MACROECONOMICS, TRADE AND INVESTMENT MACROECONOMICS, TRADE AND INVESTMENT EQUITABLE GROWTH, FINANCE & INSTITUTIONS INSIGHT A Review of Fiscal Policy Responses to COVID-19 Eric Lacey Joseph Massad Robert Utz ABSTRACT Countries around the world adopted a wide range of fiscal measures in 2020 to mitigate the health and economic impacts of the COVID-19 pandemic. At the beginning of 2021, the outlook on the evolution of the pandemic remains uncertain. COVID-19 cases are at high levels in many countries, but effective vaccines have been approved and are being rolled out. The occurrence of new variants of the virus that spread more easily and more quickly and that may be associated with an increased risk of death adds to the uncertainty as to how quickly the pandemic can be brought under control. © 2021 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved. This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO), http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: World Bank. 2021. “A Review of Fiscal Policy Responses to COVID-19”. MTI Insight. World Bank, Washington, DC. License: Creative Commons Attribution CC BY 3.0 IGO. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third- party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@ worldbank.org. 4 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >>> Contents 1. Introduction 7 2. Overview of Fiscal Policy Interventions 8 3. Characteristics of Fiscal Interventions 17 4. A Look Forward 26 5. Summary 28 6. Appendixes 29 Appendix A: Methodology 29 Appendix B: Country-level Policy Dimension Scores 37 Appendix C: Country Group Scores 45 Appendix D: Pairwise Correlations of Policy 48 Dimensions Appendix E: Logistic Regression Results 49 Appendix F: Assessment of Options for Fiscal Policy 53 Measures >>> Acknowledgements This note was prepared by Robert Utz, Eric Lacey, and Joseph Massad under the guidance and supervision of Chiara Bronchi. It builds on an earlier note on “The fiscal impact and policy response to COVID-19.” Work on the assessment of policy action was initiated by Marijn Ver- hoeven and the fiscal policy action data base was updated by Olawunmi Ifetola Ola Busari. Massimo Mastruzzi, Sebastian James, and Jaffar Al Rikabi provided helpful discussion and feedback during the preparation of the note. The note also benefitted from feedback of staff from the Fiscal Policy and Sustainable Growth Unit at a presentation of the draft note. Brittany Walters and Mark Volk edited the report. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 5 7. MTI INSIGHT >>> A Review of Fiscal Policy Responses to COVID-19 >> INTRODUCTION Countries around the world adopted a wide range of fiscal measures in 2020 to mitigate the health and economic impacts of the COVID-19 pandemic. At the beginning of 2021, the outlook on the evolution of the pandemic remains uncertain. COVID-19 cases are at high levels in many countries, but effective vaccines have been approved and are being rolled out. The occurrence of new variants of the virus that spread more easily and more quickly and that may be associated with an increased risk of death1 adds to the uncertainty as to how quickly the pandemic can be brought under control. Policymakers face difficult choices in an environment where fiscal space may be narrowing and additional spending requirements emerge, including for the purchase and distribution of vaccines as well as for measures to support an economic recovery. Which of the policy measures already in place need to be maintained and extended and which ones can be phased out? Which new measures need to be added? Where are opportunities for greater efficiency and better targeting? What needs to be done to recover funds from measures such as tax deferrals and loans? This note and the associated COVID-19 fiscal policy measures dashboard are intended to help inform answers to these questions by providing analysis of the countries’ portfolios of fiscal policy measures adopted in 2020 and their key characteristics such as reversibility of policy measures, possibility of cost recovery, or targetability. This note allows policymakers to better understand which policies and types of policies follow international best practice and which measures should be carefully monitored in the implementation or closure phases. In all, about 4000 policy actions adopted by 203 economies have been grouped into seven categories and 47 sub-categories. About half of these policy actions have been judged as to whether they met the criteria across nine dimensions, including targetability, speed, abuse resistance, affordability, predictability and cost control, scalability, reversibility, administrative complexity, and feasibility considering social distancing and contagion risks. The dashboard offers a flexible tool to analyze country portfolios of fiscal policy measures by categories, sub-categories, and characteristics of measures and it allows comparison with other country or country groupings. 1. CDC. 2021. New Variants of the Virus that Causes COVID-19. Updated Feb. 2, 2021. https://www.cdc.gov/coronavi- rus/2019-ncov/transmission/variant.html. Accessed on Feb. 14, 2021. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 7 >>> ORGANIZATION OF THE PAPER The paper consists of three parts. The first part provides an overview of the fiscal response to the COVID-19 pandemic from its onset until September 2020. We look at the different types of interventions that countries have adopted to mitigate the health impacts of the pandemic and to support households and businesses. The second part of the paper reviews the characteristics of fiscal policy interventions, using the policy assessment framework from Fiscal Impact and Policy Response to COVID-19 (World Bank 2020) as a guide. The third part of the paper concludes by drawing lessons for continued efforts to mitigate the health and economic impacts of the pandemic and the implications for transitioning from the current set of crisis response measures to support for economic recovery and fiscal stabilization. Six annexes describe the methodology used to assess fiscal policy interventions and offer a deep look into topics that are referenced throughout the paper, including a post-scoring diagnostic of the scoring methodology, a list of country-level scores on the nine policy dimensions of the policy assessment framework, ANOVA statistical analysis of variance of scores across country groups, logistic regression analysis assessing the marginal impact of policies and country characteristics on policy dimension scores, pairwise correlation tables of policy scores, and the full table of Assessment Options for Fiscal Policy Measures from the fiscal policy paper. >> OVERVIEW OF FISCAL POLICY INTERVENTIONS >>> A FAST AND SIZABLE RESPONSE and revenue measures and 2.5 percent of GDP in support through other instruments. Low income developing countries On March 11, 2020, the World Health Organization declared (LIDCs) had the smallest fiscal response at about 1.8 percent COVID-19 a pandemic. Most countries followed the advice of GDP which was almost entirely in the form of expenditure of health experts and quickly introduced strict containment and revenue measures (1.6 percent of GDP). measures to mitigate the spread and adverse health impact of the virus. The Oxford Stringency Index captures the > > > combined restrictiveness of measures such as business F I G U R E 1 - Global Average Oxford Stringency Index closures and travel restrictions for each country for each day by Day of calendar year 2020. Between March 9, 2020 and April 1, 2020, the world became serious about controlling the spread of the virus (Figure 1). Although these containment measures were necessary from a health perspective, they contributed to a sudden reduction of income for many businesses and individuals, led to a sharp increase in unemployment worldwide, and raised serious concerns not just about the short-term liquidity of businesses, but their long-term solvency. The need for measures to protect livelihoods and secure business assets was clear. Most governments quickly introduced fiscal policy measures to provide financial support to businesses and households, and Source: University of Oxford2, December 2020. to improve the capacity of the health sector to respond to the pandemic. The magnitude of the fiscal response and the type of instruments differs significantly by country groupings (Figure 2). Advanced economies’ (AEs) crisis response amounted to more than 9 percent of GDP in expenditure and revenue measures and another 11 percent in support through equity and loans, guarantees, and quasi-fiscal activities. Emerging markets and middle-income economies’ (EMMIEs) response is more muted, but they also deployed all instruments for the fiscal response, with about 3.4 percent of GDP in expenditure 2. Hale, Thomas, Noam Angrist, Emily Cameron-Blake, Laura Hallas, Beatriz Kira, Saptarshi Majumdar, Anna Petherick, Toby Phillips, Helen Tatlow, Samuel Webster (2020). Oxford COVID-19 Government Response Tracker, Blavatnik School of Government. 8 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > for nearly all of the fiscal response to the crisis. For example, F I G U R E 2 - Breakdown of Fiscal Support, by Type Guatemala’s cash transfers alone account for a third of its (As of September 11, 2020, percent of GDP) fiscal package at 1.2 percent of GDP, while Eswatini’s relief fund for laid-off workers costs 0.04 percent of GDP, or about 1.5 percent of the total cost of its fiscal policy response at 2.8 percent of GDP.5 Larger and higher-income economies tended to adopt more policies (Figure 4, panels B and C). This may reflect, inter alia, greater resource availability and more sophisticated economies that require complex responses. There appears to be a weak yet significant relationship between the cumulative number of cases per 100,000 and the cumulative number of fiscal policies implemented (Figure 4, panel D), indicating that the number of fiscal policy interventions tended to be slightly higher in countries with more infections. However, the relationship between these two indicators is complex. High numbers of cases could prompt a stringent Source: IMF Fiscal Monitor (October 2020). lockdown, requiring a robust fiscal policy response to support households and businesses. Alternatively, low numbers of The number of policies3 introduced by each country varies cases could reveal that a strict response is working, which widely, with most countries introducing anywhere from eight also requires a robust fiscal policy response. Furthermore, to twenty measures as of September 1, 2020. Many recently economies can be quite different, and the way that the health introduced measures have been revisions or additions to crisis exacerbates the economic crisis can vary by country, as earlier policies.4 Most countries adopted between one to well as the response deemed appropriate by authorities. three interventions in support of their health sector, two to five interventions targeted at supporting households, and two to The data show no relationship between the gross five interventions to support businesses. government debt stock and the number of fiscal policies implemented (Figure 4, panel E). However, countries with The total cost of the fiscal package has only a weak greater government effectiveness tended to adopt a higher relationship with the number of policies that make up the fiscal number of fiscal measures (Figure 4, panel F). Seeing some package (Figure 4, panel A). Although, on average, having alignment between the number of policies adopted and more policies is associated with a higher overall cost, the countries capacity is important, especially since a plethora of variation in policies only explains 22 percent of the variation policies could also reflect divided governments and competing in cost. In some cases, a stand-alone measure can account bureaucratic forces. 3. The data on policy responses to COVID-19 used in this report were taken from the Fiscal Policy and Sustainable Growth Unit’s (FPU) Fiscal Policy Matrix, last updated on September 1, 2020. The data cover 203 economies and comprise nearly 4,000 fiscal policy measures launched or announced in response to the COVID-19 pan- demic. The policies were gathered in two waves: from the onset of the pandemic through May 1, 2020 (amounting to 2,400 policies), and then from May 1, 2020 until September 1, 2020 (totaling 1,591 policies). The main sources referenced to compile our dataset were the IMF, OECD, IBDF, and the Doing Business policy tracker databases. Other data featured in this report are from the World Development Indicators, the ICTD Government Revenue Dataset, the World Health Organization, the University of Oxford, and the October 2020 IMF Fiscal Monitor. 4. About 1,500 of the nearly 4,000 fiscal measures accounted for in our database were of the exact same type as measures previously implemented by the same country. Some of these measures were not exactly repeats; rather, they were benefits which had been previously offered to one demographic extended to new demographics. Many were revisions to early policies, extending benefits beyond the originally planned end date, or scaling up spending on health measures as need evolved further. 5. IMF policy tracker and IMF Fiscal Monitor (October, 2020). A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 9 > > > F I G U R E 3 - Number of Fiscal Policy Actions per Country Source: World Bank Fiscal Policy data base, January – September 2020. 10 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > F I G U R E 4 - Scatters of the Number of Fiscal Measures Against Selected Indicators Source6: WB Fiscal Measures Database (2020), IMF Fiscal Monitor (October 2020), WDI (2020), WHO (2020), WB Fiscal Space Database (2020), WGI (2020). 6. Data show cumulative cases as of September 1, population in 2019, cumulative cases per 100,000 as of September 1, GDP per capita in 2019 (constant 2010 USD), and government debt in 2019. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 11 >>> A CORE SET OF FISCAL POLICY RESPONSES WAS ADOPTED BY MOST COUNTRIES The objectives of fiscal policy actions adopted by countries fall into three broad groups: (a) to deal with the health pandemic, (b) to support households and (c) to bolster businesses. Governments pursue these objectives by using the whole range of fiscal instruments, including tax and expenditure measures, credits, and guarantees, according to which we group policy actions into seven broad categories and 47 sub-categories. > > > T A B L E 1 - Classification of Fiscal Measures7 FISCAL MEASURES TO SUPPORT BUSI- FISCAL MEASURES TO SUPPORT FISCAL MEASURES TO SUPPORT THE NESSES HOUSEHOLDS HEALTH SECTOR Revenue measures to protect Revenue measures to protect Revenue measures to promote businesses individuals availability of medical items Accelerated asset depreciation (CIT) Deferral of tax filing (PIT, payroll Lower tax rates for medical items Extend loss carry-forward for losses taxes, property tax, etc.) (import duties, VAT and other incurred during the crisis (CIT) Deferral of tax payments (PIT, indirect taxes) Broaden tax deductibility (e.g., to payroll taxes, property tax, etc.) and/ Tax exemptions/waiver/suspension all business expenses related to or interest and penalty payments Other COVID-19) Tax rate reduction (PIT, payroll taxes, Introduce tax credits property tax, etc.) Deferral of tax filing (CIT, PIT for self- Tax amnesty (including for overdue employed, VAT, other business taxes) taxes and penalties) REVENUE Deferral of tax and/or interest and Broaden tax deductibility (e.g., for penalty payments contributions to health care) (PIT) Tax rate reduction (CIT, PIT for self- Introduce tax credits employed) Tax exemptions/waiver/suspension Tax amnesty Other Accelerating refunds (VAT) Lower advance payment (CIT, PIT for Revenue measures to boost self-employed) consumption / demand Suspend debt collection activities Lower tax rates (import duties, VAT Suspend audit activities and other indirect taxes and levies) Tax exemptions/waver/suspension Tax exemptions/waiver/suspension Other Other Direct support to businesses Expenditure measures for Health expenditure measures One-off grants to industries in individuals Supply of low-cost medical items distress Direct cash transfers for individuals (masks, gloves, testing kits, gowns, Income support Expansion of unemployment benefits face shields, etc.) both in terms of compensation and Supply of high-cost medical items length (ventilators, etc.) EXPENDI- Temporary expansion of existing Targeted infrastructure investments benefits such as pensions and health to expand health care capacity TURE insurance Expansion of human resources Supplementary ad hoc programs General (feeding programs, utility waivers) Wage compensation subsidies and enhanced paid/sick leave allowances Other Preferential loans to firms (and Preferential loans to households CREDIT/ industries) in distress Other (includes support measures for EQUITY households and businesses) Revenue increase Other revenue increase measures OTHER Other measures not falling into any of the categories nobit voluptae Source: World Bank 7. Note the three objectives (each column), the seven categories of measure (color coding), and the 47 distinct sub-categories. 12 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 Figure 5 shows the number of categories addressed by each country’s fiscal policy response. Some high- and upper-middle income countries stand out for their multi-dimensional responses, engaging multiple of the following categories: revenue measures for businesses, revenue measures for the health system, revenue measures for households, expenditure measures for businesses, expenditure measures for the health system, expenditure measures for households, and credit and equity measures. Europe, North and South America, and Oceanic Southeast Asia demonstrate a tendency to approach the fiscal response from multiple angles. Sub-Saharan Africa and Central America favored generally narrow approaches, with the notable exceptions of Niger and South Africa. Readers will notice a pronounced but imperfect relationship between income and the multi-dimensionality of the fiscal policy response. > > > F I G U R E 5 - Multi-dimensionality8 of the Fiscal Policy Reponse as of September 1, 2020 Source: World Bank Fiscal Policy data base, January – September 2020. 8. The seven categories correspond to the seven identified in Table 2. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 13 > > > T A B L E 2 - The Five Most Frequent Fiscal Inverventions by Country Groupings and Regions (% of Countries Implementing Measure in Category) Source: Original calculations for this publication. Of the 47 different types of fiscal policy measures identified in this study, most countries and country groups implemented 8-12 types of measures, and often a similar mix. The most common measures were typically found in all country groups and include general health spending, loans and deferral of tax payments for businesses, and direct cash transfers and supplementary ad-hoc programs for households (Table 2). Readers may also wish to consider Table 3, which shows detailed information on the share of countries implementing each type of policy by country group. Some country groups gravitate strongly toward particular policies more than other country groups do. For instance, low- and lower-middle income countries, FCVs, countries collecting tax revenue below 15 percent of GDP, and Sub- Saharan Africa favored health expenditure measures far more than other measures. This was likely an effort to prioritize capacity building in the health sector. Although necessary given the nature of the crisis, such measures have significant implications for fiscal space. There is also considerable variation across individual countries. The reader is invited to look up details on policy implementation for any country or country group using the “Assessment of Fiscal Policy Responses to COVID-19 Dashboard.” 14 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > F I G U R E 3 - Fiscal Measures, Percentage9 of Countries Implementing as of September 1st MEASURE ALL LIC/ EAP ECA LAC MENA SAR SSA FCV <15% T/ LMIC GDP REVENUE MEASURES TO PROTECT BUSINESSES Accelerated asset depreciation (CIT) 5% 0% 13% 5% 7% 0% 0% 0% 0% 7% Extend loss carry-forward (CIT) 9% 7% 10% 14% 7% 0% 0% 6% 3% 7% Broaden tax deductibility 17% 7% 35% 24% 7% 0% 13% 9% 0% 11% Tax credits 9% 7% 6% 14% 10% 0% 13% 6% 7% 4% Deferral of Tax Filing 33% 29% 19% 33% 45% 39% 50% 23% 23% 39% Deferral of Tax Payments 73% 65% 52% 89% 71% 72% 75% 64% 53% 75% Tax Rate Reduction 27% 18% 42% 37% 19% 22% 13% 19% 17% 21% Tax Amnesty 16% 17% 16% 19% 12% 17% 0% 19% 17% 21% Accelerating Refunds 24% 21% 26% 25% 14% 22% 13% 30% 10% 27% Lower Advance Payment 15% 8% 13% 28% 14% 0% 13% 6% 7% 9% Suspend Debt Collection 14% 15% 0% 16% 19% 17% 13% 13% 0% 16% Suspend Audit Activities 16% 24% 6% 18% 10% 11% 13% 23% 20% 13% Tax Exemption/Waiver/Suspension 44% 38% 55% 54% 40% 44% 25% 30% 17% 41% Other 39% 35% 32% 49% 29% 56% 38% 34% 30% 36% REVENUE MEASURES TO PROTECT INDIVIDUALS Deferral of Tax Filing 13% 10% 10% 14% 24% 0% 25% 4% 7% 11% Deferral of Tax Payments 27% 21% 29% 40% 26% 33% 38% 4% 10% 21% Tax Rate Reduction 6% 4% 6% 12% 2% 0% 0% 4% 0% 5% Tax Amnesty 3% 1% 0% 5% 7% 0% 0% 2% 0% 4% Broaden Tax Deductability 5% 4% 10% 11% 0% 0% 0% 2% 0% 4% Tax Credits 2% 1% 0% 4% 2% 6% 0% 2% 0% 0% Tax Exemption/Waiver/Suspension 23% 26% 26% 26% 21% 11% 0% 26% 10% 25% Other 16% 10% 13% 23% 21% 11% 0% 9% 7% 11% REVENUE MEASURES TO PROMOTE AVAILABILITY OF MEDICAL ITEMS Lower Tax Rates for Medical Items 14% 8% 16% 21% 14% 11% 13% 4% 3% 11% Tax Exemption/Waiver/Suspension 41% 46% 35% 51% 38% 11% 38% 47% 23% 43% Other 18% 25% 13% 18% 10% 22% 0% 30% 17% 18% REVENUE MEASURES TO BOOST CONSUMPTION/DEMAND Lower Tax Rates 20% 17% 6% 33% 21% 17% 13% 15% 13% 14% Tax Exemption/Waiver/Suspension 15% 17% 16% 16% 14% 11% 0% 17% 17% 14% Other 11% 6% 13% 16% 10% 6% 13% 6% 7% 5% 9. Percentages show the share of countries in the country group (column head) implementing the measure indicated (row head). A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 15 MEASURE ALL LIC/ EAP ECA LAC MENA SAR SSA FCV <15% T/ LMIC GDP HEALTH EXPENDITURE MEASURES Supply of Low Cost Medical Items 3% 6% 3% 2% 0% 6% 0% 6% 13% 4% Supply of High Cost Medical Items 6% 11% 0% 2% 12% 6% 0% 13% 13% 9% Targeted Infrastructure Investments 11% 13% 10% 12% 12% 0% 13% 13% 17% 13% Expansion of Human Resources 9% 14% 6% 9% 5% 0% 0% 19% 7% 7% General 76% 96% 77% 74% 57% 67% 100% 94% 97% 91% EXPENDITURE MEASURES FOR INDIVIDUALS Direct Cash Transfers for Individuals 63% 67% 61% 68% 62% 56% 100% 53% 53% 71% Expansion of Unemployment Benefits 31% 17% 29% 47% 33% 44% 25% 4% 10% 20% Temporary Expansion of Existing 33% 38% 39% 37% 26% 33% 38% 28% 10% 34% Benefits Supplementary Ad Hoc Programs 71% 71% 71% 67% 69% 67% 100% 70% 63% 73% Wage Compensation/Enhanced Paid 49% 29% 42% 77% 36% 56% 50% 23% 20% 32% Leave Other 15% 14% 19% 23% 12% 11% 0% 9% 7% 13% CREDIT AND EQUITY MEASURES Preferential Loans to Firms 64% 54% 55% 81% 52% 56% 100% 55% 27% 63% Preferential Loans to Households 13% 13% 13% 11% 19% 11% 13% 6% 0% 11% Other 25% 17% 26% 35% 19% 28% 13% 15% 10% 20% REVENUE MEASURES TO RAISE REVENUE Revenue Increase 6% 6% 0% 7% 7% 11% 13% 6% 7% 11% Other 2% 3% 0% 2% 2% 0% 0% 4% 3% 2% EXPENDITURE MEASURES FOR BUSINESSES Income Support 41% 32% 39% 54% 43% 28% 13% 34% 23% 25% One-Off Grants 17% 6% 13% 33% 14% 6% 0% 4% 7% 4% OTHER Other 37% 35% 35% 51% 24% 22% 63% 32% 27% 38% Source: Original calculations for this publication. 16 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >> CHARACTERISTICS OF FISCAL INTERVENTIONS >>> THE FISCAL POLICY ASSESSMENT Some of the dimensions are in tension with one another. FRAMEWORK For instance, devoting time to differentiate which populations to target may make policies more abuse resistant; however, We assess a sample of about 2000 policies introduced this comes at a cost of then needing to validate beneficiary by countries against nine dimensions (Box 1), based on status, which may slow disbursements, especially in countries inherent qualities of policies themselves and stipulations with weak PFM systems, low quality of governance, and outlined by those who design the policies. The assessment low transparency and accountability arrangements. Again, focuses on fiscal interventions aimed to support businesses policy evaluation is more fully understood in light of the and households and revenue measures to support the health country context. Nonetheless, aggregating these ratings response. We do not assess expenditure measures aimed for a particular country or region does allow an assessment at controlling the pandemic, where the medical rationale of overall tendencies and variations across countries and will typically be the main driver in addition to economic regions. Regional and income-group averages presented in considerations. These scores, as well as trends, are discussed this report are flat averages of all policies across samples, in the following section. which is essentially an average at the country level weighted by policy number. This is to avoid cases where a country picks The assessment criteria for each policy dimension are a single policy and influences sample averages as much as discussed in detail in Annex A. Policy dimensions were scored countries in the same sample that implemented 30 policies. It with either a 0 (indicating the policy does not meet the criteria) also offers a convenient interpretation: each dimension score or a 1 (meets the criteria). The policy assessment does not of a country group is the share of that group’s policies that met account for country context in any way. It simply provides the criteria of the given dimension. For example, the score on information on the characteristics of a policy action, while the the “abuse resistance” policy dimension for the whole world appropriateness of a specific action can only be assessed by (pictured in Figure 6) is .61, which means that 61 percent of considering the country context. In this way, a score of 1 does all policies met the criteria of the abuse resistance dimension. not necessarily indicate that a particular policy action is better. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 17 >>> BOX 1: FRAMEWORK FOR ASSESSING FISCAL POLICY MEASURES IN RESPONSE TO COVID-19 Fiscal policy measures in this report are assessed in a framework that derives from the traditional timely-targeted-tempo- rary model for assessing responses to crisis situations, with a focus on the following aspects: Efficiency. The efficiency of a specific fiscal instrument to achieve particular objectives in a cost-effective way will be influenced by: • Targetability – the extent to which the instrument allows to directly target specific business or population groups or activities • Speed – the time elapsed between the adoption of the instrument and the desired impact • Abuse resistance – the ease with which abuse by eligible beneficiaries and other parties involved with the measure can be controlled Cost and fiscal sustainability. Containing the cost of fiscal measures is another important aspect of the fiscal response. This will also involve consideration of costs and benefits of specific instruments and their interactions. For example, measures that aim at reducing lay-offs may generate benefits in terms of reduced unemployment and social security pay- ments. • Affordability – the extent to which the use of the instrument impacts on fiscal stability. For example, instruments that provide support in the form of credits or through the deferral of payments will have lower cost implications than instru- ments in the form of outright grants and expenditure. • Predictability and control of cost – the extent to which upper limits for the cost of a program can be established and can the actual cost be reasonably well predicted. Flexibility. The high uncertainty regarding the duration of the pandemic and the intensity with which individual countries will be affected puts a premium on the flexibility with which an instrument can be deployed, including the ability to scale up the instrument or to stop its use as needed. • Scalability – the extent to which the instrument can be expanded or replicated for additional groups of beneficiaries in accordance with needs • Reversibility – the ease with which the response can be withdrawn, without causing economic and behavioral distor- tions Feasibility. Measures may not have their intended effect if they are difficult to implement because of administrative con- straints or impact is blunted by health measures, such as social distancing and lockdowns. • Administrative ease – the extent to which the instrument can be implemented within existing administrative capabili- ties • Impacts of the pandemic and containment measures* – the COVID-19 pandemic has direct impacts on the deploy- ment of fiscal instruments. For example, scaling up of health expenditure may be constrained by a lack of qualified personnel; measures that involve human contact, especially in groups, will be less desirable than instruments that limit such exposure; and scaling up of consumption and investment may face supply side constraints as suppliers and contractors may be in lockdown mode *This paper will refer to this dimension by its name in the April 2020 Fiscal Policy Note: “Resilience to Health Measures.” 18 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 Our findings come with two caveats. First, information on > > > policy actions differed widely across countries with respect F I G U R E 6 - Global Average Performance on Fiscal to detail and specificity, which in some cases may reflect Policy Dimensions some vagueness of the policies themselves, while in others it may be a matter of limited reporting. Consequently, the categorizing and rating of individual policy actions required some judgment and assumptions, described in detail in Annex A. Second, we assess policies against our framework in a vacuum, divorced from important factors that often determine the appropriateness of a fiscal policy, such as fiscal space, implementation capacity, pre-existing spending and coverage gaps, and the cost of the policy. This was a deliberate choice that allows us to standardize and compare policy scores across countries, but it comes with a tradeoff. We can only offer a one-dimensional look at how policies perform against our framework, and this information must be paired with familiarity with each country context in order to determine whether policies are appropriate. Our findings are not the final word on Source: Original calculations for this publication. the quality of a country’s fiscal response to COVID-19, rather they offer policymakers a previously unavailable benchmark against which to make their own assessment. For instance, learning that a country’s fiscal policy response to the crisis was ranked 150th on cost recoverability may raise a red flag and lead to a closer assessment that would not otherwise occur. A country’s performance will also depend on other factors, such as administrative capacity, political commitment, and procurement bottlenecks, yet the design and choice of policies remain an important consideration. >>> OVERALL ASSESSMENT In response to COVID-19, it was common to select policies that did not conflict with social distancing requirements, brought relatively fast relief, were scalable in terms of time, magnitude, or targeted beneficiaries, and where possible to discontinue at the intended time (reversibility). On the other hand, most countries chose policies for which the benefit felt by beneficiaries constituted an unrecoverable cost. Additionally, performance on targetability, administrative complexity, abuse resistance, and predictability and cost control varied significantly by country. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 19 > > > T A B L E 4 - Descriptive Statistics Table with Kernel Densities at the Country Level Source: Original calculations for this publication. Table 4 shows descriptive statistics of policy dimension scores at the country level. On each dimension, a country’s score reflects the share of its policies that met the criteria, from 0 to 1 (indicating 100 percent). The kernel density distributions give a sense of how countries fared on each policy dimension. Targetability, abuse resistance, and predictability and cost control have broad distributions centered slightly above 50 percent, showing that it was common for countries to have slightly more targeted, abuse resistant, and predictable policies, with moderate numbers of countries having country scores slightly above and below the average. Most other dimensions had high average country scores, with gradual tapers toward 0, indicating that most countries have a high share of policies which meet the criteria of those dimensions and only a few countries that do not. Speed and Resilience to Health Measures were concentrated around 1, indicating that most countries had every policy meeting the criteria of these dimensions. Recoverability is the only dimension with country scores distributed around low averages, indicating most countries had a low share of policies that had recoverable costs, although the distribution shows a fair number of countries with very low shares of cost-recoverable policies, and fair numbers of countries with moderate shares of cost-recoverable policies. Countries with country scores that fall below one standard deviation from the mean on any dimension may want to examine what they did differently from peers. See Annex B for each country’s score on each dimension, as well as its world rank. 20 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > F I G U R E 7 - Policy Assessment Scores of Selected Countries Source: Original calculations for this publication. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 21 Figure 7 demonstrates that countries can vary significantly CHARACTERISTICS >>> OF FISCAL in scoring profiles. Venezuela had considerably more targeted INTERVENTION BY C O U N T RY policies than Brazil, but all relief was direct and unrecoverable. GROUPINGS Rwanda’s policies were generally consistent with WB advice on fiscal policy, with the strong exception of recoverability and This section discusses characteristics of fiscal interventions partial exception of reversibility and predictability. Chad had for different country groupings, including lending categories, a below-average share of policies on all but one dimension. income groups, regions, and between FCV and non-FCV Mozambique implemented many quick and resilient to health countries. The section highlights notable differences and measures policies (two dimensions that had high average attempts to attribute the differences to either policy choice or scores for the whole world), and excelled on recoverability policy design. Significant differences across groups that are (tied with Monaco and Suriname for the highest score in the due to policy choice arise because different groups favored world), while Niger chose only policies with no prospect of policies that are inherently more likely to meet criteria along recovering the cost, such as tax exemptions. Mozambique’s certain policy dimensions. For instance, cash transfers have score can be explained by several “low cost” policies, including unrecoverable costs since beneficiaries are not expected to tax deferrals, accelerated VAT refunds, and other activities reimburse their governments once the crisis has ended. Many designed to alleviate the tax compliance burden without countries have lower average scores for recoverability simply permanently foregoing revenue. because they concentrated on direct cash transfers. Policy choice is usually the strongest determining factor in a country The UAE and Switzerland are not typical Bank clients, but or country group’s average score along policy dimensions. their scoring profiles offer a lesson. The UAE favored practical, The second consideration is policy design. Here, policy design straightforward policies that largely benefited all citizens and/ refers to specific stipulations of policies that determine their or businesses, and above half of its policies had recoverable score along the policy dimensions, independent of any intrinsic costs. Switzerland chose more complicated policies on nearly characteristic of the policy. For instance, cash transfers are not all dimensions: it targeted more carefully, had more complex inherently well-targeted—they can be universal or available measures which were more prone to abuse, and chose more to specific vulnerable groups. A cash transfer measure can policies for which the long-run costs are difficult to predict and only be well-targeted if specifically stipulated in the policy. See difficult to recover. However, with exceptional governance Annex E for a detailed discussion on the logistic regressions and a positive fiscal balance, Switzerland is a case that as well as a table of results. Through logistic regression shows complicated and pricy policies may be appropriate in analysis, we hold policy choice (among other relevant factors, the right context. However, as indicated earlier, high scores e.g., GDP per capita) constant, in order to determine whether on policy dimensions are not necessarily indicative of ideal there is a relationship between country characteristics and policy design for all contexts. Rather, the fiscal space and policy dimension scores that are independent of the particular implementation capacity of each country may determine its policy mix a country or country group chose. optimal policy design. The Policy Assessment Dashboard offers figures on fiscal space and implementation capacity for all selectable samples that may inform interpretation of policy dimension scores. A list of country scores (aggregates of policy scores at the country level) across all nine policy dimensions can be found in Annex B. Aggregate scores of income levels, lending categories, regions, and FCV status, along with ANOVA analysis of variance across categories, can be found in Annex C. Annex D presents country- and policy-level pairwise correlations. There is preliminary evidence at both the country- and policy-levels that as targeting increased, responses were more administratively complex with no pattern emerging for abuse resistance at the country-level. The next section discusses notable findings. 22 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 Through logistic regression analysis, we hold policy choice >>> ABUSE RESISTANCE (among other relevant factors, e.g., GDP per capita) constant, in order to determine whether there is a relationship between About three quarters of South Asia’s policies were abuse country characteristics and policy dimension scores that are resistant, while only about half of Sub-Saharan Africa’s independent of the particular policy mix a country or country policies were (see Figure 9). Some of these differences can group chose. be explained by particular country groups favoring abuse- prone policies, such as tax exemptions, non-uniform tax rate > > > reductions, the introduction of cash transfers to new groups of F I G U R E 8 - Targetability by Region beneficiaries, suspended tax debt collection, and suspended audit activities. Sub-Saharan Africa, in particular, tends to favor such policies in slightly higher shares than do other regions. After accounting for policy choice, IDA/Blends were no more likely than IBRDs or non-IDA/IBRDs to adopt abuse-prone policies. Regionally, a few differences remained significant after introducing controls. Holding policy choice and income constant, East Asia and the Pacific and South Asia were more likely to make their policies abuse-resistant than the rest of the world, and Sub-Saharan African countries were less likely to make their policies abuse-resistant than the rest of the world. > > > F I G U R E 9 - Abuse Resistance by Region Source: Original calculations for this publication. >>> TARGETABILITY The Middle East and North Africa stood out with lower average targetability scores than the rest of the world, due to policy design. Holding income and policy choice constant, countries in this region are significantly less likely to target their policies than the rest of the world. On the other end of the spectrum, countries in Europe and Central Asia were significantly more likely than the rest of the world to design Source: Original calculations for this publication. targeted policies, often making benefits available only to affected companies and sectors, MSMEs, the unemployed, and other vulnerable groups. >>> SPEED There was little to no variation in speed scores across country groups, largely due to limitations in the scoring team’s ability to assess speed based on brief policy descriptions, with no data on follow-through. Our methodology shows acceptable potential for adequately fast implementation for countries with moderate implementation capacity, although many factors could still cause delays. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 23 >>> COST RECOVERABILITY regions with the lowest shares of policies that had predictable costs were North America and Europe and Central Asia. Recoverability scores generally improve as incomes rise (Figure 10), although some regions with many high-income Holding policy choice constant, IDA/Blends are more likely countries (North America and East Asia and the Pacific) had than IBRDs and non-IDA/IBRDs to have predictable costs. the lowest average scores, while Europe and Central Asia had This is in large part a result of how countries designed their the highest share of cost-recoverable policies. cash transfer programs. Many IBRD cash transfer measures tied the duration of the policy, its eligible beneficiaries, or the The significant variation in recoverability scores across magnitude of the transfers to the impact of the pandemic. categories is almost completely determined by policy choice. This impact was considered inherently unpredictable, LICs and IDA/Blends have lower recoverability scores causing policies that scale with it to be scored as having simply because they choose a greater share of policies with an unpredictable cost. IDA/Blends chose more policies with inherently unrecoverable costs. This can be substantiated inherently unpredictable costs but where possible, designed by looking at the most popular policies which generally their policies to have more predictable and controllable costs. have recoverable costs — deferral of tax payments and The net effect on scores caused IDA/Blends to have similar preferential loans for firms. Although neither of these policies scores to other countries, despite having more predictable has completely recoverable costs (tax deferrals forego the policies when holding policy choice constant. In this case, collection of interest on delayed payments, for instance), most policy choice and policy design were factors countervailing of the cost can be ultimately recovered. For IDA/Blends, these against each other. two policies only made up 17 percent of the total policy mix. For IBRDs, that figure is 21 percent. For non-IDA/IBRDs, that For similar reasons, countries in East Asia and the Pacific figure is 26 percent. For regions which scored poorly on cost tended to have a higher share of predictable policies than the recoverability relative to other regions, these policies made up rest of the world, holding policy choice and log of GDP per only 14 percent (EAP), 19 percent (NA), and 17 percent (SSA) capita constant, despite average scores that did not deviate of all policies. Other cost-recoverable policies explain the rest from other regions. Conversely, countries in Europe and of the variation in recoverability. Central Asia tended to implement more unpredictable policies than the rest of the world. > > > F I G U R E 1 0 - Recoverability by Income Group > > > F I G U R E 1 1 - Predictability and Cost Control by Lending Category Source: Original calculations for this publication. >>> PREDICTABILITY AND COST CONTROL Source: Original calculations for this publication. Predictability and cost control scores (Figure 11) have the opposite tendency of policy dimension scores discussed so far, with low-income countries presenting a higher share of policies meeting the criteria than high-income countries. The 24 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >>> REVERSIBILITY from Sub-Saharan Africa’s, with a greater share of scalable policies. However, this is not necessarily due to particular As a broad average, shares of reversible policies in IDA/ attention to scalability on the part of South Asia (controlling Blend countries and LMICs are significantly lower than in for policy choice, South Asian countries did not score better non-IDA/IBRD countries and HICs. Sub-Saharan Africa also than other regions); rather, it is due to South Asian countries stands out with average reversibility scores significantly below gravitating toward inherently scalable policies. those of other regions, such as Europe and Central Asia or North America. However, holding policy choice constant, IDA/ >>> ADMINISTRATIVE COMPLEXITY Blends and countries in Sub-Saharan Africa do not design less reversible policies than do other regions and categories, Administrative complexity of fiscal policy interventions is although a slight relationship emerged from regression lowest for IDA/Blends among the lending categories, for analysis between GDP per capita and reversibility scores. The Middle East and North Africa among the regions, and for FCVs difference in average reversibility scores is mostly a function compared to non-FCVs. Despite the high share of measures of policy choice, and partly a function of policy design. meeting the criteria among IDA/Blends, LICs scores were not significantly different from those of other income groups. Lower-middle income countries did not shy from reversible policies but simply chose them in smaller shares than other This was not determined by policy choice alone; IDA/ income groups. LMICs chose policies, which together make Blends were more likely than IBRDs and non-IDA/IBRDs to up about 22 percent of policies, that are more likely to be make their policies simpler, holding policy choice constant. reversible (tax payment deferrals for businesses, preferential For cash transfers in particular, policy language reined in the loans to firms in distress, and enhanced paid leave). In sophistication level of policies by targeting beneficiaries of high-income countries, a total of 36 percent of policies were existing systems, targeting easy-to-distinguish beneficiaries, dedicated to tax deferrals, preferential loans, and enhanced or by not targeting at all. paid leave, which explains their better reversibility scores overall. The Middle East and North Africa was a standout among the regions for being significantly more likely to make policies administratively simple than the rest of the world, holding policy > > > choice and GDP per capita constant. Israel, Lebanon, Qatar, F I G U R E 1 2 - Reversibility Scores by Income Group the UAE, and the West Bank and Gaza all ranked among the top 50 countries on administrative simplicity, with almost three- quarters of countries in the region ranking among the top half of countries in the world with respect to this policy dimension. To some extent, this is due to choosing inherently simple policies, such as the suspension of audits and tax deferrals. But even among policies that all regions were prone to choose, Middle Eastern and North African countries tended to simplify their policies, with a strong tendency to leave policies untargeted. Source: Original calculations for this publication. >>> SCALABILITY Variation along scalability was minor, with no income group or lending group scoring differently from other groups, on average. South Asia stands well above the world average in scalability scores and has a performance statistically distinct A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 25 >>> RESILIENCE TO HEALTH MEASURES Nearly every policy scored well on this policy dimension, and as a result, the average scores of all lending categories, income groups, and regions are above 0.9 out of 1. Still, some significant differences emerge, with IDA/Blends, LICs, and developing regions scoring statistically significantly lower than other categories. All variation can be attributed to a single policy: supplementary ad hoc programs. Most policies saw no variation in scores (were all scored “1”), and several saw some variation but only supplementary ad hoc programs were significantly less likely to meet the criteria for resiliency to health measures than other policies. LICs scored worse than other income groups on this dimension because supplementary ad hoc programs were their favorite policy (16 percent of total), while LMICs only chose them 8 percent of the time, UMICs chose them 8 percent of the time, and HICs chose them 5 percent of the time. >> A LOOK FORWARD Even with the rollout of effective vaccines, the COVID-19 policy challenges in dealing with the impacts of the COVID-19 crisis is far from over and fiscal policy instruments will remain pandemic include the following: critical to deal with the pandemic and support an economic • Sustaining and adjusting fiscal policy measures to recovery. The World Bank (2021) projects that economic mitigate the health and economic impacts of the crisis; growth in EMDEs will recover to 5 percent in 2021, mostly • Transitioning to fiscal policy measures in support of driven by a strong recovery in China. However, aggregate economic recovery; EMDE output is projected to still be 6 percent below its • Phasing out fiscal policy measures and recovering pre-pandemic projection in 2022. These projections are funds; and underpinned by the assumption that in advanced economies • Introducing reforms aimed at fiscal sustainability. and major EMDEs pandemic control measures will reduce the daily numbers of infections in the first half of 2021 and In the following, we briefly discuss the first three challenges widespread vaccination will be achieved by the end of 2021. and how the present stocktaking and assessment of fiscal Progress in the control of the pandemic in other EMDEs and policy measures can help inform policy choices. Issues related LICs is projected to follow with a lag of two to four quarters, to funding options and fiscal sustainability will be discussed in partly because of delayed vaccine distribution. Such progress a separate note. will rely on continued adequate fiscal policy interventions and comes with large downside risks. These include uncertainties > > > with respect to the progression of the pandemic and the F I G U R E 1 3 - Key Fiscal Policy Challenges in Dealing with the COVID-19 Pandemic deployment of vaccines as well as the risk of widespread corporate and sovereign defaults if financial stress persists. Under this macro-economic outlook, fiscal challenges will remain significant at a time where extraordinary fiscal and monetary policy efforts in 2020 have left many countries with reduced fiscal space. Even though growth is projected to recover in 2021, real GDP in EMDEs will be lower at the end of 2021 than it was at the end of 2019 in all regions except East Asia and the Pacific and Europe and Central Asia. For LICs the pandemic represented a particularly severe setback with projections of rising poverty and economic activity at levels 5.2 percent below the pre-pandemic trend. Source: World Bank While real GDP remains below pre-pandemic levels in many countries, government revenue has fallen even lower, and will recover more slowly.10 At the same time, public expenditure requirements due to the pandemic remain elevated. Key fiscal 10. Empirical work on short-term tax buoyancy and elasticity (Dudine P. and J. Tovar Jalles, 2017) suggests that short-term buoyancy tends to be larger during contractions than during times of expansions, which would imply that recovery of government tax revenue will be slower than the recovery of GDP. However, country specific tax buoyancy depends on a range of structural, demographic, and institutional factors and macro-economic conditions. 26 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >>> S U S TA I N I N G AND ADJUSTING predictable costs. Just as informative are cases in which FISCAL POLICY MEASURES TO countries which probably should perform above average but M I T I G AT E THE H E A LT H AND do not. IDA/Blend countries should probably aim to design ECONOMIC IMPACTS OF THE CRISIS measures that are more abuse-resistant and affordable than measures designed by IBRD and non-IDA/IBRD countries, but Many of the measures adopted to contain the pandemic and there is no evidence that this is done. Shining a light on these mitigate its economic impacts will remain relevant as long as tendencies can help give the attention necessary to improve the pandemic is not sufficiently contained. As vaccines are policy design and highlight measures that need to be carefully becoming available, countries will increase expenditure for managed. their procurement and distribution. >>> TRANSITIONING TO FISCAL Since the outbreak of the pandemic, a large amount of POLICY MEASURES IN SUPPORT OF experience and knowledge on the effectiveness of fiscal policy EC O N O M I C R EC OV E RY interventions has been accumulated. It is beyond the scope of this note to summarize this literature. However, from a fiscal The design principles and fiscal policy assessment perspective the policy assessment framework can provide framework were established with emergency response in some indications as to how intervention approaches can be mind. A new set of desirable characteristics is needed as strengthened. countries pivot from crisis response to economic recovery. During the recovery phase, there will be a greater focus on The policy scores presented in this paper are not the final investment spending. Regular public investment management say on whether a policy is fit for purpose and should be approaches will be appropriate for the assessment and sustained or not. Some countries may deliberately adopt prioritization of investment proposals. The impact of fiscal complex policies, which are prone to abuse, with costs that intervention on stimulating economic activity, employment, are unrecoverable and difficult to predict, if the support the and expanding the government revenue base becomes policy provides is worthwhile. Individual country fiscal policy central. The contribution of an intervention in supporting the scores – both aggregated at the country level (Annex B) and green recovery and mitigating climate risks would be another at the policy level are available upon request) – can be used to important design characteristic against which to assess fiscal assess how policies performed against objective standards of policy interventions. Work on fiscal policy interventions to fiscal policy design. If a country performed below benchmark support a sustainable green recovery is underway, with a focus or worse than anticipated, perhaps a closer look with a new on fiscal policy interventions that help advance the transition lens could reveal areas of improvement. These scores must to a low carbon economy. be combined with knowledge of the country context in order to draw appropriate conclusions. However, it is important to note that even during the recovery phase, interventions to support households and businesses After controlling for policy choice (which was the major will still be necessary in many countries to prevent excessive determining factor in dimension scores) and GDP per capita, hardship. The World Bank (2021) suggests that the crisis may some country groups demonstrated a strong, statistically result in a permanently lower trajectory of economic activity significant tendency to perform differently from the rest of the compared to the pre-crisis trajectory. In addition, recovery from world on certain policy dimensions. For instance, countries in the crisis will be accompanied by structural change as some the Middle East and North Africa are significantly less likely businesses and some jobs will not be preserved or restored, to target their measures than the rest of the world. If this while others may emerge. The speed and extent of recovery comes as a surprise, policy makers may wish to reevaluate will also show some regional and sectoral heterogeneity, with their policies, looking carefully for ways to improve targeting. some sectors, such as tourism, expected to recover more Additionally, Sub-Saharan African countries are significantly slowly than others. Social protection measures to support less likely to make their policies abuse-resistant than the rest people who lose incomes and employment in this transition of the world and less likely to make their policies affordable together with investments in retraining will thus need to be an than the rest of the world. Countries in Europe and Central important element of recovery spending. Asia are significantly less likely to design policies which have A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 27 >>> PHASING OUT FISCAL POLICY With respect to the potential for recovery of funds, data M E AS U R E S A N D R EC OV E RY O F F U N D S on the magnitude of the fiscal response (Figure 2) show that for advanced economies, interventions through loans, As the pandemic is being brought under control and equity amounted to about 8 percent of GDP, i.e., about half economies recover, many of the fiscal emergency measures of their fiscal response. However, for Low Income Countries, adopted during the crisis will need to be phased out. In addition, such interventions amount only to 0.2 percent of GDP, i.e., deferred payments, short term loans, and other recoverable just over 10 percent of their fiscal response. In addition to interventions will need to be recouped. The assessment of loans, deferred tax payments for businesses were among reversibility and cost recoverability of fiscal interventions the most frequent policy interventions by all country groups. provides some insights into the challenges individual countries Other measures such as deferred tax filing for businesses will face and also highlights countries that are likely to face the and individuals, deferred tax payment for individuals, and biggest challenges in this area. suspended debt collection also will require action to recover funds. Here it is important to note that the recovery of funds is Our analysis indicates that while for LICs about 77 percent of by no means automatic or simple but will entail in most cases actions are reversible, for high HICs this percentage increases significant administrative and political effort. to 85 percent. While well-designed policy actions have inbuilt expiry dates, many actions lack such provisions and will thus require attention by the authorities to terminate such actions at the right time, against likely resistance from beneficiaries and other stakeholders. > > S U M M A RY Countries around the world have adopted numerous fiscal be used to inform countries’ policy designs. The assessment policy actions to mitigate the health and economic impacts of in this note is carried out for individual policy actions as well the COVID-19 pandemic. Our fiscal policy data base covers 203 as for the entire “policy package” adopted by a country. In the economies and contains nearly 4000 policy actions, adopted assessment, an important distinction between policy choice between the outbreak of the pandemic and September 1, (i.e., relating to the “inherent” or “typical” characteristics 2020. The number of individual actions adopted by countries associated with a policy actions) and policy design (i.e., the is as high as 58, although the average number is 20 and the specific formulation of a policy action by the authorities) is median is 17. This note groups these actions into 47 different discussed. The analysis reveals significant differences in types. However, there is a small number of interventions that the design characteristics of policy packages adopted by have been adopted by a large majority of countries. These countries, where most of these differences are due to policy include general health expenditures, loans and deferred tax choice and to a much lesser extent due to policy design. payments for businesses, and direct cash transfers and ad- hoc interventions for the benefit of individuals and households. The findings on the types of interventions and their Country size, income, the number of COVID-19 cases and the characteristics can help inform key policy decisions in the overall size of the fiscal response to the pandemic are positively next stages of the fiscal policy response, which include the correlated with the number of policy actions adopted. Country continuation of fiscal policy measures to mitigate the health indebtedness, on the other hand, shows no correlation with and economic impacts of the COVID-19 pandemic (which will the number of actions adopted. have to be augmented with measures to fund the procurement and deployment of vaccines), the transition of measures to The note then reviews a subset of these policy actions support economic recovery, and finally, the phasing out of against a set of design characteristics (targetability, speed, measures and recovery of deferred payments and loaned abuse resistance, cost recoverability, predictability and control funds. of cost, scalability, reversibility, administrative ease, and resilience to health measures). These characteristics were set forth at the onset of the pandemic and have proven to be a useful metric to assess policy responses. In the future, they can 28 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >> APPENDIXES >>> APPENDIX A: METHODOLOGY description (for example, a measure can state an intended >>> SCORING demographic of MSMEs, making it targeted). In order to apply the same standard to all countries, country context was not To ensure transparency, the assessment team developed considered. This has a serious implication for learning from a systematic scoring process and documented the criteria these scores: the final assessment of whether policy actions necessary to score each policy along each dimension. These are fit for purpose will require considering the country context. are listed in table A1 below. The further discussion section features the scoring team’s assessment of its own ability to Although health sector measures have scoring guidelines in score consistently as well as further analysis of groups of table A1, they were excluded from analysis, as policy choices policies that tend to be unscored or to be scored as “likely are primarily driven by medical requirements. This paper meets/does not meet the criteria” or “missing” rather than be focuses thus primarily on fiscal policy measures to mitigate scored with a score of “certainty.” the economic impact of the pandemic, where policymakers could choose among a wide range of measures. Additionally, Each policy was assessed on the basis of the inherent some health measures come with significant additional qualities of the policy itself (for example, tax deferrals have health management costs which may or may not have been an inherently recoverable cost for tax administrations of presented as part of the response packages. moderate capacity), as well as details stipulated in the policy > > > T A B L E A 1 - Fiscal Policy Scoring Criteria SCORING KEY: [L] - LIKELY POLICY DIMENSION CASE M - MEETS THE CRITERIA N - DOES NOT MEET THE CRITE- RIA A - TOO AMBIGUOUS TO SCORE Explicit statement that a policy is universal N No mention of intended beneficiaries. [L] N No mention of details on how beneficiaries will be identified. [L] M Example: “for businesses affected.” Strong details about targeting criteria. M Targeted population is vulnerable during the pandemic (the homeless, the jobless, the elderly, people with health M conditions, businesses in distress, SMEs, etc.) TARGETING When the intended beneficiary is a resilient population (example: taxpayers receiving dividends from other N countries). “Unincorporated businesses” and the self-employed will be M considered SMEs. Targeting the elderly, even if they are well-off. M Ambiguity. Example: “Special arrangements to support taxpayers who are in the process of settling overdue taxes.” A It is unclear whether the intended beneficiaries are low- income or simply non-compliant A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 29 Actual benefit is months into the future but understood and internalized at policy announcement. Beneficiaries can budget accordingly, freeing up cash. Example: tax payment [L] M deferrals that delay a payment already due months into the future Benefit realized within 1 month. Example: “Unemployment benefits equal to at least minimum wage for three months M starting in April.” Benefit realized in a few months or sooner. [L] M SPEED Benefit realized over six months. [L] N Benefit realized in more than one year. Example: extended N loss carry-forward. Hospital procurement, low- and high-cost. N Expanding manufacturing capacity and infrastructure. [L] N Expanding human resources (health expenditures); labor market considered inflexible as health workers are in high [L] M demand worldwide. Targeted tax rate cuts where taxpayers do not have distinct tax regimes (potentially inviting arbitrage). Example: rate [L] N cuts for health companies. Targeted tax rate cuts where taxpayers do have distinct tax regimes (identifying to whom the rate applies is easier). [L] M11 Example: rate cuts for SMEs. Tax rate cuts which apply evenly across taxpayers. [L] M12 Tax exemptions. N New benefits programs. N Increasing the benefits to known beneficiaries under existing [L] M programs. ABUSE RESISTANCE Adding beneficiaries to an existing program (changing the [L] N definition of who is eligible). Supplementary ad hoc programs which distribute physical N13 goods. Broadened tax deductability: new deduction options create [L] N new opportunities for abuse. Tax credits. Distinguishing eligible beneficiaries requires [L] N above-average enforcement capability. Suspending audit and debt collection activities compromises N enforcement. Large-scale procurement, including low- and high-cost medical items and for health-sector related infrastructure [L] N spending (vulnerable to leakages and corruption). Policy has a prospect of repayment on the part of beneficiaries or is revenue neutral. Examples: tax deferrals, M deferred payments, flexible payment arrangements, credit COST RECOVERABILITY programs, loan guarantees, accelerated VAT refunds, etc. Partial prospect of repayment on the part of beneficiaries. [L] M Example: lost interest from tax deferrals. 11. This is a new assumption that will be applied in the update. The previous assumption ruled that all targeted tax rate cuts (non-uniform) invited arbitrage. Scoring still reflects this. 12. During scoring, these circumstances were treated as not meeting the criteria of the abuse resistance dimension, and current scores reflect this. This new guideline will apply going forward. 13. This is a new guideline which we will apply going forward. Current scoring reflects the old assumption that physical goods had less scope for abuse. 30 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 No prospect of repayment on the part of beneficiaries. Examples include cancelled tax debt, tax exemptions, cash N transfers, broadened tax deductibility, extended loss carry forward, and expansion of benefits. Suspended audit and enforcement activities. Unclear whether foregone revenue will be recoverable later or if A14 retroactive auditing will be less effective once activities resume and uncollected revenue will stay uncollected. Tax amnesty would likely be “unrecoverable” without the N policy anyway. Ambiguity. Example: Extension of the seasonal suspension of A evictions, where it is unclear who bears the cost. Predicting full cost of policy requires estimating number of beneficiaries – case: ultimate beneficiaries easy to estimate. M Example: benefits for college students. Predicting full cost of policy requires estimating number of beneficiaries – case: ultimate beneficiaries hard to estimate. [L] N Example: benefits for laid off workers. No specific detail on unit cost. Assume this was known by Ignore - judge other factors implementers but not stated. Benefits tied to outcomes of the pandemic (such as unemployment, increase in childcare needs, spike in N beneficiaries, and duration of the pandemic) have costs as unpredictable as those outcomes. Aggregate expenditure cap explicitly stated. Example: upper [L] M spending limits for preferential loan program. Aggregate expenditure cap exception: existence of political pressure likely to force government to spend past its aggregate expenditure cap. Example: health spending to increase capacity of health sector. As the pandemic evolves, [L] N PREDICTABILITY AND COST health emergencies could develop (which governments CONTROL cannot predict), and governments will be forced to spend further in response. Related and unpredictable: low- and high-cost medical procurement. Any policy that undermines enforcement, facilitation, or trust hurts tax compliance and therefore future collections, which has an ultimately unknowable impact on revenues. [L] N Examples include exemptions, unbalanced rate reductions, audit activity suspensions, and debt forgiveness. Expanding existing pandemic-response measures, revealing them to have been poorly predicted and poorly controlled. Ruling: treat as new measure and judge based on its Ignore - judge other factors characteristics, not implications based on earlier version of the measure. Example: a one-off cash transfer followed by another one-off cash transfer to the same beneficiary group. Large-scale infrastructure spending (including health infrastructure and public works programs); typically affected N by time delays and cost over-runs. Expansion of human resources in medical sector – it is considered easier to plan intended expenditures and predict [L] M costs than medical procurement. 14. Current scoring reflects ambiguity (no score); however, this policy will be considered [L] N going forward, as consequently tax arrears tend to increase, and this in prac- tice is what has often led to large political pressures for tax amnesties. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 31 Policy includes a sunset date and is not salient or popular enough to expect resistance as the end date approaches. M Example: tax filing and payment deferral for six months. Policy includes an end or sunset date but is salient and popular among citizens and may face resistance as the end date approaches. Example: cash transfers to households for [L] M two months. Ruling: give them credit for establishing an end date. Policy does not explicitly include an end or sunset date REVERSIBILITY and is not salient or popular enough to expect resistance in the event that policymakers would like to discontinue it. [L] M Example: suspension of audit activities. Ruling: will not face much resistance. Policy does not explicitly include an end or sunset date and is salient and popular among citizens and may face resistance [L] N in the event that policymakers would like to discontinue it. Example: indefinite increase in pension payouts. Ad hoc programs and preferential loans to businesses and individuals (these policies are assumed to have strong [L] M implicit end dates). Policy cost is listed and could be increased. Example: unemployment benefits increased from $200/mo to $250/ M mo. Policy cost is not listed (neither aggregate nor unit cost) but could likely be increased. Example: Expansion of social safety [L] M net programs. Policy has an unknown cost of benefit (e.g., exemptions). Ignore - judge other factors Policy duration is stated and could be extended. M SCALABILITY Policy duration could not be extended (e.g., a one-off cash N (unless cost or beneficiaries transfer). can be scaled) Policy could be replicated to different group of beneficiaries (e.g., from unemployed individuals to other vulnerable [L] M populations) or products and retain its essence. Policy’s expansion to a different group of beneficiaries or products would fundamentally change its essence (e.g., from [L] N (unless cost or duration supporting SMEs to targeting all firms); effectively making can be scaled) it a new policy, not a scaled-up version of a current policy. Policy would be relatively easy to identify the targeted [L] M beneficiaries. Example: universal benefit to all businesses. Policy would be relatively difficult to identify the targeted [L] N beneficiaries. Example: transfers to informal workers. Policy would be relatively easy to enforce. Example: universal [L] M tax rate cuts. Policy would be relatively difficult to enforce. Example: tax [L] N rate reductions for which eligibility is hard to distinguish. ADMINSTRATIVE COMPLEXITY Policy eligibility is determined using an impartial metric (e.g., poverty line, employment status, etc.), is intentionally widespread, or is a blanket exemption. Example: Moratorium [L] M on pension contributions for the hospitality sector; one-off payment to income support recipients. Eligibility is determined on a case-by-case basis and is therefore subject to discretion. Example: banks providing [L] N flexible solutions to customers on an individual basis. Policy adds a new business line (non-blanket exemptions). Example: A new program to compensate entrepreneurs and [L] N their employees. 32 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 Requesting funds/lending from multi-lateral institutions. [L] M Implementing ad hoc programs. M Re-introducing an old program. M Ambiguity. Example: housing scheme for the homeless. A Policy explicitly reduces in-person contact. M Policy does not state whether face-to-face contact is [L] M RESILIENCE TO HEALTH MEA- required. SURES Policy requires person-to-person contact (complicating factor: this may obviate the need for the beneficiary to N subject themselves to other transmission risks). Example: delivery of in-kind goods. If a policy has multiple components, the assessment is based on the single lowest-performing score for each dimension. Example: “Creation of the ‘Stay at Home’ program with CROSS-CUTTING ISSUES AND different elements, including (i) top-up to safety net of For each dimension, the worst- 5,000 pesos (USD 92) for two months to 811,000 beneficiary performing score was recorded MISCELLANEOUS families that have the Solidarity payment card to purchase food and basic necessities; (ii) horizontal expansion at the same amount 5,000 pesos (USD 92) to 690,000 nonpoor and vulnerable non-beneficiary families in the social registry.” Distinct occurrences of the same policy. Example: one- off cash transfer to civil servants and taxi drivers; and a Judge each independently (second) one-off cash transfer to taxi, tourist, and rickshaw drivers and tourist guides. State guarantees for private companies. This type of policy called was considered a preferential loan as it increases the company’s access to financing, while the government bears - the risk of default. It was scored as if it were a standard credit and equity measure. Policies beyond the framework. It would not be appropriate to judge policies designed to raise revenue with this framework and were therefore exempted from scoring along A the dimensions. Example: Creation of a solidarity tax of relatively unaffected public officials. Sub-national responses. Provincial and local policy responses tended to be grouped together and lacked the same depth as national government responses. Example: “State and Territory governments also announced fiscal stimulus A packages, together amounting to A$11.5 billion (0.6 percent of GDP), including payroll tax relief for businesses and relief for households, such as discount utility bills, cash payments to vulnerable households, and support for health spending.” Source: World Bank. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 33 >>> FROM FRAMEWORK TO REAL POLICIES full cost of each policy. Additionally, judging each policy against the country context means that each policy is The assessment team applied reasoning featured in the judged against a different standard. This is ideal for tailoring COVID-19 fiscal policy note (World Bank 2020) whenever a policy to the country context but not ideal for building a possible. However, there were some areas of the original standardized, cross-sectional dataset. Instead, the scoring framework that needed to be changed when applying it to real- team replaced the concept of “Affordability” described in the world policies as they appear in the compiled database. This framework with “Cost Recoverability.” This switch allowed was necessary partly to make the scoring exercise feasible— the researchers to ask: “Was the government able to provide not all of information necessary to assess the criteria discussed a benefit to beneficiaries and later recoup some of the cost in the original framework was available to the assessment of the benefit?”15 In this manner, “cost recoverability” can be team. This was also necessary in order to create a cross- assessed across countries by an objective standard, allowing sectional dataset. The original framework assessed policies for meaningful comparison. Consequently, some of the full against the country context, which is a standard that varies by meaning of “affordability” is lost. Individual policies did not country. In order to create a consistent standard, policies were leave a lot of room for customization with respect to whether scored in a vacuum. Finally, it is also worth noting that some the benefit will be repaid by beneficiaries. Measures either policies may have lost detail when copied over in the data provide a benefit that will not be repaid, as with a cash transfer, collection stage, and some assumptions about missing detail or repayment is implicit, as with a subsidized loan. As a result, were necessary to score them. This section summarizes the regression analysis attributed no tendency to making policies judgment calls made in converting a conceptual framework for with recoverable costs to any particular country characteristic, practical application. when controlling for policy choice and GDP per capita. Targetability – The original definition of targetability Predictability and Cost Control – Similar to the issue of emphasized the ability of a policy to directly target specific abuse resistance above, policies were judged against an businesses or populations. Analysts revised the criteria of objective standard (independently of government capacity), this dimension to only consider a targeted demographic that since it would be difficult to gather and apply data on the is in any way vulnerable during the pandemic. For example, forecasting capabilities of each country. This makes policies targeting SMEs counts as good targeting, but targeting more convenient to score and standardized across countries. investors with foreign-earned income from capital investments does not. Furthermore, analysts did not score follow-through Reversibility – The policy summaries in the database may but rather a declaration to target. Regardless of capacity to not have included a specified end-date, which could be by target, countries which stipulated that their policies were for design, mistake, or omission. For consistency, analysts specific groups received credit for targeting. followed the breakdown described in Table A1, which assumes that countries without an end-date do not have one. Speed – The concept of speed was changed to include benefits that a beneficiary can anticipate and therefore budget Scalability – The most persistent issue for this dimension for, effectively realizing a benefit earlier than it is received. was the question of at which point a policy is scaled beyond the scope of the original policy, and therefore does not qualify as Abuse Resistance – Policies were judged in a vacuum, a scaled-up version of itself but rather a new policy altogether. ignoring government capacity so that in-depth country The analysts decided that when a policy’s expansion to a new knowledge would not be necessary. This creates a consistent group of beneficiaries would fundamentally change the nature standard across countries, allowing for meaningful comparison. of the policy, it would not be considered “scalable” (e.g., cash transfers to unemployed workers could not be scaled up to Cost Recoverability – The Fiscal Policy Assessment include employed workers), unless some other characteristic Framework, as written, defines affordability as “the extent to of the policy (duration, amount) could be scaled. which the use of the instrument impacts on fiscal stability.” Unfortunately, assessing the impact on fiscal stability is a Administrative Complexity – The assessment of resource-intensive exercise, which requires more detail administrative complexity required the same approach as than was available for both the policies and the countries for “Abuse Resistance.” Policies would be judged without implementing them, as well as post-hoc estimation of the considering the country’s implementation capacity in order 15. Example: tax deferrals usually involve waiving interest on tax owed. This is costly, but is a small cost relative to the full value of the tax payment that taxpayers can hold on to during the deferral period. Therefore, such policies are considered “cost-recoverable.” 34 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 to make them more convenient to score and standardized (e.g., Albania’s “Lk11bn (0.6 percent of GDP) sovereign across countries. guarantee fund for companies to access overdrafts in the banking system to pay wages for their employees for up to Resilience to Health Measures – This policy called for three months with an interest rate capped at 2.85 percent for judging whether the policy measure itself required in-person a maturity of up to two years”). The discrepancy in policy detail contact. The feasibility of business continuity would be necessitated a consistent framework, for certain, likely, and another appropriate concept of “resilience,” and one that missing data points. future research teams should consider. A comparison of the likelihood that a policy went unscored Further considerations – Ambiguity that stemmed from or was flagged as “likely meets the criteria” or “likely does a lack of detail or clarity in how policies are worded were not meet the criteria” across each of the policy dimensions more likely to be scored “missing” or “likely” than “certain.” In is presented in table A2 below. By reviewing the number of general, the assessment team was unsure whether a policy missing scores for the policy dimensions, it is apparent that lacking detail was due to 1) the policy itself being light on detail most were close to the sample-wide average of 4.8 percent. The as written in official documents in the implementing country, average number of missing scores for “Speed” and “Resilience 2) the policy’s summarization at the time of publication, to Health Measures” are both two or more percentage points either by government spokespersons or newspapers, or 3) below the average, while the average number of missing for summarization of the policy at the time of transcription into the “Administrative Complexity” is well above the sample average database. Three dimensions were severely affected by this at almost 12 percent of measures, indicating a high level uncertainty: “Targetability,” “Predictability and Cost Control,” of ambiguity with respect to this dimension. The inherently and “Reversibility.” Meeting the criteria of these dimensions qualitative aspect of “Administrative Complexity” caused requires that policies specify beneficiaries, indicate an analysts to create more scoring guidelines for this dimension aggregate spending cap, or indicate a sunset date. All may than for any other, and it still had significantly more missing be casualties of incomplete transcription, since there are values. seemingly minor details that might be trimmed when trying to reduce word count that may have strong implications for the policy assessment. >>> POST-SCORING ASSESSMENT OF METHODOLOGY Ensuring a consistent scoring system was the priority of the assessment team. After deciding on the criteria for scoring dimensions, the reviewers independently scored the same samples of policy measures, compared results, and created further guidelines. The first iteration of this self-assessment produced a similarity of 82 percent across all policy dimension scores. Each of the scores of the approximately 130 policies were discussed, and a second testing sample was selected. Scoring consistency improved to 84 percent across all dimensions. Policies’ detail ranged immensely. At one end, some policies were vague (e.g., Australia’s “State and Territory governments also announced fiscal stimulus packages, together amounting to A$11.5 billion (0.6 percent of GDP), including payroll tax relief for businesses and relief for households, such as discount utility bills, cash payments to vulnerable households, and support for health spending”), while others were specific A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 35 > > > T A B L E A 2 - Policy Dimension by Missing and Likely/Unlikely Score POLICY DIMENSION MISSING # (%) LIKELY # (%) TARGETABILITY 118 (4.9%) 1241 (51.7%) SPEED 54 (2.3%) 1374 (57.3%) ABUSE RESISTANCE 101 (4.2%) 2170 (90.4%) RECOVERABILITY 135 (5.6%) 569 (23.7%) PREDICTABILITY 124 (5.2%) 2075 (86.5%) REVERSIBILITY 101 (4.2%) 1830 (76.3%) SCALABILITY 78 (3.3%) 629 (26.2%) ADMINISTRATIVE COMPLEXITY 286 (11.9%) 1924 (80.2%) RESILIENCE TO HEALTH MEASURES 48 (2.0%) 41 (1.7%) AVERAGE 116 (4.8%) 1317 (54.9%) Source: Original calculations for this publication. The percentage of data scored as “likely meets the criteria” or “likely does not meet the criteria” (uncertain) was more varied across the dimensions: “Abuse Resistance,” “Predictability,” “Reversibility,” and “Administrative Complexity” all had many more instances of assumed data than the overall average. It should also be noted that almost 55 percent of scores were marked as “likely meets the criteria” or “likely does not meet the criteria,” reflecting various sources of ambiguity previously discussed. The second lens of analysis for missing data and data scored with uncertainty across lending groups. There are three cuts for the purposes of this analysis: IDA/Blend countries, IBRD countries, and neither. Table A3 below presents missing and uncertain data by these lending categories. > > > T A B L E A 3 - Total of Lending Category by Missing and Uncertain Scores LENDING CATEGORY - TOTAL MISSING # (%) LIKELY # (%) IDA/BLEND - TOTAL 322 (6.6%) 2607 (53.6%) IBRD - TOTAL 418 (4.9%) 4666 (55.0%) NEITHER - TOTAL 305 (3.7%) 4580 (55.5%) Source: Original calculations for this publication. Examining policy dimensions through the lens of lending category reveals that the proportion of missing and uncertain data does not statistically significantly differ among World Bank lending categories. This suggests the policies that are scored are not a misrepresentative sample, making for meaningful comparison across categories. The slight variation in missing and uncertain scores is related to types of policies chosen by lending category (see Analysis of Significance Difference for more detail). In general, some policy types which are more likely to be scored as missing or assumed are also disproportionately favored by IDA/Blend countries, relative to IBRDs and others. For example, IDA/Blend countries introduced “Supplementary ad hoc programs (feeding programs, utility waivers)” for about 11 percent of all policy measures while IBRDs and others only introduced such policies about 6.5 percent of the time. Consider also that this policy measure was significantly more likely to receive a missing score (2.3 percent) than other policy types (0.8 percent) for the dimension “Resilience to Health Measures.” Since this type of policy measure is more likely to receive a missing score in this category and IDA/Blend countries rely relatively more on this policy, it results in a higher number of scores marked “missing” for IDA/Blends, on average. 36 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A P P E N D I X B : C O U N T RY- L E V E L >>> >>> A NOTE ON SCORE INTERPRETATION POLICY DIMENSION SCORES >>> INTRODUCTION These scores are best understood by reading the methodology section and findings section of this report. In Before using or reporting these scores, please carefully read summary, scores for any policy are either 0 (does not meet the the methodology and findings sections of this report, which criteria of the policy dimension listed) or 1 (meets the criteria). discuss limitations in data quality and availability, as well as The scores listed here are flat averages of each listed country’s trends behind the scores. For a full dataset of scores (at the policy scores. Effectively, the scores represent the share of policy level, 2,400 policies, certain and likely scores flagged), a country’s policies that met the criteria of each dimension. please contact Eric Lacey (elacey@worldbank.org) and This reporting does not disaggregate scores that were flagged Joseph Massad (jmassad1@worldbank.org), who compiled as likely from those that had a higher degree of confidence. the dataset. Finally, please find the Policy Assessment Please see the accompanying report and dashboard for Dashboard (an excel file circulated with this report), which more information. Finally, note that the following country-level features visuals, analysis, and sample-selection tools in a scores exclude scores on health expenditures. point-and-click format. > > > T A B L E A 4 - Country-level Policy Dimension Scores and World Ranking A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 37 38 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 39 40 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 41 42 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 43 Source: Original calculations for this publication. 44 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > A P P E N D I X C : C O U N T RY G RO U P S C O R E S >>> INTRODUCTION This section presents policy scores aggregated across income levels, regions, lending categories, and FCV status. For brevity, this annex omits a discussion of scores and significant differences across categories (see “findings” in the main body of the report for that discussion). For statistical analysis that controls for relevant factors (logistic regression analysis), see Annex E. For a list of country-level scores, see Annex B. For a discussion of how scores were derived, see methodology in the main body of the report. For more detail on the original assessment framework, see Annex F. Additionally, the full list of policies assessed, the original fiscal policy framework, the database of scores at the policy level, and a dashboard for easy navigation of visuals and analysis can be requested (see Annex B). > > > F I G U R E A 1 - Policy Scores Across Income Groups Note: Stars show results of ANOVA test of variation among samples (*** p<0.01, ** p<0.05, * p<0.1). Source: Original calculations for this publication. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 45 > > > F I G U R E A 2 - Policy Scores by Region Note: Stars show results of ANOVA test of variation among samples (*** p<0.01, ** p<0.05, * p<0.1). Source: Original calculations for this publication. > > > F I G U R E A 3 - Policy Scores by Lending Category Note: Stars show results of ANOVA test of variation among samples (*** p<0.01, ** p<0.05, * p<0.1). Source: Original calculations for this publication. 46 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 > > > F I G U R E A 4 - Policy Scores by FCV Status Note: Stars show results of ANOVA test of variation among samples (*** p<0.01, ** p<0.05, * p<0.1). Source: Original calculations for this publication. A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 47 > > > A P P E N D I X D : PA I RW I S E C O R R E L AT I O N S O F P O L I C Y D I M E N S I O N S * denotes significance at the 1% level > > > T A B L E A 5 - Country-level Dimension Pairwise Correlations (194 < n < 197) [1] [2] [3] [4] [5] [6] [7] [8] [9] TARGETABILITY [1] 1.00 SPEED [2] 0.09 1.00 ABUSE RESISTANCE [3] -0.04 0.20* 1.00 RECOVERABILITY [4] -0.29* 0.07 0.29* 1.00 PREDICTABILITY [5] -0.09 0.05 0.54* 0.28* 1.00 REVERSIBILITY [6] 0.07 0.21* 0.34* 0.30* 0.20* 1.00 SCALABILITY [7] 0.05 0.16 0.35* 0.01 0.30* 0.30* 1.00 ADMINISTRATIVE COMPLEXITY [8] -0.30* -0.05 0.09 0.02 0.16 0.05 0.07 1.00 RESILIENCE TO HEALTH MEASURES [9] -0.16 0.04 -0.16 0.23* -0.17 -0.05 -0.11 -0.16 1.00 Source: Original calculations for this publication. > > > T A B L E A 6 - Policy-level Dimension Pairwise Correlations (1818 < n < 2127) [1] [2] [3] [4] [5] [6] [7] [8] [9] TARGETABILITY [1] 1.00 SPEED [2] 0.02 1.00 ABUSE RESISTANCE [3] -0.08* 0.08* 1.00 RECOVERABILITY [4] -0.31* 0.07* 0.48* 1.00 PREDICTABILITY [5] -0.16* 0.03 0.59* 0.56* 1.00 REVERSIBILITY [6] 0.00 0.08* 0.22* 0.20* 0.17* 1.00 SCALABILITY [7] -0.05 -0.02 0.27* 0.13* 0.24* 0.35* 1.00 ADMINISTRATIVE COMPLEXITY [8] -0.36* -0.01 0.21* 0.24* 0.20* -0.03 0.03 1.00 RESILIENCE TO HEALTH MEASURES [9] -0.05 0.13* -0.05 0.09* -0.06* 0.00 -0.05 -0.04 1.00 Source: Original calculations for this publication. 48 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >>> APPENDIX E: LOGISTIC variables were automatically dropped in these regressions R E G R E S S I O N R E S U LT S because they did not explain any variation (most policies met the criteria of these two dimensions). Thus, sample sizes for The following table (A7) show the results of logistic the Speed and Resilience to Health Measures regressions are regressions attempting to disaggregate the marginal effects too small to be robust. of policy choice and policy chooser. That is to ask, to what extent are policy dimension scores determined by the type >>> DATA NOTE of policy a country favored (which often inherently perform well or poorly along policy dimensions) and to what extent All medical sector expenditure policies were removed are scores determined by design tendencies common among from consideration. Their scores are not included in country country groups? average scores, they are not considered in the regression results below, nor are their scores included in the comparison >>> PRESENTATION OF RESULTS “base case” group when considering the performance of other policies in isolation. The reason for this is that these policies Conventionally, regression results are presented with one are almost always necessary responses to the pandemic, regression per column in a table. In this annex, in order and do not make sense to be included in an assessment of to make the presentation concise, each cell of the table choices a country made as a part of its response. Furthermore, represents the results of a unique regression, with the cells in these policies tend to perform poorly against our assessment each column sharing the policy dimension featured at the top framework, strictly speaking, since they are expensive, have of that column as the dependent variable. Most regressions costs that are difficult to predict (as the pandemic’s impact control for the same factors but feature a different isolated on the health sector is difficult to predict), slow to take effect, variable in order to test its performance relative to the entire and prone to leakages. This means that countries with great rest of the sample (not relative to a base case that may not be health sector needs would score more poorly in their general relevant for each category). The IDA/Blend variable’s results fiscal response, which may not accurately reflect the care with come from regressions that control for policy choice (fixed which they supported businesses and households. The results effects created using dummy variables for each policy) and associated with country groups below are limited in scope to regional dummies, and show the performance of IDA/Blend policies over which governments had control. countries compared to a base case of IBRDs and non-IDA/ IBRDs. The results in the rows of regional variables each As a robustness test conducted but not featured here, come from a separate regression in which there is a control medical sector expenditure measures (225 measures) were for the log of GDP per capita and policy choice (fixed effects included in logistic regression analysis. Significant tendencies for each policy); in each separate regression, the single among country groups (for example, IDA/Blend countries’ regional variable captures the intangible quality associated tendency to have more administratively simple measures than with policies that come from the given region, as compared those of IBRDs and non-IDA/IBRDs) were the same as those with the rest of the world (the default, as no other regional reported below, although coefficients varied a trivial amount. dummies are included). This means the standard is slightly different for each region (the definition of the “rest of the >>> INTERPRETATION world” depends on the region being excluded), but results are intuitive to interpret (example: countries in the Middle East The results presented in each cell are marginal effects, which and North Africa are about 17 percent less likely to target represent the increased likelihood of success with respect to their policies than the rest of the world, controlling for policy the given policy dimension attributable to the variable indicated choice and GDP per capita). Finally, the results for each policy at the head of the row. For example, the “IDA” variable shows a variable come from separate regressions that control for all value of -0.0177 for “Abuse Resistance”; this is not statistically regional fixed effects and the log of GDP per capita in order to significant, which means that, holding policy choice constant, assess whether a policy is generally associated with a score IDA/Blend countries are not generally better or worse than of 1 (indicating that it meets the criteria of the given policy IBRDs and non-IDA/IBRDs at designing abuse-resistant dimension), after accounting for the marginal effects of GDP policies. Compare this result with an ANOVA test comparing per capita and regional tendencies. Results for Speed and the abuse resistance of lending categories, which resulted in Resilience to Health Measures are presented, although many a statistically different score for IDA/Blends (a lower score) A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 49 with a P-value below 0.05. This result is not contradictory. IDA/Blends did perform worse on abuse resistance but did so because they chose policies inherently prone to abuse, not because IDA/Blends are worse at policy design with respect to abuse resistance. For the “IDA” variable’s coefficient under “Predictability and Cost Control,” the value is 0.137***, which indicates that IDA/Blends are about 13.7 percent more likely than IBRDs and non-IDA/IBRDs to score well on predictability and cost control, holding region and policy choice constant. > > > T A B L E A 7 - Logistic Regression Results To save space in the regression results table, the following policy categories have been given an abbreviated tag. TAG MEANING R FOR B Revenue measures to protect businesses R FOR I Revenue measures to protect individuals MEDICAL ITEMS Revenue measures to promote availability of medical items CONSUMPTION Revenue measures to boost consumption / demand E FOR I Expenditure measures for cash transfers to individuals E FOR B Expenditure measures to protect businesses CREDIT/EQUITY Credit and equity measures OTHER Measures which do not fit into any of the above categories 50 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 51 Source: GDP per capita controls are from the WDI (2020); IDA eligibility according to IDA19; Measures collected from the IMF’s Policy Tracker (2020), the OECD’s policy tracker (2020), IBDF (2020), and the Doing Business policy tracker database (2020). Note: Numbers of observations are averages of all regressions in the column. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 52 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 >>> APPENDIX F: ASSESSMENT OF OPTIONS FOR FISCAL POLICY MEASURES > > > T A B L E A 8 - Assessment of Options for Fiscal Policy Measures (Originally from Fiscal Impact and Policy Response to COVID-19) A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 53 54 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 55 56 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19 <<< 57 Source: World Bank. 16. This measure was later reclassified as an expenditure measure for businesses. 58 >>> A REVIEW OF FISCAL POLICY RESPONSES TO COVID-19