target for the structural primary balance, it is necessary to consider a stable or permanent value for it (psb or psb*). The real interest rate and growth rate are assumed to remain unchanged so that the discount factor ( ) is also constant and the wealth equation becomes18: ( ) Where one plus the discount factor represents the ratio between one plus the real interest rate and one plus the GDP growth rate: : For the discount factor to be positive, the real interest rate must be higher than the GDP growth rate. The base equation for fiscal net worth in the long run can be derived by developing the equation for public net worth for subsequent periods: the long-term fiscal net worth (b (t + N)) is a function of the initial fiscal net worth (b (t-1 )), the permanent primary surplus (psb *) and the discount factor (). This relationship defines the fiscal target for the primary surplus based on the long-term goal for the public sector net worth and other conditions. ( ) [ ( )] ( ) [ ( ) ( ) ] ( ) ∑( ) [ ] The target for the primary structural fiscal surplus (psb*) depends on the initial fiscal net worth (bt), the long-term goal (bt+N), the time to reach this goal (N) and on the discount factor (). The specific value of the fiscal target psb* will be greater the lower the initial value of the fiscal net worth (b t-1), the higher the fiscal net worth long-term goal (bt + N), and the lower the number of periods to reach the objective (N). A particular goal for wealth as a ratio of GDP within a specified period of N years in the future bt + N results in the structural fiscal target in its most complete expression: [ ] ∑ [ ] A simple case could be to consider by setting as the long-term goal to maintain the public net worth as a percentage of GDP unchanged at its initial level b t-1. Then, the structural 18 See Croce and Juan Ramon (2003) 39 fiscal target is simpler and only depends on the initial level of public net worth and the discount rate: [ ( ) ] ∑ [ ] Simulations for the Fiscal Target and the Public Wealth Goal In order to perform simulations of future fiscal balance and public net worth scenarios, real trend GDP was assumed to grow at a sustained rate of 4.5 percent for the next 25 years. It needs to be acknowledged that there may be different opinions regarding the future growth of the Paraguayan economy; our estimate of 4.5 percent sustained growth, while it represents an estimate based on a well-defined methodology, may seem high by historical standards of this country. This rate was derived using the estimated production function and projections of key variables obtained from different sources. However, the trend growth rate is endogenous to the policies implemented and to other uncontrollable conditions, which may reduce or increase the rate of sustained GDP growth. GDP growth may be altered by effects on productivity growth —such as incentives or disincentives on innovation--, capital accumulation —such as incentives or disincentives on investment, employment growth —such as incentives or disincentives on participation in the labor force, migration, or labor market functioning. It is clear then that different views regarding the Paraguayan economy and its future trend GDP growth may exist; it is necessary to obtain an agreement for this key parameter, which must be periodically reviewed. In order to calculate the fiscal target and as shown in the above equation, there are several other information requirements. First, on the level of the public sector net debt, which was obtained from ECLAC; second, on the discount factor, for which the real interest rate and the real growth rate are a requisite. A long-term goal for public net worth needs to be added —the negative of the net public debt— and the horizon within which this objective is to be achieved. The relevant interest rate for Paraguay’s sovereign debt was estimated based on Uruguay’s interest rate plus a margin in consideration of the higher risk assessment for Paraguay. Considering that Paraguay is a net debtor economy, its estimated relevant real interest rate amounts to 6.5 percent; this estimation is obtained using as a reference a study for the relevant interest in the Uruguayan economy (5.75 percent) 19 and adding 75 bp, considering that Paraguay’s sovereign debt is rated three grades below Uruguay (S&P); each grade was estimated to correspond to 25 bp. This approach was chosen since there is no data available regarding the financing cost for Paraguay. In case Paraguay adopts a rule such as the one suggested in this report, this interest rate is likely to fall. Net public debt —the negative of public net worth— is endogenous and generates a direct impact on the risk that markets allocate to the repayment of sovereign debt. With the establishment of a structural balance target, the Paraguayan economy may enter into a virtuous circle leading to lower deficits, lower debt and increased economic stability; while, the absence 19 See Le Fort (2012) 40 of a structural target can mean, as it did in the past, a vicious circle where cyclical deficit increases lead to expansions in the level and the cost of the debt and the transmission of financial instability. Ranges were used for other parameters, including the goal for public financial wealth between -20 percent of GDP and +10 percent of GDP, and the time horizon to achieve it, between 5 and 25 years. Since it is not possible to determine a unique fiscal objective and horizon to achieve it, a double-entry table was used so as to present the fiscal targets resulting from different long-term goals and different time horizons. The calculations were performed on the basis of the equation for the primary fiscal balance target and considering various alternatives for the long-term goal (b (t + N)) and for the time horizon to reach it (N). The discount factor ψ was considered to be constant and equal to the estimated value of 1.02153120. Regarding the goal of future wealth of the public sector, a wide range of options was considered: from -20 percent of GDP, similar to the net public debt in 2009, to 10 percent of GDP (negative net debt), similar to the level achieved by Chile in 2011. Different alternatives for the relevant horizon were included, from 5 to 25 years, considering as a minimum one governmental period and as a maximum five government periods. In order to define a target for long periods of time a strong political consensus is required so as to maintain the structural fiscal policy across the different government periods. The results are presented in Table 1.8. Table 1.8: Targets for the Primary fiscal balance as a function of the Long Term Goal for Fiscal Net Worth and Time Horizon to achieve them. Variables as a percentage of GDP HORIZONTE/ PATRIMONI -20 -15 -10 -5 0 O percent percent percent percent percent 5 percent 10 percent 0,5 1,4 2,3 3,3 4,2 5,2 5 percent percent percent percent percent percent 6,1 percent 0,4 0,9 1,3 1,8 2,2 2,7 10 percent percent percent percent percent percent 3,1 percent 0,4 0,7 1,0 1,3 1,6 1,8 15 percent percent percent percent percent percent 2,1 percent 0,4 0,6 0,8 1,0 1,2 1,4 20 percent percent percent percent percent percent 1,6 percent 0,4 0,6 0,7 0,9 1,0 1,2 25 percent percent percent percent percent percent 1,3 percent Source: LE&F based on data from ECLAC and the World Bank The simulation results show a wide range of targets for the primary fiscal balance: from a small surplus of 0.43 percent of GDP to a huge primary surplus of 6 percent of GDP. To maintain public debt as a percent of GDP constant at its current level —the equivalent to a net worth of -20 percent of GDP— a primary surplus of 0.43 percent of GDP is required, regardless of the time horizon. If, instead, a zero net debt level or zero net worth is pursued, the fiscal effort needs to be much more substantial; in order to reach this goal within 25 years, the primary fiscal balance must achieve and sustain a surplus of 1 percent of GDP. If, instead, the same zero ( ) ( ) 20 The discount factor used in the simulations was: ( ) ( ) ( ) 41 indebtedness goal is intended to be achieved in just 10 years, the required primary surplus would be of 2.2 percent of GDP for the 10 years; and, in order to reach a zero debt in a period of five years, the required structural primary surplus would amount to 4.21 percent GDP. (Figures 1.34 and 1.35) Figure 1.34: Public debt and interest spending as percent of GDP with a fiscal target of 1.23 percent and 20 years’ time Source: LE&F based on data from CEPAL and the World Bank. Figure 1.35: Public debt and interest spending as percent of GDP with a fiscal target of 2.2 percent and 10 years’ time Source: LE&F based on data from ECLAC and the World Bank 42 With a more challenging fiscal policy, Paraguay could reach a positive net worth position, assets greater than liabilities, equivalent to 10 percent of its GDP in 20 years. With a sustained primary surplus of 1.6 percent of GDP, Paraguay could achieve over a period of 20 years a level of positive public net worth of 10 percent of GDP. (Figure 1.36) Figure 1.36: Public debt and interest spending as percent of GDP with a fiscal target of 1.62 percent and 20 years’ time Source: LE&F based on data from ECLAC and the World Bank In Paraguay, according to available information, there is no domestic debt market. It can be concluded that the country is heavily exposed to shocks on the real exchange rate considering the composition of its sovereign debt and since the development of a domestic local currency public debt market has not been possible. The impact of the RER on the level of public debt is high, about one-to-one, and represents a significant risk considering its historical instability. It is on this basis that our main recommendation is that, at least, a long-term goal of zero net debt should be established; this would require a structural primary surplus of 1.23 percent per annum, sustained for the next 20 years. According to our estimates it is feasible for Paraguay to get the fiscal balance to evolve from the 0 percent recorded in 2011 to the balance target of 1.23 percent of GDP until the 2030s. It needs to be emphasized that this recommendation is intended only as a reference, since the fiscal target should be adopted under a new fiscal institutional framework, including a review of the structural fiscal target as well as the corresponding adjustments. In addition, it must be considered that exogenous shocks can generate significant changes in the exchange rate and in the value of the debt, even in the value of GDP, and thus, in the debt as a share of GDP. If that is the case, the structural fiscal target can be adjusted in order to achieve the desired goal in a longer or shorter time horizon. 43 Under the target of a structural primary balance of 1.23 percent, interest spending should amount to 0 percent of GDP by 2030. Under this objective, interest expenses should converge gradually from 1.23 percent of GDP in 2011 to 0 percent in 2029. It should be noted that interest expenses were calculated based on the real interest rate, leaving aside any effect on nominal interest rates. Our proposal uses the Primary Structural Balance that, despite its calculation complexity, has the advantage —over Total Structural Balance— to be achieved through a single target. (Figure 1.37) Figure 1.37: Fiscal Balance Evolution ( percent of GDP) Source: LE&F based on data from ECLAC and the World Bank The main benefit of a structural fiscal target is to avoid crises and fiscal insolvency. The rationale for the application of a structural fiscal policy in Paraguay is not only to reduce the cyclical fluctuations of fiscal spending but rather to consider a broader objective such as minimizing the vulnerability of the economy by building a political commitment regarding the fiscal balance. The increased vulnerability of the fiscal position and the risk of an insolvency episode are on its own sufficient justification for the establishment of a structural fiscal policy. No government or political group may by its own assume the cost of initiating a process of long- term fiscal consolidation; this, since all the efforts to reduce debt and deficit can cut down their chances of remaining in power, and open the door for other political groups seeking to increase future levels of spending. Fiscal consolidation must be part of an agreed political effort, a national or State policy built on the basis of permanent institutions that can assure a sustained effort that would result in benefits for all Paraguayans in the coming years. The projected scenarios assume that in the future there are no contingencies affecting the evolution of GDP. It should be made clear that the simulations were performed on the basis of no future contingencies. Negative contingencies, such as banking crises and significant devaluations, clearly result in increases in debt levels which require counter-cyclical policy measures; in turn, this would imply giving up the established objective for a while and resuming it when the crisis eases off. In the event of positive contingencies, such as a positive shock on 44 prices of major exports or increased hydropower generation due to increased rainfall, the country could raise its saving rate, a condition that should lead to a recalculation of the annual target. Changes in the fiscal net worth base conditions, as well as reviews on the macroeconomic parameters, may result in the need of a revision of the primary surplus target. It seems advisable that such revisions are made at regular intervals, for example every 5 years. It also seems advisable to keep the long-term fiscal net worth objective as given, but, accepting adjustments in the time horizon considered to achieve it. Adverse events which increase the level of public debt and deteriorate the fiscal initial net worth could justify an extension in the deadline for achieving the net worth objective. On the other hand, favorable contingencies that reduce the level of debt or increase public net worth as a share of GDP could justify achieving the final goal in a shorter period. It is necessary to review the annual fiscal target once the initial long term goal is achieved planning to maintain indefinitely a level of public assets. If, for example, after maintaining for 20 years a primary fiscal surplus, the goal of a 0 percent of GDP net worth has been achieved, it would be appropriate to consider reviewing the annual fiscal target. It would be reasonable to consider keeping the same goal indefinitely, for which it would be necessary to establish a new target for the structural primary balance of 0 percent of GDP. Alternatively, an objective of reaching a positive or more ample financial position, similar to Chile, could be contemplated, for which the new target would be a surplus whose value will depend, among other things, on the deadline to do so. But, in 20 or 25 years the discount rate may be very different, either because the GDP growth rate is lower because the economy has matured or since the interest rate is down due to the lower risk corresponding to a zero public debt and a fiscal rule in operation. It would be risky to intend such calculation now. Some Recommendations for Implementation The process of determining the structural fiscal framework requires a periodic update of the estimates for a number of macroeconomic variables. A periodic estimation and projection of a number of macroeconomic and fiscal variables is required in order to determine the goal for the public financial wealth which in turn is defined on the basis of a structural fiscal target; those estimates and projections will allow to determine, for the long-term objective and within the defined time horizon, the target for the structural primary balance that must be achieved every year. Once the target is defined, the expenditure budget is obtained after projecting the structural revenue; the latter requires projections for the trend GDP and for the gap. Among the main macroeconomic variables to be estimated is the long-term real interest rate relevant to the Paraguayan debt and the trend GDP growth. The estimates of the macroeconomic variables should be performed by a technical agency, a committee of independent experts that offers assurance of independence and technical capacity. One possibility is to organize a specialized public institution in charge of these projections, a kind of technical agency for the fiscal framework. This institution should be headed by a technical committee of several members, all of which should be well-known economists and represent different views on Paraguay’s fiscal policy. On the basis of the projections of the relevant variables, the committee should monitor the estimates of the primary balance target. Based on this macroeconomic framework and the 45 fiscal target, the economic authorities should present their estimates of the structural revenue and the annual expenditure budget, or the annual reviews of the already executed budget. The committee should evaluate and present to the authority a document with the results of its deliberations on these proposals, the value and process by which the fiscal target was determined, and the points of agreement and discrepancies. This should serve as the basis for the authority to calculate the value of the target and thus complete all the procedures for obtaining, from the projected GDP trend and structural fiscal revenue, the space available for primary and total expenditure. Then, the authority should submit to Parliament and the public a technical report explaining the methodology and the results obtained in the estimation of structural revenues, of the expenditure budget and of the structural balance. The availability of data and information as well as fiscal transparency are essential requisites for the proper implementation of the fiscal rule. Greater transparency is a prerequisite for the success of a structural fiscal policy. A fiscal policy of this type should be based on the availability of quality data on the public sector: aggregated fiscal revenue and expenditures, operations, fiscal assets and liabilities in addition to contingencies. Data must be reliable and adequately backed up; in addition, the information should be timely available through public access media (eg, web pages). Deficiencies in Paraguay’s data are significant, the series cover relatively short periods, and there appears to be some lack of consistency between national accounts and data on assets and liabilities of the public sector. In addition, a more thorough insight of contingent liabilities of the public sector is necessary. Fiscal transparency requires a major effort for the collection, compilation and coverage of fiscal statistics. We recommend the creation of an external and domestic Paraguayan public debt market which could contribute to the dissemination of information on the policies followed, and to obtaining full benefits from the adoption of a structural fiscal policy. The transaction of Paraguayan debt instruments denominated in foreign currency (dollars or euros) in the international markets could establish a reference for Paraguay’s country risk, either a margin as measured by the "Emerging Bond Index" (EMBI) or a premium as in the Credit Default Swaps (CDS); the latter operate as a default insurance, are traded in the international markets, and establish a measure of the risk perception of a particular debt instrument. The sovereign debtor pays for external funding through the international reference rate plus a country risk margin. Establishing a reference country risk would allow for a more refined calculation and projection of the real interest rate relevant to Paraguay. But, more important, this combined with the introduction of a structural fiscal policy would allow measuring the effects of such policy on the risk perception by international markets, and the improvements accomplished in this regard. This would benefit not only the public sector by reducing its funding cost, but also the Paraguayan private sector that could get increased access to external financial markets and a reduction in its cost since risk margins for private debtors are set on the basis of the country risk margin. The development of a domestic government bond market could have a very positive effect on the diversification of the fiscal financial risk and facilitate the development of a local capital market. One of the disadvantages of Paraguay's debt is that it is denominated in dollars, which produces an important exposure of the public sector to the exchange rate. A real devaluation would proportionally increase the value of the public debt since virtually all of it is denominated in dollars. One way to diversify currency risk would be to issue domestic currency 46 debt indexed by inflation, but this requires the development of a domestic market for local currency public debt. Public debt is the ideal instrument for the founding of the basis of the development of financial markets for private instruments. It would also be beneficial for the private sector since this also provides the basis for the development of local capital markets, including long-term financing in inflation indexed instruments, which can be as important in the stimulus of investment in general and of housing in particular. 4. Concluding Remarks The central objective of this paper is to present a proposal for a structural fiscal target for Paraguay starting from estimations of an aggregate production function. The estimation of variables such as trend GDP and the GDP gap was done through a Cobb-Douglas type of function. Among the estimates, a 4.5 percent to 4.8 percent annual trend GDP growth was concluded. On the other hand, the fiscal revenue estimates were obtained through cointegration econometric models and error correction models. The estimated Total Public Revenue elasticity with respect to GDP amounts to 1.8 percent. With this value it can be concluded that economic growth leads to increases in fiscal revenues. It is also noted that indirect taxes have great significance in total income; the role of direct taxes, such as income tax, should be strengthen, thereby moving closer to the tax systems of other countries in the region21. Total revenues present pro-cyclical behavior, unlike total fiscal expenditure. The variables were analyzed in relation to the economic cycle; the results indicate that public expenditure has a cycle-neutral behavior, unlike the behavior of Fiscal Revenue, variable that moves with the cycle. Real GDP was assumed to grow 4.5 percent annually over the long-term, while the relevant interest rate was assumed to be 6.5 percent per year. Estimated values were assumed in the calculation of the structural fiscal target, such as a real GDP growth of 4.5 percent. It must also be considered that since Paraguay does not have a developed domestic debt market, the determination of a benchmark interest rate is quite complicated; that is why Uruguay’s rate was assumed as the basis for the determination of the relevant interest rate; to Uruguay’s rate, of 5.75 percent, 75p.b. were added in consideration to the difference of three grades between the risk rating of the two economies. In order for the current debt level to be maintained constant as a percentage of GDP, it is necessary to establish as a structural fiscal target a primary surplus of 0.43 percent, regardless of the time horizon. Alternatively, the objective may be to progress from the current situation of a fiscal wealth or net financial position of -20 percent of GDP to a 0 percent of GDP. To achieve this goal within five years a primary structural surplus of 4.21 percent of GDP per year is required, or, if the period is extended to 20 years, a primary structural surplus of 1.23 percent of GDP per year. 21 See Annex N°2 47 The evolution of the debt could lead to a reduction in the interest rate which would, in turn, require reviewing the fiscal target. The estimates consider a constant interest rate throughout the entire forecast period, but, the fact remains that as Paraguay improves its borrowing conditions, the relevant interest rate should decrease to the extent its risk rating improves. A reduction in the relevant interest rate would eventually allow recalculating the structural fiscal target; in such case, it would be possible to define a less strict target, but, that equally assures the country’s financial sustainability. The proposed scenarios do not consider any shocks on GDP throughout the entire forecast period. It should be noted that the proposed scenarios consider a constant growth rate, so, they do not incorporate the effects on the fiscal target of any shocks on the level of domestic production. In the case of negative shocks, the authority will be compelled to implement counter- cyclical policies that would require a re-assessment of the fiscal target. As well, positive shocks that enable the country to raise temporarily its savings level would lead to a recalculation of the target. 5. Technical Appendix Regressions This section explains the models estimated but finally dismissed justifying the final choice of the one used and presented in the body of this work. The selection and dismissal was performed on the basis of the statistical significance of the parameters, of the goodness of fit of the models and of information criteria. As explained in previous chapters, the estimations were performed by ordinary least squares with HAC variance-covariance matrix. Aggregate Error Correction Models The development of the Error Correction Model was based on the following cointegration regression: ln( I _ Tot) t   0  1 ln( PIB) t  t Where: ln( I _ Tot ) t = Natural logarithm of Total Revenue in period t. ln( PIB) t = Natural logarithm of Trend GDP in period t. Following the results: Table 1.9: OLS estimation Log of Total Fiscal Revenue 1990-2010 (T = 21) HAC standard errors, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -15.64522 1.466162 -10.67087 0*** Ln(PIB) 1.844291 0.088857 20.75583 0*** 48 Media de la vble. dep. 14.78506 D.T. de la vble. dep. 0.274289 Suma de cuad. residuos 0.063559 Suma de cuad. regresión 0.057838 R-cuadrado 0.957759 R-cuadrado corregido 0.955536 Estadístico F 430.8046 Prob (Estadístico F) 0 Log-verosimilitud 31.10554 Criterio de Akaike -2.771956 Criterio de Schwarz -2.672478 Crit. de Hannan-Quinn -2.750367 Thus, the model found is ln(I_Tot)t  15.6452  1.844291ln(PIB)t Considering the above, the models developed are: Model first difference of total revenue, explained by its lagged variable and GDP: ˆt 1  t  ln( I _ Tot)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3 ln( I _ Tot)t 1   4u Table 1.10: OLS Estimate First Difference logarithm Total Revenue 1992-2010 (T = 19) HAC standard errors, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,027565 0,017572 1,57E+00 0,139 D Ln(PIB) 1,038253 0,298204 3,481683 0,0037** D Ln(PIB) (-1) -0,083164 0,530721 -1,57E-01 0,8777 D Ln(I_Tot) (-1) 0,057589 0,235089 2,45E-01 0,81 U_hat (-1) -0,713623 0,257128 -2,78E+00 0,0149* Media de la vble. dep. 0,053155 D.T. de la vble. dep. 0,064629 Suma de cuad. residuos 0,032337 Suma de cuad. regresión 0,04806 R-cuadrado 0,569888 R-cuadrado corregido 0,446999 Estadístico F 0,569888 Prob (Estadístico F) 0,013586 Log-verosimilitud 33,61193 Criterio de Akaike -3,011782 Criterio de Schwarz -2,763245 Crit. de Hannan-Quinn -2,969719 First difference model with error correction ˆt 1  t  ln( I _ Tot)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3u Table 1.11: OLS Estimation First Difference Total Revenue logarithm 1991-2010 (T = 20) HAC standard errors, with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,023377 0,018166 1,286841 0,2165 D Ln(PIB) 1,033637 0,31092 3,324448 0,0043** 49 D Ln(PIB) (-1) -0,007607 0,432653 -0,017583 0,9862 U_hat (-1) -0,60516 0,22972 -2,634335 0,018* Media de la vble. dep. 0,04979 D.T. de la vble. dep. 0,06468 Suma de cuad. residuos 0,040637 Suma de cuad. regresión 0,050397 R-cuadrado 0,488753 R-cuadrado corregido 0,392894 Estadístico F 5,098675 Prob (Estadístico F) 0,011497 Log-verosimilitud 33,60922 Criterio de Akaike -2,960922 Criterio de Schwarz -2,761776 Crit. de Hannan-Quinn -2,922047 ˆt 1  t  ln( I _ Tot) t   0  1 ln( PIB) t   2  ln( I _ Tot) t 1   3u Table 12: OLS Estimate First Difference Total Revenue logarithm 1992-2010 (T = 19) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,026407 0,015415 1,713009 0,1073 D Ln(PIB) 1,046141 0,284208 3,680902 0,0022*** D Ln(I_Tot) (-1) 0,034419 0,176727 0,19476 0,8482 U_hat (-1) -0,693906 0,216826 -3,200291 0,006** Media de la vble. dep. 0,053155 D.T. de la vble. dep. 0,064629 Suma de cuad. residuos 0,032394 Suma de cuad. regresión 0,046471 R-cuadrado 0,569133 R-cuadrado corregido 0,48296 Estadístico F 6,604519 Prob (Estadístico F) 0,004615 Log-verosimilitud 33,59528 Criterio de Akaike -3,115293 Criterio de Schwarz -2,916463 Crit. de Hannan-Quinn -3,081643 Error Correction Models with prices of export products The Error Correction Model was based on the following cointegration regression: ln( I _ Tot) t   0  1 ln( PIB) t   2 ln( P _ carne) t   3 ln( P _ soja) t  t Where: ln( I _ Tot ) t = Natural logarithm of Total Revenue in period t. ln( PIB) t = Natural logarithm of Trend GDP in period t. ln( P _ carne) t = Natural logarithm of international meat price in period t22. ln( P _ soja) t = Natural logarithm of the international price of soybean in period t 23. 22 Price in Guaraníes/kg (Base=1994) 23 Price in Guaraníes/Metric Ton (Base=1994) 50 The results obtained are: Table 1.13: OLS Estimate Total Revenue logarithm 1990-2010 (T = 21) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -16,62072 1,896502 -8,76388 0*** Ln (PIB) 1,916494 0,121473 15,77717 0*** Ln (P_carne) 0,068889 0,078697 0,875371 0,3936 Ln (P_soja) -0,092502 0,112917 -0,819197 0,424 Media de la vble. dep. 14,78506 D.T. de la vble. dep. 0,274289 Suma de cuad. residuos 0,060553 Suma de cuad. regresión 0,059682 R-cuadrado 0,959757 R-cuadrado corregido 0,952655 Estadístico F 1351445 Prob (Estadístico F) 0 Log-verosimilitud 3161421 Criterio de Akaike -2629925 Criterio de Schwarz -2,430968 Crit. de Hannan-Quinn -2586746 Then, the model found is: ln( I _ Tot ) t  -16,62072  1,916494 ln( PIB) t  0,068889 ln( P _ carne) t  0,092502 3 ln( P _ soja) t As can be seen, the parameters associated with the two prices, meat and soybeans, are not significant, so, no further progress was made in estimating the error correction model Total Revenue model, considering the price of soybeans with the GAP The Error Correction Model was based on the following cointegration regression: ln( I _ Tot) t   0  1 ln( PIB) t   2 ln( P _ soja) t  t Where: ln( I _ Tot ) t = Natural logarithm of Total Revenue in period t. ln( PIB) t = Natural logarithm of Trend GDP in period t. ln( P _ soja) t = Natural logarithm of the international price of soybean in period t 24 The results obtained were: Table 1.14: OLS Estimation Total Revenue logarithm 1990-2010 (T = 21) HAC standard errors with bandwidth 3 (Kernel de Bartlett) 24 Price in ($/mt) Guaraníes (Base=1994) 51 Coeficiente Desv. Típica Estadístico t Prob Constante -15,62225 1,505291 -10,37823 0*** Ln (PIB) 1,854394 0,097962 18,92964 0*** Ln (P_soja) -0,022061 0,078699 -0,280321 0,7824 Media de la vble. dep. 14,78506 D.T. de la vble. dep. 0,274289 Suma de cuad. residuos 0,063283 Suma de cuad. regresión 0,059293 R-cuadrado 0,957943 R-cuadrado corregido 0,95327 Estadístico F 204,9955 Prob (Estadístico F) 0 Log-verosimilitud 31,15128 Criterio de Akaike -2,681074 Criterio de Schwarz -2,531857 Crit. de Hannan-Quinn -2,64869 Then the model found is: ln( I _ Tot) t  -15,62225  1,854394 ln( PIB) t - 0,022061ln( P _ soja) t As can be seen, the parameter associated with the price of soybeans is not significant, so, no further progress was made in estimating the error correction model. Total Revenue model explained by the price of meat with the gap The Error Correction Model was based on the following cointegration regression: ln( I _ Tot) t   0  1 ln( PIB) t   2 ln( P _ carne) t  t Where: ln( I _ Tot ) t = Natural logarithm of Total Revenue in period t. ln( PIB) t = Natural logarithm of Trend GDP in period t. ln( P _ carne) t = Natural logarithm of the international price of meat in period t25 The results obtained were: ln( P _ soja) t = t26 Table 1.15: OLS Estimation Total Revenue logarithm, 1990-2010 (T = 21) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -16,00222 1,723768 -9,283276 0*** Ln (PIB) 1,854232 0,093888 19,74934 0*** Ln (P_carne) 0,022946 0,054701 0,419483 0,6798 25 Precio en (cents/kg) Guaraníes (Base=1994) 26 Price in ($/mt) Guaraníes (Base=1994) 52 Media de la vble. dep. 14,78506 D.T. de la vble. dep. 0,274289 Suma de cuad. residuos 0,062944 Suma de cuad. regresión 0,059134 R-cuadrado 0,958168 R-cuadrado corregido 0,95352 Estadístico F 206,1482 Prob (Estadístico F) 0 Log-verosimilitud 31,20769 Criterio de Akaike -2,686447 Criterio de Schwarz -2,537229 Crit. de Hannan-Quinn -2,654063 Thus, the model found is: ln( I _ Tot) t  -16,00222  1,854232 ln( PIB) t  0,022946 ln( P _ carne) t As can be appreciated, the parameter associated with the price of meat is not significant, so, no further progress was made in estimating the error correction model. Error Correction Models with Income from Taxes on Goods and Services The development of these Error Correction Models was based on the following cointegration regression: ln(Impto_BsyS s) t   0  1 ln( PIB) t  t Where: s) = Natural logarithm of Taxes on goods and services in period t ln(Impto_BsyS ln( PIB) t = Natural logarithm of trend GDP in period t The results are: Table 1.16: OLS Estimation logarithm of Taxes on Goods and Services 1992-2010 (T = 19) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -42,93481 6,158638 -6,971477 0*** Ln(PIB) 3,137738 0,372715 8,418593 0*** Media de la vble. dep. 8,9107 D.T. de la vble. dep. 0,460886 Suma de cuad. residuos 0,739699 Suma de cuad. regresión 0,208595 R-cuadrado 0,806538 R-cuadrado corregido 0,795158 Estadístico F 70,87271 Prob (Estadístico F) 0 Log-verosimilitud 3,876706 Criterio de Akaike -0,197548 Criterio de Schwarz -0,098133 Crit. de Hannan-Quinn -0,180723 Thus, the model found is: s) t  -42,93481  3,137738 ln( PIB) t ln(Impto_BsyS 53 Models of first differences and error correction Considering the above, the models developed are: ˆt 1  t s)t  0  1 ln( PIB)t   2u  ln(Impto_BsyS Table 1.17: Estimation logarithm of First Difference of Taxes on Goods and Services 1993-2010 (T = 18) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -0,025609 0,08271 -0,30962 0,7611 D Ln(PIB) 3,562126 1,549227 2,299292 0,0363* U_hat (-1) 1,073254 0,704221 1,52403 0,1483 Media de la vble. dep. 0,120426 D.T. de la vble. dep. 0,23618 Suma de cuad. residuos 0,698305 Suma de cuad. regresión 0,215763 R-cuadrado 0,263607 R-cuadrado corregido 0,165422 Estadístico F 2,684785 Prob (Estadístico F) 0,100767 Log-verosimilitud 3,704343 Criterio de Akaike -0,07826 Criterio de Schwarz 0,070135 Crit. de Hannan-Quinn -0,057799 Model of first differences and error correction s)t 1   4u s)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3 ln(Impto_BsyS  ln(Impto_BsyS ˆt 1  t Table 1.18: OLS Estimation logarithm of First Differences of Taxes on Goods and Services 1994-2010 (T = 17) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -0,023359 2,62E-02 -8,92E-01 0,3897 D Ln(PIB) 2,373409 4,87E-01 4,88E+00 0,0004*** D Ln(PIB) (-1) 0,541651 5,47E-01 9,91E-01 0,3412 D Ln(Impto_BsySs) (-1) 0,034857 8,18E-02 4,26E-01 0,6776 U_hat (-1) 0,301141 2,41E-01 1,25E+00 0,2346 Media de la vble. dep. 0,069785 D.T. de la vble. dep. 0,101098 Suma de cuad. residuos 3,98E-02 Suma de cuad. regresión 5,76E-02 R-cuadrado 0,75661 R-cuadrado corregido 0,67548 Estadístico F 9,325895 Prob (Estadístico F) 0,001152 Log-verosimilitud 27,36297 Criterio de Akaike -2,630938 Criterio de Schwarz -2,385875 Crit. de Hannan-Quinn -2,385875 Model of taxes on goods and services with error correction ˆt 1  t s)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3u  ln(Impto_BsyS 54 Table 1.19: OLS Estimation logarithm of First Differences of Taxes on Goods and Services 1993-2010 (T = 18) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -0,05386 0,094447 -0,570263 0,5775 D Ln(PIB) 3,671356 1,587546 2,312598 0,0365* D Ln(PIB) (-1) 1,216057 1,833288 0,66332 0,5179 U_hat (-1) 1,022836 0,72176 1,417142 0,1783 Media de la vble. dep. 0,120426 D.T. de la vble. dep. 0,23618 Suma de cuad. residuos 0,677028 Suma de cuad. regresión 0,219907 R-cuadrado 0,286046 R-cuadrado corregido 0,133055 Estadístico F 1,869699 Prob (Estadístico F) 0,181103 Log-verosimilitud 3,982842 Criterio de Akaike 0,001906 Criterio de Schwarz 0,199767 Crit. de Hannan-Quinn 0,029189 Model of first differences and error correction for taxes on goods and services s)t 1   4u s)t  0  1 ln( PIB)t   2  ln(Impto_BsyS  ln(Impto_BsyS ˆt 1  t Table 1.20: OLS Estimation logarithm of First Differences of Taxes on Goods and Services 1994-2010 (T = 17) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -0,009602 0,022178 -0,432975 0,6721 D Ln(PIB) 2,218141 0,460608 4,815676 0,0003*** D Ln(Impto_BsySs) (-1) 0,073114 0,072071 1,014472 0,3289 U_hat (-1) 0,248989 0,234636 1,061171 0,3079 Media de la vble. dep. 0,069785 D.T. de la vble. dep. 0,101098 Suma de cuad. residuos 0,04306 Suma de cuad. regresión 0,057553 R-cuadrado 0,736688 R-cuadrado corregido 0,675924 Estadístico F 12,12369 Prob (Estadístico F) 0,000457 Log-verosimilitud 26,69424 Criterio de Akaike -2,66991 Criterio de Schwarz -2,47386 Crit. de Hannan-Quinn -2,650423 Error Correction Model Revenue from International Trade taxes The estimation of the Error Correction Model was based on the following cointegration regression: ln(Impto_Comex) t   0  1 ln( PIB) t  t Where: 55 ln(Impto_Comex)t = Natural logarithm of International Trade Taxes in period t. ln( PIB) t = Natural logarithm of trend GDP in period t The results are: Table 1.21: OLS Estimation logarithm International Trade Taxes 1990-2010 (T = 21) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -0,143033 4,204272 -0,034021 0,9732 Ln(PIB) 0,494627 0,254799 1,941241 0,0672 Media de la vble. dep. 8,018168 D.T. de la vble. dep. 0,176959 Suma de cuad. residuos 0,522631 Suma de cuad. regresión 0,165852 R-cuadrado 0,165511 R-cuadrado corregido 0,165852 Estadístico F 3,768415 Prob (Estadístico F) 0,06721 Log-verosimilitud 8,983022 Criterio de Akaike -0,66505 Criterio de Schwarz -0,565571 Crit. de Hannan-Quinn -0,64346 Thus, the model estimated is: ln(Impto_Comex) t  -0,143033  0,494627 ln( PIB) t  t Error Correction Model Revenue from taxes on Corporations and Companies The development of the Error Correction Model was based on the following cointegration regression: ln(Impto_Corp) t   0  1 ln( PIB) t  t Where: ln(Impto_Corp)t = Natural logarithm of Taxes on Corporations and Companies in period t. ln( PIB) t = Natural logarithm of trend GDP in period t The results are: Table 1.22: OLS Estimation logarithm Taxes on Corporations and Companies 1990-2010 (T = 21) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -29,58101 4,069103 -7,269664 0*** Ln(PIB) 2,279713 0,246607 9,2443 0*** 56 Media de la vble. dep. 8,033602 D.T. de la vble. dep. 0,366845 Suma de cuad. residuos 0,489565 Suma de cuad. regresión 0,16052 R-cuadrado 0,818107 R-cuadrado corregido 0,808534 Estadístico F 85,45708 Prob (Estadístico F) 0 Log-verosimilitud 9,669272 Criterio de Akaike -0,730407 Criterio de Schwarz -0,630929 Crit. de Hannan-Quinn -0,708818 Thus, the model found is: ln(Impto_Corp) t  -29,58101  2,279713 ln( PIB) t  t Considering the above, the models developed are ˆt 1  t  ln(Impto_Corp)t  0  1 ln( PIB)t   2u Table 1.23: OLS Estimation logarithm First Differences Taxes on Corporations and Companies 1991-2010 (T = 20) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,066394 0,036173 1,835428 0,084 D Ln(PIB) 0,17045 0,771988 0,220794 0,8279 U_hat (-1) -0,623833 0,189387 -3,293957 0,0043* Media de la vble. dep. 0,066927 D.T. de la vble. dep. 0,15623 Suma de cuad. residuos 0,282878 Suma de cuad. regresión 0,128996 R-cuadrado 0,390019 R-cuadrado corregido 0,318256 Estadístico F 5,434857 Prob (Estadístico F) 0,014969 Log-verosimilitud 14,20594 Criterio de Akaike -1,120594 Criterio de Schwarz -0,971234 Crit. de Hannan-Quinn -1,091438 ˆt 1  t  ln(Impto_Corp)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3 ln(Impto_Corp)t 1   4u Table 1.24: OLS Estimation logarithm First Differences Taxes on Corporations and Companies 1992-2010 (T = 19) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,00854 0,042649 0,200229 0,8442 D Ln(PIB) 0,27523 0,738711 0,372582 0,715 D Ln(PIB) (-1) 2,31948 1,192711 1,944712 0,0722 D Ln(Impto_Corp) (-1) 0,125101 0,21856 0,572387 0,5761 U_hat (-1) -0,541532 0,225377 -2,402785 0,0307* Media de la vble. dep. 0,068982 D.T. de la vble. dep. 0,160233 Suma de cuad. residuos 0,168841 Suma de cuad. regresión 0,109819 57 R-cuadrado 0,634655 R-cuadrado corregido 0,530271 Estadístico F 6,079993 Prob (Estadístico F) 0,004729 Log-verosimilitud 17,91089 Criterio de Akaike -1,359041 Criterio de Schwarz -1,110505 Crit. de Hannan-Quinn -1,110505 ˆt 1  t  ln(Impto_Corp)t  0  1 ln( PIB)t   2  ln( PIB)t 1  3u Table 1.25: OLS Estimation logarithm First Differences Taxes on Corporations and Companies 1991-2010 (T = 20) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante -7,15E-03 4,10E-02 -0,17453 0,8636 D Ln(PIB) 0,463743 6,67E-01 6,96E-01 0,4966 D Ln(PIB) (-1) 2,78E+00 1,02E+00 2,723185 0,015* U_hat (-1) -4,01E-01 1,81E-01 -2,214397 0,0417* Media de la vble. dep. 0,066927 D.T. de la vble. dep. 0,15623 Suma de cuad. residuos 1,93E-01 Suma de cuad. regresión 1,10E-01 R-cuadrado 0,583199 R-cuadrado corregido 0,505049 Estadístico F 7,462546 Prob (Estadístico F) 0,002411 Log-verosimilitud 18,01414 Criterio de Akaike -1,40E+00 Criterio de Schwarz -1,202267 Crit. de Hannan-Quinn -1,362538 ˆt 1  t  ln(Impto_Corp)t  0  1 ln( PIB)t   2  ln(Impto_Corp)t 1  3u Table 1.26: OLS Estimation logarithm First Differences Taxes on Corporations and Companies 1992-2010 (T = 19) HAC standard errors with bandwidth 3 (Kernel de Bartlett) Coeficiente Desv. Típica Estadístico t Prob Constante 0,063109 0,03497 1,804676 0,0912 D Ln(PIB) -0,224349 0,7541 -0,297505 0,7702 D Ln(Impto_Corp) (-1) 0,34407 0,203957 1,686975 0,1123 U_hat (-1) -0,808188 0,194748 -4,149919 0,0009*** Media de la vble. dep. 0,068982 D.T. de la vble. dep. 0,160233 Suma de cuad. residuos 0,214452 Suma de cuad. regresión 0,119569 R-cuadrado 0,443155 R-cuadrado corregido 0,443155 Estadístico F 5,77499 Prob (Estadístico F) 0,007856 Log-verosimilitud 15,63921 Criterio de Akaike -1,22518 Criterio de Schwarz -1,026351 Crit. de Hannan-Quinn -1,19153 Unit Root Test for Cointegration Model and Error Correction Model 58 In order to determine the existence of cointegration in the error correction models a unit root test was performed for both variables, Total Revenue and trend GDP, obtaining the following results27: Table 1.27: Augmented DickeyFuller Test Total Revenue and GDP (1990-2010) Constante Constante + Tendencia PIB Estadístico 0.687778 -2.97611 Test criticalvalues: 1 percent level -3.808546 -4.667883 5 percent level -3.020686 -3.7332 10 percent level -2.650413 -3.310349 D PIB Estadístico -3.879164 -3.845221 Test criticalvalues: 1 percent level -3.831511 -4.532598 5 percent level -3.02997 -3.673616 10 percent level -2.655194 -3.277364 Ingresos Totales Estadístico -0.19664 -1.758268 Test criticalvalues: 1 percent level -3.808546 -4.498307 5 percent level -3.020686 -3.658446 10 percent level -2.650413 -3.268973 D Ingresos Totales Estadístico -4.526943 -4.393646 Test criticalvalues: 1 percent level -3.831511 -4.532598 5 percent level -3.02997 -3.673616 10 percent level -2.655194 -3.277364 Source: LE&F based on data from ECLAC Dickey Fuller Test A Dickey Fuller test was conducted to determine the existence of unit root28 Table 1.28: Dickey Fuller Augmented Test Estimated Errors Estadístico -3.039409 27 The Phillip-Perron test (1987) and Kwiatkowski-Phillips-Schmidt-Shin test (1992) for unit root in the errors were performed for purposes of more accuracy and confirmation. 28 The Phillip-Perron (1987) and Kwiatkowski-Phillips-Schmidt-Shin (1992) test for unit root in the errors were conducted for purposes of more accuracy and confirmation 59 1 percent level -3.808546 5 percent level -3.020686 10 percent level -2.650413 Source: LE&F bases on data from ECLAC Tax Revenue Comparison with other countries in the region Fiscal revenue generated during the 2010 fiscal year represented a 22 percent of GDP with taxes arising from the collection of Value Added Tax (VAT) as one of the main sources of fiscal revenue. One aspect to be considered corresponds to the non-tax revenues which are mostly generated by power generation activities, especially by the Yacyretá and Itaipu plants, which directly depend on the annual precipitation in the Paraná River basin. It is also necessary to mention that in 2010 Congress again delayed the introduction of a Personal Income Tax; thereby, Paraguay remains as the only country in Latin America that does not apply a tax of this kind; this collection is precisely the one that could considerably raise revenue levels, counterbalancing the dependence on indirect tax revenues. Figure 1.38: Tax revenue as percent of GDP in 2010 Source: LE&F based on information from the IMF and ECLAC. 6. Paraguay’s Data Base BASE TRABAJO.xlsx Copia de Base_expertos-Paraguay(1)_update-2.xls Copia de Copia de Proyecciones MCE Con correcciones.xlsx Copia de Datos Nuevos2.xlsx Copia de Proyecciones MCE Correccion.xlsx Copia de Proyecciones MCE Segundo Método.xls Copia de Proyecciones MCE Segundo Método-1.xlsx EXPORTACIONES.XLS 7. Bibliographical References 60 Altar Moisa, Necula Ciprian and Bobrica Gabriel (2010). Estimating the Cyclically Adjusted Budget Balance for the Romanian Economy. A Robust Approach. Romanian Journal of Economics Forecasting. Anderson Barry and Minarik Joseph (2006), Design Choices for Fiscal Policy Rules, OECD Journal on Budgeting Bezdek Vladimír, Dybzak Kamil and Kreejdl Aleš (2007), Cyclically Adjusted Fiscal Balance OECD and ESCB Methods, Czech Journal of Economics and Finance. Croce, Enzo and Hugo Juan Ramón (2003), Assessing Fiscal Sustainability: A Cross Country Comparison, IMF Working Paper WP/03/145, July 2003. Cromin David and McCoy Daniel (1999), Measuring Structural Budget Balances in a Fasto Growing Economy: The Case of Ireland. Technical Paper Central Bank of Ireland. Dirección de Presupuestos (Dipres) Gobierno de Chile. Acta Comité Consultivo PIB Tendencial 2002-2012. http://www.dipres.gob.cl/572/propertyvalue-16157.html Dirección de Presupuestos (Dipres), Gobierno de Chile. Acta Resultados del Comité Consultivo del Precio de Referencia del Cobre 2002-2012. http://www.dipres.gob.cl/572/propertyvalue-16158.html Dirección de Presupuestos (Dipres), Gobierno de Chile (2011), Propuestas para Perfeccionar la Regla Fiscal. Comité Asesor para el Diseño de una Política Fiscal de Balance Estructural de Segunda Generación para Chile Dos Reis Laura, Manasse Paolo y Panizza Ugo (2007), Targeting the Structural Balance. BID Working Paper #598 Du Pelssis Stain and Boshoff Wimpie (2007). A Fiscal Rule to Produce Counter – Cyclical Fiscal Policy in South Africa. Stellenbosch Economic Working Papers. Engel Eduardo, Marcel Mario y Meller Patricio (2007), Meta de superávit estructural: elementos para su análisis. Dirección de presupuestos Chile. Fedelino Annalisa, Ivanova Anna and Horton Mark (2009), Computing Cyclically adjusted balances and automatics stabilizers. Fiscal Affairs department, IMF. Franken Helmut, Le Fort Guillermo y Parrado Eric (2006), Business cycle responses and the resilience of the chilean economy, External Vulnerability and Preventive Policies. Banco Central de Chile. Fondo Monetario Internacional. (2011). Paraguay Consulta del Artículo IV Correspondiente a 2011. Informe del País del FMI N° 11/239. Ford Benjamin (2005), Structural fiscal indicators: an overview. Economic Roundup Giorno Claude, Richardson Pete, Roseveare Deborah and van den Noord Paul (1995), Estimating potential output, output gaps and structural budget balances . OCDE economics department working papers no. 152 Girouard Nathalie and André Cristophe (2005) Measuring Cyclically adjusted budget balances for OECD countries. OECD Economics department, working paper 434. Hagemann Robert (1999), The structural budget balance the IMF methodology. IMF Working paper Le Fort V. Guillermo (2006), Política Fiscal con Meta Estructural en la experiencia chilena. Presentado en la Segunda Reunión Anual del Grupo Latinoamericano de Especialistas en Manejo de Deuda Pública (LAC Debt Group) Cartagena, Colombia Le Fort V. Guillermo y Budnevich, Carlos (1997), “La Política Fiscal y el Ciclo Económico”, Revista de la CEPAL N°61, Abril 1997. http://www.cepal.org/cgi-bin/getProd.asp?xml= 61 /revista/noticias/articuloCEPAL/5/19165/P19165.xml&xsl=/revista/tpl/p39f.xsl&base=/revista/tp l/top-bottom.xslt Le Fort V. Guillermo (2011), Structural Fiscal Policy in an Oil Based Economy: A Proposal for Trinidad and Tobago, LADB Research Proyect Preconditions for Stablishing a Structrual Fiscal Balance Rule in Latinamerican and Caribbean Countries. Le Fort V. Guillermo (2012), Structural Fiscal Policy in Uruguay: A Proposal, LADB Research Proyect Preconditions for Stablishing a Structrual Fiscal Balance Rule in Latinamerican and Caribbean Countries. Marcel Mario, Tokman Marcelo, Valdés Rodrigo y Benavides Paula (2001), Balance fiscal: La base para la nueva regla de política fiscal chilena. Economía Chilena Marcel Mario, Tokman Marcelo, Valdés Rodrigo y Benavides Paula (2001), Balance estructural del gobierno central: Metodología y estimaciones para Chile 1987-2000. Dirección de presupuestos, Ministerio de Hacienda, Chile. Marcel Mario (2009), La Regla De Balance Estructural En Chile Diez Años, Diez Lecciones . CIEPLAN. Ministerio de Finanzas de Dinamarca (2008) ,Danish Fiscal Policy in 2009 in view of the European Economic Recovery Plan (Addendum to Denmark’s Convergence Programme 2008) Price Robert and Nluller Patrice, Structural Budget indicators and the interpretation of fiscal policy stance in OECD economies Rodríguez Jorge, Tokman Carla and Vega Alejandra (2007), Structural Balance Policy in Chile. OECD Journal on Budgeting Schick Allen (2005), Sustainable Budget Policy: Concepts and Approaches. OECD Journalon Budgeting Ter-Minassian Teresa (2009), Preconditions for a successful introduction of structural fiscal balance-based rules in Latin America and the Caribbean: a framework paper. World Economic Outlook 2012, 2011, 2010. IMF. 62 Chapter 2. Paraguay: Agriculture Commodity Prices and Tax Revenue Collection29, by Edgardo Favaro, Friederike (Fritzi) Koehler-Geib, Nathalie Picarelli, Agustin Inaci Executive Summary This paper aims at improving the understanding of the relationship between soybean and beef commodity prices with tax revenue collection in Paraguay. For this, the authors use monthly observations of beef and soybean exported over fourteen years, and the canonical Nerlove model (1959) of partial adjustment as an underlying framework. The model can be interpreted as an optimal farmer response when there are adjustment costs and farmers build their price expectations rationally. Using Dynamic OLS with distributed lags, the paper finds a statistically significant relationship between tax revenue collection and the value of soy and beef exports. While the characteristics of the data imply serious limitations to the economic interpretation of statistical results, they are not trivial. The absence of direct taxes on agriculture income in Paraguay, seem not to be a limitation for a positive relationship working through the Value Added Tax. Further understanding the channels through which this relationship takes places, calls for a detailed analysis of the value-chain of beef and soybean export-production in Paraguay. Introduction This paper aims at improving the understanding of the relationship between soybean and beef commodity prices with tax revenue collection in Paraguay. Agricultural commodities have become one of Paraguay’s strong engines of growth in the past decade, with soybean and beef production at the top of the list. Given Paraguay’s low tax base and the structure of its tax system, understanding this relationship is central in order to improve tax revenue forecasts as well as for the design of economic policy. Indeed, the existence of a statistically significant soybean (beef) supply response to changes in price of soybeans (beef), would imply that information on future prices of these commodities could be used to improve tax revenue forecasts. To this aim, two questions arise. First, since Paraguay does not levy direct taxes on agriculture income30 or exports, the presence of a statistically significant relationship between soybean and beef production with tax revenue collection begs analyzing the channels through which such a link occurs. Secondly, if a relationship between soybean and beef production with tax revenue collection indeed exists, it remains key to distinguish the degree of the relationship for each of the commodities, and the underlying reasons for any possible observed difference. The study is based on monthly observations of beef and soybean exported over fourteen years, using the canonical Nerlove model (1959) of partial adjustment as underlying 29 This paper was prepared as a background paper for the World Bank Public Expenditure Review of Paraguay, by a team composed of Edgardo Favaro, Fritzi Koehler-Geib, Nathalie Picarelli and Agustin Indaco. 30 There is a tax on presumed income (IMAGRO) but revenue from this tax is not very significant as a percentage of total revenue. 63 framework. The model can be interpreted as an optimal farmer response when there are adjustment costs and farmers build their price expectations rationally. One important note: the characteristics of the data imply serious limitations to economic interpretation of statistical results. For instance, monthly production of either beef or soybeans is less elastic than monthly exports; the elasticity of response of beef and soybean production increases over time as producers have time to increase (or decrease) the amount of land and agriculture inputs allocated to crops and livestock production. The main results obtained in the study indicate that (1) there is a strong supply response of beef and soybean exports to prices, and (2) that there is a positive and statistically significant relationship between tax revenue collection and the value of beef and soybean exported. The first results need however to be assessed carefully as the meaning of estimated elasticities are based on exports. These are thus not estimates of the supply of beef and soybeans produced, but rather of the quantity of soybean and beef exported. It is not surprising therefore to find that holders of inventories (soybean seeds and beef herd) respond rapidly to price changes. The second results, which are the main interest of the analysis, seem to indicate that the main underlying reason for the positive relationship is the working of the value added tax (VAT). Beef and soybean production generates income that is spent, for the most part, inside Paraguay. Part of this expenditure generates tax revenue through the VAT system and other part generates revenue through the corporate tax and other tributes. This result supports the view that being exempt from legally paying taxes does not mean that the income generated by agriculture activity is tax exempt. In addition, it suggests that it is worthwhile for the Government to think about whether to insulate fiscal revenues from commodity price fluctuations or factors that impact the volume of production, such as climate shocks. The rest of the paper is organized in four sections: the first section gives a qualitative background of the agricultural sector and its taxation, and then focuses on the soy and beef subsectors in particular. The second section analyzes the methodology of estimation of agriculture supply of soybeans and beef and estimates of price elasticities. Section three then explores the relationship between economic activity and tax revenue collection. It finds, as expected, a significant and positive relationship for the broad economic index. It continues by analyzing the relationship between the level of economic activity and exports of beef and soybeans, as well as the relationship between these and tax revenue collection. Conclusion is carried out in the last section. 1. The Agricultural Sector in Paraguay with its Soy and Beef Subsectors The relevance of agriculture Paraguay’s economy depends strongly on agriculture. In 2011, the agricultural production covered around 50 percent of the country’s total surface and employed around 25 percent of the labor force. At the same time, agriculture represents the largest share of total real value added after the trade and services sector and has continuously increased over the past decade (Table 2.1). The agricultural sector amounted to 32 percent of total value added in 2011 including cattle, forest and fishing, up from 25 percent in 2000. Growth in the sector has explained over 80 percent of the variation of real GDP growth since the early 90s (Figure 2.1). 64 Table 2.1: Sectoral shares in total value added Figure 2.1: Annual real GDP growth versus agricultural value growth 0,2 Shares in total value y = 0,37x + 0,01 R² = 0,82 added (percent) 0,15 Annual real GDP growth 2000 2011 0,1 Agriculture 16 24 0,05 Cattle, Forest and Fishing 9 8 0 Industry and Mining 17 13 -0,2 -0,1 0 0,1 0,2 0,3 0,4 Gas, electricity, water 2 2 -0,05 Construction 4 4 -0,1 Trade and Services 52 53 Annual real agricultural value added growth Source: Central Bank of Paraguay Source: WB staff, Central Bank of Paraguay The sector is concentrated in a few products, soy and beef primarily. These two products accounted for 56 percent of total exports in 2011 up from an average of 48 percent between 2004 and 2011. Also these products were key in terms of contribution to export growth. In 2011, soy explained 16 percentage points of export growth up from an average of 8 percentage points on average between 2004 and 2011. In contrast to previous years, meat contributed negatively to growth in 2011 due to an outbreak of foot and mouth disease late in the year. The average contribution in prior years had been as high as 5 percentage points. This study focuses on soy and beef for which disaggregate data about production and export volumes, values, and international prices are available. The sector’s link to tax revenue collection Despite its importance in the economy, the agricultural sector has historically contributed little to tax revenue collection (Appendix 3). IMAGRO is the income tax in agriculture (Impuesto sobre la renta del sector agropecuario). Prior to the 2004 tax reform, IMAGRO collection was based on 0.9 percent of the total cadastral value of the land used for agricultural production. With cadastral values inferior to the actual surfaces, the sector’s contribution to tax revenue collection was low. Since the 2004 tax reform, IMAGRO has been determined by the net income of agricultural businesses, at a rate of 10 percent. Nonetheless, between 2005 and 2011 tax collection under IMAGRO fell due to exemptions. In particular, since 2005, VAT paid on inputs of goods and services in agricultural and cattle production could be used as a fiscal credit against the payment of IMAGRO. As a consequence, VAT credits mostly paid IMAGRO; until 2012 and despite the credit system having been eliminated by decree in September 2008, the VAT balance to pay IMAGRO has not been totally consumed. For that reason, in 2012 a new decree (8279/12) suspended the accreditation of VAT credit against IMAGRO as of 2014. Without further changes, this means that the agricultural sector would start contributing to direct taxes beginning in 2013. 65 Table 2.2: Product Contribution to Total Export Growth Average 2004-2011 2011 contribution to contribution to annual growth Share in total export growth annual growth Share in total export growth rate (percent) (percent) (percentage rate (percent) (percent) (percentage points) points) Cotton -6 2 0 -30 0 0 Soy 31 32 8 44 42 16 Vegetable oils 32 8 2 23 6 1 Flour 22 10 2 19 8 1 Cereals 30 10 2 11 11 1 Beef Meat 54 54 16 55 -18 -18 14 14 -4 -4 Wood 7 3 0 -5 2 0 Other 22 18 4 37 18 6 Source: WB staff, Central Bank of Paraguay The soy subsector Soy has been an important crop in Paraguay historically, and has played an increasing role over the past decade. The surface used for the production of soy, which represents 7 percent of the country’s total surface (2,957,408 hectares), has doubled since 2000, and even increased 5 - fold since 1990 (according to data from CAPECO—Camera Paraguaya de Exportadores de Cereales y Oleaginosas and from the Ministry of Agriculture). Most of the soy production is located in the eastern part of Paraguay and is realized by farms that exceed 100 hectares in size (of which 43 percent are farms of 100 to 1000 hectares and 44 percent farms with more than 1000 hectares) according to the 2008 Census. Simultaneously, soy production in tons has increased by a factor 2.4 since 2000 and by a factor 8 since 1990. In 2011, the total volume of production amounted to 8.3 million tons. Based on this extraordinary increase soy has surpassed other historically more important crops, such as wheat and corn. At the same time, increases in yields were more modest and volatile relative to the expansion of surface used for soy and the increase in production. Reaching 2.9 tons per hectare in 2011 yields increased 14 percent relative to 2000 and 60 percent relative to 1990. A maximum yield of 2962 kg per hectare was achieved in 2010 in contrast to a minimum of 1500 kg per hectare in 2009. In the period relevant for the 2010 main harvest, weather conditions in Paraguay were ideal while a severe drought occurred during the growing phase of soy harvested in 2009. Overall, the standard deviation of yields has reached 447 kg per hectare since 1990. Soy export occurs quickly after the harvest, due to storage capacity and financial constraints. In most years, there are two soy harvests. The first one can be as early as February or as late as April and the second one falls into the months of July-August. In some years there is even a third harvest around October-November. This schedule is reflected in the monthly export volumes. Once harvested, soy beans are processed in silos, where they are treated, dried and cleaned for optimal quality. In Paraguay, CAPECO estimates that the static capacity of silos in 66 2011 was close 6 million tons while the production was significantly higher at 8 million tons. The constraint of storage capacity is one factor explaining why soy beans are normally exported within a month after harvesting. Another factor is financial constraints. Financial cushions in the sector are relatively small. Producers have therefore little room to schedule soy exports with view to the best price they could get for their beans. The beef subsector The second sector that is highly relevant for agricultural production in Paraguay is beef which has also seen a significant increase in the past decade. According to the Ministry of Agriculture (MoA), around 40 percent of the country’s surface is used as permanent pastures to support cattle breeding. Most of the production occurs in the country’s western part. Also in this sector there are imbalances: Farms between 1 and 20 hectares own less than 15 percent of the livestock and represent more than 70 percent of all producers. In contrast, producers with more than 1000 hectares (1 percent of the farms) own 77 percent of the farm land and 60 percent of the livestock population. Overall, the herd size expanded from around 8 million in the beginning of the 1990s, to almost 10 million in the late 1990s and early years of the new millennium, and exceeded 12 million in 2010 (Figure 2.3). Representing around 25 percent increases respectively. Also, in terms of export volumes Paraguay has seen a strong increase. In particular since 2003, export volumes have multiplied 6 fold to reach over 236,000 tons. Due to an outbreak of the foot and mouth disease late in 2011 however, the total export volume was 40 percent lower than in the previous year. It is expected that beef exports will recover when Paraguay regains the status of a country without foot and mouth disease by the World Organization for Animal Health. On the one hand, a surge in the international price of beef, on the other hand, the country’s strategy to open up new export markets, may have driven the observed growth of production and export volumes. Relative to 2000, average annual international beef prices were up 50 percent in 2011, reaching almost USD 4000 (Figure 2.2). Following the Central Bank of Paraguay, we use the price of Australian beef to proxy the price of Paraguayan beef. In addition, volatility has decreased again after a period of heightened volatility between October 1999 and January 2003 where the Inclan Tiao (1994) test for volatility breaks detects two significant shifts (see Figure 2.2). In terms of export destinations, Paraguay has achieved a considerable diversification over time with 90 percent of exports going to Chile, Russia, Venezuela, Brazil, and Israel and among others Angola in 2010 (see Arce 2012 and Table 2.3). 67 Figure 2.2: International beef price (USD per ton Figure 2.3: Cattle herd size in million Australian beef) and Inclan Tiao (1994) volatility break points 4500 500 13 450 Monthly USD beef price per ton 4000 12 Stddev. of beef prices 400 Cattle herd size in million 350 11 3500 300 10 3000 250 200 9 2500 150 8 100 2000 7 50 1500 0 6 1997M01 1998M01 1999M01 2000M01 2001M01 2002M01 2003M01 2004M01 2005M01 2006M01 2007M01 2009M01 2010M01 2011M01 2008M01 5 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Source: WB staff, Central Bank of Paraguay Source: Ministry of Agriculture of Paraguay Significant export growth has by far exceeded the expansion of the cattle herd due to a change in the composition of production for internal versus external use and technological change. For example in 2000, 49 percent of the cattle slaughtered to be traded as frozen carcasses were produced for domestic use and 51 percent for export. This compares with 1 percent and 99 percent respectively in 2010. According to Arce (2012), Paraguayan beef has been sold to increasingly high value markets based on technological change which has led to improvement of the genetic pool of the cattle and a more efficient processing of the cattle throughout the production chain. Table 2.3: Main International Export Markets for Paraguayan Beef (2010) Country Chile Russia Venezuela Brazil Israel Other Share in total 49.8 25.1 5.6 3.9 3.0 12.5 beef exports Source: Central Bank of Paraguay. Beef producers have some room to time the export of beef and sell a relatively diversified product. Annual beef exports peak occur mostly in the months of May through July, yet a pattern is not strong. The reason is that beef producers can decide to sell their cattle at different maturities, trading off additional weight gain with expected price developments. In addition, beef products are diversified, i.e. frozen versus fresh meat, or different cuts of the cattle. And different export markets which consume different products have different peak demand periods throughout the year. 68 QS 2. Agriculture supply elasticities (soybean and beef) 1,400,000 1,200,000 Description of the data 1,000,000 As described in more depth in section one, monthly soybean exports (QS) exhibit a marked seasonal pattern; beef exports (QB) much less so800,000 (Figures 2.4). Seasonality is associated to 600,000 nature: a crop with specific harvest periods in the case of soybeans and more abundant pastures in the fall and spring (in the case of beef). Both series exhibit an upward trend (soybean, HPSOY 400,000 and beef HPBEEF) estimated using the Hodrick-Prescott filter (Figure 2.5).31 200,000 0 Figure 2.4: Soybean and Beef Exports (in tons per month) 1998 2000 2002 2004 2006 2008 2010 QS QB 1,400,000 24,000 1,200,000 20,000 1,000,000 16,000 800,000 600,000 12,000 400,000 8,000 200,000 4,000 0 1998 2000 2002 2004 2006 2008 2010 0 1998 2000 2002 2004 2006 2008 2010 QB 24,000 Note: Monthly exports of soybeans (QS) and beef (QB) 20,000 Source: Authors. 16,000 Figure 2.5: Soybean and Beef Export Trends 12,000 20,000 450,000 8,000 400,000 16,000 4,000 0 350,000 12,000 1998 2000 2002 2004 2006 2008 2010 300,000 8,000 250,000 4,000 200,000 0 150,000 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 QSTREND QBTREND 31 The HP trend for beef exports was estimated using monthly information from January 1997 to September 2011. We truncated the sample in September 2011 (instead of using information until December 2011) to avoid giving undue weight to the impact of the outbreak of hoof and mouth disease on exports of beef during October 2011 (it brought them to zero). The inclusion of such an extreme observation in the sample would seriously bias the estimated HP trend during 2011. We are indebted to Hannah Nielsen for having brought to our attention this issue in a previous draft. 69 Prices of soybeans (PSOY) and beef (PB) also exhibit an upward trend (Figure 6). Paraguay is a price-taker in the international market for soybean and beef exports therefore prices can be assumed to be exogenous for the rest of the analysis. This fact greatly simplifies the interpretation of patterns of export-supply response observed in the past decade (in technical jargon changes in the international price of beef and soybeans can be used to econometrically identify a supply function). An increase in international commodity prices is expected to increase supply for two reasons: in the short run farmers have incentives to use more inputs in production (fertilizers etc.); in the medium term, more land and capital in the form of tractors, more efficient harvesting equipment and pesticides are added to the production process. Figure 2.6: Trends in Price of Soybeans and Beef Hodrick-Prescott Filter (lambda=14400) Hodrick-Prescott Filter (lambda=14400) 4,500 600 4,000 500 3,500 3,000 400 800 2,500 200 300 2,000 400 150 1,500 200 100 0 1,000 50 100 -400 0 -800 -50 -1,200 -100 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 PB Trend Cycle PS Trend Cycle Source: Authors. The short run supply response to an increase in prices is what is described in intermediate microeconomics a ‘move along a supply function; the medium-term supply response is a ‘shift to the right’ of the supply function (Box 1). Box 2.1 A Simple Short-run Model of Agriculture Supply We observe monthly exports but we know that exports are constrained by total production. To simplify the interpretation of the data we present below a simple model where producers can export or and total exports ( ) are constrained by overall production ( ): . The problem solved by firms is to maximize profits ( ) given external prices subject to the constraint that exports cannot be higher than production. ( ) Embodies the cost of exporting at time or To further simplify we assume the cost of exporting immediately after production is zero. The cost of exporting at time is ( ) () . The function ( ) embodies transportation, interest rates, storage costs, among others. ( ) ( ) [ ] 70 The equilibrium conditions resulting from this simple model are: ( ) ( ) ( ) ( ) This implies that in equilibrium: ( ) ( ) Equation (1) shows the link between international prices and Paraguay’s production possibilities: an increase in international prices implies an increase in the shadow price of capacity ( ), which has the units of a marginal cost. Secondly (4) is an arbitrage condition: under the assumptions of this simplified model it implicitly defines exports at as a function of the difference between current and expected prices and production costs: ( ) (5) Defines a move along a supply function. The time between production and exports will be a function of the expected path of prices and cost of storage etc. A simple extension of the model making ( ) where is capital used in production (a vector of inputs) would define and implicit demand for capital as a result of changes in international prices and therefore gives content to a medium- long-term shift of the supply of agricultural goods. The series analyzed in what follows are monthly soybean and beef prices called , and monthly soybean and beef exports called , respectively. Further, we clean QS for a statistical outlier by estimating the coefficient of the dummy variable (Du=1 for 2005M07 and 0 otherwise). The new series clean from the outlier effects is QS1 A simple econometric model of soybean and beef supply The canonical model to estimate agriculture supply elasticity is Nerlove (1959) 32. The model consists in two equations: equation (1) posits the desired quantity supplied as a function of the market price; equation (2) posits a partial adjustment model where differences between the actual ( ) and desired supply ( ) adjust gradually over time (in each period only a fraction of the gap adjusts. ( ) ( ) [ ] If we assume that expectations are formed rationally: 32 Nerlove, Marc, 1958, The Dynamics of Supply: Estimation of Farmers' Response to Price. Baltimore: The Johns Hopkins Press. 71 And Plugging (1) into (2): ( ) +( ) Equation (3) can be also written as an Error Correction Model (ECM) if is (1); for instance if follows a random walk: ( ) ( ) [ ( )] The ECM is consistent with inter-temporal optimization when there are quadratic adjustment costs (Nickell, 1985)33. Moreover, Nerlove’s specification is consistent with a form of the Error Correction Model. We start by testing the case of soybeans. To proceed with the estimation of an ECM we start by running an Augmented Dickey-Fuller Test (ADF) on PS, QS1 and on their logarithms, following with the Phillips-Perron test34 on the same variables. The latter yield mixed results. On the one hand, we cannot reject the hypothesis of a unit root or unit root with a drift for either PS or QS1 but we can reject the hypothesis against the alternative of a time trend. However, on the other hand, we cannot reject the hypothesis of a unit root for PS against either alternative based on a Phillips-Perron test. These results are however intuitively appealing. Changes in world demand for soybeans and beef have resulted in an upward shift of these prices in the latter years of the period under analysis. We can thus interpret these results as a structural change in prices. This idea is also consistent with the volatility breaks and hence the significant changes in volatility of the series that we detected in Section One. We thus proceed to estimate equation (10) and ECM of (3) using two alternative methods: (Method 1) the Engle-Granger Method and (Method 2) the Dynamic Ordinary Least Squares (DOLS) as advocated by Stock and Watson (1993)35 (Appendix 1a). Arguably the DOLS method provides the most reliable estimates of supply elasticities among the methods used in this section. We have used alternative ones given that the stochastic context is one where several variables exhibit unit roots; we will privilege DOLS method in the following sections. In the case of soy beans, we find that all Nerlove-Error Correction models exhibit a high and significant correlation between change in price and quantity response (Figure 2.7). The response (elasticity) is higher in the long run than in the short run, and ranges between 1.8 to 11.8, and 0.67 to 0.94, respectively. When interpreting these results, it is important to keep in mind that they are not elasticities of supply of production but rather elasticities of supply of exports vis-à-vis price. 33 Nickell, Stephen (1985), Error Correction, Partial Adjustment and all that: An Expository Note, Oxford Bulletin of Economics and Statistics, 47, 2. 34 Phillips, P.C.B. and P. Perron (1988), “Testing for a Unit Root in Time Series Regression,” Biometrika, 75, 335- 346. 35 The cointegrating regression is estimated adding leads and lags of (see EVIEWS User’s Guide II, pp230). 72 Figure 2.7: Short- and Long-run Soybean Price Elasticities Engle-Granger Dynamic OLS First Second First Second EQ05 EQ06 step: step: step: step: EQ01 EQ02 EQ03 EQ04 Dependent QS1 QS1 LQS1 LQS1 QS1 LQS1 Variable/independent variables PS 565.8 601.0 (124.9) (196.3) RES01 -0.63 (0.21) LPS 0.90 0.94 (0.25) (0.35) RES03 -0.57 (0.17) Elasticity SR 0.67 (min) 0.90 0.71 0.94 (max) Elasticity LR 1.8 (min) 2.09 11.8 2.93 (max) Source: Authors We proceed on the same manner specified above using the data for beef. As such, we start the estimation of an ECM by running the ADF on PB, QB and on their logarithms. In this case we also find mixed results. On the one hand, we cannot reject the hypothesis of a unit root for PB. On the other hand, results are mixed regarding a unit root in quantities (QB). As a second step, we proceed to estimate equation (10) and ECM on equation (3) using the before-mentioned alternative methods for the case of beef and adding Method 3 – Partial Adjustments Method (Appendix 1b). In summary, similarly as in the case of soybeans, we find that all Nerlove-Error Correction models exhibit a significant elasticity of response of quantity of beef exported to price (Figure 2.8). Identically, the response (elasticity) is higher in the long run than in the short run; it ranges between 1.98 to 2.82, and 0.28 to 0.31, respectively. Figure 2.8: Short-and Long-run Beef Price Elasticities Engle-Granger Dynamic OLS Partial Adjustment First Second First Second EQ15 EQ16 EQ09 EQ10 step: step: step: step: EQ13 EQ12 EQ11 EQ12 Dependent QB QB LQB LQB QB LQB QB LQB Variable/independent variables PB 5.95 6.36 0.84 73 (0.58) (1.23) (0.31) LPB 2.49 2.58 0.31 (0.19) (0.40) (0.12) Speed of adjustment: -0.12 RES11(-1) (0.04) Speed of adjustment: -0.07 RES13(-1) (0.04) Speed of adjustment: -0.11 EQ15 (0.04) Speed of adjustment: -0.14 EQ16 (0.05) QB(-1) 0.88 (0.03) LQB(-1) 0.89 (0.03) Elasticity SR ------------- -------- ------ ------- 0.28 0.31 (min) (max) Elasticity LR 1.98 (min) 2.49 2.12 2.58 2.33 2.82 (max) Source: Authors 3. Tax Revenue collection, agriculture and economic activity Tax Revenue collection and level of economic activity This section starts by estimating the relationship between tax revenue collection at constant prices (TAX REAL) and the level of economic activity measured through the IMAE index (IMAEP) (Figure 2.9). Given the importance of the VAT in the Paraguayan tax system, this relationship is central to understand the volatility of tax revenue collection. It is the first step to estimate the cointegration between agriculture commodities of soybeans and beef with tax revenue collection. Figure 2.9: Tax Revenue Collection and Level of Economic 10,000 8,000 180 6,000 160 4,000 140 2,000 120 0 100 80 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 IMAEP TAXREAL Source: Authors 74 We approach the study of this relationship following the same methodology presented in Section Two. As can be anticipated, the Augmented Dickey-Fuller Test (ADF) cannot reject the hypothesis that the IMAEP or TAXREAL have a unit root. Using the DOLS method and a further check of the model using the Johansen (1991, 1995) 36 framework, we find that there is high response of tax revenue collection to the level of economic activity (Figure 2.10) with the elasticity being both at 1.85. However, it is important to pinpoint that there may be other factors than a high response at work, namely the introduction of measures to improve the efficiency of the tax administration during the period under study. Figure 2.10: Summary Elasticity of response: TAXREAL-IMAEP Dynamic OLS Johansen EQ11 Dependent TAXREAL Variable/independent variables IMAEP 65.8 66.0 (3.26) (3.16) Speed of adjustment: RES11(-1) -0.62 (0.24) Elasticity of response 1.85 1.85 Source: Authors Agriculture commodities, cointegration with economic activity and tax revenue collection In order to estimate the correlation between agriculture commodities and tax revenue collection, we start by analyzing the relationship of cointegration between economic activity and exports of soy and beef, following the same two methods described above (DOLS and Johansen). Here we find a one-cointegration equation which shows that exports of beef and soybeans have a significant impact on level of economic activity (Figure 2.11 & 2.12). We then proceed to estimate the relationship between tax revenue collection and agriculture commodities (using exports of soy and beef), with a lags in a DOLS model (Appendix 2). The results show a significant positive correlation of exports of soy and beef with tax revenue collection, with a higher elasticity for beef exports (1.29) than for soy exports (0.15). While the results are expected to be stronger given that we have worked with exports data, they remain quite significant to understand tax revenue collection given the absence of direct taxes on agriculture commodities. Figure 2.11: Johansen Method. Agriculture (XSOY, XBEEF) and Level of Economic Activity (IMAEP) Sample (adjusted): 1997M06 2011M12 Included observations: 175 after adjustments 36 Johansen, Soren (1991), “Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models,” Econometrica, 59, 1551-1580 Johansen, Soren (1995), Likelihood-based Inference in Cointegrated Vector Autoregressive Models, Oxford: Oxford University Press. 75 Trend assumption: Linear deterministic trend Series: IMAEP XBEEF XSOY Lags interval (in first differences): 1 to 4 Unrestricted Cointegration Rank Test (Trace) Hypothesized Trace 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None * 0.313343 75.07475 29.79707 0.0000 At most 1 0.050738 9.288781 15.49471 0.3394 At most 2 0.001008 0.176412 3.841466 0.6745 Trace test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegration Rank Test (Maximum Eigenvalue) Hypothesized Max-Eigen 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None * 0.313343 65.78596 21.13162 0.0000 At most 1 0.050738 9.112369 14.26460 0.2769 At most 2 0.001008 0.176412 3.841466 0.6745 Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I): IMAEP XBEEF XSOY -0.087231 0.001268 0.000852 -0.065688 0.002657 -0.000223 -0.108963 0.001641 -5.03E-05 Unrestricted Adjustment Coefficients (alpha): D(IMAEP) 0.427641 0.525674 0.094465 D(XBEEF) -45.29141 -33.91471 5.674888 D(XSOY) -732.0977 66.30062 1.710752 1 Cointegrating Log -3167.238 76 Equation(s): likelihood Normalized cointegrating coefficients (standard error in parentheses) IMAEP XBEEF XSOY 1.000000 -0.014541 -0.009766 (0.00203) (0.00108) Adjustment coefficients (standard error in parentheses) D(IMAEP) -0.037304 (0.02636) D(XBEEF) 3.950833 (1.67168) D(XSOY) 63.86190 (7.72935) 2 Cointegrating Log Equation(s): likelihood -3162.682 Normalized cointegrating coefficients (standard error in parentheses) IMAEP XBEEF XSOY 1.000000 0.000000 -0.017153 (0.00246) 0.000000 1.000000 -0.508002 (0.14205) Adjustment coefficients (standard error in parentheses) D(IMAEP) -0.071834 0.001939 (0.03269) (0.00088) D(XBEEF) 6.178610 -0.147551 (2.07219) (0.05587) D(XSOY) 59.50677 -0.752439 (9.65891) (0.26041) Source: Authors 77 Figure 2.12: Vector Error Correction Estimates Vector Error Correction Estimates Sample (adjusted): 1997M04 2011M12 Included observations: 177 after adjustments Standard errors in ( ) & t-statistics in [ ] Cointegrating Eq: CointEq1 IMAEP(-1) 1.000000 XBEEF(-1) -0.010349 (0.00399) [-2.59429] XSOY(-1) -0.012876 (0.00176) [-7.31506] C -81.49847 Error Correction: D(IMAEP) D(XBEEF) D(XSOY) CointEq1 -0.022415 0.017794 31.80050 (0.01475) (0.96130) (4.59600) [-1.51942] [ 0.01851] [ 6.91917] D(IMAEP(-1)) -0.466133 -1.317970 -17.40743 (0.07490) (4.88061) (23.3344) [-6.22357] [-0.27004] [-0.74600] D(IMAEP(-2)) -0.257894 -7.664256 -13.15721 (0.07390) (4.81580) (23.0245) [-3.48961] [-1.59148] [-0.57144] D(XBEEF(-1)) -0.001384 -0.077079 0.327454 (0.00125) (0.08144) (0.38935) [-1.10740] [-0.94648] [ 0.84102] D(XBEEF(-2)) -0.002967 -0.151614 0.429337 (0.00128) (0.08311) (0.39736) [-2.32640] [-1.82425] [ 1.08049] D(XSOY(-1)) 3.90E-05 0.036772 0.372552 (0.00023) (0.01495) (0.07149) [ 0.16977] [ 2.45919] [ 5.21127] 78 D(XSOY(-2)) -0.000192 0.001906 0.078945 (0.00025) (0.01608) (0.07687) [-0.77939] [ 0.11854] [ 1.02704] C 0.669991 7.338784 0.338658 (0.30274) (19.7275) (94.3182) [ 2.21309] [ 0.37201] [ 0.00359] R-squared 0.239529 0.078157 0.272092 Adj. R-squared 0.208031 0.039974 0.241942 Sum sq. resids 2689.628 11420891 2.61E+08 S.E. equation 3.989356 259.9601 1242.881 F-statistic 7.604404 2.046907 9.024599 Log likelihood -491.9614 -1231.272 -1508.217 Akaike AIC 5.649281 14.00308 17.13240 Schwarz SC 5.792836 14.14663 17.27595 Mean dependent 0.419209 4.932203 -2.020339 S.D. dependent 4.482792 265.3170 1427.507 Determinant resid covariance (dof adj.) 1.55E+12 Determinant resid covariance 1.35E+12 Log likelihood -3225.057 Akaike information criterion 36.74640 Schwarz criterion 37.23090 Sample (adjusted): 1997M04 2011M12 Included observations: 177 after adjustments Trend assumption: Linear deterministic trend Series: IMAEP XBEEF XSOY Lags interval (in first differences): 1 to 2 Unrestricted Cointegration Rank Test (Trace) Hypothesized Trace 0.05 Prob. No. of CE(s) Eigenvalue Statistic Critical Value ** 0.000 None * 0.237746 58.14091 29.79707 0 0.273 At most 1 0.055082 10.08981 15.49471 9 0.804 At most 2 0.000347 0.061487 3.841466 1 Trace test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level 79 **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegration Rank Test (Maximum Eigenvalue) Hypothesized Max-Eigen 0.05 Prob. No. of CE(s) Eigenvalue Statistic Critical Value ** 0.000 None * 0.237746 48.05110 21.13162 0 0.210 At most 1 0.055082 10.02832 14.26460 1 0.804 At most 2 0.000347 0.061487 3.841466 1 Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegrating Coefficients (normalized by b'*S11*b=I): IMAEP XBEEF XSOY -0.049197 0.000509 0.000633 -0.074089 0.002605 -0.000133 -0.086767 0.000997 -2.25E-05 Unrestricted Adjustment Coefficients (alpha): D(IMAEP) 0.455610 0.516167 0.058023 D(XBEEF) -0.361688 -43.68580 3.221340 D(XSOY) -646.3934 -9.570805 6.813275 1 Cointegrating Equation(s): Log likelihood -3225.057 Normalized cointegrating coefficients (standard error in parentheses) IMAEP XBEEF XSOY 1.000000 -0.010349 -0.012876 (0.00399) (0.00176) Adjustment coefficients (standard error in parentheses) D(IMAEP) -0.022415 (0.01475) D(XBEEF) 0.017794 (0.96130) D(XSOY) 31.80050 (4.59600) 80 2 Cointegrating Equation(s): Log likelihood -3220.043 Normalized cointegrating coefficients (standard error in parentheses) IMAEP XBEEF XSOY 1.000000 0.000000 -0.018997 (0.00240) 0.000000 1.000000 -0.591523 (0.11421) Adjustment coefficients (standard error in parentheses) D(IMAEP) -0.060657 0.001576 (0.02643) (0.00079) D(XBEEF) 3.254437 -0.113969 (1.71189) (0.05108) D(XSOY) 32.50959 -0.354022 (8.30815) (0.24792) D(TAXREAL) -1.090047 (0.13561) D(IMAEP) 0.000528 (0.00064) Source: Authors 4. Conclusions The purpose of this study was to measure the relationship of beef and soybean prices (measured with international prices) with tax revenue collection. The basis for the analysis is the canonical Nerlove model of partial adjustment; the model can be interpreted as an optimal farmer response when there are adjustment costs and farmers build their price expectations rationally. The statistical results show a significant elasticity of response of both beef and soybeans to prices. The meaning of estimated elasticities has to be assessed carefully. First, these are not estimates of the supply of beef and soybeans produced but rather of the quantity of soybean and beef exported. It is not surprising therefore the finding that holders of inventories (soybean seeds and beef herd) respond rapidly to price changes. The case is clearer with soybeans because there is a limit in time that brokers can hold the crop rather than commercialize it in the market. In the case of beef it is possible to withhold supply when facing an increase in prices by way of reducing the supply of cows to the market so as to increase production of calves next year. Second, part of beef and soybeans exports may have been produced in Argentina rather than in Paraguay which 81 would bias upwards the estimated supply elasticity and downwards the elasticity of tax revenue response to agriculture exports.37 In the following sections, we explore the relationship between an index of economic activity and tax revenue collection, as well as between economic activity and exports of beef and soybeans, and find, as expected, a significant and positive relationship. These are the basis for finally analyzing the relationship between tax revenue collection and soybean and beef exports. We find a positive statistically significant response. The result is not trivial. What are the channels for such a positive correlation given the absence of direct taxes on agricultural production in Paraguay? Evidence points at the working of the VAT. Beef and soybean production generates income that is spent, for the most part, inside Paraguay. Part of this expenditure generates tax revenue through the VAT system and another part generates revenue through the corporate tax and other tributes. This result supports the view that being exempt from legally paying taxes does not mean that the income generated by the activity is tax exempt. It would be interesting to continue this analysis by comparing the correlation and volatility that a direct tax on agriculture income would have compared to the current baseline situation. In addition, a detailed study of the value-chain of soybeans and beef production would further complement such an analysis by identifying the different stages for taxation. 5. Bibliography Arce, L. (2012): “La Industria Carnica en Paraguay”, Observatorio de la Economia Internacional, Centro de Análisis y Difusión de la Economia Paraguaya, January 2012. Inclán, C. and G. Tiao (1994): “Use of Cumulative Sums of Squares for Retrospective Detection of Changes of Variance,” Journal of the American Statistical Association, Vol. 89, No. 427 (Sep.), pp. 913-23. USDA (2005): “Crop Production Annual Summary, 2005”, United States Department of Agriculture, National Agricultural Statistics Service, http://usda.mannlib.cornell.edu/MannUsda/viewDocumentInfo.do?documentID=1047, 01.12.2005. USDA (2010): “Crop Production Annual Summary, 2010”, United States Department of Agriculture, National Agricultural Statistics Service, http://usda.mannlib.cornell.edu/MannUsda/viewDocumentInfo.do?documentID=1047, 01.12.2010. World Bank (2010): “Paraguay—Estudio de Pobreza: Determinantes y Desafíos para la Reducción de la Pobreza”, October, The World Bank, Washington D.C. World Bank (2008): “Rising Food Prices—The World Bank’s Latin America and Caribbean Region Position Paper”, The World Bank, Washington D.C. 6. Appendix Appendix 1a- Methods for estimating Soybean Supply 37 We are indebted to John Nash for making this point. The issue can be further explored introducing a measure of changes in constrains and taxation in Argentina in analyzing the Paraguayan data. 82 Method 1: Engle-Granger Method Case 1: Dependent variable is QS1 Engle-Granger Method: QS1 (EQ01 and EQ02) Dependent Variable: QS1 Method: Least Squares Sample: 1997M01 2011M12 Included observations: 180 Variable Coefficient Std. Error t-Statistic Prob. C 79223.82 37300.57 2.123931 0.0351 PS 565.7505 124.9362 4.528316 0.0000 R-squared 0.103300 Mean dependent var 237174.1 Adjusted R-squared 0.098262 S.D. dependent var 186730.3 S.E. of regression 177318.8 Akaike info criterion 27.02033 Sum squared resid 5.60E+12 Schwarz criterion 27.05581 Log likelihood -2429.830 Hannan-Quinn criter. 27.03472 F-statistic 20.50565 Durbin-Watson stat 0.689788 Prob(F-statistic) 0.000011 Dependent Variable: DQS1 Method: Least Squares Sample (adjusted): 1998M02 2011M12 Included observations: 167 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C 4375.026 8593.336 0.509119 0.6114 RES01(-1) -0.630539 0.213663 -2.951082 0.0037 DQS1(-1) 0.252886 0.216010 1.170713 0.2435 DQS1(-2) 0.169613 0.205554 0.825148 0.4106 DQS1(-3) 0.133712 0.188409 0.709688 0.4790 DQS1(-4) 0.071912 0.175445 0.409885 0.6825 DQS1(-5) 0.034883 0.158729 0.219765 0.8263 DQS1(-6) -0.018444 0.146699 -0.125728 0.9001 DQS1(-7) -0.019702 0.134485 -0.146500 0.8837 DQS1(-8) -0.182240 0.122136 -1.492105 0.1377 DQS1(-9) -0.114766 0.110949 -1.034401 0.3026 DQS1(-10) -0.233164 0.099563 -2.341865 0.0205 DQS1(-11) -0.103913 0.089559 -1.160279 0.2477 DQS1(-12) 0.308523 0.081501 3.785517 0.0002 R-squared 0.509831 Mean dependent var 859.9945 Adjusted R-squared 0.468183 S.D. dependent var 149680.7 S.E. of regression 109155.9 Akaike info criterion 26.11905 Sum squared resid 1.82E+12 Schwarz criterion 26.38044 Log likelihood -2166.941 Hannan-Quinn criter. 26.22514 F-statistic 12.24133 Durbin-Watson stat 2.116973 Prob(F-statistic) 0.000000 Breusch-Godfrey Serial Correlation LM Test: F-statistic 1.343211 Prob. F(12,141) 0.2009 Obs*R-squared 17.13225 Prob. Chi-Square(12) 0.1447 83 Case 2: Dependent variable is LQS1 Engle-Granger Method: LQS1 (EQ03 and EQ04) Dependent Variable: LQS1 Sample (adjusted): 1997M02 2011M12 Included observations: 173 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C 6.966472 1.385108 5.029551 0.0000 LPS 0.897993 0.248165 3.618528 0.0004 R-squared 0.071125 Mean dependent var 11.96813 Adjusted R-squared 0.065693 S.D. dependent var 1.213266 S.E. of regression 1.172737 Akaike info criterion 3.168052 Sum squared resid 235.1785 Schwarz criterion 3.204506 Log likelihood -272.0365 Hannan-Quinn criter. 3.182841 F-statistic 13.09374 Durbin-Watson stat 0.937929 Prob(F-statistic) 0.000390 Dependent Variable: DLQS1 Sample (adjusted): 2000M03 2011M12 Variable Coefficient Std. Error t-Statistic Prob. C 0.064500 0.049834 1.294291 0.1979 RES03(-1) -0.568072 0.165633 -3.429706 0.0008 DLQS1(-1) 0.139292 0.168810 0.825141 0.4108 DLQS1(-2) 0.100480 0.158027 0.635841 0.5260 DLQS1(-3) 0.076569 0.147452 0.519281 0.6045 DLQS1(-4) 0.046705 0.135088 0.345739 0.7301 DLQS1(-5) -0.053341 0.123144 -0.433164 0.6656 DLQS1(-6) -0.018964 0.116057 -0.163400 0.8705 DLQS1(-7) -0.000436 0.107052 -0.004069 0.9968 DLQS1(-8) -0.229763 0.097022 -2.368156 0.0194 DLQS1(-9) -0.130907 0.093206 -1.404499 0.1626 DLQS1(-10) -0.194172 0.084852 -2.288356 0.0238 DLQS1(-11) -0.034244 0.082332 -0.415924 0.6782 DLQS1(-12) 0.350045 0.066969 5.226992 0.0000 R-squared 0.673589 Mean dependent var 0.038404 Adjusted R-squared 0.640438 S.D. dependent var 0.975923 S.E. of regression 0.585197 Akaike info criterion 1.859651 Sum squared resid 43.83436 Schwarz criterion 2.151071 Log likelihood -118.0352 Hannan-Quinn criter. 1.978072 F-statistic 20.31878 Durbin-Watson stat 1.897714 Prob(F-statistic) 0.000000 84 Method 2: DOLS Case 3: Dependent variable is QS1 2.1 DOLS Method: QS1 (EQ05) Dependent Variable: QS1 Method: Dynamic Least Squares (DOLS) Sample (adjusted): 1997M03 2011M11 Included observations: 177 after adjustments Cointegrating equation deterministics: C Fixed leads and lags specification (lead=1, lag=1) Long-run variance estimate (Bartlett kernel, Newey-West fixed bandwidth = 5.0000) Variable Coefficient Std. Error t-Statistic Prob. PS 601.0248 196.2522 3.062512 0.0025 C 73269.59 58100.66 1.261080 0.2090 R-squared 0.113683 Mean dependent var 240117.1 Adjusted R-squared 0.093071 S.D. dependent var 186759.2 S.E. of regression 177856.1 Sum squared resid 5.44E+12 Durbin-Watson stat 0.688885 Long-run variance 7.23E+10 Cointegration Test - Engle-Granger Date: 07/19/12 Time: 15:50 Equation: EQ05 Specification: QS1 PS C Cointegrating equation deterministics: C Null hypothesis: Series are not cointegrated Automatic lag specification (lag=13 based on Schwarz Info Criterion, maxlag=13) Value Prob.* Engle-Granger tau-statistic -4.358547 0.0028 Engle-Granger z-statistic 82.69746 0.9999 *MacKinnon (1996) p-values. 85 Case 4: Dependent Variable is LQS1 2.2 DOLS Method LQS1 (EQ06) Dependent Variable: LQS1 Method: Dynamic Least Squares (DOLS) Date: 07/19/12 Time: 15:54 Sample (adjusted): 1997M03 2011M11 Included observations: 171 after adjustments Cointegrating equation deterministics: C Fixed leads and lags specification (lead=1, lag=1) Long-run variance estimate (Bartlett kernel, Newey-West fixed bandwidth = 5.0000) Variable Coefficient Std. Error t-Statistic Prob. LPS 0.935536 0.354301 2.640512 0.0091 C 6.772602 1.975278 3.428683 0.0008 R-squared 0.078445 Mean dependent var 11.97575 Adjusted R-squared 0.056239 S.D. dependent var 1.216802 S.E. of regression 1.182091 Sum squared resid 231.9584 Durbin-Watson stat 0.937490 Long-run variance 2.631735 86 2.3. Cointegration test LQS1 Cointegration Test - Engle-Granger Equation: EQ06 Specification: LQS1 LPS C Cointegrating equation deterministics: C Null hypothesis: Series are not cointegrated Automatic lag specification (lag=12 based on Schwarz Info Criterion, maxlag=13) Value Prob.* Engle-Granger tau-statistic -4.276098 0.0038 Engle-Granger z-statistic -1367.002 0.0000 *MacKinnon (1996) p-values. Intermediate Results: Rho – 1 -0.684719 Rho S.E. 0.160127 Residual variance 0.296738 Long-run residual variance 46.20052 Number of lags 12 Number of observations 160 Number of stochastic trends** 2 **Number of stochastic trends in asymptotic distribution. Engle-Granger Test Equation: Dependent Variable: D(RESID) Method: Least Squares Sample (adjusted): 2000M03 2011M12 Included observations: 142 after adjustments Variable Coefficient Std. Error t-Statistic Prob. RESID(-1) -0.684719 0.170934 -4.005750 0.0001 D(RESID(-1)) 0.263797 0.172485 1.529392 0.1286 D(RESID(-2)) 0.212623 0.161069 1.320070 0.1891 D(RESID(-3)) 0.191917 0.150533 1.274918 0.2046 D(RESID(-4)) 0.154585 0.137039 1.128034 0.2614 D(RESID(-5)) 0.038234 0.123896 0.308601 0.7581 D(RESID(-6)) 0.064176 0.116023 0.553131 0.5811 D(RESID(-7)) 0.070997 0.105855 0.670705 0.5036 D(RESID(-8)) -0.178100 0.094982 -1.875103 0.0630 D(RESID(-9)) -0.094296 0.091029 -1.035896 0.3022 D(RESID(-10)) -0.158992 0.083610 -1.901598 0.0595 D(RESID(-11)) -0.011303 0.081343 -0.138952 0.8897 D(RESID(-12)) 0.366219 0.066012 5.547747 0.0000 R-squared 0.674141 Mean dependent var 0.033311 Adjusted R-squared 0.643829 S.D. dependent var 0.974363 S.E. of regression 0.581501 Akaike info criterion 1.840675 Sum squared resid 43.62047 Schwarz criterion 2.111279 Log likelihood -117.6879 Hannan-Quinn criter. 1.950637 Durbin-Watson stat 1.898010 87 Appendix 1b-Methods for estimating Beef Supply Method 1: Engle-Granger Method Case 1: Dependent variable is QB 1.1 Engle-Granger Method: QB (EQ13 and EQ14) Dependent Variable: QB Method: Least Squares Sample: 1997M01 2011M12 Included observations: 180 Variable Coefficient Std. Error t-Statistic Prob. C -9004.368 1803.042 -4.993988 0.0000 PB 5.954493 0.578460 10.29370 0.0000 R-squared 0.373152 Mean dependent var 9107.247 Adjusted R-squared 0.369630 S.D. dependent var 6656.416 S.E. of regression 5284.918 Akaike info criterion 19.99415 Sum squared resid 4.97E+09 Schwarz criterion 20.02963 Log likelihood -1797.474 Hannan-Quinn criter. 20.00854 F-statistic 105.9602 Durbin-Watson stat 0.233010 Prob(F-statistic) 0.000000 Dependent Variable: DQB Sample (adjusted): 1998M02 2011M12 Included observations: 167 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C 102.6457 184.2170 0.557200 0.5782 RES13(-1) -0.070441 0.036869 -1.910567 0.0579 DQB(-1) -0.122357 0.083636 -1.462965 0.1455 DQB(-2) -0.122140 0.084723 -1.441635 0.1514 DQB(-3) -0.019303 0.092655 -0.208332 0.8352 DQB(-4) -0.117742 0.098931 -1.190146 0.2358 DQB(-5) -0.205947 0.103277 -1.994118 0.0479 DQB(-6) -0.186670 0.101223 -1.844153 0.0671 DQB(-7) -0.248941 0.098957 -2.515635 0.0129 DQB(-8) -0.062551 0.097087 -0.644276 0.5204 DQB(-9) -0.025516 0.100562 -0.253730 0.8000 DQB(-10) -0.042449 0.102447 -0.414354 0.6792 DQB(-11) 0.053232 0.101654 0.523657 0.6013 DQB(-12) 0.191070 0.097109 1.967587 0.0509 R-squared 0.199525 Mean dependent var 42.33413 Adjusted R-squared 0.131511 S.D. dependent var 2467.373 S.E. of regression 2299.412 Akaike info criterion 18.39880 Sum squared resid 8.09E+08 Schwarz criterion 18.66019 Log likelihood -1522.300 Hannan-Quinn criter. 18.50490 F-statistic 2.933583 Durbin-Watson stat 1.982043 Prob(F-statistic) 0.000777 Breusch-Godfrey Serial Correlation LM Test: F-statistic 1.006724 Prob. F(12,141) 0.4460 Obs*R-squared 13.17916 Prob. Chi-Square(12) 0.3562 88 Case 2: Dependent variable is LQB 1.2 Engle-Granger Method: LQB (EQ11 and EQ12) Dependent Variable: LQB Method: Least Squares Sample: 1997M01 2011M12 Included observations: 179 Variable Coefficient Std. Error t-Statistic Prob. C -11.07806 1.552283 -7.136627 0.0000 LPB 2.485542 0.194206 12.79850 0.0000 R-squared 0.480636 Mean dependent var 8.778851 Adjusted R-squared 0.477702 S.D. dependent var 0.910855 S.E. of regression 0.658276 Akaike info criterion 2.012726 Sum squared resid 76.69901 Schwarz criterion 2.048340 Log likelihood -178.1390 Hannan-Quinn criter. 2.027167 F-statistic 163.8017 Durbin-Watson stat 0.431961 Prob(F-statistic) 0.000000 Dependent Variable: DLQB Method: Least Squares Sample (adjusted): 1998M02 2011M09 Included observations: 164 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C 0.029170 0.027802 1.049213 0.2958 RES11(-1) -0.118125 0.044141 -2.676061 0.0083 DLQB(-1) -0.441526 0.082833 -5.330311 0.0000 DLQB(-2) -0.018478 0.090475 -0.204234 0.8384 DLQB(-3) -0.046529 0.089806 -0.518103 0.6052 DLQB(-4) -0.089609 0.090131 -0.994200 0.3217 DLQB(-5) -0.053327 0.089808 -0.593793 0.5535 DLQB(-6) -0.184899 0.085004 -2.175173 0.0312 DLQB(-7) -0.345395 0.084664 -4.079598 0.0001 DLQB(-8) -0.103292 0.088619 -1.165580 0.2456 DLQB(-9) 0.059745 0.088572 0.674531 0.5010 DLQB(-10) -0.036129 0.088531 -0.408090 0.6838 DLQB(-11) -0.038504 0.088350 -0.435808 0.6636 DLQB(-12) 0.095858 0.077626 1.234872 0.2188 R-squared 0.353719 Mean dependent var 0.015574 Adjusted R-squared 0.297707 S.D. dependent var 0.410146 S.E. of regression 0.343715 Akaike info criterion 0.783491 Sum squared resid 17.72096 Schwarz criterion 1.048113 Log likelihood -50.24623 Hannan-Quinn criter. 0.890917 F-statistic 6.315154 Durbin-Watson stat 1.979246 Prob(F-statistic) 0.000000 Breusch-Godfrey Serial Correlation LM Test: F-statistic 1.271325 Prob. F(12,138) 0.2422 Obs*R-squared 16.32542 Prob. Chi-Square(12) 0.1768 89 Method 2: DOLS Case 3: Dependent variable is QB 2.1 Engel-Granger QB (EQ15) & Cointegration Test: Engle-Granger QB Method: Dynamic Least Squares (DOLS) Date: 07/20/12 Time: 10:23 Sample (adjusted): 1997M03 2011M11 Included observations: 177 after adjustments Cointegrating equation deterministics: C Fixed leads and lags specification (lead=1, lag=1) Long-run variance estimate (Bartlett kernel, Newey-West fixed bandwidth = 5.0000) Variable Coefficient Std. Error t-Statistic Prob. PB 6.358035 1.227968 5.177686 0.0000 C -10116.75 3830.538 -2.641077 0.0090 R-squared 0.394011 Mean dependent var 9194.568 Adjusted R-squared 0.379918 S.D. dependent var 6668.695 S.E. of regression 5251.284 Sum squared resid 4.74E+09 Durbin-Watson stat 0.218802 Long-run variance 1.13E+08 Cointegration Test - Engle-Granger Date: 07/20/12 Time: 10:22 Equation: UNTITLED Specification: QB PB C Cointegrating equation deterministics: C Null hypothesis: Series are not cointegrated Automatic lag specification (lag=0 based on Schwarz Info Criterion, maxlag=12) Value Prob.* Engle-Granger tau-statistic -3.155096 0.0828 Engle-Granger z-statistic -20.01492 0.0490 *MacKinnon (1996) p-values. Intermediate Results: Rho - 1 -0.111815 Rho S.E. 0.035440 Residual variance 6163361. Long-run residual variance 6163361. Number of lags 0 Number of observations 179 Number of stochastic trends** 2 **Number of stochastic trends in asymptotic distribution. Engle-Granger Test Equation: Dependent Variable: D(RESID) Method: Least Squares Date: 07/20/12 Time: 10:22 Sample (adjusted): 1997M02 2011M12 Included observations: 179 after adjustments Variable Coefficient Std. Error t-Statistic Prob. RESID(-1) -0.111815 0.035440 -3.155096 0.0019 90 Case 4: Dependent variable LQB 2.2 DOLS LQB Engel-Granger QB (EQ16) Dependent Variable: LQB Method: Dynamic Least Squares (DOLS) Sample (adjusted): 1997M03 2011M11 Included observations: 176 after adjustments Cointegrating equation deterministics: C Fixed leads and lags specification (lead=1, lag=1) Long-run variance estimate (Bartlett kernel, Newey-West fixed bandwidth = 5.0000) Variable Coefficient Std. Error t-Statistic Prob. LPB 2.583970 0.398142 6.490068 0.0000 C -11.84672 3.184806 -3.719761 0.0003 R-squared 0.501517 Mean dependent var 8.791210 Adjusted R-squared 0.489857 S.D. dependent var 0.909627 S.E. of regression 0.649695 Sum squared resid 72.17979 Durbin-Watson stat 0.413693 Long-run variance 1.632440 2.3 Cointegration Test Engle-Granger LQB Cointegration Test - Engle-Granger Date: 07/20/12 Time: 10:33 Equation: EQ16 Specification: LQB LPB C Cointegrating equation deterministics: C Null hypothesis: Series are not cointegrated Automatic lag specification (lag=1 based on Schwarz Info Criterion, maxlag=12) Value Prob.* Engle-Granger tau-statistic -2.977623 0.1210 Engle-Granger z-statistic -17.79714 0.0786 *MacKinnon (1996) p-values. Intermediate Results: Rho - 1 -0.138596 Rho S.E. 0.046546 Residual variance 0.143211 Long-run residual variance 0.076235 Number of lags 1 Number of observations 176 Number of stochastic trends** 2 **Number of stochastic trends in asymptotic distribution. Engle-Granger Test Equation: Dependent Variable: D(RESID) Method: Least Squares Date: 07/20/12 Time: 10:33 Sample (adjusted): 1997M03 2011M09 Included observations: 175 after adjustments Variable Coefficient Std. Error t-Statistic Prob. RESID(-1) -0.138596 0.046680 -2.969054 0.0034 D(RESID(-1)) -0.370603 0.070826 -5.232622 0.0000 R-squared 0.231076 Mean dependent var -0.002680 Adjusted R-squared 0.226632 S.D. dependent var 0.431565 S.E. of regression 0.379524 Akaike info criterion 0.911567 Sum squared resid 24.91871 Schwarz criterion 0.947736 Log likelihood -77.76213 Hannan-Quinn criter. 0.926238 Durbin-Watson stat 2.031517 91 Method 3: Partial Adjustment Case 1: Dependent variable is QB 1.3 Partial Adjustment Model QB (EQ09) Dependent Variable: QB Method: Least Squares Sample (adjusted): 1997M02 2011M12 Included observations: 179 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C -1393.179 813.6832 -1.712188 0.0886 DUBEEF -11167.99 2212.484 -5.047718 0.0000 QB(-1) 0.883404 0.030955 28.53843 0.0000 PB 0.836188 0.307485 2.719440 0.0072 R-squared 0.893600 Mean dependent var 9146.169 Adjusted R-squared 0.891776 S.D. dependent var 6654.516 S.E. of regression 2189.160 Akaike info criterion 18.24252 Sum squared resid 8.39E+08 Schwarz criterion 18.31374 Log likelihood -1628.705 Hannan-Quinn criter. 18.27140 F-statistic 489.9144 Durbin-Watson stat 2.085262 Prob(F-statistic) 0.000000 And reject the hypothesis of no autocorrelation of residuals: Breusch-Godfrey Serial Correlation LM Test: F-statistic 0.741432 Prob. F(2,173) 0.4779 Obs*R-squared 1.521254 Prob. Chi-Square(2) 0.4674 92 Case 2: Dependent variable is LQB 2.1 Partial Adjustment Model LQB (EQ10) Dependent Variable: LQB Method: Least Squares Sample (adjusted): 1997M02 2011M12 Included observations: 177 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C -3.007366 1.050019 -2.864106 0.0047 LQB(-1) 0.778105 0.043381 17.93647 0.0000 LPB 0.621724 0.158821 3.914616 0.0001 R-squared 0.820826 Mean dependent var 8.788678 Adjusted R-squared 0.818766 S.D. dependent var 0.910898 S.E. of regression 0.387783 Akaike info criterion 0.960064 Sum squared resid 26.16540 Schwarz criterion 1.013897 Log likelihood -81.96563 Hannan-Quinn criter. 0.981896 F-statistic 398.5611 Durbin-Watson stat 2.598695 Prob(F-statistic) 0.000000 Dependent Variable: LQB Method: Least Squares Sample (adjusted): 1997M03 2011M09 Included observations: 175 after adjustments Convergence achieved after 10 iterations MA Backcast: 1997M02 Variable Coefficient Std. Error t-Statistic Prob. C -1.539374 0.734191 -2.096696 0.0375 LQB(-1) 0.894515 0.033502 26.70027 0.0000 LPB 0.309868 0.115582 2.680934 0.0081 AR(1) -0.443502 0.164740 -2.692141 0.0078 MA(1) 0.016010 0.193020 0.082945 0.9340 R-squared 0.848294 Mean dependent var 8.794876 Adjusted R-squared 0.844724 S.D. dependent var 0.910933 S.E. of regression 0.358954 Akaike info criterion 0.816908 Sum squared resid 21.90410 Schwarz criterion 0.907330 Log likelihood -66.47944 Hannan-Quinn criter. 0.853586 F-statistic 237.6465 Durbin-Watson stat 1.984197 Prob(F-statistic) 0.000000 Inverted AR Roots -.44 Inverted MA Roots -.02 We cannot reject the hypothesis of no autocorrelation of residuals. Breusch-Godfrey Serial Correlation LM Test: F-statistic 0.112022 Prob. F(2,168) 0.8941 Obs*R-squared 0.233068 Prob. Chi-Square(2) 0.8900 93 Appendix 2 An alternative method to explore the effect of exports of beef and soybeans and tax revenue collection uses distributed lags. To test whether XBEEF and XSOY have any impact on tax revenue collection at constant prices (TAXREAL), we run the following regression: Dependent Variable: TAXREAL Method: Least Squares Date: 07/06/12 Time: 15:15 Sample (adjusted): 1997M04 2011M12 Included observations: 177 after adjustments Variable Coefficient Std. Error t-Statistic Prob. C 2829.191 132.4115 21.36666 0.0000 XBEEF(-1) -0.026936 0.286278 -0.094091 0.9251 XBEEF(-2) 0.307507 0.379407 0.810493 0.4188 XBEEF(-3) 1.010793 0.285841 3.536209 0.0005 XSOY(-1) 0.060007 0.057942 1.035636 0.3018 XSOY(-2) 0.183009 0.079876 2.291148 0.0232 XSOY(-3) -0.090690 0.058992 -1.537319 0.1261 R-squared 0.641217 Mean dependent var 4536.861 Adjusted R-squared 0.628554 S.D. dependent var 1528.639 S.E. of regression 931.6498 Akaike info criterion 16.55054 Sum squared resid 1.48E+08 Schwarz criterion 16.67615 Log likelihood -1457.722 Hannan-Quinn criter. 16.60148 F-statistic 50.63739 Durbin-Watson stat 1.672192 Prob(F-statistic) 0.000000 The effect of XBEEF and XSOY on TAXREAL can be approximated as: An alternative approximation is to estimate the same tax revenue collection equation using the Almon lag (polynomial distributed lag) procedure38: Dependent Variable: TAXREAL Method: Least Squares Date: 07/06/12 Time: 15:29 Sample (adjusted): 1997M04 2011M12 Included observations: 177 after adjustments 38 See EVIEWS7: User’s Guide (2009), pp24. 94 Variable Coefficient Std. Error t-Statistic Prob. C 2832.901 137.6035 20.58741 0.0000 PDL01 -0.161362 0.192425 -0.838573 0.4029 PDL02 -0.137448 0.209411 -0.656354 0.5125 PDL03 0.373911 0.183415 2.038603 0.0430 PDL04 0.131335 0.035817 3.666862 0.0003 PDL05 0.091364 0.040187 2.273485 0.0242 PDL06 -0.094944 0.037095 -2.559454 0.0114 R-squared 0.644064 Mean dependent var 4536.861 Adjusted R-squared 0.631502 S.D. dependent var 1528.639 S.E. of regression 927.9458 Akaike info criterion 16.54257 Sum squared resid 1.46E+08 Schwarz criterion 16.66818 Log likelihood -1457.017 Hannan-Quinn criter. 16.59351 F-statistic 51.26909 Durbin-Watson stat 1.716901 Prob(F-statistic) 0.000000 Lag Distribution of XBEEF i Coefficient Std. Error t-Statistic . * | 0 0.35000 0.24369 1.43624 *. | 1 -0.16136 0.19242 -0.83857 .* | 2 0.07510 0.18864 0.39811 . *| 3 1.05938 0.24911 4.25262 Sum of Lags 1.32312 0.10202 12.9697 Lag Distribution of XSOY i Coefficient Std. Error t-Statistic * . | 0 -0.05497 0.04920 -1.11744 . *| 1 0.13133 0.03582 3.66686 . *| 2 0.12776 0.03535 3.61394 * . | 3 -0.06571 0.04983 -1.31871 Sum of Lags 0.13841 0.04906 2.82142 The SUM of LAGS estimate is very close to the approximation of the effect of XBEEF and XSOY using three lags estimated above. 95 Appendix 3 Table on Taxes on production by sector % of GDP 2003 2004 2005 2006 2007 2008 2009 2010 Agriculture 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Cattle 0,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 Forestry 0,0 0,0 0,1 0,1 0,0 0,0 0,0 0,0 Fishing 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Minery 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Meat production 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Production of vegetable oils 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Dairy production 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Milling and baking 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Sugar 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Other foods 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Beverages and Tobacco 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Textile and clothes 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Leather and footwear 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Wood industry 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Paper and paper products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Petroleum refinery 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Chemical products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Manufacturing on non-metal products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Manufacturing of common medal products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Manufacturing of common medal products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Manufacturing of other products 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Electricity and Water 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Construction 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Trade 0,1 0,1 0,2 0,2 0,1 0,1 0,1 0,1 Transportation 0,2 0,2 0,4 0,3 0,2 0,2 0,2 0,3 Communication 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Financial intermediation 0,3 0,3 0,7 0,6 0,3 0,4 0,4 0,5 Housing rental 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Services to businesses 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 restaurants and hotels 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Services to households 0,0 0,0 0,1 0,1 0,1 0,1 0,1 0,1 Government services 0,0 0,0 0,0 0,0 0,0 0,0 0,0 0,0 Sectoral total 0,7 0,8 1,8 1,5 0,8 0,9 0,9 1,1 Taxes on products 8,2 9,1 8,8 9,0 9,2 9,4 9,7 10,5 TOTAL 9,0 9,9 10,6 10,5 10,0 10,3 10,7 11,6 Source: Central Bank of Paraguay. 96 Chapter 3. Evolution and Composition of Tax Revenue in Paraguay, Effects of the Tax Reform of 2004, by Osvaldo Schenone Executive Summary This study analyzes why tax revenue in Paraguay barely exceeded 13 percent of GDP in 2011— only a modest increase from the 10 percent collected in 1994—despite the tax reforms of 1992 and 2004. Tax revenue today in Paraguay is, along with Ecuador and Guatemala, among the lowest in Latin America as a share of GDP. The analysis focuses on the design of tax policy. Evasion and deficient administration are other important factors to consider, but are beyond the scope of this study. 1. Situation prior to the tax reform of Law 2421/04. 2012 is the 20th anniversary of the 1992 tax reform (Law 125/91), which defined the tax system until the passage of Law 2421/04. Following Law 125/91, four taxes generated more than 90 percent of total revenue: the value-added tax (VAT), corporate income tax (personal income tax did not exist), selective consumption taxes and customs import duties. Total revenue never exceeded 11.5 percent of GDP, and stabilized at around 10.5 percent of GDP between 1996 and 2003. A) Agricultural income tax (impuesto sobre la renta del sector agropecuario—IMAGRO). The IMAGRO rate was 0.9 percent of the official land valuation. The largest sector of the Paraguayan economy paid almost nothing, since official land values were much lower than the real values. For example, a 100 hectare property in Oveido had a 1997 official value of G. 116,691 per hectare and paid on 80 hectares, meaning that the annual IMAGRO payment was G. 101,517, or less than US$3 per month. B) Commerce, industry and services income tax (impuesto sobre la renta de las actividades de comercio, industrias y servicios—IRACIS). The general tax rate was 30 percent, but re- invested profits were subject to a 10 percent rate. Foreign companies paid 35 percent. In 1994, the recently-created IRACIS generated 2 percent of GDP in revenue. Ten years later, revenue had declined to 1.6-1.7 percent of GDP, from a maximum of 2.22 percent in 1995. C) Value-added Tax (VAT). Almost immediately after its adoption, the VAT of Law 125/91 produced more than 40 percent of total tax revenues from a rate of 10 percent, one of the lowest rates in Latin America, exceeding only Panama’s 5 percent rate. D) Selective consumption taxes (impuestos selectivos al consumo—ISC). These taxes applied to tobacco, alcohol, alcohol for fuel and fuels derived from petroleum. The immediate effect of the adoption of these taxes was approximately 1 percent of GDP, two-thirds of which was generated by fuel taxes. By the end of the 1990s, these taxes generated revenues equal to about 1.3 percent of GDP, half of which came from fuels. 97 2. The reform of Law 2421/04. Law 2421, passed on 5 July 2004, was an ambitious and broad reform, modifying exemptions to the VAT, ISC and customs duties, creating a personal income tax, replacing the Single Tax with an income tax for small payers and modifying both the IRACIS and IMAGRO. A) Creation of the personal income tax (PIT). The creation of a PIT according to Law 2421 was repeatedly postponed and did not enter into effect until 1 January 2013, according to Law 4.064/10, Art. 38. Individuals and small associations will pay the tax, and the taxable base will be on activities generating Paraguayan personal income. The rate is 10 percent on net income above 10 minimum monthly salaries and 8 percent on net income between three and 10 minimum monthly salaries. Income below three minimum monthly salaries is exempt. B) Creation of income tax for small payers (impuesto sobre la renta para pequeños contribuyentes—IRPC). The IRPC taxes net income equal to 30 percent of gross income with a rate of 10 percent. This tax substitutes the Single Tax (which itself replaced the VAT and PIT), freeing single-person companies from the PIT and permitting access to a simplified VAT. It establishes an annual maximum income of G. 100 million to remain in this system. C) Agricultural income tax (impuesto sobre la renta del sector agropecuario—IMAGRO). The official valuation of a property is no longer used as the basis for determining the tax and a net income criterion is used instead, although a system for properties with a surface area of less than a certain number of hectares is still used in some areas of the country. In 2005-2011 revenue practically disappeared. D) Commerce, industry and services income tax (impuesto sobre la renta de las actividades de comercio, industrias y servicios—IRACIS). Law 2421/04 reduced the tax rate on company profits to 20 percent in 2005 and 10 percent as of 2006. A rate of 5 percent is additionally applied to distributed profit, or 15 percent on profits distributed abroad. IRACIS revenue rose from 2 percent to 2.5 percent of GDP (reaching a maximum of 3 percent in 2009) after enacting Law 2421, despite the rate reduction. Although not as dramatic as the VAT performance, IRACIS also showed a favorable trend, with revenues rising 25 percent as a share of GDP in seven years. Law 2421/04 also modified Law 60/90, establishing that exonerations of taxes on profits, sending dividends abroad, interests, commissions and capital will only apply to foreign investment of at least US$5 million, and the taxes on these dividends and profits would not be a fiscal credit by the investor in the country of origin. In the previous tax system, no conditions on minimum investment amount or lack of fiscal credit in country of origin existed. E) Value-added Tax (VAT). The VAT has continued to increase collections (40 percent in seven years) and is, as a result, the tax that caused total collections to rise from 12 percent of GDP before the reform to 13.2 percent of GDP in 2011. 98 While the 10 percent rate is maintained, a differential rate of 5 percent is now also used for: a) transfer of assets and the sale of property; b) sale of goods making up the family consumption basket (rice, noodles, yerba mate, cooking oils, milk, eggs, uncooked meat, flour, salt and pharmaceutical products); and c) interests, commissions and charges on loans and financing. The Fuel Regime also incorporates a differential rate and an exemption from VAT. In import and sales transactions of fuel alcohol, pure alcohol, biodiesel, gasoline and gasoil with sulfur content above 2000 ppm, a VAT rate of 20 percent is applied. For petroleum derivatives, import and sale of fuels and crude continue to be exempt from VAT since the VAT that was supposed to be applied since 1 January 2009 is still suspended. F) Selective consumption taxes (impuestos selectivos al consumo—ISC). Law 2421/04 modified the tax rates and expanded the list of goods covered by the ISC. G) Customs Import Duties. Customs duties are a part of the country’s external trade policy and trade agreements, such as Mercosur, and as such should be considered as separate from tax policy. Consequently, the 2004 reform did not introduce changes and collections remained between 1.5 and 2 percent of GDP before and after the reform. 3. Conclusions and Recommendations While the reform of Law 125/91 increased collection from 8 percent of GDP in 1991 to nearly 12 percent of GDP in 1995 due to the introduction of the VAT—an increase of nearly 50 percent in five years—the reform of Law 2421/04, by contrast, only raised collection from 12 percent to 13.2 percent in seven years. While the reform of 1991 exploited the opportunity of introducing the VAT, the reform of 2004 did not exploit any new revenue-generating opportunities. The most important opportunity to improve collection and overall economic efficiency, without which reform efforts will necessary have few significant results, is taxation of the agricultural sector. This sector contributes 5 to 6 percent of tax collections, despite generating more than 25 ercent of GDP. Without an IMAGRO that collects a share of tax proportionate to the size of the sector in the economy, and without eliminating the VAT exemption on sector products, tax collections as a share of GDP will continue to evolve slowly, as has occurred in the past 20 years. The main recommendations and an estimation of the collection impacts that available information permits are presented in the Executive Summary Table. The collection impacts that could not be estimated due to the lack of data are indicated with a positive or negative sign, according to whether they would increase or reduce tax collection. 99 Executive Summary Table Estimated Recommendations Collection Impact (percent of GDP) PIT Applying Law 4064/10—two impacts on collection: 1) the direct impact of PIT collection itself; and 2) the indirect impact of greater collection in other 1.0 taxes due to the higher degree of formalization in the economy. IMAGRO 1) Eliminate the categories of large and medium properties. Establish for all contributors (apart from small contributors with less than 20 hectares in the + eastern region or 100 hectares in the western region) the obligation of paying the tax through a mechanism of gross income minus expenses. + 2) Eliminate the deduction for cattle ranching loss of life up to 3 percent of the value of the cattle without requiring proof. + 3) Eliminate the deduction of personal and family expenses and investments. + 4) Eliminate the deduction for expenses on neighboring properties. IRACIS - 1) Permit the deduction of fiscal losses during the subsequent five years 0.01 2) Replace the 1 percent regimen for import-assembly businesses (maquila) and tax profits identical to those of other businesses. + 3) Eliminate the benefits of Law 60/90 for dividends and profits and interest payments abroad. VAT 0.30 1) Terminate the suspension of VAT application to fuel imports. 0.29 2) Eliminate the 5 percent VAT. 0.33 3) Eliminate VAT exemptions for the agricultural sector. 0.25 4) Eliminate VAT exemptions on interest charges. ISC 1) Determine the tax base on public sale price and not on the factory price, at + least in the case of cigarettes. - 2) Eliminate the tax on assets now charged at the 1 percent rate. + 3) Tax auto imports at 10 percent. Import Duties Eliminate the exemptions on capital imports (Law 60/90) and primary 0.15 materials (primary materials regime). 100 Introduction This study analyzes why tax revenue in Paraguay barely exceeded 13 percent of GDP in 2011, a modest increase from the 10 percent collected in 1994, despite the tax reforms of 1992 and 2004. Tax revenue today in Paraguay is, along with Ecuador and Guatemala, among the lowest in Latin America as a share of GDP (Table 3.1). The analysis focuses on the design of tax policy. Evasion and deficient administration are other important factors to consider, but are beyond the scope of this study. While Paraguay and Ecuador had similar GDP per capita levels during 2000-2011, only above Bolivia, and face similar tax pressures (less than Bolivia) over the same period, opportunities exist in Paraguay that have not been taken advantage of, as will be discussed in this study. Table 3.1: Tax Revenues in Latin America (percent of GDP) and Per Capita GDP (in US$) GDP per Before Law After Law 2421/04 GDP per capita (2000) 2421/04 (2000) capita Country (2008) (2010) (2011) Argentina* 7733 18.1 25.5 25.9 10945 Bolivia 998 12.3 19.0 24.5 2315 Brazil** 3762 20.0 26.7 25.5 12789 Chile 5174 16.3 18.6 19.9 14278 Colombia 2480 11.2 13.5 17.7 7132 Ecuador 1275 10.2 12.5 n.d. 4424 Paraguay 1335 10.8 12.4 13.5 3252 Peru 2116 12.3 15.6 15.6 5782 Venezuela 4845 8.6 13.5 n.d. 10610 Uruguay 6914 15.2 17.1 n.d. 13914 Average 13.5 17.4 20.4 *Not including social security **Including sub-national governments Source: (1) Zarate, W., Efectividad de la Política Tributaria en Paraguay. Antes y después de la reforma (CADEP, January 2010) Annex VII, p. 39. (2) Zarate, W., Análisis del sistema tributario Paraguayo (CADEP, March 2011) Table 1, p. 6. GDP from IMF World Economic Outlook. 101 1. Situation Prior to Tax Reform Law 2421/04 2012 is the 20th anniversary of the tax reform implemented in 1992 (Law 125/91). This law defined the Paraguayan tax system until the passage of Law 2421/04. Following Law 125/91, four taxes generated more than 90 percent of total revenue: the value- added tax (VAT), corporate income tax (personal income tax did not exist), selective consumption taxes and customs import duties. Total revenue never exceeded 11.5 percent of GDP, and stabilized at around 10.5 percent of GDP between 1996 and 2003. The value-added tax (VAT) was adopted with Law 125/91, and immediately became the highest collecting tax in Paraguay (about half of revenue), replacing the sales tax that had never achieved even 1 percent of GDP in revenue. The reform of Law 125/91 did not include a personal income tax, the incorporation of the agricultural sector and fuels in the VAT, nor the application of the VAT on all sectors of the economy. Despite the omissions in Law 125/91, tax collection rose from 8 percent of GDP in 1990 to 11.45 ercent in 1995 (Table 3.2). This increase was almost completely due to revenue increases from the VAT. However, by 1995 collections had stagnated and never again in the 1990s reached 11 percent of GDP, and was around 10 percent during 2000-2003. Table 3.2: Paraguay: Tax Collection 1990-2003 (percent of GDP) 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 IRACIS 1.14 1.14 1.37 1.43 1.97 2.22 2.16 1.94 1.93 n.d. 1.80 1.60 1.70 1.60 IMAGRO - - - - - - 0.06 0.09 0.09 n.d. 0.10 0.10 0.10 0.00 IVA - - 1.35 3.42 3.98 4.33 4.11 4.54 4.51 n.d. 4.70 4.60 4.40 4.70 ISC (Fuels) 0.88 0.89 0.90 0.84 0.81 0.84 0.98 0.80 0.67 n.d. 1.30 1.60 1.20 1.70 ISC (Other) 0.56 0.56 0.35 0.26 0.34 0.36 0.41 0.50 0.61 n.d. 0.60 0.60 0.60 0.40 Customs 1.70 1.78 1.65 1.63 1.85 2.70 2.16 2.22 2.16 n.d. 1.90 1.80 1.60 1.60 Other taxes 3,76 4.35 2.93 1.08 1.04 1.00 0.87 0.83 0.77 n.d. 0.50 0.40 0.40 0.30 Total 8.04 8.72 8.55 8.66 9.99 11.45 10.75 10.92 10.74 n.d. 10.90 10.70 10.00 10.3 Sources: (1) Secretaría de Estado de Tributación. (2) Shome, P., Haindl, E., Schenone, O. and Spahn, P. “Paraguay: Estrategia de la Re forma del Sistema Tributario” (IMF, Public Finance Department, April 1999), Table 2, p. 19. (3)Varsano, R, Fenochietto, R. and Ago stini, C., Paraguay: 102 Agricultural income tax (impuesto sobre la renta del sector agropecuario—IMAGRO) Prior to the reform of Law 125/91, the agricultural sector had been practically exempt from income tax. While the law established a legal basis for taxing the sector, for several reasons it was not applied until 1996. The law established tax payments based on official property valuations. For all properties small than 100 hectares, the first 20 hectares were exempt from income tax. The total number of contributors was estimated in 1996 at 5,862, an amount susceptible to manageable oversight.39 Gross income was presumed to be equal to 12 percent of official property valuation (not counting forest and lake areas), and expenses were assumed to be 70 percent of gross income (of which 30 percent was to be demonstrated with receipts containing VAT on inputs). In consequence, net income was equivalent to 3.6 percent of official valuation (assuming that the farmer or rancher had sufficient VAT input receipts). Net income was taxed at 25 percent; hence the IMAGRO was equal to 0.9 percent of the property’s official valuation. Estimates made with data from 1997 (when IMAGRO collections were 0.09 percent of GDP) demonstrate the deficient collection performance of this tax. That year the agricultural sector generate value added equivalent to 21.4 percent of GDP. Discounting salaries, workers’ own income and profits declared by agricultural firms, agricultural income was almost 15 percent of GDP. With IMAGRO collections at 0.09 percent of GDP, the average taxation in the sector was approximately 0.6 percent of agricultural income (Table 3.3).40 Table 3.3: Paraguay: IMAGRO Collection 1996-2003 ( percent of GDP) Year 1996 1997 1998 1999 2000 2001 2002 2003 IMAGRO 0.06 0.09 0.09 0.10 0.10 0.10 0.00 Source: Table 2. The reason why the largest sector of the Paraguayan economy was not contributing significantly is that official valuations are far from the actual economic value of properties. A property of 100 hectares in Oviedo, one of the best in the country, had a 1997 official valuation of G. 116,691 per hectare and paid tax on 80 hectares, totaling G. 9,335,280, resulting in an annual IMAGRO of G. 101,517, or less than US$3 per month. A 2,000 hectare property in Villa Ygatimi, in the Canindeyu department, had a 1997 official valuation of G. 126,777 per hectare, or a total valuation of G. 253,554,000, with an IMAGRO of G. 2,281,986 per year or a bit more than US$67 a month. In 1999 the official valuation of all registered rural properties, with a total surface area of 36.37 million hectares, was less than US$650 million, or about US$18 per hectare.41 39 Spahn, P. Haindl, E. and Schenone, O. “Paraguay: Perfeccionamiento del Sistema Tributario” (IMF Public Finance Department, March 1996), Table 2, p. 29. 40 Ibid., p. 30. 41 Ibid. Table 8, p. 32. 103 Commerce, industry and services income tax (impuesto sobre la renta de las actividades de comercio, industrias y servicios—IRACIS) IRACIS42 taxes Paraguayan income from commercial, industrial and service business that are not personal in nature. Applying a territorial principal on the source of income reduces the tax base, by not taxing income by businesses or people in Paraguay obtained in another country, and also due to the difficulty of controlling evasion as it incentivizes declaring income actually generated in Paraguay as foreign-earned. Contributors are single-person companies, associations with or without legal standing, including public companies and people domiciled abroad, and branches, agencies or establishments in the country and cooperatives as defined by law.43 The general tax rate is 30 percent, while re-invested income is subject to a 10 percent rate. Foreign companies pay 35 percent. In 1994, with the IRACIS recently implemented, collection amounted to 2 percent of GDP. Ten years later collection had declined to 1.6-1.7 percent of GDP, with a maximum collection of 2.22 percent in 1995 (Table 3.4). Table 3.4: Paraguay: IRACIS Collection 1994-2003 (percent of GDP) Year 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 IRACIS 1.97 2.22 2.16 1.94 1.93 1.80 1.60 1.70 1.60 Source: Table 2. Laws 60/90 (on capital investment) and 1064 of 1997 (on tax regime for assembly plants, or maquilas) are the two sources of erosion to IRACIS collections. The first establishes an income tax discount of 95 percent for five years on all investments defined by Law 60/90. During the 1990s, the Investment Council approved more than 3,100 projects to benefit from this regime.44 The maquila regime replaced 30 percent IRACIS payments with a 1 percent rate on the difference in value between exports and imports of maquila companies. Value-added tax (VAT) The VAT of Law 125/91 produced more than 40 percent of total tax collections almost immediately after being introduced with a rate of 10 percent, one of the lowest rates in Latin America, above only Panama (5 percent). However, from the start the VAT has been subject to exemptions that have limited revenue collection as well as economic efficiency. The most significant exemptions are applied to: (1) fuels derived from petroleum and the import of crude oil; (2) unrefined agricultural products; (3) investments protected by Law 60/90 as fiscal incentives; (4) goods imported by investors for the 42 IRACIS is regulated by Law 195/91, Art. 1 to 25, with modifications introduced by Law 2421/04 and Decree 6359/05. 43 Law 438/94. 44 Shome, P., Haindl, E., Schenone, O. and Spahn, P., op. cit., p. 22. 104 initial installation process; and (5) the import-for-export regime (known as “the tourism regime”), which applies a fraction of the VAT rate (50, 20 or 2 percent) for imported goods supposing that 50, 80 or 98, respectively, of those goods are exported without any VAT credit. The exemptions with the greatest negative impact on collections are the first two, as the other three replace (imperfectly) normal procedures for applying VAT on sales and recognize, simultaneously, a VAT credit to the purchaser (when this is subject to taxation; that is, is not the final consumer or exporter).45 The first two exemptions lead to collection losses proportional to the value of exempted goods (agricultural and fuel) that are sold to final consumers. Since crude oil is not a final consumption product, this exemption did not generate any collection loss (VAT that would be paid on crude imports would be immediately credited when paying the VAT on fuels, without any collection impact). A 1999 estimate indicated that the collection loss attributed to the agricultural sector equaled approximately 0.5 percent of GDP in 1997.46 Selective consumption taxes (impuestos selectivos al consumo—ISC) These taxes cover tobacco, alcoholic beverages, fuel alcohol and petroleum-based fuels. The rates established by Law 125/91 for tobacco products were 8 and 7 percent, according to whether they were cigarettes or other products; 8 percent for beverages with low alcohol content (less than 2 percent) and beer, and 10 percent for other alcoholic beverages. Alcohol for fuel was taxed at 5 percent, and alcohol for unspecified uses at 10 percent. Petroleum-based fuels were taxed at the following rates: Etanol 0 percent Unleaded gasoline 45 percent Regular gasoline 41 percent Premium gasoline 39.5 percent Gas oil 10.57 percent Fuel oil 10 percent Kerosene 10 percent The immediate revenue impact of the adoption of these taxes by Law 125/91 was approximately 1 percent of GDP, two-thirds of which was generated by fuel taxes. At the end of the 1990s these taxes generated approximately 1.3 percent of GDP in revenue, half of which came from fuel taxes. 45 The tourism regime is a simplified version of the draw-back or temporary admission systems, although it is an alternative of dubious quality. The tourism regime is, in practice, equivalent to implicitly recognizing a type of geographically unlimited free zone in the entire Paraguayan territory. A true free zone, with strictly limited territory in which goods are admitted from and to other countries without customs duties or VAT, would avoid needing to assume the fraction of which imports are consumed locally. As these values are necessarily inexact, the system allowed in consumption products without VAT to varying degrees according to different products. 46 Shome, P., Haindl, E., Schenone, O. and Spahn, P., op. cit., Table 22, p. 52. 105 2. Tax reform of law 2421/04 Law 2421, approved 5 July 2004, was an ambitious and far-reaching reform that modified VAT exemptions, consumption taxes and customs duties, created a personal income tax, replaced the Single Tax with a small contributors’ income tax and modified IRACIS and IMAGRO. This reform was presented as a simplification of Paraguay’s tax system using a “10 -10-10” formula: 10 percent VAT for all products (including soya, which is a commercial product) and services; 10 percent income tax for income above 10 minimum salaries and 10 percent corporate income tax (agriculture, industry and services). The revenue collection increase on implementing these reforms completely was estimated at 1.5 percent of GDP.47 Despite the many aims of the law, revenue collection of the main taxes has stayed practically flat (Table 3.5), with the exception of the VAT, which has continued increasing collection (up 40 percent in seven years) and is, as a result, the cause of total collection rising from 12 percent of GDP prior to the reform to 13.2 percent in 2011. Although less notable than the performance of the VAT, IRACIS has also showed a favorable evolution, increasing as a share of GDP by 25 percent in seven years. Table 3.5: Paraguay: Tax Revenue 2004-2010 (percent of GDP) 2004 2005 2006 2007 2008 2009 2010 2011 IRACIS 2.0 2.0 1.9 2.0 2.1 3.0 2.4 2.5 IMAGRO 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.1 IVA 4.9 5.4 5.6 5.7 6.0 6.2 6.9 6.9 ISC (Fuels) 2.0 1.5 1.6 1.6 1.4 1.5 1.3 1.3 ISC (Other) 0.5 0.5 0.6 0.5 0.5 0.6 0.6 0.6 AA Taxes 1.9 1.9 1.8 1.4 1.4 1.5 1.8 1.7 Other Taxes 0.4 0.5 0.6 0.2 0.1 0.1 0.1 0.1 Total 11.8 11.8 12.1 11.4 11.5 12.9 13.1 13.2 Sources: (1) Departamento de Estudios y Estadísticas Tributarias –DPTT--SET. (2) Zarate, W., Análisis del sistema tributario Paraguayo (CADEP, March 2011) Table 3, p. 10 and Annex 3, p. 23. 47 Borda, D., “Paraguay: Resultado de las reformas (2003-2005) y sus perspectivas”. Informes y estudios especiales (CEPAL, Santiago de Chile, January 2007). 106 One important characteristic of the period after the 2004 reform is the increase in the number of taxpayers. Between 2004 and 2011 the number of small contributors doubled, while the number of large contributors stayed roughly flat (Table 6). Creation of the personal income tax (PIT) The creation of a PIT according to Law 2421 has been repeatedly postponed48 and, according to Law 4.064/10, Art. 38, will come into effect on January 1, 2013. Table 3.6: Small and Large Taxpayers, 2000-2011 Year Small Taxpayers Large Taxpayers Total 2000 235,234 748 235,982 2001 249,363 794 250,157 2002 267,222 823 268,045 2003 283,236 872 284,108 2004 301,459 903 302,362 2005 328,495 947 329,442 2006 383,509 971 384,480 2007 428,889 992 429,881 2008 470,528 990 471,518 2009 513,011 987 513,998 2010 556,964 980 557,944 2011 603,629 973 604,602 Source: Sub-Secretaría de Estado de Tributación (SET). PIT taxpayers are individuals and simple associations, and income from Paraguayan sources that come from the realization of activities that generate personal income will be subject to taxation. These include: a) Remuneration (habitual or accidental, of any type) for the exercise of professions, arts or other occupations and providing personal services of any type, independently or in relation to a company, whatever the denomination of the benefit or remuneration or type of contract. Also taxed is all remuneration received by the owner, shareholders, managers, directors and other higher-level personnel of corporations or entities for services rendered, and remuneration from public service, elective or not, habitual, occasional, permanent or temporary. b) Fifty percent of dividends, profits and excesses obtained by shareholders or partners in entities that undertake activities covered by IRACIS or IMAGRO, distributed or accredited. c) Capital gains derived from the occasional sale of property, rights, shares, titles or capital quotas. d) Interests, commissions or returns on capital. 48 Law 248 of 1971 is possibly the most remote ancestor of a PIT in Paraguay. It could not be applied because the law was only in force for only a few months before being suspended indefinitely. 107 The following deductions are permitted: a) Legal discounts for contributions to the Institute of Social Security or the Pension and Retirement Funds created by law or decree. b) Donations to the state, municipalities and religious entities recognized by competent authorities and entities with legal character for social, educational, cultural, charity or beneficial assistance, recognized by the administration as organizations of public benefit. c) All expenses and investments directly related to the taxed activity in the case of physical individuals, when this is a real, documented expenditure at market prices. d) All personal and family expenses and investments incurred by individual taxpayers for sustenance, education, health, clothing, housing and holidays, as long as the expense is legally documented according to existing tax regulations. e) For individuals not paying into obligatory social security, up to 15 percent of gross income each tax period. This deduction is by legal assumption and requires no justification from the taxpayer. For the first year the tax is in force, individuals with net income below 10 minimum salaries a month (or 120 minimum monthly salaries a year) are not subject to the tax. This exemption will gradually be reduced by one minimum salary per year until a level of three minimum salaries per month (or 36 per year). The 10 percent rate is applied on net income above 10 monthly minimum salaries, while an 8 percent rate is applied to net income between 3 and 10 monthly minimum salaries and net income below three monthly salaries is exempt. The most novel characteristic of the tax, compared to PIT in most countries, is that the deduction indicated in d) means that, automatically, net income taxed by PIT is equivalent only to the part of personal income that has not been consumed (neither by occupational activity or in the family of the taxpayer) and properly documented. This means that taxable net income will only be positive in the case of families that do not consume (and properly document) all of their income, independent of each family’s wealth. Unlike most countries, in which PIT revenue derives from workers employed by companies (who cannot evade taxes as they are withheld from their salaries), in Paraguay the majority of revenue comes from taxpayers who cannot properly document their consumption, independent of their status as a dependent or independent worker. The explicit purpose of this peculiar PIT design in Paraguay is, above all, to induce formalization in the economy, combat evasion, and generate greater revenue from all taxes, not necessarily the PIT in particular. The evasion that has been quantified in Paraguay corresponds to the VAT. This remains high, as much as 54 percent of potential revenue collection (or 117 percent of actually collected revenue) for 2006. 49 If economic 49 Garzón, Hernando; Paraguay: La brecha de Evasión del IVA, November 2007, mimeo. 108 formalization promoted by PIT reduced evasion only by a 20th, VAT revenue would increase by almost 6 percent, or 0.4 percent of GDP. Although solid data is not available to precisely estimate PIT revenue, available information suggests collection of approximately 0.65 percent of GDP (Table 3.7). Table 3.7: Estimated PIT Revenue Collection (billions of 2011 G. and percent of GDP) Data and Assumptions Estimated Revenue I. Source of Taxable Net Income: Distribution of Profits and Dividends (a) IRACIS collected: 2540 (b) Profits subject to IRACIS: 2540/0.145 = 17517 (c) Net profits of IRACIS: (b)-(a)= 14 977 (d) Taxable gross income: 50 percent of (c)= 7488 (e) Taxable net income: Assuming deduction of 1/3 value of (d) with documentation= 5000 Estimated PIT collection 500 (0.50 percent of (10 percent rate) GDP) II. Source of Taxable Net Income: Personal Work (f) Participation of work remuneration in GDP: percent GDP = 30000 (g) Taxable net income: Assuming deductions of 95 percent of the value of (f) with documentation=1500 Estimated PIT collection 150 (0.15 percent of (10 percent rate) GDP) Sources: (1) Central Bank of Paraguay, Sistema de Cuentas Nacionales (November 2011), Table 3.5.1, p. 84, (2) Secretaría de Estado de Tributación and own calculations. Estimated revenue collection from the adoption of the PIT, taking into account the direct effects calculated in Table 7 (0.65 percent of GDP) and the indirect effect of reducing VAT evasion (0.4 percent of GDP), is approximately 1 percent of GDP. Creation of income tax for small taxpayers (impuesto sobre la renta para pequeños contribuyentes—IRPC) 109 The IRPC taxes at 10 percent presumed net income equal to 30 percent of gross income. This tax substitutes the Single Tax (which had a rate of 4 percent on gross income and replaced the VAT and income tax). It frees single-person companies from the income tax regime and allows access to the simplified VAT regime. It increases the annual limit of income to stay within this regime from G.52 million to G.100 million. The IRPC of Law 2421/04 requires that all small taxpayers also pay under the regular VAT regime, thus incorporating a great number of small taxpayers without adding large taxpayers (Table 6), which makes tax administration more difficult. As well, those not using the Single Tax should do so with the aim of being able to print official receipts. While the number of taxpayers subject to the Single Tax grew at almost 4 percent a year over the final years the system was in force, the number of IRPC contributors grew at about 10 percent per year between 2006 and 2011 (Table 3.8). Table 3.8: Number of Single Tax and IRPC Contributors, 2003-11 (thousands) 2003 2004 2005 2006 2007 2008 2009 2010 2011 Single 118.0 123.4 127.0 133.1 - - - - - Tax IRPC - - - - 150.6 169.0 184.3 199.6 215.7 Source: Secretaría de Estado de Tributación. Agricultural income tax (impuesto sobre la renta del sector agropecuario—IMAGRO) After the 2004 reform, several different tax-paying systems exist: (1) The so-called “Large Properties”, that individually or jointly have a usable agricultural area (superficie agropecuaria útil—SAU)50 greater than or equal to 300 hectares in the eastern region or 1500 hectares in the western region can choose to pay using either a “simplified” or “accounting” regime. Taxpayers using less than 30 percent of their property can only pay the tax using the “presumed regime”. (2) Properties with an SAU less than 300 hectares or 1500 hectares, according to region, can choose pay the tax under any of the three regimes (simplified, accounting or presumed). (3) Individuals with property equal or less than 20 or 100 hectares SAU in the eastern or western regions, respectively, are exempt from IMAGRO. The simplified regime calls for paying the tax on the difference between income and expenses. The IMAGRO rate for this regime is 10 percent. 50 To determine the SAU, deduct from total property area the following surfaces:  Spaces occupied by natural forest, permanent or semi-permanent cultivation and wetlands.  Areas not appropriate for productive use, such as rocky areas, estuaries or salt flats.  Protected wildlife areas.  Areas occupied by routes, roads or thoroughfares.  Areas destined for environmental services declared by authorities. 110 The accounting regime consists in paying IMAGRO based on accounting results. In this regime, apart from incurred and documented expenses, payers can also deduct:  Deprecation of female cattle as of the second year of life, equal to 8 percent of the animal’s value.  Die offs in cattle ranching up to 3 percent of the cattle value without the need for proof.  Depreciation of machinery and installation improvements.  In the case of individuals, all personal and family expenses.  The direct costs for assistance to individual property owners on small neighboring farms up to 20 percent of gross income.  The regime’s rate is 10 percent. The presumed regime calls for paying the tax based on multiplying SAU by a “index of soil productivity” (which depends on geographic location of the property and the goods produced) by the market production price (which is determined annually by the Ministry of Agriculture and Ranching). Taxpayers under this regime pay 2.5 percent. Table 3.9 summarizes the different IMAGRO regimes. The VAT corresponding to the purchase of goods and services directly related to production constitutes a fiscal credit that is charged directly to the IMAGRO since 2005. The amount of VAT from one fiscal year is carried over to the next until it is completely used up. As of 2012 the accumulated VAT had not been used up, even though the VAT fiscal credit against IMAGRO was eliminated by Decree 238 in September 2008. Table 3.9: IMAGRO Payment Regimes Simplified Regime Accounting Regime Presumed Regime Taxpayer (IMAGRO rate=10 (IMAGRO rate=10 (IMAGRO rate=2.5 percent) percent) percent) Large Properties Accepted only if less SAU>300 hectares in eastern than 30 percent of region or Accepted * Accepted the property is used SAU>1500 hectares in western region Medium Properties 20 50 y>4 Total Tekoporã $0.32 $0.37 $0.35 $0.40 $0.36 $0.47 $0.80 $0.50 $0.52 $0.02 Other Direct Transfers $0.71 $0.67 $0.69 $0.68 $0.69 $0.20 $0.62 $0.00 $0.25 $0.00 Tekoporã + Other Direct $0.38 $0.39 $0.38 $0.08 $0.18 $0.43 $0.78 $0.50 $0.48 $0.03 Social Tariff for Electric Energy $0.01 $0.01 $0.01 $0.01 $0.01 $0.01 $0.02 $0.04 $0.02 $0.08 Education: primary $0.81 $0.86 $0.84 $0.72 $0.79 $0.72 $0.76 $0.84 $0.73 $0.76 Education: secondary $1.67 $2.18 $1.95 $2.14 $2.03 $1.96 $2.25 $2.54 $2.07 $2.06 Education: tertiary $16.65 $13.96 $15.46 $6.53 $9.62 $9.55 $10.93 $13.97 $10.81 $10.75 Health $1.99 $0.80 $1.06 $22.02 $4.77 $1.16 $1.46 $1.54 $1.34 $1.41 Pensions $6.33 $2.65 $3.23 $9.48 $4.08 $2.79 $7.06 $40.95 $8.30 $8.18 Average Income $0.72 $1.88 $1.45 $3.23 $2.29 $6.70 $18.14 $114.40 $14.14 $10.77 Population Shares by group 5.6 % 9.4 % 15.0 % 13.4 % 28.4 % 38.9 % 31.0 % 1.7 % 71.6 % 100.0 % Source: For Paraguay, authors’ calculations using Encuesta Permanente de Hogares (2010) and National Accounts; for Argentina, Lustig and Pessino (2013); for Bolivia, Paz et al. (2013); for Brazil, Higgins and Pereira (2013); for Guatemala, Morán and Cabrera (2013); for Mexico, Scott (2013); for Peru, Jaramillo (2013); for Uruguay, Bucheli et al. (2013). Note: Benefits are in purchasing power parity (PPP) adjusted dollars of 2005. Groups indicate income groups; for example, “1.25 < y < 2.5” indicates individuals with household per capita income between $1.25 and $2.50 PPP per day. As a result of the above discussion, two policy measures can be recommended on the spending side if they are within the fiscal capacity of the government: first, to seek to expand coverage of direct transfer programs among the poor, and second, to increase the transfer sizes paid to the beneficiaries of targeted anti-poverty programs. On the tax side, taxes should be made more progressive. Overall taxes are slightly regressive, mainly due to the fact that regressive indirect taxes make up a large component of overall taxes. 187 7. Conclusions We presented results of applying a standard incidence analysis of taxes and social spending in Paraguay using the Encuesta Permanente de Hogares (2010). The analysis was conducted for a benchmark case, in which pensions were considered part of market income, and a sensitivity analysis, in which they were considered a government transfer. The results were placed in comparative perspective with seven other countries for which a comparable incidence analysis has been undertaken as part of the Commitment to Equity project. The main results are as follows. 1. Paraguay achieves only a small reduction in inequality and poverty when direct and indirect taxes, direct and in-kind transfers, and indirect subsidies are considered. In comparison with seven other Latin American countries, it performs worst or among the worst in terms of poverty reduction, inequality reduction and poverty reduction effectiveness, and closer to the middle of the pack in terms of redistributive effectiveness. 2. Direct transfers are progressive, indirect taxes are somewhat regressive and overall taxes are slightly regressive. 3. Social spending is progressive in relative terms, but less so than in any of the other countries analyzed. (In most of the other countries, social spending is progressive in absolute terms.) Education spending and health spending are each progressive in relative terms, but also less progressive than in other countries. Spending on tertiary education is regressive, which only occurs in Paraguay and Guatemala. 4. The small reduction in extreme and moderate poverty is not a result of a large proportion of direct transfer benefits going to the non-poor. Instead, it is a result of low coverage among the poor by direct transfer programs, and low per capita transfers to those who are covered. A larger reduction in poverty might be achieved by attempting to expand coverage and increase transfer sizes, if these policy measures are fiscally possible. 8. References Banco Central del Paraguay. Informe Económico Mensual (Central Bank of Paraguay. Monthly Economic Report). December 2012. BID (Banco Interamericano de Desarrollo). 2009. Equidad Fiscal en Brasil, Chile, Paraguay y Uruguay. Serie de Equidad Fiscal en América Latina del BID. Bucheli, Marisa, Nora Lustig, Máximo Rossi and Florencia Amábile. 2013. “Social Spending, Taxes and Income Redistribution in Uruguay.” Public Finance Review, forthcoming. Galeano, Luis. El Impacto del Programa Tekoporã de Paraguay en la Nutrición, el Consumo y Economía (The Impact of the Tekoporã Program of Paraguay on Nutrition, Consumption and the Economy). 2008. http://www.rlc.fao.org/es/prioridades/seguridad/ingreso3/pdf/mejorando.pdf Guillen, S. El gasto social en Paraguay: una mirada detallada del periodo 2002/2010 (Social spending in Paraguay: a detailed view of the period 2002-2010). Asunción. CADEP- OFIP: December 2010. 188 Higgins, Sean and Claudiney Pereira. 2013. “The Effects of Brazil’s Taxation and Social Spending on the Distribution of Household Income.” Public Finance Review, forthcoming. Imas, Victor. Las Transferencias Monetarias con Corresponsabilidad y la Disminución de la Pobreza en el marco de las Políticas de Protección Social (Conditional Cash Transfer Programs and Poverty Reduction within the framework of Social Protection Policies). 2011. http://idlbnc.idrc.ca/dspace/bitstream/10625/47110/1/133485.pdf. Jaramillo, Miguel. 2013. “The Incidence of Social Spending and Taxes in Peru.” Public Finance Review, forthcoming. Lindert, Kathy, Emmanuel Skoufias, and Joseph Shapiro. 2006. “Redistributing Income to the Poor and Rich: Public Transfers in Latin America and the Caribbean.” Social Protection Discussion Paper 0605. Washington, D.C.: The World Bank. Lustig, Nora and Sean Higgins. 2012. “Commitment to Equity Assessment (CEQ): Estimating the Incidence of Social Spending, Subsidies and Taxes Handbook.” Tulane University Department of Economics Working Paper 1219. Lustig, Nora and Carola Pessino. 2013. “Social Spending and Income Redistribution in Argentina During the 2000s: the Rising Role of Noncontributory Pensions.” Public Finance Review, forthcoming. Lustig, Nora, Carola Pessino, and John Scott. 2013. “The Impact of Taxes and Social Spending on Inequality and Poverty in Argentina, Bolivia, Brazil, Mexico and Peru: An Overview.” Public Finance Review, forthcoming. Morán, Hilcías and Maynor Cabrera. 2013. “Comparing the incidence of taxes and transfers on inequality and poverty in rural and urban areas of Guatemala.” Prepared for the International Fund for Agricultural Development. O‘Donnell, Owen, Eddy van Doorslaer, Adam Wagstaff and Magnus Lindelow. 2008. “Analyzing Health Equity Using Household Survey Data: A Guide to Techniques and Their Implementation.” WBI Learning Resources Series. Washington, DC: The World Bank. Paz Arauco, Verónica, George Gray Molina, Wilson Jiménez Pozo, and Ernesto Yáñez Aguilar. 2013. “Explaining Low Redistributive Impact in Bolivia.” Public Finance Review, forthcoming. Ribas, Rafael Perez, Guilherme Issamu Hirata, and Fábio Veras Soares. April 2010. “El programa Tekoporã de transferencias monetarias de Paraguay: un debate sobre métodos de selección de beneficiarios.” CEPAL Review. http://www.eurosocialfiscal.org/uploads/documentos/centrodoc/6449893fc621b9dbdf16d f6e7a3ca293.pdf Scott, John. 2013. “Redistributive Impact and Efficiency of Mexico's Fiscal System.” Public Finance Review, forthcoming. Scott, John. 2011. “Gasto Público y Desarrollo Humano en México: Análisis de Incidencia y Equidad.” Working Paper for Informe sobre Desarrollo Humano México 2011. Mexico City: UNDP. 189 Souza, Pedro, Rafael Osorio, and Sergei Soares. 2011. “Uma metodología para simular o Programa Bolsa Família.” Instituto de Pesquisa Econômica Aplicada (IPEA) Working Paper 1654. Wagstaff, Adam. 2012. “Benefit-incidence analysis: Are government health expenditures more pro-rich than we think?” Health Economics 21: 351-366. 9. Appendix A1. Definitions of Income Concepts79 As usual, any incidence study must start by defining the basic income concepts. In our study we use five: market, net market, disposable, post-fiscal and final income. One area in which there is no agreement is how pensions from the contributory system should be considered. Some authors treat them as part of market income and others place them under government transfers, and others exclude them altogether. Since this is an unresolved issue, in our study we defined a benchmark case in which contributory pensions are part of market income. We also did a sensitivity analysis where pensions are classified under government transfers. In what follows, we present the precise definitions of each income concept used in the benchmark case and the sensitivity analysis. Market income is defined as: Im = W + IC + AC + IROH + PT + SSP (benchmark) Ims = W + IC + AC + IROH + PT (sensitivity analysis) Where, Im, Ims = market income80 in benchmark and sensitivity analysis, respectively. W = gross (pre-tax) wages and salaries in formal and informal sector; also known as earned income. IC = income from capital (dividends, interest, profits, rents, etc.) in formal and informal sector; excludes capital gains and gifts. AC = autoconsumption; also known as self-production. IROH = imputed rent for owner occupied housing; also known as income from owner occupied housing. PT = private transfers (remittances and other private transfers such as alimony). 79 For more details on concepts and definitions, see Lustig and Higgins (2012). 80 Market income is sometimes called primary income. 190 SSP = retirement pensions from contributory social security system. Net Market income is defined as: In = Im – DT – SSC (benchmark) Ins = Ims – DT – SSCs (sensitivity analysis) Where, In, Ins = net market income in benchmark and sensitivity analysis, respectively. DT = direct taxes on all income sources (included in market income) that are subject to taxation. SSC/ SSCs = respectively, all contributions to social security except portion going towards pensions81 and all contributions to social security without exceptions. Disposable income is defined as: Id = In + GT (benchmark) Ids= Ins + GTs (sensitivity analysis) Where, Id, Ids = disposable income in benchmark and sensitivity analysis, respectively. GT = direct government transfers; mainly cash but can include transfers in kind such as food. GTs = GT + SSP Post-fiscal income is defined as: Ipf = Id + IndS – IndT (benchmark) Ipfs = Ids + IndS – IndT (sensitivity analysis) 81 Since here we are treating contributory pensions as part of market income, the portion of the contributions to social security going towards pensions are treated as ‘saving.’ 191 Where, Ipf, Ipfs = post-fiscal income in benchmark and sensitivity analysis, respectively. IndS = indirect subsidies (e.g., lower electricity rates for small-scale consumers). IndT = indirect taxes (e.g., value added tax or VAT, sales tax, etc.). Final income is defined as: If = Ipf + InkindT – CoPaym (benchmark) Ifs = Ipfs+ InkindT – CoPaym (sensitivity) Where, If , Ifs = final income in benchmark and sensitivity analysis, respectively. InkindT = government transfers in the form of free or subsidized services in education and health; urban and housing. CoPaym = co-payments, user fees, etc., for government services in education and 82 health. Because some countries do not have data on indirect subsidies and taxes, we also defined Final income* = If* = Id + InkindT – CoPaym. A2. Construction of Income Concepts i. Allocating Taxes and Transfers at the Household Level83 Unfortunately the information on direct and indirect taxes, transfers in cash and in-kind, and subsidies cannot always be obtained directly from household surveys. Thus, one of the most important aspects of CEQ is a detailed description of how each component of income is calculated (for example, directly drawn from the survey or simulated) and the methodological assumptions that are made while calculating them. When taxes and transfers can be obtained directly from the household survey, we call this the Direct Identification Method. When the direct method is not feasible, one can use the inference, simulation, imputation or alternate 82 One may also include participation costs such as transportation costs or foregone incomes because of use of time in obtaining benefits. In our study, they were not included. 83 Taken from Lustig and Higgins (2012). 192 survey methods (described in more detail below). As a last resort, one can use secondary sources: e.g., incidence or concentration shares by quintiles or deciles that have been calculated by other authors as is done by Goñi et al. (2011) for instance. Finally, if none of these options can be used for a specific category, the analysis for that category will have to be left blank. The six methods one can use to allocate taxes and transfers are described below. Direct Identification Method On some surveys, questions specifically ask if households received cash benefits from (paid taxes to) certain social programs (tax and social security systems), and how much they received (paid). When this is the case, it is easy to identify transfer recipients and taxpayers, and add or remove the value of the transfers and taxes from their income, depending on the definition of income being used. Imputation Method The imputation method uses some information from the survey, such as the respondent reporting attending public school or receiving a direct transfer in a survey that does not ask for the amount received, and some information from either public accounts, such as per capita public expenditure on education by level, or from the program rules. Inference Method In some cases, transfers from social programs are grouped with other income sources (in a category for “other income,” for example). In this case, it might be possible to infer which families received a transfer based on whether the value they report in that income category matches a possible value of the transfer in question. Simulation Method In the case that neither the direct identification nor the inference method can be used, transfer benefits can sometimes be simulated, determining beneficiaries (taxpayers) and benefits received (taxes paid) based on the program (tax) rules. For example, in the case of a conditional cash transfer that uses a proxy means test to identify eligible beneficiaries, one can replicate the proxy means test using survey data, identify eligible families, and simulate the program’s impact. However, this method gives an upper bound, as it assumes perfect targeting and no errors of inclusion or exclusion. In the case of taxes, estimates usually make assumptions about informality and evasion. The four methods described above rely on at least some information taken directly from the household survey being used for the analysis. As a result, some households receive benefits, while others do not, which is an accurate reflection of reality. However, in some cases the household survey analyzed lacks the necessary questions to assign benefits to households. In this case, there are two additional methods. 193 Alternate Survey When the survey lacks the necessary questions, such as a question on the use of health services or health insurance coverage (necessary to impute the value of in-kind health benefits to households), an alternate survey may be used by the author to determine the distribution of benefits. In the alternate survey, any of the four methods above could be used to identify beneficiaries and assign benefits. Then, the distribution of benefits according to the alternate survey is used to impute benefits to all households in the primary survey analyzed; the size of each household’s benefits depends on the quantile to which the household belongs. Note that this method is more accurate than the secondary sources method below, because although the alternate survey is somewhat of a “secondary source,” the precise definitions of income and benefits used in CEQ can be applied to the alternate survey. Secondary Sources Method When none of the above methods are possible, secondary sources that provide the distribution of benefits (taxes) by quantile may be used. These benefits (taxes) are then imputed to all households in the survey being analyzed; the size of each household’s benefits (taxes) depends on the quantile to which the household belongs. ii. Construction of Income Concepts: Paraguay The methods used in Paraguay are presented in Table A1. Table A1. Construction of Income Concepts in Paraguay. MARKET INCOME Autoconsumption Not included Imputed rent for owner occupied housing Included Earned and Unearned Incomes of All Possible Included. Pensions are only included in market Sources Including Social Security Pensions and income in the benchmark case. Excluding Government Transfers NET MARKET INCOME=MARKET INCOME - (DIRECT TAXES AND EMPLOYEE CONTRIBUTIONS TO SOCIAL SECURITY) Direct Taxes Subtracted from Market Income to generate Net Market Income. Direct Identification Method. The survey question for this variable appears on the survey as, “Algún miembro del hogar pagó por impuesto inmobiliario, tasas municipales, 194 asfalto, tasa de cementerio, etc?. Cuánto?” Employee contributions to social security Not included. DISPOSABLE INCOME = NET MARKET INCOME + DIRECT GOVERNMENT TRANSFERS Tekoporã (Flagship CCT) Inference and Simulation Methods. See explanation in section 3. Other direct transfers Inference and Simulation Methods. This variable was calculed as follows: (Other Regular Monthly Income) minus (Estimated Tekopora Contributions), where the latter is estimated as described in section 3. Zeroes are placed in all observations where the (Other Reg…) is smaller than (Estimated Tek…) Pensions Direct identification. In the benchmark case, pensions were already included in market income so they are not added here. In the sensitivity analysis, they were not included in market income, and are added here. As described in section 3, there is no reliable way to separate contributory from non-contributory pensions in EPH. POST-FISCAL INCOME = DISPOSABLE INCOME + INDIRECT SUBSIDIES - INDIRECT TAXES Indirect subsidies Imputation Method. This is equivalent to Paraguay's Tarifa Social, a public program that discounts electricity bills for families with low usage. Using the current price of electricity per 100kw/h in Paraguay, and the variable in our survey for "money spent on electricity," we estimate who would be eligible as a Tarifa recipients. Then working backwards from how much they spent, we estimate the likely benefits they have received. Indirect taxes Secondary Sources. There are two Indirect Taxes in Paraguay. There is the Value Added Tax and a tax on combustibles. Incidence for 195 both comes from BID (2009). FINAL INCOME = POST-FISCAL INCOME + GOVERNMENT IN-KIND TRANSFERS/FINAL INCOME* = DISPOSABLE INCOME + GOVERNMENT IN-KIND TRANSFERS In-kind education Imputation Method. The education benefit is based on cost per student by level. This benefit is applied to students who report attending public school. If they attend Primary School (imputed by their age, and if they said "yes" to currently attending school), they are assigned an in-kind benefit equal to the government’s per student spending on Primary school. We did the same for Secondary and Tertiary. In-kind health Imputation Method. There are two types of in- kind health: free and paid. If the individual said "no" to having insurance and "yes" to "received a medical service within the last 3 months", they are assigned a proportion of National total spent on free health services. Otherwise, if the individual received medical service and said "yes" to having insurance then they are assigned a proportion of the National sum spent on recipients of IPS health care (Instituto Publico de Salud), which is the public health administration system. SCALED-UP INCOMES, TAXES AND TRANSFERS FOR INCIDENCE ANALYSIS INCLUDING GOVERNMENT IN-KIND TRANSFERS Scaling up factor and method All variables on Taxes, Spending are scaled proportionally to national totals. That is, the original (unscaled) variables provides the proportion of the National total amount. Scaled totals are used for calculating inequality, effectiveness, incidence, concentration shares, and progressivity. Non-scaled totals are used for calculating poverty and transfer sizes. (For an explanation of why, see Lustig and Higgins, 2012.) 196 A3. Effectiveness Indicators84 In mathematical notation, let ( ) be the inequality or poverty measure of interest (e.g., the Gini coefficient or headcount index), which is defined at each benchmark case income concept j = (market income, net market income, disposable income, post-fiscal income and final income) and each sensitivity analysis income concept j = . Let be total public spending on the direct transfer programs captured by the survey or otherwise estimated by the authors, measured by budget size in national accounts (note that in the sensitivity analysis this concept includes spending in social security pensions), and let and be total public spending on health, education, and (where included) housing programs, respectively. Then the effectiveness indicator for direct transfers is defined as: ( ( ) ( )) ( ) and the effectiveness indicator for direct and in-kind transfers is defined as: ( ( ) ( )) ( ) ( ) Note that in the sensitivity analysis, when contributory pensions are considered a government transfer, they are not part of net market income but are part of disposable income, thus some of the change between ( ) and ( ) is attributable to contributory pensions, and therefore in the sensitivity analysis must include spending on contributory pensions. In the benchmark case, however, contributory pensions are already included in net market income, so does not include any spending on contributory pensions. Also note that should only be included in the denominator of the effectiveness indicator for direct and in-kind transfers if housing programs 84 Taken from Lustig and Higgins (2012). 197 Chapter 6. Equality of Opportunities and Public Spending in Paraguay, by Jose Cuesta and Pablo Suárez Becerra Abstract Previous studies in Paraguay have found that circumstances one is born to and has no control over—such as gender, parents’s socioeconomic status, or location—determine whether or not children have access to critical opportunities in education and basic services such as water and sanitation. This report extends such previous analyses in four directions. First, it updates the previous estimates of the Human Opportunties Index (HOI) for 2009 and 2010. Second, it shows the evolution of opportunities between 2003 and 2010. Third, it analyzes a health opportunity, timely and affordable access to heath care. Fourth, it links the analysis of opportunties with public spending. This analysis partly confirms results from previous studies. Gender is not found to be a critical circumstance, while speaking only Guarani at home and departmental location of residence play a larger role than previously acknowledged. Between 2003 and 2010, the HOI improved for almost all opportunities considered, although at different paces and with different dynamics. These improvements are better explained as coverage increases rather than equalizing effects (that is, from increases across the board rather than disproportionate gains for the most disadvantaged). Finally, public spending on education for children age 5 to 17 is neither pro-poor nor pro-rich and slightly progressive, with distinct distributive incidences for elemental and secondary education spending. Public spending on health care is neither progressive nor regressive and concentrates on middle-income groups. This analysis also suggests a simple method to identify circumstance groups that would benefit the most from targeted additional public spending on specific opportunities. Key words: Opportunities, public spending, incidence analysis, Paraguay Introduction Paraguay has experienced impressive growth rates since 2003, the strongest period of economic performance since the 1970s (World Bank 2010). The country has weathered the global financial crisis and is the holder of the world’s fourth largest growth rate in 2010 (World Bank 2013). Substantive poverty reduction has accompanied this economic growth. In fact, poverty incidence declined by 10 percentage points between 2002 and 2007. Yet, by 2008, almost two out of five Paraguayans remained poor, and one out of five lived in extreme poverty (World Bank 2010). Moreover, extreme poverty has declined much more modestly in that period, and the distribution of incomes and assets remains very unequal. With a Gini coefficient exceeding 0.52, only Honduras, Colombia, and Brazil are more unequal than Paraguay in a region that is already the most unequal in the world (World Bank 2013). More than 40 percent of total income is in the hands of the richest 10 percent of the population, and just 2 percent of the agricultural establishments in the country own almost 82 percent of the agriculturally exploited land (World Bank 2009). 85 Similarly, welfare is marked by urban and rural disparities, 86 as 85 About 6,400 farms account for 20 million of the 24 million hectares in agricultural use, or half of Paraguay’s total area of 40 million (World Bank 2009). 198 observed in the incidence of poverty and the access to drinkable water, safe sanitation and public health care, among others. Previous studies in Paraguay have shown that circumstances like parental education, household income, number of siblings, and geographical location still determine access to basic services.87 Existing estimates of the status of equal opportunities in Paraguay portay a mixed picture: the HOI (see below) in Paraguay is below the regional average (for 19 countries) in Latin America, but has grown faster than the regional average. This confirms some degree of catching up. However, World Bank projections indicate that Southern Cone countries will not achieve universal access to education (school enrollment and completion of sixth grade on time) before 2046, possibly longer in Paraguay, because the country continues to have the lowest educational ranking on the HOI among its neighbors (Molinas et al. 2010, 50). As in other cases, future poverty and inequality reduction in Paraguay will depend critically on the extent to which the generation of the current poor—particularly their children—can unshakle themselves from circumstances such as their region of birth and residence and parents’ educational attainments and socioeconomic status, which may limit their ability to realize their full potential in life. These limitations on potential could take the form of barriers to education up to a certain level, the inability to work in an occupation befitting their level of human capital, or simply restriction of the ability to migrate to search for better economic opportunities. What is the extent to which an individual, irrespective of the circumstances to which he or she is born, has access to some of the most basic opportunities to realize ones productive potential? That is the central question addressed in this paper. The analysis here addresses three specific queries: (i) are children in Paraguay given equal opportunities early in their lives to allow them to build a dignified and productive life of their choosing?; (ii) Do their gender, parents’ socioeconomic status, family structure, and the region where they grew up as children hinder access to these basic services? If so, which of these circumstances plays a stronger role in limiting opportunities? (iii) Going forward, what is the most efficient way to expand the provision of basic services to best improve the equality of opportunities in Paraguay? To answer these questions, this analysis expands existing studies on the equality of opportunities in Paraguay—World Bank (2010) and Molinas et al. (2010)—in four directions. First, previous HOI estimates for 2008 are updated to include 2009 and 2010. Second, the number of opportunities analyzed is extended, and now includes public health care—alternative definitions for some educational opportunities are also used. Third, using 86 According to World Bank’s (2010) latest Paraguay Poverty Assessment, over half of the poor and more than two - thirds of the extreme poor are located in rural areas of Paraguay. With only 41.4 percent of the population, rural households had a disproportional amount of the poor (53.5 percent) and the extreme poor (67.5 percent) in 2008. 87 In fact, different circumstances have distinctive impacts across opportunities. Parental education, household income, and number of siblings have reportedly the largest impacts on school attendance, while parents’ education and child’s gender matter the most for completion of sixth grade. Yet, the urban or rural location is the most important driver behind disparities in access to sanitation, water, and electricity (see the latest World Bank Poverty Assessment, World Bank [2010]). 199 multiple rounds of the Encuesta Permanente de Hogares (EPH, Permanent Household Survey), analysis includes the dynamics of the HOI and its key contributors since 2003. Fourth, the analysis of opportunities, specifically the HOI, is linked with public spending in education and health care in 2004 and 2009. The rest of the paper is organized as follows: section 2 lays out the concepts and methodology of the basic human opportunities framework. Section 3 describes the application of the methodology to Paraguay, while section 4 reports key results on the status and evolution of the HOI for educational, health, and housing opportunities. Section 5 presents the results of the benefit incidence analysis from the opportunity perspective, and section 6 summarizes the findings and implications for the future. 1. The Human Opportunity Index: Concepts and Measurement A large body of social science literature has been concerned for some time with equality of opportunity. Amartya Sen (1977, 2001) has been deeply influential in arguing for an equitable distribution of “capabilities,” which essentially are a person’s ability and efforts to convert resources into outcomes they have reason to enjoy. John Roemer’s (1998) Equality of Opportunity was the first to formalize an equality of opportunity principle and remains the most relevant piece of academic literature underpinning this analysis of Paraguay. “Opportunity” in Roemer’s context, and in the context used throughout this report, is understood as the set of basic services or goods that make it possible for an individual to lead a life with dignity and freedom of choice. Circumstances are attributes of individuals for which society believes individuals should not be held accountable, and which affect their ability to achieve access to the advantage (opportunity) that is being sought. Roemer argues that policy should work to equalize opportunities independent of circumstances and that outcomes should depend only on effort. The World Bank’s (2006) World Development Report: Equity and Development argues that inequality of opportunity, both within and among nations, results in wasted human potential and weakens prospects for overall prosperity. Conducting an analysis of inequality of opportunity, however, requires a measure or a set of measures that provide a practical way to track a country’s progress toward equalizing opportunities for all its citizens. To be u seful to analysts and policy makers alike, such a measure must combine a few attractive properties: intuitive appeal, simplicity, practicality (especially in relatively data-scarce environments), and sound microeconomic foundations to ensure that it has an interpretation that is consistent with its objective. Much of the empirical work in developing countries until recent times has focused on measuring (and comparing) average rates of access to goods or services in health and education for the general population and different subgroups within. What has been lacking for the most part is an intuitive and unified framework to address a range of questions across different types of opportunities, such as: How far away is a country from universalizing each type of opportunity? How unequally are available opportunities distributed across different subgroups of the population? How important are circumstances to which an individual is born 200 into in determining access to opportunities? Which circumstances matter for access, and in that sense, contribute the most to inequality in access? What would it take, in terms of resources, to reduce inequality in opportunities when providing universal access is clearly not possible in the near term? The HOI measures how far a society is from universal provision of basic services and goods, such as sanitation, clean water and education, and the extent to which those goods and services are unevenly distributed.88 A key feature of the HOI is that it not only takes into account the overall coverage rates of these services, but also how equally the coverage is distributed—by measuring the extent to which those without coverage are concentrated in groups with particular circumstances (for example, economic status, gender, parental education, ethnicity, and so on), which are conditions a child is typically born into. More specifically, HOI is an inequality-sensitive coverage rate that incorporates: (i) the average coverage of a good or service that society accepts should be universal (which implies that the individual is not held responsible for lack of access) and (ii) whether it is allocated according to an equality of opportunity principle. The HOI is defined as the difference between two components: the overall coverage rate of the opportunity (C); and a “penalty” for the share of access to opportunities that is allocated in violation of the equality of opportunity principle (P). HOI = C – P, which implies that the maximum value on the HOI that a particular opportunity can take is the average access (or coverage) rate for that service. It also implies that an HOI value of 1 would be possible only when access is universal (C is equal to 1 and P is equal to 0). Figure 6. 1: Graphical Interpretation of HOI Source: Adapted from Molinas et al. (2010). 88 This discussion draws from three sources: Barros et al. (2009) and Molinas et al. (2010). 201 Figure 6.1 shows a simple graphical interpretation of the HOI. It plots the probability of a child of a particular circumstance (for example, percentile of per capita income or wealth) completing sixth grade on time, with circumstance (on the horizontal axis) improving from left to right. The horizontal line is the average coverage rate for the entire population of children. The curved line shows access rates for different levels of circumstance. There is no equality of opportunity in this case, since the probability of access to the opportunity is positively correlated with circumstance, which is shown by the fact that the curved line does not coincide with the horizontal line. Opportunities allocated in the red area above the average coverage violate the equality of opportunity principle: they show dependence of the access to education on income or wealth. There is an intuitive interpretation of the red area: it is the share of the total number of opportunities that are “misallocated” across groups of different circumstances, which is to say allocated to children with better circumstances so that they have higher than average access to the opportunity.89 The HOI corresponds to the blue area in the graph, which is the area below the curved line discounted by the red area above the average coverage rate. A second interpretation of the HOI invokes an index (D), equivalent to (P/C), which is known as the “inequality of opportunity” or “dissimilarity” index. The D-index corresponds to the share of opportunities that would have to be reallocated across groups—for an unchanged rate of overall coverage—to achieve equality of opportunity, out of the total amount of opportunities available in society; HOI = C – P = C × (1 – D). Box 6.1 outlines a simple example of how the HOI is measured, using a hypothetical situation with two countries with identical population of children and average coverage rate of primary school enrollment. The example demonstrates how the HOI is sensitive to inequality in coverage and how it would change in response to an increase in overall coverage or reallocation favoring the more disadvantaged group. Box 6.1: The HOI—A Simple and Intuitive Example Consider two countries, A and B, each with a total population of 100 children. Each country has two groups of children, I and II, which consist of the top 50 percent and bottom 50 percent by per capita income, respectively. Coverage rate of school enrollment (or the average enrollment rate) for both countries is 0.6, that is, 60 children attend school in each country. The table below shows the number of children going to school in each group for each country. Number of children age 6–10 years enrolled in primary school Group of circumstances Country A Country B (100 children) (100 children) Group I (top 50 percent by income) 40 35 Group II (bottom 50 percent by income) 20 25 Total 60 60 Given the total coverage rate, the principle of equality of opportunity will hold true for each country if each of the two groups in each country has the same rate of coverage, that is, if each group has 30 children going to school. But in reality, group II has 20 enrollments in country A and 25 in country B. This suggests firstly that opportunities are unequally distributed, and secondly, that the inequality of opportunities is higher in country A. The D-index is the share of total 89 This also implies that the red area is the share of total opportunities that would have to be reallocated to children with lower than average opportunities to achieve equality of opportunities for a given level of coverage. 202 enrollments that is “misallocated,” namely 10/60 and 5/60 for A and B, respectively. Therefore, HOI A = C0 (1 - D) = 0.6 × (1 - 10/60) = 0.50; HOIB = C0 (1 - D) = 0.6 × (1 - 5/60) = 0.55. Thus, even though both countries have equal coverage rates for enrollment, the higher inequality of opportunity in country A leads to the D-index being higher for A than for B, and the HOI being higher for B than for A. It is also easy to see that the HOI will increase in a country if: (i) the number of enrollments in each group increases equally (in proportionate or absolute terms); (ii) if enrollment for any group increases without decreasing the coverage rates of the other group; and (iii) if enrollment for group II increases, keeping the total number of children enrolled unchanged (implying enrollment in group I reduces by an equivalent amount). These three features relate to the “scale,” “Pareto improvement,” and “redistribution” properties of the HOI, respectively—properties that are intuitively appealing. Source: Author’s compilation. The HOI is an inequality sensitive coverage rate in the sense that it improves when inequality decreases with a fixed number of opportunities in a society, or when the number of opportunities increases and inequality stays constant. In more formal terms, the properties of the HOI guarantee that the improvement in the index is sensitive to: (i) the overall coverage — when the coverage for all groups increases by factor k the HOI increases by the same factor; (ii) Pareto improvements—when the coverage for one group increases without decreasing the coverage rates of other groups, the HOI increases; (iii) redistribution of opportunities—when the coverage rate of a vulnerable group increases for a constant overall coverage rate, there is decrease in inequality and an increase in the HOI. To compute an HOI for a particular opportunity for the children of a country, household survey data are essential. To allow computation of an HOI for education and health opportunities, the survey must have a minimum set of information at the individual (child) or household level, as appropriate. Examples of information needed would include whether the child is attending school or not, grade level, last grade completed, and health indicators such as weight and height of child and whether the child has been immunized. Computing an HOI for access to basic infrastructure, like safe water, electricity and sanitation, would require that household-level information on these indicators be available. With regard to circumstances, the minimum information needed to make the analysis meaningful would be gender, age and location (urban/rural and/or regional) of the child; demographic characteristics of the household (size and composition); characteristics of parents (gender, age, and education); and some measure of household income, consumption, or wealth. In practical terms, computing an HOI for a particular opportunity when the number of circumstances is relatively large (more than three) requires an econometric exercise, which involves obtaining a prediction of the D-index from observed access to opportunities and circumstances among children. In simple terms, the exercise consists of running a logistic regression model to estimate the relationship between access to a particular opportunity and circumstances of the child, on the full sample of children for whom the HOI measure will be constructed. The estimated coefficients of the regression are used to obtain, for each child, his/her predicted probability of access to the opportunity, which is then in turn used to estimate the D-index, the coverage rate, and eventually the HOI.90 90 See appendix 1 as well as Molinas et al. (2010) for more technical details on the econometric exercise. 203 Change in the HOI over time can be used to assess progress in access to opportunity in a society, taking into account both universality of access and inequality in access among different circumstance groups. To help understand the factors that contribute to a change in the HOI, a decomposability property of the HOI is useful. A change in HOI can be decomposed into: (i) a composition effect, which refers to changes in the distribution of circumstances (for example, if the distribution of wealth improves, chances of accessing opportunities are likely to increase); (ii) a scale effect, which refers to proportional change in the coverage rate of all groups (for example, if there is policy directed toward increasing coverage of an opportunity across all groups); (iii) an equalization effect, which refers to change in the coverage of vulnerable groups (groups with coverage below the national average), with the average coverage rate held unchanged—in other words, a move toward greater (or less) inequality for the same average level of coverage. Interpretations of the three decomposed components are quite intuitive. A positive composition effect shows whether the underlying circumstances that children are born into are improving over time, as a result of demographic changes, economic growth, or social progress. A positive scale effect shows whether opportunities are improving for all groups in the society, perhaps as result of public policy or social progress, for example, increased awareness among all households. The equalization effect in essence indicates the trend in equity in a society, showing whether available opportunities are distributed more equitably among its members, so that the circumstances a child is born into begin to matter less for access to basic goods and services. The HOI—as described above—is an indicator of coverage discounted by the level of inequality in access along the dimensions of the circumstances of the child . While this composite indicator is informative in itself, what is of paramount interest, particularly from the point of view of policy, is information on circumstances that are most salient in explaining the observed inequality. This analysis uses a method that allows the unraveling of the sources of inequality for a given level of an HOI using a technique known as the Shapley value decomposition. The basic idea is to find the increase in inequality that would occur if a given circumstance was added to a set of preexisting circumstances.91 2. Choice of Opportunities and Circumstances for Paraguay The HOI methodology focuses on opportunities to improve a person’s ability to expand his or her future production possibility frontier by investing in human capital in the early stages of his or her life cycle. For this reason, this analysis focuses on a number of basic services that are critical early in life to provide the opportunities to allow a child to grow up in a 91 Implementing this method is complicated by the fact that the circumstances can be correlated with each other. Therefore to identify the contribution of any specific circumstance, an analysis would need to consider the addition this circumstance will make to the inequality for all possible subsets and permutations of the rest of the circumstances. Once these specific contributions are known, a weighted average across all permutations is used to obtain the overall impact of a circumstance on inequality (Hoyos and Narayan 2011). 204 reasonably healthy environment, receive education, and access affordable health services to function productively in Paraguay’s society. Of course, the variables that can be analyzed as opportunities and the variables that can be used as circumstances in an analysis of this sort are also determined by available information. Most of the HOI analysis in Paraguay relies on the subsequent rounds of the Encuesta Permanente de Hogares, the Permanent Household Survey (EPH), between 2003 and 2010. The EPH is an annual survey of approximately 20,000 individuals 92 typically conducted between October and January each year.93 This sample is representative of the country and of each of the seven large regional groups within the country: that is, Asunción, Concepción, Caaguazú, Itapúa, Alto Paraná, Central and “others” which includes the departments of San Pedro, Cordillera, Guairá, Caazapá, Misiones, Paraguarí, Ñeembucú, Amambay, Canindeyú, and Presidente Hayes.94 By computing the HOI for a comparable set of opportunities using identical circumstances over time, the analysis can identify trends in Paraguay’s HOI for a meaningful length of time. In addition, the analysis also focuses on the relationship between public spending and opportunities for education (2004 and 2009) and medical attention (for 2009). However, this paper does not include a fiscal incidence analysis for 2010, because the 2010 EPH does not report the location for all households—nor for 2011, because the 2011 EPH does not report households’ out-of-pocket spending on education and health care activities. The analysis includes only children under the age of 18 years for all the opportunities studied. In addition to the intrinsic value of measuring access of key goods and services by children, focusing on the young children also obviates the need to make the distinction between access and utilization related to effort, attitudes, or preferences of the child or the child’s parents. What this implies is that as long as society agrees on universalizing an opportunity, it must ensure utilization by children, independent of the preferences of the child or the child’s family. For example, a child may have access to a school at a reasonably close location, but may not attend school because the parents do not value education or because the school is of a low quality or too distant. In such instances, that child will be treated as having no access to school. If this is a basic service, society must ensure that the child uses the service, which might entail not only having a school nearby, but also maintaining schools at a level of quality or requiring obligatory attendance. It is worth noting, however, that assuming that children do not take any part on decisions associated with his or her health care or education is not free of caveats. Personal maturity and family dynamics may make this generalization troublesome, more so as one approaches the age of majority. Furthermore, education in Paraguay is mandatory only until the age of 14, increasing the probability of teenagers engaging in labor activities, and therefore perhaps having more say in household decision making. These considerations are assumed away for simplicity. 92 Sample sizes in 2003 and 2004 were larger, around 43,000 and 34,000 individuals respectively. 93 Exceptions are the surveys of 2003 and 2004 which started in August, the survey of 2005 which started in November and ended in February 2006, and the survey of 2006 which started in October and ended in March 2007. 94 Data are not collected for the departments of Alto Paraguay and Boqueron, the least populated of the country, with about 1 percent of Paraguay’s total population. 205 Another consideration is the quality of service. Basic goods and services are usually not homogeneous: their quality varies tremendously. This is particularly true of opportunities analyzed in this study, such as education, health care, and housing. A relatively simple approach to measuring education quality is to focus on timely progression through school. While going to school provides a sense of inclusion, timely progression may reflect children’s adequate progress. This is, of course, no substitute for a more direct measuring of learning, such as standardized test scores. Unfortunately, information on test scores in Paraguay could not be readily matched to EPH data. For this reason, analysis is limited to disparities in attendance for a broad group of children aged 5 to 17; timely start and completion in the first and second cycles; and completion of third cycle of primary education. Table 6.1 below summarizes these opportunities. For health care opportunities, quality is partly accounted for in the definition of the opportunity itself, which—as explained below—considers timely and affordable access to medical attention.95 Furthermore, EPHs allow discriminating across different types of safety of water and sanitation services. This is not the case, however, for electricity provision; EPHs do not capture quality issues such as frequency and severity of blackouts or disruptions in service. Opportunities considered This analysis focuses on three broad categories of opportunities that are critical for an individual, especially during early childhood. They are: (i) the opportunity to receive adequate education; (ii) the opportunity to receive required medical attention96 in a timely and affordable way; and (iii) the opportunity to grow up in a household with housing conditions that are sufficient to provide a safe, stable, and a stimulating childhood (table 5.1). Assessing the opportunity to acquire adequate education uses a broad subset of indicators related to current status of attendance in schools (ages 5–17); timely start of primary education (age 6-7); timely completion of sixth grade (age 13); and completion of ninth grade (ages 16–17) 97 . These are relevant opportunities in a country like Paraguay, where education attendance between grades first and sixth is almost universal, but after sixth grade, 95 Interestingly, only about 2 percent of Paraguayans who reported being ill or having accident in the last 90 days of 2010 purposefully opted out of public medical services because they considered them either too expensive, unavailable or of too low quality. Another reported reason for not seeking medical attention when suffering an illness or accident is that respondents did not consider the illness or ailment to be serious or they were “too busy” to seek attention (about 5 percent of the sick). These cases imply that, in effect, the condition was perceived as not severe enough for medical attention. In that case, these individuals were not included in this analysis of access. Therefore, we consider that people demand health care if they were ill or had an accident and they did consider it as serious enough to seek attention. 96 The focus here is on medical attention that can be individualized. Thus public health interventions are not considered here because these interventions are public goods. We also face limitations of the survey to determine their beneficiaries. 97 The official age to start first grade in Paraguay is 6 years. Given that the surveys are typically collected between October and January and the academic course starts in April, the analysis considers starting first grade on time if a 6 or 7 year old child attends or has completed first grade at the time of the survey. For consistency, finishing sixth grade on time is analyzed across children aged 13 as by that age, anyone who started on time first grade –at age 6– and did not drop out or repeat would have finished sixth grade. 206 enrollment rates markedly start declining. Another area widely acknowledged to being critical for educational policy, preschool, is not compulsory in Paraguay. However, this analysis focuses on access, disparities, and trends in preschool, the last stage of initial education in Paraguay and directed toward children age 5. It is important to note that the analysis is an overestimation of the disparities that underline access differentials if parents purposefully decide to not send their children to preschool. In effect, evidence shows that in Paraguay, children age 5 do not attend preschool because parents consider them as not having the appropriate age (65 percent of those not attending). Families reporting economic issues or school availability and quality reasons are less than 5 percent.98 Table 6.1: Opportunities in Paraguay Opportunities Description I. Education Attend school Children age 5–17 attend school Attend preschool Children age 5 attend preschool Start school on time Children age 6-7 attend first grade of primary Finish sixth grade on time Children age 13 have finished sixth grade Finish ninth grade Children age 16–17 have completed ninth grade II. Health Timely and affordable access to Children age 0–17 suffering an illness or accident in the last 90 days who demand health care medical attention access timely and affordable health care services III. Housing Access to drinkable water Children age 0–17 live in households with access to drinkable water a Access to safe sanitation Children age 0–17 live in households with access to safe sanitationb Access to electricity Children age 0–17 live in households with access to electricity Source: Authors’ compilation. a. Sources of drinkable water considered are ESSAP, SENASA, community network, private provider, artesian wells, protected and unprotected ground wells, and bottled water. Unsafe sources of drinkable water are unprotected springs, rainfall, and surface water. b. Sources of sanitation considered safe include sewage and septic tank. Unsafe sources are considered latrines and surface disposal. For the opportunity related to timely and affordable public or private health care, the focus is on whether the child is growing up in a household that can afford health care when needed. More specifically, the analysis identifies households in which a child suffered an illness or accident that was considered serious enough to seek medical attention . Among those cases, the analysis further identifies those who responded that although willing to receive attention, they did not seek it because medical services are not close by, are too expensive, or are not sufficiently good to effectively demand them. This group is categorized as being excluded from public and private health services. Those who self-medicated are also considered as excluded of health services99 as well as those who visited a healer100 or a relative. Those who 98 Data come from the 2010 EPH. It is not clear, however, if parents consider their children too young at age 5 to attend preschool or if the children were not old enough to enter preschool. 99 We also run the analysis assuming that self-medication is considered as not seeking health services. Results are presented in Appendix 6. 207 suffered an illness or accident in the last 90 days, considered it sufficiently important to seek attention and were not subject to restrictions of supply, costs, and quality101 are categorized as enjoying the opportunity of timely and affordable access to health care services. 102 Public provision of health care refers to services provided by the Ministry of Health and Social Welfare (Ministerio de Salud Publica y Bienestar Social) and the Institute of Social Security (Instituto de Prevision Social). 103 Public care services are further divided into those provided in a health center and those provided in hospitals. No further disaggregation is possible given the available administrative and household survey data. 104 Private health care corresponds to services provided by pharmacies and private professionals and medical institutions. Finally, the analysis also considers a set of indicators of housing services or amenities that capture the opportunity to grow up in a household with conditions sufficient to provide a safe, stable, and a stimulating childhood. These indicators include access to infrastructure facilities such as drinking water, safe sanitation facilities, and electricity in the household.105 Circumstances considered The analysis of opportunities in Paraguay considers nine circumstances: gender of the child; gender of the household head; education of the household head; 106 household’s per capita income; presence of the household head’s spouse/partner in the household; number of siblings; urban versus rural area of residence; regional residence (grouped in seven large regions); and primary language or languages spoken in the household. 100 It is possible that the family preferred the attention of a healer instead of that of standard health services, even if the latter were available and affordable. In this case we would be overestimating exclusion to health care. However, the percentage of the ill who visited a healer or a relative wass 1.49 percent in 2010 101 In the survey these represent those who reportedly visited a health professional or a pharmacist. 102 No matter how much care is taken in terms of differentiating needs, demands and access to health services, the discussion of access to medical care requires further caution. This is because the opportunity is defined on the subset of children requiring medical care as opposed to the entire population, and this induces biases related to the selectivity of the sample. However, for the form of bias that would be most worrisome, one can argue that the inequality estimated in the selected sample would be an underestimate of the actual inequality. The reasoning is as follows: even if the likelihood of getting ill or requiring some form of medical attention were to be distributed randomly among the population, it is likely that those with circumstances that make them more likely to be able to secure the necessary care would be precisely the ones more likely to be reporting to have needed it in the first place. 104 Under health centers we consider primary and secondary health services provided by postas de salud and clinicas de salud. Secondary and tertiary health services provided by regional and central hospitals and by MSPyBS and IPS hospitals are grouped under hospital services. 105 Water and sanitation are primary drivers of public health and improvements in these services have been shown to reduce the incidence of diarrhea and its serious long-term consequences such as malnutrition, pneumonia, and physical and mental stunting. In that sense, these opportunities could have just as well have been categorized under the health opportunities. 106 While education and gender of household head need not necessarily be the same as that of the parents of a child living in the household, there is a large overlap between household heads and parents. Using the information of household heads allows for easier analysis, given the way the data are reported in the surveys. Replacing the household head characteristics with those of actual parents leads to similar results in the HOI estimation, but with a smaller sample size, since parental information is available for a smaller set of children. 208 This set maintains the “core” circumstances analyzed in previous studies in Paraguay (World Bank 2010; Molinas et al. 2010) and adds three variables to that core: regional location, urban or rural location, and language of the household head. By maintaining core circumstances, the analysis controls for all socioeconomic aspects, demographic and family structure circumstances, spatial disparities and some degree of discrimination (or gender preference), all present in the previous studies in Paraguay. By adding regional variables, the analysis may be controlling for additional sociodemographic issues that go beyond urban and rural location. It is believed that language is a practical proxy for ethnicity in a country whose population by and large self-identifies with a Guarani origin. Specifically, the analysis controls for the language the household head speaks at home most of the time, whether it is only Spanish, Spanish and Guarani (mixed), or only Guarani. Other circumstances that would merit further considerations such as intrahousehold bargaining or migration conditions are not directly captured. However, the presence of the household head spouse is expected to partially control for decision-making issues because it captures a key feature of the family structure that provides clues on how and who makes decisions (that is, collectively or, rather, unilaterally by the household head). Table 6.2 presents the key summary statistics for each of these circumstances over the period of analysis, 2003–10. Table 6.2: Summary Statistics of Circumstances Used in Analysis Circumstance 2003 2004 2005 2006 2007 2008 2009 2010 Child's sex: male 50.3 51.7 51.1 50.7 51.5 51.0 52.0 51.8 Household head's sex: male 75.9 76.0 74.5 74.9 73.4 73.1 69.0 73.7 None 42.2 41.3 39.0 37.7 38.0 38.1 34.6 35.0 Household head's 6th grade 25.1 26.8 24.8 27.7 27.3 24.6 26.4 26.0 education 7th to 9th 13.1 12.9 13.1 12.5 12.3 13.3 13.3 13.9 More 9th 19.6 19.0 23.1 22.0 22.4 24.0 25.7 25.1 Household head living with couple 80.3 77.5 76.9 77.6 76.4 77.7 76.5 78.7 Household head’s main language : 53.7 55.1 54.4 49.1 47.2 49.0 49.7 51.2 Guarani only Household income per capita 404 389 420 360 411 426 450 463 (Guarani-thousands in 2005 values) Region of Residence: Urban 51.9 52.8 53.9 54.5 53.4 54.3 54.0 54.2 Number of other children aged 17 or less 2.7 2.7 2.5 2.6 2.5 2.4 2.3 2.2 Asunción 7.7 6.9 6.5 6.8 6.4 6.1 6.3 6.3 San Pedro 7.6 6.9 6.7 6.5 6.7 6.6 6.5 6.4 Caaguazú 9.0 9.2 8.8 8.7 8.8 8.2 8.9 8.5 Department of Caazapá 10.5 9.3 9.2 9.6 9.2 9.2 9.5 8.7 Residence Alto Paraná 11.5 11.3 11.9 12.6 12.8 11.6 12.3 12.1 Central 23.9 27.3 28.6 28.7 28.5 31.0 29.7 30.8 Others 29.8 29.0 28.3 27.1 27.5 27.3 26.8 27.3 Source: Authors’ estimates from EPHs 2003–10. 209 3. Description of Results Figure 6.2 presents a snapshot of the most recent status of educational, health, and housing opportunities in Paraguay. For every opportunity, the plotted dots denote the overall coverage rate, while the corresponding bars denote the HOI. Recall that the HOI is the inequality adjusted coverage rate, which means that the gap between the coverage and the HOI can be interpreted as a measure of the “penalty” due to inequitable access along the dimensions of the circumstances used in the analysis. The relatively high penalty or “D-Index” suggests that these opportunities— or their lack thereof—are distributed fairly unevenly across circumstance groups. By definition, these inequalities are unfair because society considers that they should be universally distributed. Results of this analysis confirm that children in 2010 in Paraguay still faced substantive differences in terms of equal access to a number of opportunities, such as completing ninth grade, accessing health care and accessing safe sanitation. Their gaps between coverage rate and HOI are statistically significant, as shown in Figure 6.2. This is, however, not the case for other opportunities that display an almost universal access or are close to being universally distributed. Thus, the playing field is more leveled in the case of attending school and accessing water and electricity (and, to a lesser extent, starting school on time). As it would be expected, inequities in those opportunities that are close to be universal are very limited. In contrast, completing ninth grade and access to sanitation services have large differentials in access associated with circumstances. These are precisely those opportunities that have lower coverage rates to start with and therefore are far from being universally distributed: they have large penalties for unequal distribution across circumstances that range between 8 and 15 percentage points. The largest breach between coverage and the HOI is observed for access to sanitation, followed by finishing ninth grade and finishing sixth grade on time (figure 6.2). Figure 6. 2: Coverage and HOI in Paraguay, 2010 100 HOI UB(95) LB(95) 97.0 95.6 UB(95) LB(95) Coverage 90 87.9 80 78.5 75.0 Percentage 70 69.3 93.9 95.2 85.2 63.1 60 61.2 59.9 72.2 73.6 50 61.1 56.0 51.0 47.7 40 Attend Start Finish Finish6th Finish 9th Finish Attend Housing: Access to Access to Housing: Access Housing: Access to School school School school Grade 6th grade 9th grade Pre-School Grade Health prechool health Water care clean water Sanitation to safe Electricity electricity (5-17) on Time on time on Time on time (16-17) (5) Care (0-17) (0-17) (0-17) sanitation (0-17) (6-7) (13) (0-17) (0-17) Source: Authors’ estimates from EPH 2010. 210 What are the main drivers behind the observed access differentials and how they change across opportunities? One way to answer this question is to estimate the individual contribution attributable to each specific circumstance included in the analysis in the access of the opportunity. As described in appendix 1, the contribution of specific circumstances to the D- index can be estimated using a technique called Shapley decomposition. Figure 6.3 presents the results of this decomposition for 2010. The contribution of each circumstance across opportunities is presented as shares in the colored bars. Appendix 2 compares the decomposition for 2003 and 2010. Figure 6. 3: Contribution of Circumstances to Overall Inequality (Shapley Decomposition), 2010 Attend school (5-17) 6 22 6 18 8 19 5 14 Start school on time (6-7) 9 3 4 7 21 3 12 41 Finish 6th grade on time (13) 6 24 3 22 11 8 18 9 Finish 9th grade (16-17) 9 29 15 10 16 9 12 Attend preschool (5) 2 5 11 13 34 16 5 14 Access to health care (0-17) 2 7 11 15 7 32 10 14 Access to clean water (0-17) 12 2 12 11 22 14 25 Access to safe sanitation (0-17) 12 20 17 26 6 17 Access to electricity (0-17) 2 22 3 14 7 19 9 22 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Gender Head: Gender Head: Education Head: Coupled Head: Language Per Cap. Income Urban Other Minors Region Source: Authors’ estimates from EPH 2010. In 2010, different circumstances turn out to be the largest contributors across educational opportunities. To be sure, there does not seem to be a dominant circumstance that drives unequal access for each educational opportunity considered here. Household head’s education is the largest contributor to school attendance for children aged 5 to 17, finish 6th grade on time and finish 9th grade. Household’s per capita income is the single most important contributor to inequalities in preschool attendance, suggesting the importance of demand and preferences in education decisions. Instead, regional location is the main contributor to differentials in starting school on time. Language spoken by the household head, urban-rural location of the household and –to a lesser extent– number of minors in the household are all relevant factors explaining disparities, but their contributions are smaller than those of the circumstances aforementioned. In contrast, the gender of the household head and that of the child do not appear to matter much in explaining differentials in educational opportunities, and the presence of the household head’s spouse does not appear to matter for most of the education opportunities.107 107 World Bank (2010) also reports similar results for a relatively low weight of child’s gender and household head’s gender across educational and housing opportunities in Paraguay. 211 The contributors for the health and housing opportunities tend to be more concentrated on urban/rural and regional residence, which jointly explain between 40 percent and 50 percent of the total disparities observed across those opportunities. This result underlines the importance of supply factors in explaining differentials in these opportunities. Socioeconomic variables—such as household heads’ education, language, and households’ per capita incomes—also appear as systematically relevant contributors. This implies that it is both supply and demand factors that explain disparities behind health and housing opportunities in Paraguay. An alternative way to look at the importance of circumstances consists of constructing “vulnerability profiles” of children across opportunities. These profiles allow identification of the underserved, their characteristics, and how their profiles compare with those who have better than average access. Figure 6.4 presents a snapshot of the vulnerability profiles in 2010 for all opportunities. For each opportunity, the dominant circumstances for children in the lowest quintile of the predicted probabilities of access are compared with the circumstances of children in the highest quintile of the distribution of predicted probabilities of access. For example, consider the opportunity of being enrolled in school for children aged 5–17. Of all the children aged 5–17 who have the lowest probability of attending school, 80 percent of them belong to households whose head has a level of education lower than sixth grade. Instead, 80 percent of those children in the top quintile of that opportunity—that is, those children who have the set of circumstances that make them more likely to enjoy the opportunity—have household heads with an educational attainment of 10th grade or higher. For the sake of presentational ease, this analysis focuses only on three main circumstances: household’s head education, language spoken at home by the household head and region of residence. Figure 6. 4: Vulnerability Profile for Educational Opportunities, 2010 Attend School (age 5-17) Start 1st grade on time (age 6-7) 100 100 Bottom 20% Top 20% Bottom 20% Top 20% 80 80 Percentage Percentage 60 60 40 40 20 20 0 0 Itapua Others Itapua 6th grade 6th grade Others Less than 6th Guarani only Caaguazu Less than 6th Guarani only Caaguazu San Pedro San Pedro Spanish or mixed Alto Parana Spanish or mixed Alto Parana 7th to 9th 10th or more Asuncion 7th to 9th 10th or more Asuncion Central Central Head's Education Language Region Head's Education Language Region 212 Percentage Percentage Percentage 100 20 40 60 80 0 100 100 20 40 60 80 20 40 60 80 0 0 Less than 6th Less than 6th Less than 6th 6th grade 6th grade 6th grade 7th to 9th 7th to 9th 7th to 9th 10th or more 10th or more 10th or more Guarani only Guarani only Guarani only Spanish or mixed Spanish or mixed Spanish or mixed Head's Education Language Head's Education Language Head's Education Language Asuncion Asuncion Asuncion San Pedro San Pedro San Pedro Caaguazu Caaguazu Caaguazu Itapua Itapua Itapua Region Region Region Attend Preschool (age 5) Alto Parana Alto Parana Alto Parana Bottom 20% Bottom 20% Bottom 20% Finish 6th grade on time (age 13) Access to drinkable water (0-17) Central Central Central Others Others Others Top 20% Top 20% Top 20% Percentage Percentage Percentage 20 40 60 80 100 0 100 20 40 60 80 0 100 20 40 60 80 0 Less than 6th Less than 6th 6th grade Less than 6th 6th grade 7th to 9th 6th grade 7th to 9th 7th to 9th 10th or more 10th or more 10th or more Guarani only Guarani only Guarani only Spanish or mixed Spanish or mixed Head's Education Language Head's Education Language Spanish or mixed Head's Education Language Asuncion Asuncion Asuncion San Pedro San Pedro San Pedro Caaguazu Caaguazu Caaguazu Itapua Itapua Itapua Finish 9th grade (age 16-17) Region Region Access to safe sanitation (age 0-17) Bottom 20% Region Alto Parana Alto Parana Bottom 20% Bottom 20% Alto Parana Central Central Central Access to health care services (age 0-17) Others Others Others Top 20% Top 20% Top 20% 213 Access to electricity (age 0-17) 100 Bottom 20% Top 20% 80 Percentage 60 40 20 0 Central Less than 6th 6th grade Itapua Others Guarani only San Pedro Caaguazu 7th to 9th Asuncion Alto Parana 10th or more Spanish or mixed Head's Education Language Region Source: Authors’ estimates from EPH 2010 The strong association between household head’s educational attainment and vulnerability profiles is observed across most of the educational opportunities. This is indeed the case for attending preschool, completing sixth grade on time, completing ninth grade, and starting primary education on time. The vulnerability analysis shows that the household head’s achievement lower than sixth grade is also associated with vulnerabilities in all of the four opportunities considered for health care and housing opportunities. Instead, educational attainment of at least 10th grade or more is associated with being in the top quintile of access probabilities for all those opportunities. A strong story also emerges for the two other circumstances reported in the vulnerability map in figure 6.4. Households whose heads speak mainly Guarani are systematically more likely to belong to the most vulnerable group in terms of all the educational, health and housing opportunities. In contrast, speaking both languages and mainly speaking Spanish is strongly associated with a higher access probability across all opportunities considered. 108 Finally, the vulnerability profile associated with regions also displays clear relationships between specific regions and vulnerability. Households residing in the “other” regional group are consistently more likely to be in the bottom quintile of the distribution of all opportunities. This is especially the case across housing and health care opportunities. It is also the case for education opportunities, but here, departments like Itapúa, Caaguazú, and Alto Paraná have a bearing on belonging to the most vulnerable group. In contrast, Central and—to a lesser extent—Asunción and Alto Paraná are associated with households in the least vulnerable 108 It is only in the case of access to timely medical services that there is a clear association between household heads speaking only Spanish and pertaining to the top quintile of the distribution of opportunity access. And speaking both and only Guarani do not have much of an impact on that probability. 214 group across most opportunities. The strongest associations, seen between regional residence and vulnerability, are found across housing opportunities. In addition to the static analysis for 2010, trends between 2003 and 2010 are also presented for all opportunities in figures 6.5 and 6.6. Figure 6.5 presents the respective annual coverage and the HOI for all educational opportunities. Three findings stand out. First, both the coverage and the HOI increase for preschool attendance, on time completion of sixth grade and completion of ninth grade while they stagnate or grow more slowly for timely start of school and school attendance. Second, some opportunities exhibit marked annual fluctuations such as attending preschool, start first grade on time, and finish sixth grade on time. In other words, annual changes are not steady. Third, finishing ninth grade, one of the most lagging opportunities in terms of low coverage and low HOI, has grown the most steadily across opportunities, but still remains well behind other educational opportunities. Attending preschool –an opportunity whose initial access rates were similar to finishing ninth grade in 2003– has been subject to a deep fluctuation. Thus, there does not appear to be a clear relationship between initial opportunity gaps and subsequent trends. Figure 6. 5: Change in the HOI across Educational Opportunities, 2003 –10 Attend school (age 5-17) Start school on time (age 6-7) 100 85 78.9 76.5 72.9 73.3 72.4 73.6 75.0 75 68.1 87.7 87.9 75.7 90 72.9 Percentage 86.4 Percentage 86.3 71.7 72.2 84.7 85.7 85.1 85.8 65 69.5 69.4 68.2 63.4 84.6 85.2 55 80 82.3 82.2 82.0 82.3 83.2 81.1 45 70 35 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 HOI Coverage HOI Coverage 215 Finish 6th grade on time (age 13) Finish 9th grade (age 16-17) 85 85 75 75 68.3 67.8 66.7 68.5 69.3 65.9 Percentage Percentage 62.2 65 60.4 65 60.3 62.3 59.9 57.6 55.7 53.5 55 60.1 61.5 61.1 55 51.1 51.4 58.9 56.5 57.2 52.6 54.0 45 51.6 45 50.8 51.0 48.5 46.8 42.7 44.0 42.1 35 35 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 HOI Coverage HOI Coverage Attend preschool (age 5) 85 75 Percentage 65 60.0 61.4 59.8 61.2 55.5 55.5 55 52.5 53.3 43.1 56.0 52.4 45 51.0 46.4 47.9 45.2 36.2 35 2003 2004 2005 2006 2007 2008 2009 2010 HOI Coverage Source: Authors’ estimates from EPH 2003–10. Figure 6.6 compares the temporal evolution of coverage and the HOI for health care and housing opportunities. Interestingly, the gap between health service coverage and its HOI has shrunk (by half) over time, as is the case for water and electricity access. In contrast, sanitation access has not changed over time. Compared to education, all housing and health care opportunities have improved over time and have shown a steadier trend than education opportunities. Changes in HOI between 2003 and 2010 are decomposed into three effects: composition — changes in the distribution of circumstances in the population; (ii) scale—changes in coverage; and (iii) equalization—changes in the distribution of disparities across groups. The decomposition analysis (not shown here) confirms two important results with regard to access inequalities. First, the equalization effect is typically a positive contributor to increasing the HOI across opportunities. This implies that there has been an overall reduction in access 216 disparities across circumstance groups for most opportunities. Exemptions to this positive contribution—that is, the opposite effect of increasing disparities—is found for finishing ninth grade. The magnitude of these negative contributions is very small—however, less than 10 percent of the total change between the two years. Second, the contribution of equalization effects is much smaller than scale and composition effects. Equalization explains only between 1 and 25 percent of the HOI inter temporal changes. This implies that there is still a lot of room forequalizing policies to effectively reduce disparities across groups for most opportunities analyzed in Paraguay. Figure 6. 6: Change in the Coverage and the HOI Gap over Time, 2003–10 Access to health care (age 0-17) Access to clean water (age 0-17) 100 100 97.3 81.5 79.5 95.4 95.4 95.6 78.5 94.2 75.0 80 95 92.6 96.4 91.1 91.0 Porcentage Porcentage 63.0 76.4 74.5 73.6 93.3 93.6 93.9 60 54.8 90 50.3 66.0 91.5 89.1 53.7 40 85 87.2 87.6 46.0 40.5 20 80 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 HOI Coverage HOI Coverage Access to health care (age 0-17) Access to clean water (age 0-17) 100 100 80 80 Porcentage Porcentage 63.4 63.4 63.1 63.4 63.4 63.1 60.3 60.3 56.1 56.3 56.1 56.3 60 50.6 52.4 60 50.6 52.4 40 48.5 47.8 47.7 40 48.5 47.8 47.7 44.9 44.9 39.8 41.5 39.8 41.5 34.8 36.9 34.8 36.9 20 20 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 HOI Coverage HOI Coverage Source: Authors’ estimates from EPH 2003–10. a. Health information not available for 2006 EPH. 217 4. Benefit Incidence Analysis of Opportunities This section links traditional fiscal incidence analysis with the new developments of measurement of equality of opportunities. The exercise consists of expanding traditional and commonly used distributive analytical methods for fiscal policy incidence, benefit incidence analysis (BIA), and relate it to the concept of equality of opportunities. The resulting opportunity BIA, or Opp-BIA, presents an incidence analysis of public education and public health care—medical attention—along the distribution of opportunities and compares it with the traditional BIA across income or consumption distributions. The distribution of opportunities is obtained by estimating the probability of each child to access a given opportunity—say, attending school—given his or her set of circumstances. Circumstances in this exercise are the same ones used in the definition of the HOI for Paraguay, that is: child’s gender, household head’s education level and gender; household per capita income; presence of household head’s spouse; number of other children; urban/rural and regional residence; and main language of household head. After estimating these probabilities for each child, children are then assigned to quintiles according to these probabilities by opportunity—in the same way that they are assigned to quintiles of income or consumption. The Opp-BIA has two main advantages over the traditional BIA. First, it allows analysis of the allocation of public resources to education and health against a direct concept of vulnerability in the access to education and health, rather than an indirect concept of vulnerability associated with incomes or consumption. In other words, it allows a sharper picture of the distribution of resources and specific vulnerabilities. Second, it provides insights on how multiple factors (all of those considered relevant circumstances) affect the distribution of education and health care resources. This is not to say that the analysis determines causality between circumstances and educational benefits (in the same way that a traditional BIA does not establish causality between household incomes and education spending), but it certainly complements the insights provided by the traditional BIA based on household per capita income or consumption.109 5. Education The gross unitary benefit—average public spending per student—for children enrolled in public schools and/or publically subsidized private schools in 2009 was estimated as ₲1,153,447 (US$232) for elementary students and ₲2,636,277 (US$531) for students in secondary and technical education.110 Average household contributions—that is, private out- of-pocket expenses of families—toward the education of their children averaged ₲ 541,157 (US$109) per student in elementary and ₲1,007,842 (US$203) in secondary. These averages 109 To identify beneficiaries used in this exercise, the analysis used the 2004 and 2009 EPHs. Numbers for unitary transfers in primary education are obtained from preliminary results of BOOST in Paraguay and the number of beneficiaries provided by the Ministries of Education (2010), Health (2011) and Finance (2013)). Transfers were attributed only to children who reported attending public schools. All values in the following analysis are expressed in Guaranis and US dollars of 2009. 110 In 2004, those averages were ₲573,665 (US$96) and ₲1,590,574 (US$266) per elementary and secondary/technical student, respectively. They represent ₲820,528 (US$165) and 2,275,036 (US$458) in 2009 values, respectively. 218 typically conceal significant differences by location and levels of education, and Paraguay is no exemption. In effect, administrative information on public education spending in Paraguay allows disaggregating that spending by two levels of education: primary (elemental) and secondary-technical (media-técnica). Elemental education includes preschool (inicial) and the first, second and third cycles of primary education, básica. It also includes special education for preschool and primary education (educación especial inicial y básica) and special permanent primary education (educación permanente básica bilingüe). Secondary-technical includes high school education (media), permanent professional secondary education (educación permanente formación profesional), alternative secondary education (educación media alternativa), and secondary distance education (educación media a distancia). Educational budgets disaggregate spending for each of those categories across departments. There is still another spending category, alcance nacional, which budgets do not disaggregate regionally. 111 This category typically includes salaries, educational material, and supplemental food. There are different ways to allocate this spending across regions: one consists of allocating those expenses according to the distribution of reported region-specific spending; another would allocate resources following the distribution of region-specific student enrollments. This analysis opts for allocating alcance nacional resources by geographical enrollment. The reason is that a substantive part of this spending category refers to educative material and food rations, which ultimately depend on numbers of enrolled students. Appendix 3 provides the specific estimates of education unitary benefits per department that result from this rule. Figure 6.7 presents the distribution of the estimated public education benefits per child aged 5–17 across quintiles of per capita household consumption and probability of attending school—that is, the results from the BIA and Opp-BIA, respectively. Each bar represents the proportion of beneficiaries from each quintile (20 percent of the distribution) and the proportion of benefits that the beneficiaries of each quintile capture. Figure 6.7a presents the shares of education benefits and beneficiaries across quintiles of incomes, while 5.7b presents the shares across quintiles of probabilities. Thus, for example, the 20 percent of children with the lowest income per capita benefit some 22 percent of public expenditure in elemental and secondary education in Figure 6.7a. Instead, the 20 percent of children with the lowest probability of attending school (explained by their set of circumstances), benefit some 21 percent of public expenditure in elemental and secondary education (Figure 6.7b). In 2009, the shares of public spending on education in Paraguay followed a near uniform distribution, with each quintile almost capturing a fifth of public benefits except for the last quintile. In this respect, spending on public education in Paraguay is not pro-poor or pro-rich, because it does not appear to favor disproportionately the poorest or the richest, respectively. However, this near uniform distribution is not the result of a neutral distribution of resources across levels of education but, rather, the result of a pro-poor spending on elemental education combined with a pro-rich spending on secondary education. In effect, figure 6.8a confirms that shares of public spending benefits decrease as the consumption of household increases for 111 A final budget item is “education and culture without discrimination,” which includes training for police and armed forces and physical investments associated with cultural assets. These items were not included in the analysis because they would typically not benefit children directly. 219 elemental education, while figure 6.9a shows the opposite occurs on secondary education. The bottom 40 percent of the distribution of beneficiaries of elemental education capture 47 percent of total public resources on elemental education (figure 6.8a) the same percentage of secondary education public spending goes to the top 40 percent of the income distribution (figure 6.9a). The Opp-BIA confirms the distinct distributional incidence patterns for aggregated, elemental, and secondary education spending (figures 6.7b, 6.8b, and 6.9b respectively). To be sure, for the combined public spending on elementary and secondary education, having a set of circumstances less favorable to attend school does not make a large difference in the public benefits that child will receive. Only children with the most favorable set of circumstances receive less public benefits than their share in the population would suggest. The main reason is that these children opt for private education. Results highlight, once again, an elemental education that is close to universal and equitable access and a slightly pro-rich distribution of public spending on secondary education. These key results are also confirmed for 2004, as shown in appendix 4. Figure 6. 7: Share of Aggregate Public Expenditure on Education (age 5-17), 2009 a. b. Share of public expenditure on Share of public expenditure on education by quintile of incomes (2009) education by quintile of probability (2009) 40 40 30 30 24 Percentage Percentage 22 22 21 21 21 20 20 20 20 20 20 20 20 20 20 20 20 20 19 16 14 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) Figure 6.1: Share of Public Expenditure on Elemental Education (age 5-14), 2009 a. b. Share of public expenditure on elemental Share of public expenditure on elemental education by quintile of incomes (2009) education by quintile of probability (2009) 40 40 30 30 25 25 24 Percentage Percentage 22 22 21 20 20 20 20 20 20 20 20 20 20 20 20 19 18 13 12 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) The distribution of spending share per group allows determination of the pro-poor nature of spending. In contrast, net unitary benefits—that is, the average benefit per student net of 220 the family’s private out-of-pocket spending toward his or her education—depict the progressivity of public spending. Progressivity in spending is understood as children in poorer household’s receiving increasingly larger net benefits from access to public education. The following figures report the distributions of net unitary spending for aggregate, elementary, and secondary education. As for the shares of spending reported before, both the incidence across quintiles of incomes and quintiles of opportunities are presented. Figure 6. 8: Share of Public Expenditure on Secondary Education (age 16-17), 2009 a. b. Share of public expenditure on secondary Share of public expenditure on secondary education by quintile of incomes (2009) education by quintile of probability (2009) 40 40 30 30 26 24 Percentage Percentage 23 22 21 21 20 20 20 20 20 20 20 20 20 20 20 20 20 20 10 10 11 12 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) Figure 6. 9: Distribution of Unitary Public Expenditures on Education Net of Private Household Contributions (age 5-17), 2009 a. b. Unitary public expenditure on education Unitary public expenditure on education by quintiles of incomes (2009) by quintiles of probability (2009) 300 300 279 282 291 272 267 200 200 260 258 265 251 236 US$ of 2009 US$ of 2009 100 100 209 218 204 177 162 148 128 103 68 36 0 0 -70 -64 -87 -95 -112 -105 -130 -149 -100 -100 -197 -199 -200 -200 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) 221 Figure 6. 10: Distribution of Unitary Public Expenditures on Elemental Education Net of Private Household Contributions (age 5-17), 2009 a. b. Unitary public expenditure on elemental education Unitary public expenditure on elemental education by quintiles of incomes (2009) by quintiles of probability (2009) 300 300 262 264 260 200 200 242 228 236 219 218 208 US$ of 2009 US$ of 2009 100 100 197 204 193 157 182 126 141 100 73 43 14 0 0 -65 -85 -60 -78 -102 -120 -95 -136 -100 -100 -175 -179 -200 -200 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) Figure 6. 11: Distribution of Unitary Public Expenditures on Secondary Education Net of Private Household Contributions (age 5-17), by Quintiles of Households, 2009 a. b. Unitary public expenditure on secondary education Unitary public expenditure on secondary education by quintiles of incomes (2009) by quintiles of probability (2009) 600 600 555 566 568 536 400 400 505 505 522 497 487 447 US$ of 2009 US$ of 2009 200 200 408 439 402 355 315 307 338 275 188 148 0 0 -147 -182 -128 -166 -190 -197 -186 -223 -300 -300 -200 -200 -400 -400 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) Results confirm the unambiguous progressivity of public spending on education in Paraguay, after including households’ private contributions. Such contributions increase alongside the level of consumption of the household. Simply put, richer households incur higher out-of-pocket expenses on education than poorer households do. This is true for elemental grades as well as for secondary education. Interestingly, gross unitary benefits per student slightly decrease as household consumption increases. This implies that the public education system is progressive because children in poorer household appear to receive slightly larger benefits from attending school while their families incur in lower out-of-pocket expenses towards their education. 222 In the context of this static incidence analysis, larger transfers mean larger costs of education because provision costs are equated one-on-one with education benefits.112 So, more beneficiaries from poor households attend public schools in departments with higher provision costs. These higher costs typically reflect higher salaries of teachers in rural and isolated schools and the allocation of centralized administrative expenses of the educational system (alcance nacional) across fewer students in smaller departments. Also, these results reflect that richer families opt out of the public system in favor of enrolling their children in private schools. Finally, richer families spend more on fees, texts, school materials, and other costs than poorer families.113 6. Health care The distributional incidence of public health care spending in Paraguay shows that it is neither pro-poor nor progressive. In fact, the share of spending benefiting middle-income groups—that is, children in households of the third quintile of the distribution—is larger than the share of low- and high-income group quintiles. Beneficiaries in the third quintile capture 31 percent of benefits of public health care (figure 6.13a). The remaining groups, the bottom 40 percent and the top 40 percent, capture 31 percent and 38 percent, respectively, that is, slightly below their proportional population shares. The disaggregation by types of attention, health centers, and hospitals confirms that the middle income quintile captures a larger share of benefits than other quintiles. It also shows that there are different distributional profiles for health care in centers and hospitals: while the bottom quintile of the distribution of incomes disproportionally benefits from health center services (figure6.14a), it is also disproportionally not benefiting from hospital care related public spending (figure6.15a). The opposite is observed for the top quintile: it does not benefit much from public spending on health care centers (figure 6.14a), while benefits from hospital care benefits more than its share of beneficiaries (figure 6.15a) Incidence results across the distribution of probability confirm by and large that the middle quintile benefit the most from public health care spending. Figure 6.13b also shows that public health care spending is not pro-poor when analyzed by quintiles of opportunities. The children with the set of circumstances most favorable –bottom two quintiles– only capture 29 percent of benefits (figure 6.13b). The disaggregation by nature of the attention, health center, or hospital related indicates that there are marked distributional differences as well. Medical attention in health centers is pro-poor when analyzed by quintiles of opportunities: the bottom 40 percent captures some 54 percent of such benefits (figure 6.14b). However, hospital care spending is pro-rich along the distribution of probabilities. The bottom 40 percent of the distribution captures 19 percent percent of all benefits associated with hospital care (figure 6.15b). Thus, the incidence of public health care spending by quintiles of opportunities reveals that health centers tend provide a pro-poor service, while hospitals provide a pro-rich service. 112 Because the analysis is static, it does not include future rates of return to education, that is, their true investment side. 113 Some 39 percent of children in the top quintile of the distribution of beneficiaries attend private schools; only 3 percent of children in each of the two bottom quintiles do. Those children from the top quintile who go to public or publically subsidized schools spend on average three times more than children from the bottom two quintiles. 223 Figure 6. 12: Share of Public Expenditures on Health Care (age 0-17), 2009 a. b. Share of public expenditure on health care Share of public expenditure on health care by quintile of incomes (2009) by quintile of probability (2009) 40 40 31 29 30 30 Percentage Percentage 22 20 20 20 20 20 20 20 20 20 20 20 20 20 20 18 17 15 14 14 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) Figure 6. 13: Share of Public Expenditures on Health Center Care (age 0-17), 2009 a. b. Share of public expenditure on health center care Share of public expenditure on health center care by quintile of incomes (2009) by quintile of probability (2009) 40 40 29 30 30 28 28 25 Percentage 24 Percentage 23 20 20 20 20 20 20 20 20 20 20 20 20 14 14 10 10 10 7 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) Figure 6. 14: Share of Public Spending on Hospital Care (0-17), 2009 a. b. Share of public expenditure on hospital care by quintile of incomes (2009) Share of public expenditure on hospital care by quintile of probability (2009) 40 40 36 32 30 30 Percentage 23 24 Percentage 20 20 20 21 20 20 20 20 20 20 20 18 19 20 20 10 10 11 8 8 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) The distribution of net unitary benefits per health care is neither progressive nor regressive in Paraguay. The average benefit from health care—without any disaggregation—neither systematically increases nor decreases along with income levels (figure 6.16a). Household out- of-pocket contributions do not seem to follow a monotone trend either although beneficiaries at 224 the bottom 40 percent incur less out-of-pocket contributions in absolute terms. The result is a net benefit for the top quintile of incomes that exceeds that of the poorest quintile, reaching the largest net benefit among those beneficiaries from the middle income group. This regressivity conceals, once again, distinct patterns for health centers and hospital-related spending profiles. Figure 6.17a and 6.18a depict nonlinear patterns along with consumption levels. They are neither progressive nor regressive. In the case of the distributive incidences depicted by the quintiles of probability, net benefits from public spending on health centers first decrease to then increase (figure 6.17b), while they remain mostly uniform across quintiles of probability for hospital care except, again, for the mid-income group (figure 6.18b). There are several possible explanations for these patterns. One option is that richer households, and/or those with circumstances making them more likely to access public health, do not so clearly opt out of the public system. In addition, and contrary to education, different medical conditions and accidents require different types of health care. The complexity and severity of required attention should not be expected to follow clear socioeconomic nor geographical lines. Outliers do not drive these results either. In fact, results do not change when two scarcely populated departments—Canindeyú and Ñeembucú—with extremely high average public transfers on public health are removed (appendix 5). Another potential source of bias in these estimates, self-medication, does not fabricate the reported trends. Appendix 6 show results after considering that all self-medicated individuals are in fact included in the definition of health care opportunity. Finally, it is also worth noting that these distributive incidences are not the result of the poor demanding less—or reporting less—health services because this analysis considers only those who demanded attention to a condition considered serious enough to demand some medical attention in the first place. Figure 6. 15: Distribution of Unitary Public Expenditures on Health Care Net of Private Household Contributions (age 0-17), 2009 a. b. Unitary public expenditure on health care Unitary public expenditure on health care by quintiles of incomes (2009) by quintiles of probability (2009) 400 400 403 370 372 361 300 300 310 301 241 US$ of 2009 200 234 US$ of 2009 200 242 219 326 312 310 283 259 100 238 100 192 179 182 189 0 0 -51 -40 -63 -60 -53 -52 -59 -51 -77 -79 -100 -100 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) 225 Figure 6. 16: Distribution of Unitary Public Expenditures on Health Center Care Net of Private Household Contributions (0-17), 2009 a. b. Unitary public expenditure on health center care Unitary public expenditure on health center care by quintiles of incomes (2009) by quintiles of probability (2009) 200 200 213 184 178 173 177 157 148 100 100 US$ of 2009 US$ of 2009 138 140 105 133 99 79 113 108 116 86 40 37 30 0 0 -46 -39 -46 -50 -56 -62 -73 -64 -69 -69 -100 -100 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) Figure 6. 17: Distribution of Unitary Public Spending on Hospital Care Net of Private Household Contributions (age 0-17), 2009 a. b. Unitary public expenditure on hospital care Unitary public expenditure on hospital care by quintiles of incomes (2009) by quintiles of probability (2009) 600 662 600 607 554 513 495 458 467 479 477 400 400 463 US$ of 2009 US$ of 2009 613 525 451 200 470 500 419 389 379 200 388 405 0 -84 -60 -49 -44 -88 0 -70 -43 -82 -62 -54 -200 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Public expenditures Household expenditures Net benefit Public expenditures Household expenditures Net benefit Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) 7. Policy implications: targeting additional spending The expanded incidence analysis provides an alternative strategy targeting public spending that complements the traditional focus on those with lowest access. This strategy consists of integrating both outcomes and opportunities in targeting decisions. This can be done by targeting additional spending to population groups with larger gaps between their share of public benefits and their share of population and with sets of circumstances that make them less likely to gain access to an opportunity by themselves. The number of groups that result from considering all nine circumstances at once exceeds 1,300 in Paraguay. For the purpose of illustration, this analysis considers simply the education level of the household head (less than sixth grade, sixth grade completed, seventh to ninth grade completed, or higher than ninth grade); language of household head (only Guarani, only Spanish/mixed); and residence (urban, rural). The combination of these three categories, –found 226 to be most critical in explaining educational and health HOIs in section 4–, defines 16 circumstance groups, as reported in table 6.3. This example focuses on two opportunities, attending secondary school and accessing public hospital health care, which are particularly notorious in distributive terms (see appendix 7 for a similar analysis on other opportunities). Table 6.3 reports the probabilities of accessing both opportunities across the 16 circumstance groups. Table 6.3: Circumstance Groups and Their Probabilities to Attend School and Access to Health Care Services (when sick) Average Average Household head: probability of Household head: probability of Education Region of accessing health Code main language attending (grade residence services when spoken at home school completed) required (age 5-17) (age 0-17) 1 5th or less 77 percent 64 percent 2 6th 88 percent 75 percent Guarani only 3 7th to 9th 85 percent 76 percent 4 10th or more 91 percent 84 percent Rural 5 5th or less 82 percent 71 percent 6 6th Spanish only or 90 percent 81 percent 7 7th to 9th Mixed 88 percent 78 percent 8 10th or more 93 percent 86 percent 9 5th or less 84 percent 77 percent 10 6th 91 percent 85 percent Guarani only 11 7th to 9th 89 percent 80 percent 12 10th or more 94 percent 88 percent Urban 13 5th or less 89 percent 83 percent 14 6th Spanish only or 94 percent 87 percent 15 7th to 9th Mixed 91 percent 85 percent 16 10th or more 96 percent 90 percent Source: Authors’ estimates from EPH 2009. After sorting these 16 circumstance groups by their probability of access to secondary school (from least to most likely) and their probability of access to public hospital care, respectively, figure 6.19 and figure 6.20 report the gap between each group’s share of population and share of benefits. For access to secondary education, groups that have a lower than average probability to access and receive a share of public resources lower than their population shares are children in households whose heads speak only Guarani or, if speak Spanish or both, live in rural areas. These groups are those coded as 13, 11, 3, 2 and 1 in figure 6.19. Interestingly, groups with household heads that speak only Guarani, reside in urban areas (and well educated)– that is, groups 10 and 14– opt out of the public education system to a considerable extent: between 20 percent and 36 percent of children in those circumstance groups are enrolled in private education (vis-à-vis between 1 percent and 5 percent of children in the 227 other mentioned groups). This would lead only five very specific circumstance groups to which target additional resources following the integrated outcome and opportunity criteria. Similarly, children with the least probability of accessing public health care in hospitals and having the largest gap—in relative terms—between their population and benefit shares live in rural households whose heads speak only Guarani and have low levels of education (sixth grade or less) as seen in groups 1 and 5. Also in this category –of groups receiving proportionally less benefits than their population shares– are children in households whose heads speak Spanish or both languages, reside in rural areas and typically have heads with low education. Those groups are coded 14, 13, 10, 11, 7 and 3 in figure 6.20. Admittedly, these groups represent a small fraction of the population –20 percent of the population for the eight groups identified– but their gaps in terms of population-benefit shares is very large in proportional terms. These groups –most of which do not appear to opt out of the public health system in favor of private attention 114 – would constitute obvious candidates for targeted interventions. It is also worth noting that these groups are not exactly the same ones by opportunity (although rural households with less educated heads speaking Guarani only are the least advantaged in both cases), which underscores the potential need for different target groups across equalizing interventions. Figure 6. 18: Share of Public Spending on Secondary Education by Circumstance Group (age 15-17), 2009 Share of public expenditure on secondary education by circumstance groups in 2009 Language: Head's education: 30 G Guarani only 0-5 Less than 6th grade S-M Spanish or mixed 25 6 6th grade Residence: 7-9 7th to 9th grade R Rural Percentage 19.3 19.6 20 10+ 10th grade or higher U Urban 18.7 15 12.7 13.4 10 11.4 8.4 8.9 7.5 6.9 6.3 3.9 4.5 4.8 6.8 6.7 5 2.6 4.5 5.2 4.8 3.8 1.9 0.8 1.9 1.8 3.5 0.4 3.4 2.6 0.8 0.7 1.6 0 0-5 6 7-9 0-5 0-5 6 7-9 10+ 6 7-9 10+ 0-5 6 10+ 7-9 10+ G G G S-M G S-M S-M G G G S-M S-M S-M G S-M S-M R R R R U R R R U U R U U U U U Population (age 5-17) Public Expenditure Groups sorted by average probability of attending school Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009) 114 Only groups 10 and 14 appear to substantially seek private health providers when demanding medical attention. In effect, these groups attend to private providers in one out of three cases of sickness and/or accidents, according to EPH 2009. 228 Figure 6. 19: Share of Public Expenditure on Hospital Health Care by Circumstance Group Share of public expenditure on hospital care by circumstance groups in 2009 40 Language: Head's education: G Guarani only 0-5 Less than 6th grade 30.7 S-M Spanish or mixed 30 6 6th grade Percentage Residence: 7-9 7th to 9th grade R Rural 10+ 10th grade or higher 23.7 U Urban 20 17.7 11.9 10.6 9.4 10 8.2 7.4 6.9 5.2 7.4 3.7 3.0 3.0 2.1 6.8 7.1 5.1 5.3 1.0 1.2 1.2 4.9 1.4 3.7 0.6 3.0 2.1 0.1 3.4 0.1 2.1 0 0-5 6 0-5 0-5 7-9 6 7-9 6 10+ 0-5 10+ 7-9 6 10+ 7-9 10+ G G G S-M G S-M S-M G G S-M S-M G S-M G S-M S-M R R U R R R R U R U R U U U U U Population (age 0-17 ill only) Public Expenditure Groups sorted by average probability of accessing health services Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) 8. Summary and Conclusion This study analyzes the state of human opportunity among children in Paraguay and the role that public spending has played in the distribution of such opportunities . The analysis focuses on educational, health care and housing opportunities, and considers socioeconomic, demographic, geographic, and ethnic-related variables as relevant circumstances. Some opportunities are close to universal distribution in Paraguay, such as access to electricity, access to water, and school attendance for children age 5 to 17. Also, Paraguay has also made significant progress toward more equitable access to basic public services since 2003. These improvements have taken place across most, although some like school attendance and timely start of school, progress has not been significant or steady. Finishing sixth grade on time and attending preschool have been also subject to large annual fluctuations, but improvements in coverage and HOI have been significant. Furthermore, improvements in the HOI have accrued from increases in coverage across the board rather than from equalizing changes, that is, changes in access disproportionally benefiting the most vulnerable. Ultimately, circumstances such as household heads’ education, household incomes and location still matter for education, health care, and housing opportunities in Paraguay; they continue to determine how vulnerable children are in terms of their access to basic services. Importantly, there is not a single dominant circumstance that matters the most across all opportunities. This underscores that one-size-fits-all interventions aiming to address disparities for all Paraguayan citizens are unlikely to be effective across different opportunities. 229 Public spending on education was found to be neither pro-poor nor pro-rich. This is not the result of universal access to education, but of a combination of a pro-poor elemental education spending and a pro-rich secondary education spending. Interestingly, public spending on (both elemental and secondary) education is progressive, because it benefits increasingly more of the poorer households and because richer households contribute increasingly more out of their pockets when attending public education centers. The patterns of public spending on health care appear less conclusive, with benefits concentrating in the middle group of the income distribution. Also, distributional patterns for care in health centers and hospital differ. These results may reflect some opting out of the public system by richer households as well as some specific patterns of illness and accidents that fail to follow simple socioeconomic patterns. These results, however, already control for socioeconomic differences in the demand of health services. Looking ahead, the analysis has identified a simple mechanism to channel public resources more equitably. Additional public spending may be better targeted to those population groups that (i) experience a large degree of vulnerability in their access to a given opportunity (that is, have a set of circumstances that make them unlikely to access the opportunity) and (ii) average large gaps between their share of total population and their share of public benefits associated with that opportunity. In Paraguay, this prioritization exercise identified children in households whose heads speak Guarani and have low educational attainments—both in urban and rural settings—as the groups that would benefit the most in terms of reducing inequitable differentials in access to secondary education and access to public hospital care, respectively. 9. References Barrs, R., F. Ferreira, J. Molinas Vega, and J. Saavedra. (2009). Measuring Inequality of Opportunities in Latin America and the Caribbean. International Bank for Reconstruction and Development/World Bank. Instituto de Previsión Social, IPS, (2011) Prestaciones de la Red de Servicios de IPS sobre Niveles de Atención por Departamento. Paraguay 2008 y 2011. Tabulaciones Especiales. Ministerio de Educación (2009) Estadística Educativa. Anuario 2009. Dirección General de Planificación Educativa. -------------. (2004) Estadística Educativa. Anuario 2010. Dirección General de Planificación Educativa. Ministerio de Finanzas (2013) Social Expenditures, Paraguay. Special Tabulations. Ministerio de Salud Pública y Bienestar Social, MSPyBS, (2009) Indicadores Básicos de Salud. Paraguay 2009. Molinas, J.. R. Paes, J. Saavedra, and M. Giugale.(2010). Do Our Children Have a Chance? The 2010 Human Opportunity Report for Latin America and the Caribbean. International Bank for Reconstruction and Development/World Bank. Hoyos, A., and A. Narayan. (2011). "Inequality of Opportunities among Children: How Much Does Gender Matter?" Background Paper for WDR 2012, World Bank, Washington, DC. Roemer, J. (1998). Equality of Opportunity. Cambridge, MA: Harvard University Press. Sen, A. (1977). "Social Choice Theory: A Re-examination." Econometrica 45(1): 53–89. ———. (2001). Development as Freedom. Oxford: Oxford University Press. 230 World Bank. (2006). World Development Report: Equity and Development. Washington, DC. ———. (2009). “Country Partnership Strategy, Paraguay.” Washington, DC. ———. (2010). “Paraguay Poverty Assessment.” Washington, DC. ———. (2013). World Bank Database. http://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG/countries?display=default. 10. Appendix 1: Computing the Human Opportunity Index from Household Survey Data To construct the HOI, one must obtain the conditional probabilities of access to opportunities for each child based on their circumstances. The analysis therefore estimated a logistic model, linear in the parameters β, where the event I corresponds to accessing the opportunity (for example, access to clean water), and x the set of circumstances (for example, gender of the child, education and gender of the head of the household, and so forth). The following is the logistic regression using survey data: (1)  PI  1 X  (x1 ,..., x m )  m Ln    1  PI  1 X  (x ,..., x )   x k  k  1 m  k 1 where xk denotes the row vector of variables representing the k-dimension of circumstances, hence, x  (x1 ,..., x m ) and    (1 ,...,  m ) a corresponding column vector of parameters. From the estimation of this logistic regression, one can obtain estimates of the parameters  k  to be denoted by  ˆ   k ,n where n denotes the sample size. Given the estimated coefficients, one can obtain for each individual in the sample his/her predicted probability of access to the opportunity in consideration: (2) ˆ i ,n   Exp x i  ˆ n   i n p . 1  Exp x  ˆ Finally, the analysis computes the overall coverage rate, C, the D-index, the penalty, P, and the ˆ and sampling weights, w: HOI using the predicted probability p (3) n n 1 C  ˆi ,n wi p D 2C wi p ˆi ,n  C i 1 i 1 P  C * D ; and HOI  C  P . Shapley Decomposition: Identifying How Each Circumstance “Contributes” to Inequality Following Barros et al. (2009), inequality of opportunities can be measured by the penalty (P) or by the D-index (D), as defined in expressions (1) and (3) above. The value of these two 231 measures—where P is just a scalar transformation of D—is dependent on the set of circumstances considered. Moreover, they have the important property that adding more circumstances always increases the value of P and D. If there are two sets of circumstances A and B, and set A and B do not overlap, then ( ) ( ) and alternatively ( ) ( ). The impact of adding a circumstance A is given by: | | ( | | ) ∑ [ ( ) ( )] ( ) Where N is the set of all circumstances, which includes n circumstances in total; S is a subset of N that does not contain the particular circumstance A. D(S) is the D-index estimated with the set of circumstances S. ( ) is the D-index calculated with set of circumstances S and the circumstance A. The contribution of circumstance A to the D-index can be defined as: ( ) ( ) ∑ . 232 11. Appendix 2: The Contribution of Circumstances to Overall Inequality (Shapley Decomposition) 2003 Attend school (5-17) 24 17 10 20 9 17 Start school on time (6-7) 8 24 7 12 12 27 9 Finish 6th grade on time (13) 6 26 10 18 6 14 18 Finish 9th grade (16-17) 3 21 20 14 14 11 14 Attend preschool (5) 3 4 18 17 15 11 12 19 Access to health care (0-17) 2 15 16 19 20 13 14 Access to clean water (0-17) 10 16 13 33 5 21 Access to safe sanitation (0-17) 14 20 16 23 8 18 Access to electricity (0-17) 16 14 16 24 10 17 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Gender Head: Gender Head: Education Head: Coupled Head: Language Per Cap. Income Urban Other Minors Region 2010 Attend school (5-17) 6 22 6 18 8 19 5 14 Start school on time (6-7) 9 3 4 7 21 3 12 41 Finish 6th grade on time (13) 6 24 3 22 11 8 18 9 Finish 9th grade (16-17) 9 29 15 10 16 9 12 Attend preschool (5) 2 5 11 13 34 16 5 14 Access to health care (0-17) 2 7 11 15 7 32 10 14 Access to clean water (0-17) 12 2 12 11 22 14 25 Access to safe sanitation (0-17) 12 20 17 26 6 17 Access to electricity (0-17) 2 22 3 14 7 19 9 22 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Gender Head: Gender Head: Education Head: Coupled Head: Language Per Cap. Income Urban Other Minors Region Source: Authors’ estimates from EPH 2003–2010. Note: When marked with (*), age refers to age on April of the year of the survey —start of school year, otherwise it refers to age at the moment of the survey. This analysis for 2003 reveals similar results to those reported for 2010, although with some changes. Thus, the education of the household head has the largest share in all five educational opportunities in 2003. Similar to 2010, however, language of the household head, 233 presence of other minors in the household, household’s urban/rural location, regional residence, and per capita income of the household also have significant stakes in explaining disparities across the educational opportunities. And as in 2010, neither do they follow a systematic pattern. Urban/rural location seems to matter for school attendance (both for those age 5–17 and preschool), which may indicate supply restrictions, while timely start and—consequently— timely completion of first and sixth grades, respectively, also seem to be associated with other minors in the household, which may suggest some internal competition for resources within the household. As in 2010, child’s gender, household head’s gender, and presence of household head’s spouse/couple are systematically not substantive contributors to inequalities in educational opportunities. In fact, these circumstances do not seem to explain much of the inequalities surrounding health services and housing amenities either. Interestingly, the contribution of all other circumstances to health services also seems to be very similar in 2003. This is not surprising given that this opportunity already shows very small disparities in access. As also expected regarding housing amenities or access to water, sanitation and electricity, location circumstances are the single most important drivers of their inequalities. Urban/rural and regional residences jointly account for 40–50 percent of observed disparities. Language and education of the head of the household, and the households’ per capita income explain most of the remaining disparities. 12. Appendix 3: Gross Unitary Expenditure on Public Education and Health Care Opportunities, 2004 and 2009 Education 2009 Public health care 2009 Education 2004 Elemental Secondary Health center Hospital Elemental Secondary Asuncion 847,271 2,462,488 171,818 112,686 942,055 1,789,581 Concepción 1,279,107 2,802,588 301,890 338,566 552,994 1,379,013 San Pedro 1,463,325 3,166,206 258,389 790,051 659,951 1,847,873 Cordillera 1,321,578 2,974,520 217,301 5,690,458 603,319 1,783,754 Guairá 1,487,230 3,142,344 210,839 261,535 633,887 1,543,038 Caaguazú 1,345,340 2,736,013 203,564 385,969 592,250 1,542,498 Caazapá 1,461,186 2,943,693 143,242 5,043,878 629,404 1,759,968 Itapúa 1,178,578 2,885,961 365,116 318,681 538,687 1,652,612 Misiones 1,440,166 3,590,737 240,807 437,915 592,290 2,292,872 Paraguarí 1,697,191 3,030,613 227,474 5,006,153 767,251 1,551,409 Alto Paraná 999,489 1,987,678 266,546 288,071 454,108 1,190,507 Central 879,597 2,055,014 264,080 4,225,978 437,471 1,300,564 Ñeembucú 1,505,655 4,478,997 31,836,992 227,091 630,249 2,292,993 Amambay 1,002,305 3,965,219 2,401,953 339,755 500,504 1,615,342 Canindeyú 1,248,840 3,159,369 9,887,237 3,427,084 499,945 1,658,734 Presidente 1,587,800 3,234,543 Hayes 616,888 220,496 714,349 2,423,521 Boquerón 503,013 2,658,032 58,188,777 3,184,255 242,609 1,711,564 234 Alto Paraguay 756,017 4,096,161 1,224,133 4,928,253 458,829 3,696,832 Total 1,153,447 2,636,277 225,900 225,900 573,665 1,590,574 Source: Authors’ estimates from EPH 2004, 2009. Benefits measured in ₲. 13. Appendix 4: Share of Public Expenditure on Education (age 5-17), 2004 Figure A4.1: Share of Public Expenditure on Education by Quintiles of Households (age 5-17), 2004 a. b. Share of public expenditure on Share of public expenditure on education by quintile of incomes (2004) education by quintile of probability (2004) 40 40 30 30 Percentage Percentage 21 21 22 21 20 20 20 20 21 20 20 20 20 20 20 20 20 20 19 18 18 18 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure Figure A4.2: Share of Public Expenditure on Elemental Education (age 5-14), 2004 a. b. Share of public expenditure on elemental Share of public expenditure on elemental education by quintile of probability (2004) education by quintile of incomes (2004) 40 40 30 30 Percentage Percentage 22 22 22 22 21 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 16 15 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure Figure A4.3: Share of Public Expenditure on Secondary Education (age 15-17), 2004 a. b. Share of public expenditure on secondary education by quintile of incomes (2004) Share of public expenditure on secondary education by quintile of probability (2004) 40 40 30 30 26 26 27 25 Percentage Percentage 22 22 20 20 20 20 20 20 20 20 20 20 20 20 16 16 10 10 10 9 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 5 to 17) Public expenditure Popuation (age 5 to 17) Public expenditure 235 Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2004) Appendix 5: Share of Public Expenditure on Health Care (age 0-17), 2009 Figure A5.1: Share of Public Expenditure on Health Care (age 5-17), 2009 (Outlier Departments Excluded) a. b. Shares of public expenditure and population Shares of public expenditure and population by quintiles of income (age group 0 to 17 - ill only) in 2009 by quintiles of probability of visiting health professional when ill in 2009 Expenditure on public health centers or hospitals Expenditure on public health centers or hospitals 40 40 35 29 30 30 Percentage Percentage 22 20 20 20 21 20 20 21 20 20 20 20 20 20 20 20 20 10 10 13 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Population (0 to 17- ill only) Public Expenditure Population (0 to 17- ill only) Public Expenditure Excluding Ñeembucú and Canindeyú Excluding Ñeembucú and Canindeyú Figure A5.2: Share of Public Expenditure on Health Center Care (age 5-14), 2009 (Outlier Departments Excluded) a. b. Shares of public expenditure and population by quintiles of income (age group 0 to 17 - ill only) in 2009 Shares of public expenditure and population by quintiles of probability of visiting health professional when ill in 2009 Expenditure on public health centers Expenditure on public health centers 40 40 30 30 Percentage 27 Percentage 26 23 23 21 20 20 21 20 20 20 20 20 20 20 20 20 20 20 20 10 10 10 11 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Population (0 to 17- ill only) Public Expenditure Population (0 to 17- ill only) Public Expenditure Excluding Ñeembucú and Canindeyú Excluding Ñeembucú and Canindeyú Figure A5.3: Share of Public Expenditure on Hospital Care (age 15-17), 2009 (Outlier Departments Excluded) a. b. Shares of public expenditure and population Shares of public expenditure and population by quintiles of income (age group 0 to 17 - ill only) in 2009 by quintiles of probability of visiting health professional when ill in 2009 Expenditure on public hospitals Expenditure on public hospitals 40 40 37 30 30 30 Percentage Percentage 23 24 20 20 20 20 20 21 20 20 19 20 19 20 20 20 20 10 10 11 8 7 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Population (0 to 17- ill only) Public Expenditure Population (0 to 17- ill only) Public Expenditure Excluding Ñeembucú and Canindeyú Excluding Ñeembucú and Canindeyú Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) 236 Appendix 6: Share of Public Expenditure on Health Care (age 0-17), 2009 (considering self-medication as part of health service access) Figure A6.1: Share of Public Expenditure on Health Care (age 0-17), 2009 a. b. Share of public expenditure on health care Share of public expenditure on health care by quintile of incomes (2009) by quintile of probability (2009) 40 40 30 30 26 26 27 Percentage Percentage 23 20 20 20 20 20 20 20 20 20 20 20 20 20 17 17 16 14 14 10 10 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Excluding self medication as part of public health service access Excluding self medication as part of public health service access Figure A6.2: Share of Public Expenditure on Health Center Care (age 0-17), 2009 a. b Share of public expenditure on health center care Share of public expenditure on health center care by quintile of incomes (2009) by quintile of probability (2009) 40 40 40 38 30 30 28 27 Percentage Percentage 20 20 20 20 20 20 20 20 20 20 20 20 19 15 10 10 12 11 7 3 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Excluding self medication as part of public health service access Excluding self medication as part of public health service access 237 Figure A6.3: Share of Public Expenditure on Hospital Care (age 0-17), 2009 a. b. Share of public expenditure on hospital care Share of public expenditure on hospital care by quintile of incomes (2009) by quintile of probability (2009) 40 40 34 30 30 30 28 25 Percentage Percentage 22 21 20 20 20 20 20 20 20 20 20 20 20 20 19 10 10 8 7 6 0 0 Q1 (Poorest) Q2 Q3 Q4 Q5 (Richest) Q1 (Least) Q2 Q3 Q4 Q5 (Most) Popuation (age 0 to 17- ill only) Public expenditure Popuation (age 0 to 17- ill only) Public expenditure Excluding self medication as part of public health service access Excluding self medication as part of public health service access Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011) Appendix 7: Share of Public Expenditure across Selected Opportunities by Circumstance Group, 2009 Table A7.1 Share of Public Expenditure on Education by Circumstance Group Share of public expenditure on education by circumstance groups in 2009 Language: Head's education: 30 G Guarani only 0-5 Less than 6th grade S-M Spanish or mixed 25 6 6th grade Residence: 21.3 7-9 7th to 9th grade R Rural Percentage 20 10+ 10th grade or higher U Urban 18.7 19.3 14.1 15 13.0 10 11.4 7.5 6.8 6.7 5.8 4.8 4.1 7.1 4.2 3.9 6.8 6.2 5 2.9 1.0 5.2 3.8 1.9 1.9 4.5 1.8 3.5 0.8 3.4 2.6 1.7 1.9 1.6 0 0-5 6 7-9 0-5 0-5 6 7-9 10+ 6 7-9 10+ 0-5 6 10+ 7-9 10+ G G G S-M G S-M S-M G G G S-M S-M S-M G S-M S-M R R R R U R R R U U R U U U U U Population (age 5-17) Public Expenditure Groups sorted by average probability of attending school Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009). 238 Table A7.2 Share of Public Expenditure on Elemental Education by Circumstance Group Share of public expenditure on elemental education by circumstance groups in 2009 Language: 30 Head's education: G Guarani only 0-5 Less than 6th grade S-M Spanish or mixed 25 23.5 6 6th grade Residence: 7-9 7th to 9th grade R Rural Percentage 20 10+ 10th grade or higher U Urban 18.7 19.3 14.4 15 10 11.4 11.1 7.8 6.8 6.7 7.5 5.7 4.1 4.1 4.8 3.6 6.4 5 3.1 5.2 5.4 2.0 1.1 1.9 1.8 3.8 3.5 0.8 3.4 3.9 2.6 1.9 1.9 1.8 0 0-5 6 7-9 0-5 0-5 6 7-9 10+ 6 7-9 10+ 0-5 6 10+ 7-9 10+ G G G S-M G S-M S-M G G G S-M S-M S-M G S-M S-M R R R R U R R R U U R U U U U U Population (age 5-17) Public Expenditure Groups sorted by average probability of attending school Source: Authors’ estimates from EPH 2009, Ministerio de Educación (2009). Table A7.3 Share of Public Expenditure on Health Care by Circumstance Group Share of public expenditure on health care by circumstance groups in 2009 30 Language: Head's education: G Guarani only 25 23.8 0-5 Less than 6th grade S-M Spanish or mixed 23.7 6 6th grade 20 Residence: Percentage 17.7 7-9 7th to 9th grade R Rural 10+ 10th grade or higher U Urban 16.9 15 10.6 9.4 9.1 10 7.1 6.8 6.8 7.4 5.1 7.1 3.7 5 5.6 3.6 3.0 6.0 2.1 5.3 1.2 1.0 3.7 1.2 0.2 3.4 0.6 2.5 2.1 0.2 1.8 1.4 0 0-5 6 0-5 0-5 7-9 6 7-9 6 10+ 0-5 10+ 7-9 6 10+ 7-9 10+ G G G S-M G S-M S-M G G S-M S-M G S-M G S-M S-M R R U R R R R U R U R U U U U U Population (age 0-17 ill only) Public Expenditure Groups sorted by average probability of accessing health services Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011). 239 Table A7.4 Share of Public Expenditure on Health Center Care by Circumstance Group Share of public expenditure on health center care by circumstance groups in 2009 50 44.7 Language: Head's education: G Guarani only 40 0-5 Less than 6th grade S-M Spanish or mixed 6 6th grade Residence: Percentage 7-9 7th to 9th grade 30 R Rural 10+ 10th grade or higher U Urban 23.7 20 17.7 10.6 9.4 10.5 10 6.8 7.1 5.1 5.3 7.1 7.4 3.7 3.0 2.1 1.0 3.7 1.2 1.2 7.5 1.3 0.5 0.4 0.4 1.3 2.1 1.1 4.6 3.4 1.0 2.7 2.6 0 0-5 6 0-5 0-5 7-9 6 7-9 6 10+ 0-5 10+ 7-9 6 10+ 7-9 10+ G G G S-M G S-M S-M G G S-M S-M G S-M G S-M S-M R R U R R R R U R U R U U U U U Population (age 0-17 ill only) Public Expenditure Groups sorted by average probability of accessing health services Source: Authors’ estimates from EPH 2009, MSPyBS (2010), IPS (2011). 240 Chapter 7. Boost Database for Paraguay, Massimo Mastruzzi, Eduardo Andrés Estrada, Renato Busquets, and Francisco Vazquez Ahued Introduction A core function of any government involves collecting and disbursing public funds and maintaining records of such expenditures. Budget execution reporting systems vary greatly in their organization, accuracy, and level of detail. For researchers, the ability to easily access and analyze public expenditure data is essential to provide thorough and timely advice to policymakers and relevant stakeholders. For government officials, rapid access to budget execution data in an easy-to-use format is essential for better decision making and increasing analytical capacity within government agencies. The purpose of the BOOST initiative (started at the World Bank in 2009) is to enhance the quality of public expenditure analysis by linking spending to results and improving access to fiscal data. Detailed government expenditure data have been collected and processed in over twenty countries across different regions, creating easy-to-use databases that have been employed by World Bank researchers in a variety of projects. Paraguay is the first country in Latin America and the Caribbean, and the fourth in the world (following Moldova, Kenya, and Togo), to release budget data to the public using BOOST. The BOOST database for Paraguay has been developed as part of the Paraguay Public Expenditure Review and in close collaboration with the team working on the report and the country management unit. It is available through the website of the Ministry of Finance of Paraguay115 and provides disaggregated budget data for all levels of government in a user-friendly format. This manual describes the database and provides guidance on its use. For help with issues that are not addressed here, please contact the BOOST team. This document is organized as follows:  Section 2 outlines the structure of Paraguay’s national budget  Section 3 presents the data sources.  Section 4 describes the particularities of the data and the organization of the database.  Section 5 explains how to use the database. The development of the database was a joint endeavor of the World Bank and the Ministry of Finance. The BOOST team would like to thank the staff at the Ministry of Finance for its excellent collaboration. We are also grateful for the support of our colleagues at the Poverty Reduction and Economic Management (PREM) Department in the Latin America and the Caribbean Region and the Argentina, Paraguay and Uruguay Country Management Unit of the World Bank. We hope that this tool is helpful in opening new avenues for analysis and providing answers to important questions regarding the efficiency, effectiveness, and equity of 115 http://www.openlooksolutions.com/boost_paraguay/. 241 government spending in Paraguay. The core BOOST team for Paraguay consists of Massimo Mastruzzi (mmastruzzi@worldbank.org), Eduardo Andrés Estrada (eestrada@worldbank.org), Renato Busquets (rbusquets@worldbank.org), and Francisco Vazquez Ahued (fvazquezahued@worldbank.org). 1. Structure of Paraguay’s National Budget Paraguay’s budget classifier explains in detail the structure of the national budget116. The budget classifier is approved annually as an annex to the budget law. According to the 2012 version, the following are the classifications of Paraguay’s national budget:  Classification of the Treasury and by entities ***  Classification by purposes and functions **  Program classification **  Classification of the expenditure by products **  Classification by source of funding ***  Classification of the revenue of the Treasury *  Classification by origin of revenue *  Economic classification of the revenue *  Classification by object of expenditure **  Economic classification of the expenditure **  Classification by origin of funding or funding agency ***  Classification by departments and municipalities ** Notes: * Denotes classification used for the revenue ** Denotes classification used for the expenditure *** Denotes classification used for the revenue and expenditure This designation was done according to the BOOST team’s interpretation of Paraguay’s budget classifier. This section will focus on the classifications used for the expenditure, starting with the classification of the Treasury and by entities. A description of the structure of the Treasury is not provided here, as that is related to the revenue side of the budget. To coincide with BOOST terminology, the classification by entities will be referred to as administrative classification. As shown in Figure 7. 1, state entities and agencies are divided in the following: (i) Central Administration; (ii) Decentralized Entities; (iii) Municipalities; and (iv) Mixed Enterprises. Figure 7. 1: Structure of the Administrative Classification 116 Unless otherwise noted, the information about Paraguay’s national budget that appears in this section comes from the budget classifier for 2012, annex to Law N° 4581/2011, available at http://www.hacienda.gov.py/web-presupuesto/index.php?c=192. 242 State Entities and Agencies Central Decentralized Mixed Municipalities Administration Entities Enterprises Figure 7. 2 provides the structure of Central Administration117 and Figure 7. 3 of Decentralized Entities. There are seventeen Departmental Governments, which appear under Decentralized Entities, and 238 Municipalities. The full list of state entities and agencies appears in Paraguay’s budget classifier. Figure 7. 2: Structure of Central Administration Central Administration Comptroller Legislative Executive Judicial Branch General of the Ombudsman Branch Branch Republic Currently there are nine State-Owned Enterprises in Paraguay. Administración Nacional de Electricidad (ANDE), Administración Nacional de Navegación y Puertos (ANNP), Dirección Nacional de Aeronáutica Civil (DINAC), Petróleos Paraguayos (PETROPAR), and Industria Nacional del Cemento (INC) are wholly owned by Paraguay. These State-Owned Enterprises appear under Decentralized Entities. In addition, Paraguay has a majority participation in Compañía Paraguaya de Comunicaciones S.A. (COPACO), Empresa de Servicios Sanitarios del Paraguay S.A. (ESSAP), Cañas Paraguayas S.A. (CAPASA), and Ferrocarriles del Paraguay S.A. (FEPASA). These appear under Mixed Enterprises. Figure 7. 3: Structure of Decentralized Entities 117 The Treasury appeared in the budget as a separate agency within central administration until 2006, and was then absorbed by the Ministry of Finance. 243 Decentralized Entities Autonomous Public Departmental Social Security State-Owned National Central Bank and Autarchic Financial Governments Public Entities Enterprises Universities Entities Institutions The classification by purposes and functions (in short, functional classification) determines the specific purposes according to the immediate objectives of government activity. The objective of this classification is to identify the final destination of the expenditure regardless of the economic nature of the agency or entity that is responsible for it. This classification shows the nature of the services (provision of services and production of goods) of the sectors that the state provides to the community. The functional classification has three levels of aggregation: finalidad (purpose), función (function) and subfunción (subfunction). Figure 7. 4 displays the hierarchy of the functional classification variables. Figure 7. 4: Hierarchy of the Variables of the Functional Classification Purpose Function Subfunction The program classification is used to order and provide information within the structure of the national budget of allocations by tipo de presupuesto (budget type):  Type 1: Programas de administración (Administration Programs)  Type 2: Programas de acción (Action Programs)  Type 3: Programas de inversión (Investment Programs)  Type 4: Programas del servicio de la deuda pública (Debt Service Programs) Programs must contain physical and financial measurement units to enable a subsequent evaluation of the process of execution of the national budget. The program classification is the 244 systematic grouping of programs, subprograms, and projects. The program classification is concurrent and related to the administrative and functional classifications. These classifications can be disaggregated by programs, subprograms, and projects; by groups, subgroups, and object of expenditure; generic or specific sources of funding, funding agencies; departments and combinations. According to the nature of the programs, they can form large groups of current, capital, and financing expenditures. Programs are assigned to Responsible Units, Administration and Finance Units, Administration and Finance Subunits, or Project Execution Units, which are responsible for managing the administrative, budgetary, and financial process. These units appear in the budget as unidad responsable (responsible unit). The structure of the program classification is given by the following components: Type 1: Budgets of Administration Programs Programs Responsible unit Product Group of object of expenditure Subgroup of object of expenditure Object of expenditure Source of funding Origin of funding or funding agency Departments Municipalities Type 2: Budgets of Action Programs Programs Subprograms Responsible unit Product Group of object of expenditure Subgroup of object of expenditure Object of expenditure Source of funding Origin of funding or funding agency Departments Municipalities Type 3: Budgets of Investment Programs Programs Subprograms 245 Projects Responsible unit Product Group of object of expenditure Subgroup of object of expenditure Object of expenditure Source of funding Origin of funding or funding agency Departments Municipalities Type 4: Budgets of Debt Service Programs Programs Responsible unit Product Group of object of expenditure Subgroup of object of expenditure Object of expenditure Source of funding Origin of funding or funding agency Departments Municipalities A quick review of the components of the program classification indicates that only the budgets of action and investment programs have subprograms. In addition, only the budgets of investment programs have projects. The program classification draws on the components from other budget classifications. The classification of the expenditure by products (in short, classification by product 118 ) identifies the production of goods and services of the programs, subprograms, or projects of the national budget. The classification by product allows to sort, identify, compare, quantify, assess, and provide other information within the structure of the national budget, in terms of programs, subprograms, or projects, but most importantly, it allows its orientation to results with the budget allocations of the various types of budget: Type 1 – Administration Programs, Type 2 – Action Programs, Type 3 – Investment Programs, and Type 4 – Debt Service Programs. 118 The classification by product is not available in the BOOST database for Paraguay. Paraguay introduced results-based budgeting in 2011. According to the Ministry of Finance, three ministries (Ministry of Health and Social Welfare, the Ministry of Public Works and Communications, and the Ministry of Education and Culture) are applying results-based budgeting in selected programs and subprograms. Its use will extend gradually to other government agencies. More information about the implementation of results-based budgeting in Paraguay is available at http://www.hacienda.gov.py/web- presupuesto/ppr/index.html. 246 The classification by object of expenditure determines the nature of the goods and services that the government acquires for its activities. This classification identifies the last three levels of public expenditure and is presented as a systematic and homogeneous arrangement of all the transactions contained in the budget of state agencies and entities such as: personnel services, non-personnel services, consumer goods and supplies, physical and financial investment, transfers and other expenses. The level of aggregation of the classification by object of expenditure allows the easy registration of all transactions with financial-economic impact made by state agencies and entities. This classification is an information tool to conduct analysis and monitoring of public financial management. As a result, it is considered the main or primary analytical classifier of the budget classification system. The classification by object of expenditure has three levels of aggregation: grupo (group), subgrupo (subgroup) and objeto del gasto (object of expenditure). Figure 7. 5 displays the hierarchy of the variables of the classification by object of expenditure. Figure 7. 5: Hierarchy of the Variables of the Classification by Object of Expenditure Group Subgroup Object of Expenditure The economic classification of the expenditure (in short, economic classification) identifies the economic nature of public sector transactions, with the purpose of evaluating the impact and repercussions of fiscal actions. The economic expenditure can occur with current or capital purposes or as financial applications. This classification has a strong relationship and integration with the different classifiers that allow the budgeting of the expenditure. The following are the most important ones:  Relationship economic classification / classification by object of expenditure: the object of expenditure, as the basic classifier, provides the items by object of expenditure that together make up the basic group structure of the economic classification.  Relationship economic classification / functional and program classification: in the program classification, the identification of the category “Project” (implies investment) allows current expenditures such as remunerations, consumer goods, non-personnel services, to be considered capital expenditures in the economic classifier. 247  Relationship economic classification / administrative classification: according to the criteria defined in the System of National Accounts, investments in the function Defense must be considered as current expenditures in the economic classifier. Expenditures are organized according to their economic nature and use on consumption, transfers, and investments, following the basic structure of the System of National Accounts. The economic classification has three levels of aggregation. Unlike other budget classifications, the budget classifier does not provide a name to each level. The methodology for the preparation of the economic classification of the expenditure is fairly complex, as it draws primarily on the classification by object of expenditure, but also the administrative, functional and program classifications119. The classification by origin of funding of funding agency identifies the primary funding of the different sources that finance the credits established in the national budget and the entities and organizations of the public and private sector that receive transfers from the state. The origin of funding makes reference to the specific sources that finance the budget credits from Resources from the Public Treasury (Source of Funding 10), Internal and External Public Credit (Source of Funding 20), and Institutional (Source of Funding 30). The primary origin of funding can be national or foreign. The beneficiary entity is related to the state entity or agency that receives the resources that have been financed with the different sources. The origin of funding is concurrent and related to the classification by source of funding. A specific origin of funding corresponds to each source of funding, as presented below: Source of Funding Origin of Funding 10 – Resources from the Treasury 001 – Genuine 20 – Resources from Public Credit 401 – Inter-American Development Bank (IADB) 30 – Institutional Resources 008 – Property Tax For simplicity purposes, and considering that the two classifications are related, they are referred to in this document as classification by source of funding. This classification has two levels of aggregation: fuente de financiamiento (source of funding) and origen de financiamiento (origin of funding). Figure 7. 6 displays the hierarchy of the variables of the classification by source of funding. Figure 7. 6: Hierarchy of the Variables of the Classification by Source of Funding 119 A description of the methodology appears in the budget classifier. 248 Source of Funding Origin of Funding The classification by departments and municipalities (in short, geographic classification) is used to identify the expenditure by geographic area or department, in order to determine the budgetary programming and records for the various regions and local governments in the country. In terms of departments, there are codes for the city of Asunción, the seventeen departments, and Alcance Nacional, that is, expenditure that occurs nationwide or in two or more departments. In terms of municipalities, a code identifies each of the 238 municipalities. 2. Data Sources Budget and expenditure data for Central Administration and Decentralized Entities come from the Sistema Integrado de Administración Financiera (SIAF), Paraguay’s Financial Management Information System (FMIS). The data were obtained with the assistance of the Vice Ministry of Economy of the Ministry of Finance. Data are available at the most disaggregated level for the period 2003-2012. It includes all the relevant budget classifications used for the expenditure. Budget and expenditure data for Municipalities are not included in the SIAF. Municipalities are legally mandated to send written reports on municipal finances for every budget cycle to the Ministry of Finance. However, the process of collecting and processing the municipal data is still not automatic. Another issue is the late submission of the information by some municipalities. Information on municipal finances was collected as part of the World Bank project on Improving the Quality of Public Expenditure at the Municipal Level in Paraguay. The Accounting Directorate and the Departments and Municipalities Unit of the Vice Ministry of Financial Administration of the Ministry of Finance provided assistance for this project. The BOOST database for Paraguay incorporates the budget and expenditure data for Municipalities collected through this project. Data are only available for the period 2006-2010. Unlike the data for Central Administration and Decentralized Entities, the information is not very detailed. The only variables available for the expenditure are the name of the municipality and the top two levels of the classification by object of expenditure: grupo (group) and subgrupo (subgroup). Since there is no functional, programmatic, and economic classification, the potential for analysis is limited. 249 In terms of coverage, the BOOST database for Paraguay does not contain information on Mixed Enterprises. As noted in Section 2, there are four Mixed Enterprises in Paraguay: COPACO, ESSAP, CAPASA, and FEPASA. These SOEs are incorporated as sociedades anónimas, corporations with limited liability. As such, their budget execution is not captured in the SIAF. Nonetheless, information about their budget and execution can be found in the Informe Financiero, the annual financial report of the Ministry of Finance120. 3. Particularities of the Data and Database Organization As mentioned in Section 3, the BOOST database for Paraguay contains budget and expenditure data at the most disaggregated level for Central Administration and Decentralized Entities for the period 2003-2012. For the most part, the variables in the database correspond to fields in the SIAF covering the various budget classifications in Paraguay’s national budget. Data for Municipalities are available for the period 2006-2010. The data for Central Administration and Decentralized Entities present the following particularities:  The administrative classification has four variables: Nivel (Level), Entidad (Entity), Entidad Nueva (Entity New), and Unidad Responsable (Responsible Unit). As will be described below, Entidad Nueva (Entity New) was created for the BOOST database.  Nivel (Level) identifies the level of government, following the structure of Central Administration (e.g., Legislative Branch, Executive Branch, and Judicial Branch) and Decentralized Entities (e.g., Departmental Governments, SOEs, and National Universities) presented in Section 2. Entidad (Entity) identifies the state entities and agencies under each level. Unidad Responsable (Responsible Unit) specifies the units responsible for individual programs.  A variable was created for the BOOST database to reflect changes to the administrative structure of Paraguay during the period 2003-2012. For instance, the Treasury appeared as a separate level and entity at the beginning of the period, but later it was absorbed by the Ministry of Finance. A variable was created, Entidad Nueva (Entity New), to reclassify the state entities and agencies into their 2012 equivalent. This is done only for Entidad (Entity).  The functional classification has three variables: Finalidad (Purpose), Función (Function), and Sub-función (Subfunction).  The variable Tipo presupuesto (Budget Type), which is part of the program classification, divides Paraguay’s budget into four types of programs: Administración (Administration), Acción (Action), Inversión (Investment), and Servicio de la deuda pública (Debt Service).  The program classification also has variables for Programa (Program), Sub-programa (Subprogram), Proyecto (Project), and Unidad Responsable (Responsible Unit). As noted in Section 2, only the budgets of action and investment programs have subprograms. In addition, only the budgets of investment programs have projects. 120 The Informe Financiero is available at http://www.hacienda.gov.py/web-contabilidad/index.php?c=306. 250  Programs are assigned to responsible units within state entities and agencies. For the purpose of the BOOST database, Unidad Responsable (Responsible Unit) is included as part of the administrative classification, not the program classification. Nonetheless, as explained in Section 2, the program classification draws on the components from other budget classifications.  The economic classification has three variables: Econ1, Econ2, and Econ3. For the purpose of the BOOST database, the variables of the classification by object of expenditure can also be classified as economic, as they encompass all transactions by state entities and agencies with financial-economic impact. These include Grupo (Group), Sub- grupo (Subgroup) and Objeto Gasto (Object of Expenditure).  Four transfer variables were created for the BOOST database: o A variable to identify intra-governmental transfers (Transferencia Consolidable). These are consolidated transfers between state entities and agencies to finance current and capital expenditure. These include funds from revenue sharing agreements and other earmarked special resources. o A variable to identify the type of transfer (Tipo de Transferencia Consolidable). o Variables that identify the parent entity, that is, the entity that submits the transfer (Entidad Madre) and the corresponding entity that receives it (Entidad Hija). The inclusion of the transfer variables is important, since some official reports show figures including transfers, while others exclude them.  The classification by source of funding has two variables: Fuente de financiamiento (Source of Funding) and Origen de Financiamiento (Origin of Funding).  There is a geographic variable, Departamento (Department), which identifies the department or geographic area where a specific budget item was executed. For expenditure that occurs nationwide or in two or more departments it is classified as Alcance Nacional.  There are four variables for the budget cycle: Presupuesto inicial (Initial Budget), Presupuesto vigente (Modified Budget), Obligado (Committed), and Pagado (Paid). Figures are available in guaraníes.  Presupuesto inicial (Initial Budget) contains the nominal amount that was approved by Congress in each year’s budget for a specific line item. Presupuesto vigente (Modified Budget) is the approved budget plus amends. Obligado (Committed) lists the amount that was incurred, while Pagado (Paid) contains actual payments associated with that line item. In practice, Obligado (Committted) is considered the executed amount. The data for Municipalities present the following particularities:  As noted earlier, the only variables available are the name of the municipality and the top two levels of the classification by object of expenditure: Grupo (Group) and Sub-grupo (Subgroup).  A variable for Nivel (Level) was added to the database to reflect that the data correspond to Municipalities. The name of the municipality was included under Entidad (Entity). 251  Unlike the data for Central Administration and Decentralized Entities, there are only three variables for the budget cycle: Presupuesto vigente (Modified Budget), Obligado (Committed), Pagado (Paid). Figures are available in millions of guaraníes. A complete list of variables for Central Administration and Decentralized Entities can be found in Table 7. 1. Each variable is labeled with the original name used by the Government of Paraguay. The English translation is provided in parenthesis. In some cases, a brief explanation of the purpose of the variable is included. The BOOST equivalent is provided for each variable. Standard BOOST labels identify the type of budget classification of the variable and the level that it belongs to (e.g., ECON1 corresponds to the top-level of the country’s economic classification). Table 7. 1: List of Variables for Central Administration and Decentralized Entities BOOST Variable Original Name Administrative Classification ADMIN1 Nivel (Level) ADMIN2 Entidad (Entity) ADMIN2_NEW** Entidad Nueva (Entity New): Entities described in their 2012 equivalent ADMIN3*** Unidad Responsable (Responsible Unit) Program Classification BUDGET_TYPE Tipo presupuesto (Budget Type) PROGRAM1 Programa (Program) PROGRAM2 Sub-programa (Subprogram) PROJECT1 Projecto (Project) Transfer Type Variables* TRANSFER** Transferencia Consolidable: 1 if transfer Tipo de Transferencia Consolidable: Variable that identifies the type of TRANSFER_TYPE** transfer TRANSFER_ORIGIN** Entidad Madre: Variable that identifies the origin of a transfer TRANSFER_DESTINATION** Entidad Hija: Variable that identifies the destination of a transfer Functional Classification FUNCTION1 Finalidad (Purpose) FUNCTION2 Función (Function) FUNCTION3 Sub-función (Subfunction) Economic Classification**** ECON1 Econ1: Top-level economic classification ECON2 Econ2: Mid-level economic classification ECON3 Econ3: Bottom-level economic classification ECON4 Grupo (Group): Top-level classification by object of expenditure ECON5 Sub-grupo (Subgroup): Mid-level classification by object of expenditure Objeto de gasto (Object of Expenditure): Bottom-level classification by ECON6 object of expenditure Classification by source of funding FIN_SOURCE1 Fuente de financiamiento (Source of Funding) FIN_SOURCE2 Origen de financiamiento (Origin of Funding) 252 BOOST Variable Original Name Geographic Classification GEO1 Departamento (Department) YEAR Año (Year) Budget cycle variables APPROVED Presupuesto inicial (Initial Budget) MODIFIED Presupuesto vigente (Modified Budget) COMMITTED Obligado (Committed) PAID Pagado (Paid) Notes: * Inter-governmental transfers only. ** Variable created by the BOOST Team. *** The budget classifier for Paraguay lists this variable as part of the program classification, although it is included here as part of the administrative classification. **** The budget classifier for Paraguay lists the economic classification and the classification by object of expenditure separately. For the purpose of the BOOST database, the latter is part of the economic classification. Table 7. 2 provides the list of variables for Municipalities, following the guidelines presented above. Table 7. 2: List of Variables for Municipalities BOOST Variable Original Name Administrative classification ADMIN1* Nivel (Level) ADMIN2 Entidad (Entity) Economic classification** ECON4 Grupo (Group): Top-level classification by object of expenditure ECON5 Sub-grupo (Subgroup): Mid-level classification by object of expenditure Other variables YEAR Año (Year) Budget cycle variables MODIFIED Presupuesto vigente (Modified Budget) COMMITTED Obligado (Committed) PAID Pagado (Paid) Notes: * Variable created by the BOOST team. ** The budget classifier for Paraguay lists the economic classification and the classification by object of expenditure separately. For the purpose of the BOOST database, the latter is part of the economic classification. 4. How to Use the BOOST Database for Paraguay As referenced in Section 1, the BOOST database for Paraguay is available through the website of the Ministry of Finance. The database is provided with the functionality of an interactive pivot table. Pivot tables are a powerful, albeit easy to use, tool for data analysis. They allow the end- user to create custom reports, which can then be used to compare categories and identify patterns 253 and trends in the data121. This section provides an overview of the interface and presents a few examples of reports that can be generated with it. As with many things in life, the best way to learn how to use the BOOST database is by using it. As shown in 121 For more information about pivot table reports, consult this helpful overview from Microsoft: http://office.microsoft.com/en-us/excel-help/overview-of-pivottable-and-pivotchart-reports-HP010177384.aspx. This applies to PivotTable reports created with Microsoft Excel, but the explanation of the uses of pivot tables is quite informative. 254 Figure 7. 7, the interface is structured around two tabs, “Central/Descentralizadas” for Central Administration and Decentralized Entities, and “Municipalidades” for Municipalities. The data are presented in spreadsheet format with values for the variables selected by the end-user. There are several buttons to control the functionality of the interactive pivot table. An Excel file with the description of the variables can be downloaded by clicking on “Descripción de variables” under the spreadsheet. 255 Figure 7. 7: Interface of the BOOST Database for Paraguay Example 1: Create a report for Central Administration and Decentralized Entities: Click on “Central/Descentralizadas” and then click on “Seleccionar los campos de datos.” A menu will open (see Figure 7. 8). Review the available variables under “Variables disponibles.” Select one of the variables. Drag and drop it into the box for “Variables de columna.” This will set the columns 256 (horizontal axis) of the spreadsheet. The interface only allows one variable in the box, thus it is advisable to select the variable Año (Year). Figure 7. 8: Variable Settings for Central Administration and Decentralized Entities To set the rows (vertical axis) of the spreadsheet, select one of the variables under “Variables disponibles.” Drag and drop it into the box for “Variables de fila.” Multiple variables are 257 allowed in this box, although they have to be different from the variable selected for “Variables de columna.” Select one of the variables under “Variables del ciclo presupuestario.” These are the budget cycle variables. Drag and drop it into the box for “Valores seleccionados.” Up to four variables are allowed in this box. Click on “Generar” to generate the report, “Reiniciar” to return the settings to the default, or “Cancelar” to return to the main screen. Additional options to filter the data appear under the “Filtros” tab. At the moment it is possible to filter the data by Año (Year), Nivel (Level), Entidad (Entity), Grupo (Group), and Sub-grupo (Subgroup). Additional filters will be added in a future update of the BOOST database. Example 2: Create a report for Municipalities Click on “Municipalidades” and then click on “Seleccionar los campos de datos.” A menu will open (see Figure 7. 9). Review the available variables under “Variables disponibles.” Select one of the variables. Drag and drop it into the box for “Variables de columna.” This will set the columns (horizontal axis) of the spreadsheet. The interface only allows one variable in the box, thus it is advisable to select the variable Año (Year). Figure 7. 9: Variable Settings for Central Administration and Decentralized Entities To set the rows (vertical axis) of the spreadsheet, select one of the variables under “Variables disponibles.” Drag and drop it into the box for “Variables de fila.” Multiple variables are 258 allowed in this box, although they have to be different from the variable selected for “Variables de columna.” Select one of the variables under “Variables del ciclo presupuestario.” These are the budget cycle variables. Drag and drop it into the box for “Valores seleccionados.” Up to three variables are allowed in this box. Click on “Generar” to generate the report, “Reiniciar” to return the settings to the default, or “Cancelar” to return to the main screen. Additional options to filter the data appear under the “Filtros” tab. It is possible to filter the data by Año (Year), Nivel (Level), Entidad (Entity), Grupo (Group), and Sub-grupo (Subgroup). The BOOST database is extremely useful for time series analysis, as the following example will indicate. Example 3: Create a report of initial budget by entities in the Executive Branch (2003- 2012) Click on “Central/Descentralizadas” and then click on “Seleccionar los campos de datos.” A menu will open. Leave Año (Year) in the box for “Variables de columna;” Entidad (Entity) in the box for “Variables de fila;” and Presupuesto inicial (Initial Budget) in the box for “Valores seleccionados.” Select Nivel (Level) under “Variables disponibles.” Drag and drop it into the box for “Variables de fila” on top of Entidad (Entity). Click on the “Filtros” tab. Under Nivel (Level) select “12 - PODER EJECUTIVO,” and click on the right arrow sign (see Figure 7. 10). Click on “Generar” to generate the report. Figure 7. 10: Example of Filter for Executive Branch by Nivel (Level) The time series for the Executive Branch will appear in the main screen. Click on “ Expandir todo” to display the time series for the individual entities (see 259 Figure 7. 11). The report can be downloaded to the end-user’s computer by clicking on “Descargar archivo CSV.” The CSV file can be opened for further manipulation in Microsoft Excel or any other spreadsheet software. Figure 7. 11: Example of Time Series Analysis: Initial Budget by Entities in the Executive Branch (2003-2012) As indicated in Section 3, not all municipalities report budget execution data to the Ministry of Finance every year. The expenditure will appear as 0 for the years in which a municipality did not report the expenditure. That does not mean that the municipality did not execute its budget. It simply means that the execution was not reported. Example 4: Create a budget execution report for the Municipality of Asunción (2006-2010) Click on “Municipalidades” and then click on “Seleccionar los campos de datos.” A menu will open. Leave Año (Year) in the box for “Variables de columna” and Entidad (Entity) in the box 260 for “Variables de fila.” Select Grupo (Group) under “Variables disponibles.” Drag and drop it into the box for “Variables de fila” below Entidad (Entity). Remove Pagado (Paid) from the box for “Valores seleccionados” and replace it with Obligado (Committed). Click on the “Filtros” tab. Under Entidad (Level) select “30.001 - ASUNCION,” and click on the right arrow. Click on “Generar” to generate the report. Click on “Expandir todo” to display the budget execution report. As Figure 7. 12 indicates, the execution for 2008 was 0. That means that Asunción did not report its execution to the Ministry of Finance that year. Figure 7. 12: Example of Execution Report for the Municipality of Asunción (2006-2010) Example 5: Create a budget execution report for the Judicial Branch combining various budget classifications (2009-2012) In this example we are interested in reviewing the top-level economic classification (Econ1) for the entities in the Judicial Branch for the period 2009-2012. Click on “Central/Descentralizadas” and then click on “Seleccionar los campos de datos.” A menu will open. Leave Año (Year) in the box for “Variables de columna” and Entidad (Entity) in the box for “Variables de fila.” Select Nivel (Level) under “Variables disponibles.” Drag and drop it into the box for “Variables de fila” above Entidad (Entity). Select Econ1 under “Variables disponibles.” Drag and drop it into the box for “Variables de fila” below Entidad (Entity). Remove Pagado (Paid) from the box for “Valores seleccionados” and replace it with Obligado (Committed). Click on the “Filtros” tab. Under Año (Year) select each year in the period and click on the right arrow sign. Under Nivel 261 (Level) select “13 - PODER JUDICIAL,” and click on the right arrow sign. Click on “Generar” to generate the report. Click on “Expandir todo” to display the budget execution report (see Figure 7. 13). Figure 7. 13: Example of Execution Report for the Judicial Branch (2009-2012) 262 263