The Office of the Chief Economist of the South Asia Region OCTOBER 2023 South Asia Development Update Toward faster, cleaner growth South Asia Development Update OCTOBER 2023 South Asia Development Update © 2023 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 1 2 3 4 26 25 24 23 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. 2023. Toward Faster, Cleaner Growth. South Asia Development Update (October 2023). World Bank, Washington, DC. doi: 10.1596/978-1-4648-2026-7. 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 re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use 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. ISBN (electronic): 978-1-4648-2026-7 DOI: 10.1596/978-1-4648-2026-7 Cover design: David Spours (Cucumber Design). The cutoff date for the data used in the report was September 25, 2023. Summary of Contents Acknowledgments ..................................................................................................................................... xiii Foreword ................................................................................................................................................... xv Executive summary .................................................................................................................................. xvii Abbreviations ............................................................................................................................................ xix Chapter 1 Regional outlook: Solid progress, but a long way to go................................................. 1 Box 1.1 Fiscal deteriorations around elections ......................................................... 21 Spotlight An ounce of prevention, a pound of cure: Averting and dealing with sovereign debt default ......................................................................................................................... 35 Box SL.1 Literature review: Costs of sovereign debt default ................................... 42 Chapter 2 Recruiting firms for the energy transition ................................................................... 57 Box 2.1 Literature review: Addressing barriers to technology diffusion in firms....... 70 Chapter 3 Stranded jobs? The energy transition in South Asia’s labor markets ........................... 95 Selected Topics ........................................................................................................................................ 125 v Contents Acknowledgments..................................................................................................................................... xiii Foreword ................................................................................................................................................... xv Executive Summary ................................................................................................................................. xvii Abbreviations ............................................................................................................................................ xix Chapter 1 Solid progress, but a long way to go............................................................................. 1 Introduction ................................................................................................................. 3 Economic activity ......................................................................................................... 5 Global developments .............................................................................................. 5 Regional developments........................................................................................... 5 Inflation ....................................................................................................................... 7 Global developments .............................................................................................. 7 Regional developments........................................................................................... 7 Financial conditions ..................................................................................................... 8 Global developments .............................................................................................. 8 Regional developments........................................................................................... 9 Outlook...................................................................................................................... 10 Country developments ......................................................................................... 11 Risks and vulnerabilities ............................................................................................. 13 Financial crises ..................................................................................................... 13 Slowdown in China ............................................................................................. 14 Climate change-related disasters ........................................................................... 15 Policy challenges ......................................................................................................... 16 Strengthening investment..................................................................................... 16 Removing trade and foreign currency restrictions ................................................. 18 Improving fiscal positions..................................................................................... 19 Managing the energy transition ............................................................................ 27 Annex 1.1.1 Methodology .......................................................................................... 28 References .................................................................................................................. 30 Spotlight An ounce of prevention, a pound of cure: Averting and dealing with sovereign debt default ........................................................................................................................ 35 Introduction .............................................................................................................. 37 The how, when, and where of sovereign debt default .................................................. 39 Features of successful debt defaults ............................................................................ 41 vii Spotlight Domestic debt: A costly mitigating factor ...................................................................45 Policy implications ......................................................................................................47 Annex SL.1 Regression analysis: Methodology and data ..............................................48 Annex SL.2 Event study: Methodology ......................................................................53 References ...................................................................................................................54 Chapter 2 Recruiting firms for the energy transition .................................................................57 Introduction ............................................................................................................... 59 Contributions to the literature.............................................................................. 60 Main findings ....................................................................................................... 61 Evolution of South Asia’s energy consumption ........................................................... 62 South Asia in global comparison ........................................................................... 62 Growth versus energy intensity ............................................................................. 62 Sectoral shifts versus firm-level energy intensity .................................................... 63 Case study: India .................................................................................................. 63 Energy-efficient technology adoption by firms ............................................................ 65 Cross country evidence ......................................................................................... 65 Case study: Bangladesh......................................................................................... 67 Policy implications...................................................................................................... 69 Regulation, taxes, and subsidies ............................................................................ 72 Information and behavioral nudges ...................................................................... 74 Access to finance, markets, and public services ..................................................... 75 It can be done; it has been done .......................................................................... 76 Annex 2.1 Methodology ............................................................................................. 77 References ................................................................................................................... 89 Chapter 3 Stranded jobs? The energy transition in South Asia’s labor markets .......................... 95 Introduction ............................................................................................................... 97 Contributions to the literature.............................................................................. 99 Main findings..................................................................................................... 100 Labor market effects of the energy transition............................................................. 101 Green and pollution-intensive jobs ..................................................................... 101 Characteristics of workers in green jobs .............................................................. 101 Characteristics of workers in pollution-intensive jobs ......................................... 103 Lessons from past structural transformations ............................................................. 103 Resource booms and busts .................................................................................. 103 Literature sources ............................................................................................... 104 viii Chapter 3 Subnational regions with rapid growth in green jobs: Past resource booms ......... 104 Subnational regions reliant on pollution-intensive employment: Past resource busts .............................................................................................. 105 Large-scale economic transformations................................................................. 106 Policy implications.................................................................................................... 106 Annex 3.1 Methodology: Quantifying green and pollution-intensive jobs ................. 108 Annex 3.2 Methodology: Meta regressions ................................................................ 115 References ................................................................................................................. 119 Selected Topics ........................................................................................................................................ 125 Boxes 1.1 Fiscal deteriorations around elections ......................................................... 21 SL.1 Literature review: Costs of sovereign debt default ....................................... 42 2.1 Literature review: Addressing barriers to technology diffusion in firms ...... 70 Figures 1.1 Overview ...................................................................................................... 4 1.2 Economic activity ......................................................................................... 6 1.3 Inflation ....................................................................................................... 7 1.4 Financial conditions ..................................................................................... 9 1.5 Outlook for output growth......................................................................... 10 1.6 Financial risks ............................................................................................. 13 1.7 Scenario: Sharper economic slowdown in China ········································· 14 1.8 Climate risks............................................................................................... 15 1.9 Investment weakness .................................................................................. 17 1.10 Restrictions on trade and foreign exchange transactions .............................. 18 1.11 Fiscal challenges.......................................................................................... 20 B1.1.1 Fiscal positions around elections in South Asia ........................................... 23 B1.1.2 Political budget cycles in EMDEs ............................................................... 24 B1.1.3 Political budget cycles in EMDE regions .................................................... 25 1.12 Managing the energy transition .................................................................. 27 SL.1 Government debt ....................................................................................... 38 SL.2 The how, when, and where of sovereign debt defaults................................. 39 SL.3 Features of successful debt defaults ............................................................ 41 SL.4 Debt and borrowing costs after successful and unsuccessful debt defaults.... 44 SL.5 Government debt composition and features of debt booms ........................ 45 SL.6 Costs of domestic debt................................................................................ 46 ix Figures SL.7 Prospects for GDP growth and government revenues .................................47 2.1 South Asia’s contribution to global emissions .............................................60 2.2 Air pollution in South Asia ........................................................................61 2.3 Energy intensity in South Asia ...................................................................62 2.4 Firm-level energy intensity ..........................................................................63 2.5 India: Within-firm reductions in energy intensity .......................................64 2.6 India: Energy intensity cuts and employment growth .................................65 2.7 Energy-efficient technology adoption by firms: Cross country evidence .....66 2.8 Unreliability of grid power and use of electricity generators ........................67 2.9 Energy-efficient technology adoption by firms: Randomized control trial ...68 2.10 Information spillovers among firms ............................................................69 2.11 Literature on policies for firm technology adoption and energy efficiency gains ............................................................................................71 2.12 Subsidies and effective carbon prices ...........................................................73 2.13 Historical comparison .................................................................................76 3.1 South Asia: GHG emissions and policy commitments ................................98 3.2 Renewable energy potential in South Asia ...................................................98 3.3 Economic activity .......................................................................................99 3.4 Labor productivity and informal employment ...........................................100 3.5 Green and pollution-intensive jobs in South Asia ......................................101 3.6 Regional distribution of jobs in South Asia ...............................................102 3.7 Green and pollution-intensive jobs in South Asia: Worker characteristics ..............................................................................102 3.8 Green and pollution-intensive jobs in South Asia: Wage premiums and discounts ............................................................................................103 3.9 Meta regression of the effects of resource booms and busts on employment and earnings ....................................................................................................105 x Tables 1.1 Growth in South Asia ................................................................................. 11 1.1.1 Election effects ........................................................................................... 28 1.1.2 Election timing effects in EMDEs .............................................................. 29 1.1.3 Election effects ........................................................................................... 29 SL.1.1 Countries and default years ........................................................................ 49 SL.1.2 Marginal probability of default ................................................................... 50 SL.1.3 Share of successful defaults ......................................................................... 51 SL.1.4 Fiscal outcomes after default....................................................................... 52 2.1.3.1 World Bank Enterprise Surveys coverage of South Asia and other EMDEs 79 2.1.3.2 World Bank Enterprise Surveys: Change in mean firm level energy intensity within sectors ............................................................................................ 82 2.1.4.1 India: Within-firm trends in energy intensity ............................................. 83 2.1.4.2 India: Within-firm trends in energy intensity, controlling for relative prices84 2.1.4.3 India: Within-firm trends in energy intensity, measuring energy intensity in percent of total sales ................................................................................... 85 2.1.4.4 India: Employment and energy intensity .................................................... 85 2.1.5.1 Average energy efficient technology usage rates: Country details ................. 86 2.1.5.2 Correlates of energy efficient technology adoption index ............................ 87 2.1.5.3 Power outages and generator use ................................................................ 88 3.1.1 Sample description ................................................................................... 109 3.1.2 Share of workers in green and pollution-intensive jobs in the main sectors 111 3.1.3 The marginal probability of being employed in a green job ...................... 111 3.1.4 The marginal probability of being employed in a pollution-intensive job.. 112 3.1.5 Earnings in green and pollution-intensive jobs ......................................... 113 3.1.6 Decomposition of earnings differential between workers in green jobs and the average worker.......................................................................................... 114 3.2.1 Sample description ................................................................................... 116 3.2.2 Studies of the effects of natural resource booms and busts on the labor market ............................................................................................. 117 3.2.3 Meta regression results.............................................................................. 118 xi Acknowledgments This report is a product of the Office of the Chief Economist for the South Asia Region (SARCE). The report was managed by Franziska Ohnsorge (Chief Economist, South Asia Region), under the general guidance of Martin Raiser (Regional Vice President, South Asia Region). Chapter 1 was written by Patrick Kirby with of Technology surveys. The chapter uses inputs contributions from Zoe Leiyu Xie. Nikita from a paper authored by Ritam Chaurey (Johns Perevalov (DECPG) provided scenario analysis. Hopkins University), Gaurav Nayyar (EFIAT), Siddharth Sharma, and Eric Verhoogen Colleagues from MTI provided country forecasts (Columbia University). The chapter was reviewed and other inputs to the country analysis in by Christopher Towe, Graham Hacche, and Chapter 1, including Kishan Abeygunawardana Ulrich Wagner (University of Mannheim). (Maldives and Sri Lanka), Erdem Atas (Maldives), Kanika Bhatnagar (India), Alice Joan Brooks Chapter 3 was written by Margaret Triyana. (Nepal), Rishabh Choudhary (India), Derek Inputs were received from Daniel Garotte- Hung Chiat Chen (Pakistan), Souleymane Sanchez (HECSP), Claire Yi Li (IMF), Mattia Coulibaly (ESADR), Aroub Farooq (Pakistan), Makovec (HLCSP), Maya Sherpa (HSAED), and Rangeet Ghosh (Bangladesh), Tobias Akhtar Emmanuel Vazquez (HSAED). The chapter was Haque (ESADR), Bernard Haven (Bangladesh), reviewed by Patrick Behrer (DEC), Thomas Nayan Krishna Joshi (Nepal), Nazmus Sadat Farole (SCADR), Graham Hacche, Monica Yanez Khan (Bangladesh), Aurelien Kruse (India), Shruti Pagans (HSAED), Josefina Posadas (HLCSP), and Lakhtakia (Maldives and Sri Lanka), Ran Li Christopher Towe. (India), Tanvir Malik (India), Peter Mousley (Nepal), Sayed Murtaza Muzaffari (Pakistan), Research assistance was provided by Ijaz Ahmed, Eduardo Olaberria (ESAC1), Dhruv Sharma Zara Ali, Juan Felipe Serrano Ariza, Sara Brolhato (India), Melanie Simone Trost (Bhutan), De Oliveira (ETIMT), Sarur Chaudhary, Muhammad Waheed (Afghanistan), and Richard Abhishek Deshwal, Mohammad Shah Naoaj, Walker (Maldives and Sri Lanka). Christopher Michael Norton, Bomi Okuyiga, Vanessa Towe, Graham Hacche, and William Shaw Olakemi Dovonou, Rully Prassetya, Utkarsh reviewed the chapter. Saxena, and Xiao’ou Zhu. Box 1.1 was written by Franziska Ohnsorge, Jakob Rana Al-Gazzaz facilitated the report’s de Haan (University of Groningen), and Shu Yu preparation, production, and dissemination. (EFIAT). It was reviewed by Christopher Towe Quinn John Sutton was responsible for the layout and Graham Hacche. and typesetting. David Spours (Cucumber Design) designed the graphics and layout. Peter The spotlight was written by Franziska Ohnsorge Milne and Graeme Littler copyedited the and Hayley Pallan (DECPG). It was reviewed by chapters. Elena Karaban, Diana Ya-Wai Chung, Christopher Towe, Graham Hacche, and Ugo Trishna Thapa, and Adnan Javaid Siddiqi (all Panizza (Graduate Institute of International and ECR) coordinated the dissemination. Ahmad Development Studies). Khalid Afridi provided administrative support. Chapter 2 was written by Siddharth Sharma with South Asia as used in this report includes contributions from Jonah Matthew Rexer. Xavier Afghanistan, Bangladesh, Bhutan, India, Cirera(ETIMT), Kyungmin Lee (ETIMT), and Maldives, Nepal, Pakistan, and Sri Lanka. e Santiago Reyes Ortega (CERER) contributed cutoff date for this report was September 25, inputs based on analysis of the Firm-level Adoption 2023. xiii Foreword At first glance, South Asia is a bright spot in the ability of the financial system to provide global economy. The World Bank is forecasting credit, and removing market distortions. that the region will grow more quickly than any other developing country region over the next few • Restore fiscal sustainability. Many countries in years. Some countries in the region are the region are carrying extra weight that continuing a trend of strong growth, while others makes this journey more difficult. Debt are recovering from a period of turmoil. While burdens around the world have gotten high inflation and interest rates have bogged heavier in the last decade, but the increase in down many emerging markets, South Asia seems South Asia is above average. High public debt to be forging ahead. crowds out private investment and limits the room for spending on critical infrastructure However, a second look reveals a more nuanced and human capital bottlenecks and on picture. The region is making progress, but at a improving resilience. Lightening this burden slower pace than in the pre-pandemic years. will require some combination of increased That’s an issue, because the region still has a long revenues, improved spending efficiency, and way to go. Per capita incomes average only about stronger fiscal rules to anchor better policies US$2,000—one-fifth of the level achieved by the over time. neighboring East Asia and Pacific region or the upper-middle income average, and one-twentieth • Speed the energy transition. South Asia has the level of high-income countries. Current been successful at adopting basic energy- growth rates may be higher than elsewhere, but saving technology but lags in the adoption of they are still not sufficient for countries in the more advanced technologies. Modernizing region to reach high-income status within a the economy and increasing energy efficiency generation. Moreover, not all countries in the will help the region keep pace in the global region are growing fast, and three—Afghanistan, energy transition. Pakistan and Sri Lanka—are in acute crisis. • Maintain a healthy labor market. The energy The region’s progress is akin to that of transition offers many new employment mountaineers at the foothills of the Himalayas. opportunities but risks leaving behind lower- Some have barely left the base camp. Others are skilled, informal workers that have pollution- moving at a brisk pace but still in low altitude. All intensive jobs. Workers have a better chance still have a long way to go. And the path will get of moving across sectors when they have more difficult ahead. access to education, training, finance, and markets. A robust labor market with strong This report provides a roadmap that policy social safety nets also make the path to new makers can use to hasten their way towards their work easier for displaced workers. goal. There are four ingredients that are particularly important: The path to prosperity requires high growth rates to be sustained over long periods of time. Many • Boost private investment. Strong private countries have found their aspirations curtailed as investment is critical for accelerating the pace growth faltered soon after an initial take-off. With of catch-up with high-income countries and the right policies and investments, South Asia can enabling the energy transition. In all but one avoid this fate and reach the summit to create country in South Asia, private investment sustainable livelihoods on a livable planet for its growth has slowed compared to the pre- people. pandemic period. Strengthening private investment will depend on many factors, Martin Raiser including improving infrastructure, the Vice President, South Asia Region institutional and business environment, the xv Executive Summary At just under 6 percent, South Asia is expected to grow faster than any other emerging market and developing economy (EMDE) region in 2024–25. However, for all countries, this will represent a slowdown from pre- pandemic averages. Several potential adverse events could derail this outlook, including risks related to fragile fiscal positions. Government debt in South Asia averaged 86 percent of GDP in 2022, above that of any other EMDE region. In some countries, outright defaults have short-circuited growth while, in others, increasing domestic borrowing by governments has driven up interest rates and diverted credit away from the private sector. Elections could add to spending pressures. An urgent policy priority for the region is, therefore, to manage and reduce fiscal risks. Over the longer term, the policy priority is to accelerate growth and job creation in a sustainable manner. The energy transition, away from fossil fuels toward sustainable sources of energy, presents an opportunity for the region to lift productivity, cut pollution, reduce its reliance on fuel imports, and create jobs. South Asia uses twice as much energy to produce each unit of output as the global average and the region lags in the adoption of advanced energy-efficient technologies. Even fiscally constrained governments can take action to support the energy transition with market-based regulations, information campaigns, broader access to finance, and reliable public power grids. With about 9 percent of the region’s workers employed in pollution-intensive activities, and these workers less educated and more often informally employed than the average worker, the energy transition will create challenging labor market shifts. This calls for measures to boost job creation and facilitate worker mobility, geographically and across sectors. Chapter 1. Regional outlook: Solid progress, but government expenditures, and government wage a long way to go. At just under 6 percent, output bills have tended to rise significantly around growth in South Asia is expected to remain election years. While primary spending increases stronger than in other regions in 2023–25, even have tended to be partially reversed in the with weak growth in the countries recovering following year, post-election reversals of primary from recent balance-of-payments crises. Foreign deficit and government wage bill increases have exchange and financial markets in these countries been more variable and at best partial. The have stabilized, in part owing to the introduction consequent ratcheting up of primary deficits of IMF-supported policy programs. But the around elections in EMDEs can erode fiscal financial systems of many countries in the region sustainability over the longer term, while the remain vulnerable and fiscal positions remain expansion of government wage bills can result in fragile. In some cases, restrictions on imports and spending rigidities. In South Asia, in particular, foreign exchange transactions have yet to be fully fiscal positions have tended to deteriorate around unwound. The outlook is subject to downside national elections, and, in some cases, there is also risks from weak financial systems and fiscal evidence of targeted fiscal actions around positions. Growth prospects would also worsen in subnational elections. While this result is true on the event of a further economic slowdown in average for the region, some countries—notably China or climate change-related natural disasters. India in its 2023 budget—have avoided the risk In the short term, policy priorities include of election-induced current spending increases. preserving financial stability and improving fiscal This points to a way forward for fiscally sustainability. In the longer term, it is important constrained governments in South Asia. to boost private investment growth, make economies more open to trade, and seize the Spotlight. An ounce of prevention, a pound of opportunities offered by the global energy cure: Averting and dealing with debt default. transition. South Asia has the highest average government debt-to-GDP ratio among EMDE regions, at 86 Box 1.1 Fiscal deteriorations around elections. percent in 2022, and four of the region’s Among EMDEs, primary fiscal deficits, primary countries are rated in or near sovereign debt xvii distress. The risk of sovereign defaults in the adopters of basic energy-saving technologies, they region is heightened not only by high levels of have lagged in the adoption of more advanced government debt but also by the increases in technologies, with smaller firms lagging global interest rates over the past two years: the particularly far behind. Policies that have been vast majority of past defaults occurred around the effective at encouraging technology adoption end of U.S. monetary policy tightening cycles among firms include market-based regulation, and in countries with above-median government dissemination of accurate information on energy debt-to-GDP ratios. Past experience also shows savings, and financial support. that more than one-third of defaults failed to lower government debt or borrowing costs in a Chapter 3. Stranded jobs? The energy transition lasting manner. Defaults that succeeded in in South Asia’s labor markets. The transition lowering debt or borrowing cost were away from fossil fuels in South Asia will have accompanied more frequently than others by significant labor market impacts, which could above-median debt restructurings, growth leave many workers stranded in lower-wage jobs accelerations, and fiscal consolidations. South in declining industries. In all South Asian Asia’s above-average economic growth mitigates countries except India, pollution-intensive jobs some of the default risks. Some South Asian outnumber green jobs and account for 6–11 countries have reduced their default risk by percent of all jobs; only in India do green jobs predominantly borrowing from domestic outnumber—and then only slightly—pollution- creditors. However, this strategy comes at a price: intensive jobs, which account for 9 percent of all high domestic shares of government debt have jobs. Pollution-intensive jobs are concentrated been associated with higher borrowing costs and among lower-skilled and informal workers, lower bank credit to the private sector. With the whereas green jobs tend to be held by higher- external environment likely to remain challenging skilled, better-paid, and formal-sector workers. over the next several years, it is all the more Experience from past major economic important to adopt policies to accelerate transformations, especially in resource sectors, sustainable growth and shore up fiscal positions. suggests that the transition away from fossil fuels will have large effects on the structure of Chapter 2. Recruiting firms for the energy employment and earnings, with lasting losses for transition. As the world presses ahead with the some workers, and will cause significant internal energy transition, new energy-saving technologies worker migration. A wide range of policies will be offer South Asian countries an opportunity to needed to facilitate the necessary adjustment in modernize their economies. Currently, the energy labor markets while protecting vulnerable intensity of South Asian economies is almost workers. These include: the provision of better twice the global average—despite a decline over access to high-quality education and training, the past two decades that was almost entirely finance, and markets; measures to facilitate labor driven by firm-level, within-sector cuts in energy mobility; and strengthening social safety nets. intensity. While the region’s firms have been early xviii Abbreviations AE advanced economy CO2 carbon dioxide CPI consumer price index EAP East Asia and Pacific ECA Europe and Central Asia EMDE emerging market and developing economy EU European Union FDI foreign direct investment FY fiscal year G20 Group of Twenty: Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, Republic of Korea, Mexico, Russia, Saudi Arabia, South Africa, Türkiye, the United Kingdom, the United States, and the European Union GDP gross domestic product GEP Global Economic Prospects GHG greenhouse gas IMF International Monetary Fund LAC Latin America and the Caribbean LIC low-income country MNA Middle East and North Africa NBFIs non-bank financial institutions OECD Organization for Economic Co-operation and Development OPEC Organization of the Petroleum Exporting Countries OPEC+ OPEC and Azerbaijan, Bahrain, Brunei Darussalam, Kazakhstan, Malaysia, Mexico, Oman, the Russian Federation, South Sudan, and Sudan PM2.5 particulate matter 2.5 PMI Purchasing Managers’ Index PPP purchasing power parity RHS right-hand scale SAR South Asia SOE state-owned enterprise SSA Sub-Saharan Africa TFP total factor productivity toe tons of oil equivalent VAR vector autoregression WDI World Development Indicators xix CHAPTER 1 Solid progress, but a long way to go SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 3 Chapter 1. Solid progress, but a long way to go At just under 6 percent, output growth in South Asia is expected to remain stronger than in other regions in 2023–25, even with weak growth in the countries recovering from recent balance-of-payments crises. Foreign exchange and financial markets in these countries have stabilized, in part owing to the introduction of IMF- supported policy programs. But the financial systems of many countries in the region remain vulnerable and fiscal positions remain fragile. In some cases, restrictions on imports and foreign exchange transactions have yet to be fully unwound. The outlook is subject to downside risks from weak financial systems and fiscal positions. Growth prospects would also worsen in the event of a further economic slowdown in China or climate change- related natural disasters. In the short term, policy priorities include preserving financial stability and improving fiscal sustainability. In the longer term, it is important to boost private investment growth, make economies more open to trade, and seize the opportunities offered by the global energy transition. Introduction South Asia’s output growth is forecast to remain broadly steady in 2023–25, slowing from 8.2 Global output growth continues to slow, but this percent in 2022 to 5.8 percent in 2023 and 5.6 has had only limited spillovers to South Asia, thus percent in 2024 and 2025 (figure 1.1). In all far. At just under 6 percent, growth in the region South Asian countries, projected growth will remains stronger than in any other EMDE region, remain below the 2015–19 pre-pandemic average, supported by faster potential output, resilient with the fading of post-pandemic rebounds exports, and increasing remittance inflows. Most accentuated by combinations of monetary countries are making solid progress, with the tightening, fiscal consolidation, and slowing global exception of a few countries recovering from demand growth. Projected growth will also be recent balance-of-payments crises. insufficient to return output to the path projected before the pandemic. Finally, current growth rates Headline inflation has been declining in the rest of are not strong enough for most countries to reach the world, but has remained elevated in South high-income thresholds within a generation. Asia. Food inflation in the region remains particularly high owing to both high global food Relative to the spring forecast, growth in 2023 has inflation and local supply disruptions. As been upgraded by 0.2 percentage points due to currencies have stabilized and some import stronger-than-expected data in India. The 0.3 controls have been relaxed, inflation is expected to percentage-point downgrade to the 2024 growth trend down throughout the region. This trend projection reflects weaker prospects for could be interrupted in a variety of ways, however, Bangladesh and Pakistan as both countries including by further commodity price increases, struggle to emerge from balance-of-payments exchange rate depreciations, or more persistent difficulties. second-round effects of past inflation pressures A number of downside risks could derail growth than currently anticipated. from the path projected in the baseline. A Financial systems in many countries are fragile, deterioration in market sentiment could re-ignite with limited capital buffers, high exposure to pressures on currencies, triggering renewed capital heavily indebted sovereigns, and high levels of outflows, currency depreciations, rebounds in nonperforming loans. Global developments have inflation, and further increases in borrowing costs. added pressure as major central banks have This risk is particularly elevated in countries with continued to raise policy rates to reduce inflation. fragile financial systems and limited foreign This has added to depreciation pressures and exchange reserves. borrowing cost increases across the region. The region would also be affected by any further slowdown in China’s economic growth, though by somewhat less than other parts of the world. In Note: This chapter was prepared by Patrick Kirby, with contributions from Zoe Xie. addition, South Asia has become increasingly 4 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.1 Overview are severely constrained and a global energy Output growth in South Asia is projected to remain stronger than in other transition is underway. This underscores the need regions. A sustained acceleration would require stronger private for policies to boost private investment growth by investment growth, which has been weak in most countries. The scope for government support is limited owing to high debt and still large deficits, an expanding access to finance, improving business important source of which is weak revenue collection. Addressing these climates, ensuring that government support is well shortcomings would free up resources to fund the region’s development targeted, and increasing competition. Greater priorities and support the growing number of people affected by natural disasters. openness to foreign direct investment (FDI) and international trade would facilitate access to A. Output growth in South Asia B. Real investment growth technology. It would also attract investment as Percent advanced economies pivot toward more diversified Percent 10 Other EMDE regions SAR 9 2018-22 Avg. EMDEs supply chains. 8 6 3 Countries in South Asia have had persistently large 6 0 fiscal deficits. Growing government debt burdens 4 -3 have become more costly to service as interest rates 2 -6 have risen. There is an urgent need for 0 -9 governments to reduce their borrowing 2022 2023e 2024f BGD MDV IND NPL PAK BTN LKA requirements and the risks of debt default by C. Government revenues, 2020-22 D. Number of people affected by natu- strengthening government revenue collection and average ral disasters per year, 2013-22 improving spending efficiency. This can include Percent of GDP 35 EMDE average Million people 70 Total affected Percent 4 broadening tax bases, reducing subsidies on fossil 30 60 Total share of population affected (RHS) fuels, and adhering to fiscal rules. 25 50 3 20 40 2 Chronically weak private investment growth and 15 30 limited fiscal space risk delaying the adoption of 10 20 5 10 1 technologies needed for the region to keep pace 0 0 0 with the global energy transition. The region’s LKA BGD PAK IND NPL MDV BTN SAR EAP SSA LAC MNA ECA energy intensity of output is twice the global Sources: International Disaster Database (EM-DAT); WDI (database); WEO (database); World Bank average. The energy transition presents an (Macro Poverty Outlook). opportunity to upgrade technologies and boost Note: (e) = estimate; (f) = forecast; Avg.=Average; BGD = Bangladesh; BTN = Bhutan; EAP = East Asia and Pacific; ECA = Europe and Central Asia; EMDEs = emerging market and developing productivity, cut pollution, reduce reliance on economies; IND = India; LAC = Latin America and the Caribbean; LKA = Sri Lanka; MDV = Maldives; MNA = Middle East and North Africa; NPL = Nepal; PAK = Pakistan; RHS = right hand side; SAR = energy imports, and increase employment. South Asia; SSA = Sub-Saharan Africa. A. Blue bars reflect the range of growth across all the other EMDE regions. Regional aggregate computed using 2015 GDP as weights. Sample includes 7 countries in SAR and 136 in other EMDE At the same time, the transition to green regions. B. Figure shows the annual average growth of total real gross fixed investment (in local currency), technologies will have important labor market over 2018-2022. EMDEs aggregate computed using 2015 GDP as weights. Sample includes 123 consequences. In most countries in the region, a EMDEs (13 in EAP, 21 in ECA, 21 in LAC, 15 in MNA, 7 in SAR, and 46 in SSA). C. EMDE average computed using 2015 GDP as weights. Bars show 2020-22 averages of larger share of workers is employed in pollution- government revenue. D. Bars show the total population affected by natural disasters, and diamonds show the share of total intensive jobs than in green jobs. The adoption of population affected, annual averages over 2013-2022. Sample includes 144 EMDEs (22 in EAP, 20 in ECA, 31 in LAC, 18 in MNA, 8 in SAR, and 45 in SSA). green technologies will disproportionately favor better-educated workers in the formal sector. About 9 percent of the region’s workers are vulnerable to natural disasters. These can have employed in pollution-intensive jobs and they substantial near-term economic and human tend to be lower-skilled and informal. Policies to impacts, and can also inflict lasting damage to encourage the use of more energy-efficient productivity, especially in the agriculture sector, technologies will need to be coupled with and food security. measures to boost overall job creation. A strong labor market can absorb workers who exit In almost all countries in the region, private pollution-intensive activities and facilitate a investment growth has weakened from its pre- smooth labor market adjustment for vulnerable pandemic pace. The slowdown in private workers. investment comes at a time when public finances SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 5 Economic activity global services activity, while cooling, has remained robust. The global manufacturing PMI Output growth in South Asia in 2023–24 is has been pointing to contraction since October projected to remain stronger than in other EMDE 2022, while the corresponding services index has regions. Nonetheless, it remains below pre- been at levels consistent with solid growth for pandemic (2015–19) averages, and several most of this year (figure 1.2). The combination of countries are suffering from the aftermath of slowing global growth and a shift away from the recent currency crises. more import-intensive manufacturing sector means that external demand is providing little Global developments support to activity in most countries. Global economic growth is projected to slow Regional developments further in 2023 and to stabilize in 2024. In the short term, most countries are grappling with South Asia faces many of the same economic continued high inflation and the effects of challenges as other regions, including elevated monetary policy tightening. In the longer term, inflation, higher interest rates, the need for fiscal output growth is projected to continue slowing in consolidation, and weak external demand. These many countries, reflecting weakening growth of are mitigated, however, by improving remittance the labor force, productivity, or investment, or all inflows and tourist arrivals since the pandemic, three (Kilic Celik et al. 2023). and by the region’s solid potential growth rate (Kose and Ohnsorge 2023). South Asia has also In the United States, there has been an extended been less affected by slowing global growth than period of robust expansion since late 2020, with other regions, with export growth remaining rapid employment growth and low relatively resilient. This may be due to differences unemployment. This is now slowing as excess in the composition of South Asia’s exports, which savings accumulated during the pandemic have tend to be more service-oriented. It may also be largely been spent. Credit conditions have due to the region’s below-average integration into tightened along with monetary policy and also as a the global economy as a result of limited transport result of the banking sector turmoil earlier this connectivity and restrictions on trade and foreign year. currency transactions (World Bank 2016). The euro area continues to struggle with above- In India, robust output growth in the first half of target inflation and is expected to face a steep 2023 was supported by a strong expansion of slowdown in growth. Confidence indicators point investment and, on a sectoral level, continued to persistent weakness and an increased risk of strength of services. Government infrastructure recession in the next few quarters. projects have supported momentum in the In China, the rebound following the post- construction sector, which has grown at year-over- pandemic economic re-opening appears to have year rates of around 10 percent in recent quarters. quickly faded. Continued fragilities in the Export growth has benefited from strong exports property sector are having widespread spillovers to of services, such as those related to information the rest of the economy, contributing to the technology and consulting, which have been little emergence of deflation in recent months. affected by the slowdown in global growth. India’s Consumer spending has been relatively buoyant services Purchasing Managers Index (PMI) this year, but this has been offset by weakness in reached 62.3 in August, nearly 10 points above the exports and investment. The government has thus global index. Employment indicators have been far avoided implementing broad stimulus weaker, however, suggesting that with appropriate measures in favor of allowing the overheated real policies the country’s economic growth could estate sector to cool. deliver more robust job creation. The weakness of global growth is particularly Activity in Bangladesh, Pakistan, and Sri Lanka has pronounced in the manufacturing sector, whereas continued to suffer from the aftermaths of recent 6 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.2 Economic activity balance-of-payments crises. All three countries Global output growth is slowing and shifting toward less trade-intensive have recently begun to implement IMF-supported services activities. South Asia is weathering this slowdown better than policy programs to stem capital outflows and other EMDE regions, supported by solid potential output growth, resilient export growth, and strong remittance inflows. In recent years, fiscal deficits improve debt sustainability. Activity in all three widened more in South Asia than in other regions. Now, however, fiscal cases has continued to be hampered by input consolidation as well as monetary tightening are contributing to the growth shortages related partly to higher import costs and slowdown. supply disruptions associated with remaining import restrictions. In all three countries, fiscal A. Global manufacturing and services B. Actual and potential output growth PMIs deficits remain large, while current account Index, 50+=expansion 60 Percent 7 Actual growth, 2023 deficits have improved amid sharp import Manufacturing activity Services business activity 6 Potential growth, 2022-30 average compressions. 55 5 50 4 Sri Lanka’s economy has suffered the most severe 3 2 contraction but appears to be past the worst of its 45 1 crisis, with shortages of essential inputs easing and 40 Jan Apr Jul Oct Jan Apr Aug 0 SAR EAP EMDE SSA MNA LAC ECA tourism recovering. The services PMI has been in 2022 2022 2022 2022 2023 2023 2023 expansionary territory since May 2023. Industrial production has been contracting since late 2021, C. Goods and services export growth, 2022 D. Change in average fiscal balance from 2017-19 to 2020-22 but more slowly recently. Percent 15 Percent of GDP 2 Pakistan’s economic situation is also fragile. The Primary balance Non-primary balance 12 1 U.S. dollar value of goods imports shrank by 26 9 0 percent in the year to August 2023 as a result of 6 -1 low demand alongside import and capital controls. -2 3 -3 Input shortages have affected production, with 0 -4 exports declining 5 percent in the year to August -3 SAR MNA SSA LAC EAP ECA -5 SAR EAP ECA EMDE LAC SSA MNA and industrial production shrinking by 15 percent in the year to June 2023. E. Monetary policy interest rates F. Remittance inflows to remittance- dependent SAR countries Bangladesh has had the strongest recent growth of Percent Index, 2019=100 the three. However, economic activity is being Bangladesh Nepal 24 Bangladesh India 160 Pakistan Sri Lanka restrained by supply disruptions from energy 21 Other EMDEs 18 Nepal Pakistan shortages and continued import and capital 15 Sri Lanka 120 controls. Limited trade credit due to low 12 9 80 quantities of foreign exchange in the banking 6 system has also reduced imports. There was a 3 Jul Jan Jul Jan Aug 40 Jan Apr Jul Oct Jan Apr Jul noticeable deceleration in both private 2021 2022 2022 2023 2023 2022 2022 2022 2022 2023 2023 2023 consumption and investment growth during FY23 as a result of high inflation and rising uncertainties Sources: CEIC; Haver Analytics; Kilic Celik et al. (2023); Oxford Economics; United Nations Conference on Trade and Development; World Bank (GEP June 2023); World Bank (Macro related to the external sector. Poverty Outlook). Note: EMDEs = emerging market and developing economies; EAP = East Asia and Pacific; ECA = Europe and Central Asia; LAC = Latin America and the Caribbean; MNA = Middle East and Maldives, Nepal, and, to a lesser extent, Bhutan North Africa; SAR = South Asia; SSA = Sub-Saharan Africa. A. Purchasing Managers’ Indexes (PMIs) come from IHS Markit and are seasonally adjusted. have been benefiting from the recovery in global PMIs above 50 (below 50) indicate expansion (contraction). Latest data are August 2023. tourism. Public investment in Maldives has been B. Potential growth is estimated based on production function approach. GDP-weighted averages. Potential growth averages cover 3 countries in EAP, 6 in ECA, 10 in LAC, 3 in MNA, 4 robust as a result of multiple ongoing projects, in SAR, and 3 in SSA. C. Bars show GDP-weighted average of goods and services export forecasts. Sample includes notably the expansion of its international airport. 6 countries in EAP, 6 in ECA, 9 in LAC, 12 in MNA, 2 in SAR, and 14 in SSA. The associated increase in external debt has D. Bars show changes in fiscal balances between the average for 2017-19 and the average for 2020-22. GDP-weighted averages. worsened the country’s financial vulnerabilities, F. Dotted line = 100. however. Growth in Nepal slowed in FY23, reflecting monetary policy tightening and import restrictions. It has improved more recently as the removal of import restrictions late last year has SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 7 increased the availability of productive inputs. FIGURE 1.3 Inflation Growth in Bhutan has been lifted by spillovers Inflation has been declining globally, but has remained elevated in South from strong growth in India, despite a contraction Asia. Falling commodity prices were dampening inflation until energy prices started to increase more recently. Historically, inflation in South Asia in the electricity sector. Construction and has tended to be less driven by global developments than elsewhere, but manufacturing activities have strengthened, and recent global shocks interacted with local vulnerabilities in many countries, the services sector has been supported by transport triggering currency depreciations and large increases in domestic inflation. Rising food prices are contributing to continued high inflation in several - and trade-related services activities. South Asian countries. Afghanistan is adjusting to a structurally lower A. Global headline consumer price B. Commodity prices level of aggregate demand following the cessation inflation of grant inflows and the breakdown of Percent Index, 2021=100 AEs Global EMDEs 200 international banking relationships in 2021. 12 Energy Agriculture Metals 10 160 Inflation peaked in July 2022 at a year-on-year 8 120 rate of 18 percent and then declined steeply, 6 80 leading to deflation since April 2023. Deflation is 4 40 likely the result of weak aggregate demand, 2 0 improved supply conditions, and the appreciation 0 Jan Jul Jan Jul Jan Jul Jan Jul Jan Jan Jan Jan Aug 2024 2020 2021 2022 2023 2023 of the exchange rate. Surveys show that about two 2020 2020 2021 2021 2022 2022 2023 2023 -thirds of Afghan families face significant C. Historical contribution of global D. Headline consumer price inflation challenges in maintaining their livelihoods (World shocks to domestic inflation variation Bank 2023a). Percentage points Percent 40 Oil price Global supply Global demand 20 EAP ECA LAC MNA SAR SSA Inflation 30 16 12 20 Inflation in South Asia rose sharply as a result of 8 rising global commodity prices and currency 10 4 depreciations. It remains above 7 percent in the 0 0 Advanced EMDEs India Jan Jul Jan Aug median country, partly due to continuing economies 2022 2022 2023 2023 increases in food prices. E. Core consumer price inflation F. Inflation in South Asian countries, Global developments latest Percent Percentage points Global headline inflation declined from a peak of 20 EAP ECA LAC MNA SAR SSA 30 Non-food inflation contribution Food inflation contribution 9.4 percent year-on-year in July 2022 to 4.9 16 20 Inflation Target percent in July 2023 (figure 1.3). Until recently, 12 10 falling commodity prices were contributing to this 8 0 decline. More recently, extended production cuts 4 -10 by OPEC and its partners and high demand from 0 Jan Jul Jan Aug -20 the transportation sector have pushed up energy 2022 2022 2023 2023 AFG MDV BTN LKA NPL IND BGD PAK prices. Global core consumer price inflation Sources: Afghanistan National Statistics and Information Authority (NSIA); CEIC; Haver Analytics; remains stubbornly elevated. World Bank. Note: AEs = advanced economies; AFG = Afghanistan; BGD = Bangladesh; BTN = Bhutan; EAP = East Asia and Pacific; ECA = Europe and Central Asia; EMDEs = emerging market and developing Regional developments economies; IND = India; LAC = Latin America and the Caribbean; LKA = Sri Lanka; MDV = Maldives; MNA = Middle East and North Africa; NPL = Nepal; PAK = Pakistan; SAR = South Asia; SSA = Sub- Saharan Africa. Historically, inflation in South Asia has tended to A.D.E.F. Median year-on-year inflation by country group. A. Sample includes 39 AEs and 101 EMDEs. Last observation is July 2023. be less driven by global developments than B. Diamonds show 2024 forecasted values from the April 2023 edition of the World Bank’s elsewhere, with domestic developments being Commodity Markets Outlook. Solid lines show the monthly values from the June 2023 edition of the World Bank Global Economic Prospect report. Last observation is August 2023. more important (Ha et al. 2019). Recent events C. Bars show the median shares of country-specific inflation variance accounted for by global shocks (global demand, global supply, and oil prices) based on country-specific factor-augmented vector are an exception. The rise in global prices resulting autoregression models estimated for 29 advanced economies and 26 EMDEs for 1971-2017. D. Sample includes 11 countries for EAP, 22 for ECA, 22 for LAC, 15 for MNA, 6 for SAR, 26 for from the end of the pandemic and Russia’s SSA. Last observation is August 2023. invasion of Ukraine worsened local vulnerabilities E. Sample includes 7 countries for EAP, 9 for ECA, 13 for LAC, 6 for MNA, 6 for SAR, 8 for SSA. Last observation is August 2023. in several countries in the region, leading to F. Latest CPI data is July 2023 for Afghanistan, Bhutan and Nepal, and August for Bangladesh, India, Pakistan, and Sri Lanka. The inflation target range for Pakistan is 5-7 percent, to be met by the end of FY25. 8 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 increases in current account deficits and currency economy. It is also transitioning to an interest rate depreciations. These were met with the imposition targeting regime from a system based on targets of import controls. As a result of these overlapping for monetary aggregates. factors, several countries saw large increases in domestic inflation, especially Bangladesh, Nepal, In Sri Lanka, inflation peaked around 70 percent Pakistan, and Sri Lanka. year-on-year in September 2022 but has since slowed sharply as the effects of last year’s currency Median headline and core inflation remain near 7 depreciation have faded. Unlike other central percent in South Asia, and inflation in the region banks in the region, the Central Bank of Sri Lanka has not slowed as rapidly as in other regions. Food has been cutting its policy rates since June, in inflation remains particularly high owing to both response to steep disinflation and economic high global food inflation and local supply contraction. disruptions. As currencies have stabilized and some import controls have been relaxed, inflation Financial conditions is expected to trend down throughout the region. This trend could be interrupted in a variety of Global financial conditions remain challenging, ways, however. Commodity prices could increase, while several South Asian countries are recovering exchange rates could depreciate, or the second- from recent balance-of-payments crises. Many round effects of past shocks could be more remain vulnerable to further shocks. persistent than currently anticipated. Global developments In India, inflation was trending down below the upper bound of the inflation target range before a Between late 2021 and early 2023, central banks disruptive monsoon caused a substantial recent in most advanced economies hiked key interest increase in food prices. To counter this, the rates at the fastest pace since the 1980s in response government has implemented an export ban on to persistent, above-target inflation. In recent most types of rice. The Reserve Bank of India months, however, the pace of increases has slowed, increased interest rates substantially last year, and and policy rates in many major economies seem has kept them steady since this February. close to peaking (figure 1.4). The effects of monetary tightening can be seen in rising In Pakistan, consumer price inflation stood at 27 borrowing costs and credit standards. Despite percent in the year to August, down from a peak these developments, volatility and risk spreads of 38 percent in May. The decline reflected the have remained low in most markets, and major stabilization of the exchange rate since the stock indexes have risen markedly this year. beginning of the year, following 18 months of substantial depreciation, as well as an unwinding Most EMDEs have weathered this period of of the food price spike caused by the widespread financial tightness without severe strain. Net damage from last year’s floods. The central bank capital inflows have been low but positive, and has tightened monetary policy to combat high bond issuance has rebounded after a severe inflation, increasing its benchmark interest rate by contraction last year. Most EMDE currencies have 100 basis points most recently in June, to 22 been stable since the beginning of the year. percent. There are, however, pockets of weakness, as many In Bangladesh, inflation soared to a decade high of countries with lower credit ratings are struggling 9.9 percent in the year to May, driven by double- with severe and unsustainable increases in digit food inflation. Headline inflation remains borrowing costs. Historically, higher interest rates elevated as a result of currency depreciation in advanced economies have often been associated alongside rising commodity prices. The central with financial stress in EMDEs, particularly those bank has tightened monetary policy in response with greater economic vulnerabilities (Arteta, and loosened an interest rate cap that was limiting Kamin, and Ruch 2023). the transmission of policy changes to the SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 9 Regional developments FIGURE 1.4 Financial conditions Monetary policy interest rates in major advanced economies are close to Recent global stresses helped to trigger balance-of- peaking. Global financial market volatility and sovereign risk spreads payments crises in several countries in South Asia. worsened substantially last year, resulting in multiple currency crises, but These countries suffered widening current account have eased in 2023. Low international reserves and fragile financial systems make some countries vulnerable to further shocks. deficits, sharp exchange rate depreciations, capital outflows, widening credit spreads, and the A. Monetary policy interest rates in B. Financial market volatility and depletion of foreign exchange reserves. In response the United States and euro area sovereign risk spreads to these developments, several countries Percent Basis points Sri Lanka India VIX Index introduced capital controls and import 9 Federal Reserve policy rates 8,000 Pakistan VIX(RHS) EMBI Global 40 ECB policy rates restrictions, many of which remain in place. The 6 6,000 30 situation stabilized earlier this year due in part to 4,000 20 the introduction of IMF-supported policy 3 2,000 10 programs. Severe underlying problems remain, however, and the financial systems of several 0 Jan Jan Dec 0 Jan Jul Feb Aug 0 countries remain vulnerable to adverse shocks. 2022 2023 2023 2022 2022 2023 2023 Financial stresses were most severe in Pakistan and C. International reserves D. Changes in exchange rate against the U.S. dollar Sri Lanka. In Pakistan, the rupee depreciated Months of imports Percent sharply between early 2022 and early 2023, and 12 20 Change of exchange rate in 2022 Change of exchange rate in 2023 has been broadly stable since. Last year’s attempts 10 10 0 to limit capital outflows through import and 8 -10 6 capital controls diverted remittance inflows from 4 -20 formal channels, contributing to shortages of 2 -30 -40 foreign currency. In Sri Lanka, the rupee has 0 -50 IND Other BGD LKA PAK LKA IND Other BGD PAK appreciated modestly since the beginning of the EMDEs EMDEs year, partially reversing last year’s depreciation of more than 40 percent against the U.S. dollar. Sources: Bloomberg; CEIC; Chicago Board Option Exchange; JP Morgan. Note: BGD = Bangladesh; ECB = European Central Bank; EMBI = emerging market bond index; Remittances have rebounded as the economy has EMDEs = emerging market and developing economies; IND = India; LKA = Sri Lanka; PAK = Pakistan. stabilized, although they remain well below 2019 A. Figure shows end of month policy rates from the U.S. Federal Reserve Board and European levels. There has also been a recovery in tourism Central Bank. Dashed lines show policy rate expectations. Last observation is September 22, 2023. B. Figure shows the sovereign spread for SAR countries and for the broader EMBI Global index, and earnings. In both Pakistan and Sri Lanka, foreign the volatility index (VIX). Last observation is September 22, 2023. C. Figure shows the number of months of imports that foreign reserves can cover. Last observation is reserve coverage is low, asset quality is weak in 2023Q2. both the bank and non-bank financial sectors, and D. Bars show the cumulative change in exchange rate—U.S. dollars per unit of local currency— during 2022 and 2023. A positive value indicates appreciation against the U.S. dollar. “Other EMDEs” buffers against future shocks are thin. is an unweighted average of 29 countries. The latest data is 2023Q2. In India, the financial sector has shown few signs of strain. Bank balance sheets and corporate leverage ratios have improved substantially in country has previously attempted to contain its recent years. The current account deficit has been current account deficit through exchange and predominantly financed by foreign portfolio import controls alongside multiple exchange rates. investment and remittances. Foreign exchange This has encouraged the growth of a substantial reserves are at a healthy level, while the currency informal exchange market. The authorities are has alternated between periods of stability and committed to unification of the exchange rate this mild depreciation. Nonperforming loans in the year, but import and capital controls are expected banking sector are low. to remain in place for an extended period. Bangladesh suffers from limited foreign reserves In Nepal, the financial system appears robust. and a reliance on administrative policies— Remittance inflows have surpassed pre-pandemic primarily import controls—to stem outflows. levels, reaching nearly 22 percent of GDP in 2022 Remittance inflows have been volatile, and the and helping to alleviate pressures on the balance currency has been depreciating steadily. The of payments. 10 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.5 Outlook for output growth In Bhutan, nonperforming loans in the banking Projected output growth in South Asia is stronger than in other regions, but sector remain elevated. Financial sector risks may below its potential rate and below pre-pandemic averages in all South be under-reported in official statistics due to Asian countries. Individual countries in the region are generally, but not ongoing forbearance measures that seek to prevent always, outperforming other EMDEs with similar characteristics. a loan from becoming nonperforming and resolve A. Output growth in South Asia B. Annual output growth in SAR the stock of existing nonperforming loans. countries Percent Percent Maldives has substantial external debt burdens 10 Other EMDE regions SAR 8 Range 2024-25 2015-19 Average related to fiscal strains from the pandemic, as well 8 6 as major infrastructure investments. Servicing this 6 debt could present challenges if growth or capital 4 4 inflows disappoint. 2 2 0 0 Outlook 2022 2023e 2024f IND BGD MDV NPL BTN PAK LKA South Asia’s growth (excluding Afghanistan) is C. Output growth in India compared to D. Output deviation between current forecast to slow from 8.2 percent in 2022 to 5.8 other large EMDEs and pre-pandemic forecasts for 2024 percent in 2023 and 5.6 percent in 2024 and Percent Percent deviation 12 IND CHN EMDE G20 importers 0 2025 (table 1.1). For most South Asian countries, excl. CHN and IND 9 -5 growth in 2023–25 will remain below the pre- 6 3 -10 pandemic (2015–19) average, with the fading of 0 -15 post-pandemic rebounds accentuated by -3 -20 combinations of monetary tightening, fiscal -6 -25 consolidation, and slowing global demand growth -9 -30 2019 2020 2021 2022 2023e 2024f BTN LKA NPL BGD IND PAK MDV (figure 1.5). F. Output growth in Bhutan and In all South Asian countries, projected growth is E. Output growth in Pakistan and Sri Lanka compared with other EMDEs Maldives compared to other small insufficient to return output in 2024 to the path around currency crises Percent Percent projected before the pandemic. Current growth 9 LKA PAK Average past currency crisis 16 BTN MDV Average small state rates are also not high enough for most countries 6 12 to reach high-income thresholds within a 3 generation. Closing both of these gaps will require 0 -3 8 additional strong reforms. 4 -6 Compared to the spring edition of this report, the -9 -1 0 1 2 0 0.3 percentage-point downgrade for 2024 is Time 2022 2023e 2024f 2025f accounted for by lower projected growth for Bangladesh and Pakistan. Both countries are Sources: Consensus Economics; Laeven and Valencia (2020); WDI (database); World Bank. Note: BGD = Bangladesh; BTN = Bhutan; CHN = China; EMDEs = emerging market and developing struggling to emerge from balance-of-payments economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan; SAR = South Asia. Peer groups growth rates are calculated using simple averages. problems. In 2025, growth is generally expected to A. Blue bars reflect the range of growth across all the other EMDE regions. Regional aggregate computed using 2015 GDP as weights. Sample includes 7 countries in SAR and 136 in other EMDE return to its underlying, potential pace. regions. B. Blue bars show the range of annual GDP forecasts for the period 2024-25. Private consumption in the region is expected to C. Sample in the comparator group includes 2 EMDE G20 commodity importers (Mexico and Türkiye) and excludes China and India. be dampened by monetary tightening in response D. Bars show the percent gap between level of output in the current forecast for 2024 and the forecast produced in 2019 for 2024. to continued inflation pressures. Import growth in E. Peer group includes 101 EMDEs experiencing currency crisis (as defined by Laeven and Valencia several countries has been constrained in recent 2020) over the period 1970–2017. Currency crises are defined as nominal depreciation of the currency vis-à-vis the U.S. dollar of at least 30 percent that is also at least 10-percentage-points years by the combination of economic crises and higher than the rate of depreciation in the year before. Shaded area represents FY2022/23 for Pakistan (which narrowly misses the definition of a currency crisis) and calendar year 2023 for Sri restrictive policy measures. This is expected to Lanka. For all other EMDEs, t = 0 represents the period in which a currency crisis occurred independently of the year. Dashed lines show GDP forecasts. rebound as currencies remain broadly stable and F. Peer group includes 19 EMDE small states (with populations between 0.25 million and 1.5 million). import restrictions are gradually relaxed. The region’s current account balances are expected to remain in deficit, with little change over the SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 11 TABLE 1.1 Growth in South Asia Revision to forecast from   Country fiscal year Real GDP growth at constant market prices (percent) April 2023   (percentage point) Calendar year basis 2022 2023(f) 2024(f) 2025(f) 2023(f) 2024(f) South Asia region (excluding Afghanistan) 8.2 5.8 5.6 5.6 0.2 -0.3 Maldives January to December 13.9 6.5 5.2 5.5 -0.1 -0.1 Sri Lanka January to December -7.8 -3.8 1.7 2.4 0.4 0.7 Fiscal year basis 21/22 22/23(e) 23/24(f) 24/25(f) 22/23(e) 23/24(f) Bangladesh July to June 7.1 6.0 5.6 5.8 0.8 -0.6 Bhutan July to June 4.8 4.6 4.0 4.6 0.1 0.9 India April to March 9.1 7.2 6.3 6.4 0.3 0.0 Nepal mid-July to mid-July 5.6 1.9 3.9 5.0 -2.2 -1.0 Pakistan July to June 6.1 -0.6 1.7 2.4 -1.0 -0.3 Sources: World Bank Macro Poverty Outlook; World Bank staff calculations. Note: (e) = estimate; (f) = forecast. GDP measured in 2015 prices and market exchange rates. Pakistan is reported at factor cost. National accounts statistics for Afghanistan are not availa- ble. To estimate regional aggregates in the calendar year, fiscal year data are converted to calendar year data by taking the average of two consecutive fiscal years for Bangladesh, Bhutan, Nepal, and Pakistan, as quarterly GDP data are not available. projection horizon. Strong growth in services government is also tightening its fiscal position exports and remittances is expected to be modestly, even as it maintains high capital essentially offset by a recovery in imports and expenditures. weak growth of goods exports. The regional outlook is predicated on the assumption of no Given the limited fiscal and external buffers of substantial worsening of balance-of-payment countries in the region, there are considerable pressures. downside risks to the baseline growth forecast, including rising import prices, stalled progress on Investment growth in parts of the region, structural reforms, and policy uncertainty. particularly India, is expected to remain robust, especially due to strong public investment. In Country developments many other countries, however, it will be constrained by a combination of factors. These In Bangladesh, output growth is expected to slow include higher borrowing costs, tighter credit further as a result of persistent balance-of- conditions associated with stressed financial payments pressures. Growth is projected to slow conditions and crowding out by public borrowing, from 6.0 percent in FY2022/23 to 5.6 percent in and restrictions on access to imported capital FY2023/24, before picking up to 5.8 percent in goods. FY 2024/25. Investment growth is expected to be particularly weak, as lending is impaired by the Fiscal policy is expected to weigh on growth. financial system’s elevated level of nonperforming Primary fiscal deficits are expected to narrow over and rescheduled loans, the tightening of monetary the projection period, particularly in Bangladesh, policy, and policy uncertainty. Import restrictions Pakistan, and Sri Lanka, as these countries and shortages of foreign exchange, which are consolidate their fiscal positions in line with their expected to continue in the near term, will also IMF-supported policy programs. The Indian continue to impede growth. While weak global 12 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 growth will weigh on exports in general, the the tourism sector will increase capacity, country’s largest export—ready-made garments— particularly the expansion of Velana International is less sensitive to changes in aggregate global Airport. demand. Consumption growth is expected to remain robust, supported by buoyant remittances In Nepal, growth is estimated to have slowed to and population growth. Inflation remains elevated 1.9 percent in FY2022/23 as the post-pandemic and is expected to stay above the central bank’s recovery waned, monetary policy tightened, and target in FY2023/24, even as the effects of last policy efforts to close the current account deficit year’s currency depreciation fade. weighed on activity. Import restrictions were put in place between April 2022 and January 2023 to Bhutan’s economy is estimated to have grown by contain the growing trade deficit. This led to 4.6 percent in FY2022/23 as borders re-opened shortages of many inputs needed for production, and hydropower exports rebounded. Growth is an expansion of informal markets, and a sharp fall expected to slow to 4.0 percent in FY2023/24, in government revenues. Growth is expected to stronger than previously forecast in part due to a rebound to 3.9 percent in FY2023/24, as these major salary increase for government workers. The shortages ease, and as tourism inflows and recovery in tourism has lagged the broader remittances continue to rise. Investment in recovery, largely owing to a sizable increase in the hydropower remains strong, while activities related tourism levy in September 2022, which has since to the reconstruction of infrastructure destroyed been partially relaxed. Bhutan’s growth is expected by the 2015 earthquake is fading. to continue to underperform other small states. Pakistan’s economy is estimated to have shrunk by Private investment growth is expected to remain 0.6 percent in FY2022/23, reflecting widespread weak. Credit supply is being restrained by a damage from the 2022 floods, elevated inflation, moratorium on new housing and hotel and difficulties with its balance of payments. construction loans and by high nonperforming Positive growth is projected to return in loans in the banking sector. FY2023/24, but at a rate of only 1.7 percent. The Although India’s post-pandemic economic economy remains dependent on capital inflows to rebound is now fading, growth is expected to finance substantial fiscal and current account remain stronger than in other large EMDEs. deficits. Import controls intended to narrow the Output is forecast to grow by 6.3 percent in trade deficit have also impeded the supply of FY2023/24 and 6.4 percent in FY2024/25— industrial raw materials and depressed growth roughly equal to the estimated pace of India’s more than expected. These controls have been potential growth. The dampening effect of removed this year as an IMF lending program has monetary policy tightening on domestic demand, stabilized the currency and boosted business particularly investment, will likely peak in the confidence. Nonetheless, the economy still faces coming year. The effects of slowing global demand substantial challenges from continued inflation and rising interest rates will be mitigated by pressures, tight fiscal policy related to debt India’s low external debt and the healthy balance repayments, and extensive flood damage. sheets of its financial and corporate sectors. Pakistan’s foreign exchange reserves remain low, Growth of merchandise exports is expected to slow leaving the country with limited buffers against as a result of weak foreign demand growth, external shocks. although this will be offset by robust services In Sri Lanka, the economy appears to have exports. bottomed out after its severe recession and is Output in Maldives is expected to grow by 6.5 showing signs of recovery. Support from the IMF percent this year before slowing to 5.2 percent in and other external lenders has helped stabilize the 2024 and 5.5 percent in 2025. Tourism has currency and ease import shortages. The economy rebounded strongly from the global pandemic. is also being supported by the recovery of tourism. The recovery is expected to continue for longer After contracting by 3.8 percent in 2023, the than in other small states because investments in economy is expected to grow by 1.7 percent in SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 13 2024 and 2.4 percent in 2025. The country’s path FIGURE 1.6 Financial risks to recovery is very narrow, however. Its limited Nonperforming loan ratios are rising or are already elevated in most South fiscal and reserve buffers leave little room for error Asian countries; further increases would erode capital buffers and could eventually alarm financial markets. Large and growing holdings of as it implements a broad set of reforms and government debt by domestic financial systems leave them vulnerable to restructures its external debt. shifts in confidence in the sovereign and risk crowding out private sector credit. No forecast has been formulated for Afghanistan, B. Nonperforming loans in non-bank as official data collection halted in 2021. The A. Nonperforming loans in banking sectors financial or microfinance bank sectors country’s economy remains fragile, but surveys Percent of total loans Percent of total loans indicate that basic food and non-food items are 14 Most recent 2021Q4 25 Most recent 2021Q4* available in sufficient quantities, and that 12 20 10 employment and wages have improved this year 8 15 (World Bank 2023a). 6 10 4 5 Risks and vulnerabilities 2 0 0 NPL IND MDV PAK BTN BGD LKA NPL IND PAK LKA BGD Risks to the baseline forecast remain tilted to the downside. The most pressing concerns are C. Financial system claims on general D. Government debt financial and fiscal stress, slowing activity in government China, and climate change. Percent of domestic credit 50 Percent of GDP 140 2010 2022 40 120 Financial crises 100 30 80 Several countries in the region are vulnerable to 20 60 financial market disruptions. Bangladesh, 10 40 Pakistan, and Sri Lanka have drawn on IMF 0 2010 2021 2010 2021 20 0 assistance to weather the global shocks of higher EMDEs SAR SAR EMDEs commodity prices and borrowing costs and (in the Sources: CEIC; IMF (various staff reports); Kose et al. (2022); national sources; World Bank. case of Sri Lanka) reduced tourism earnings, and Note: BGD = Bangladesh; BTN = Bhutan; EMDEs = emerging market and developing economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan; SAR = South Asia. to stem capital outflows and currency A. The most recent data is 2023Q1 for Bangladesh, India, Nepal, and Sri Lanka, 2023Q2 for Bhutan, depreciation. Maldives may also require assistance Maldives, and Pakistan. B. The most recent data is 2022Q2 for Bangladesh, 2022Q3 for Nepal, and 2023Q1 for India, when its interest expenses triple to a peak of about Pakistan, and Sri Lanka. 2021Q1 is used for India’s historical number. The data is for microfinance banks for Pakistan, and for non-bank financial institutions for other countries. US$1 billion in 2026. The region’s persistent C. GDP-weighted averages (at 2010-19 average prices and market exchange rates). trade deficits have averaged 4 percent of GDP D. Bars show unweighted averages (at 2010–19 average prices and market exchange rates). Yellow whiskers indicate minimum-maximum range for seven South Asian economies, and interquartile since 2015. These require financing by capital range for EMDEs. inflows, which can make countries vulnerable to adverse shifts in market sentiment. Such shifts can result from stress in either the private sector— particularly the financial system—or in against adverse shocks. In Sri Lanka, governments’ fiscal positions. Vulnerability to nonperforming assets, as a proportion of total sudden changes in investor sentiment is assets, have recently been as high as 11 percent in particularly high in countries with low foreign the banking sector and 17 percent in the non- currency reserves. banking sector. In Bangladesh, the corresponding figures have been nearly 9 and 23 percent, The financial systems of many South Asian respectively. In contrast, nonperforming loans in economies are under pressure from challenging both India and Nepal appear to have been low. domestic economic conditions and rising borrowing costs. Nonperforming loans in bank The sovereign-bank nexus is strong throughout and non-bank financial sectors have recently the region and could propagate adverse shocks exceeded 5 percent of total assets in all South (World Bank 2023b). On average, nearly half of Asian countries for which there are data, except South Asian countries’ financial systems’ assets are Nepal (figure 1.6). This leaves limited buffers claims on the government. This share is higher 14 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.7 Scenario: Sharper economic slowdown in In the longer term, capital outflows could increase China governments’ reliance on domestic financing. This In a scenario where China’s real estate sector slows down sharply before could crowd out private borrowing, and make it policy makers intervene with stimulus, global growth, commodity prices, more difficult for less connected and less formal and inflation would all be lower than in the baseline forecast. The spillover firms to access finance than it already is. It has to South Asia would be smaller than to other regions. been estimated that a 10-percentage-point increase B. Deviation of global oil price from in the share of government loans in total bank A. Deviation of global and Chinese output growth from baseline scenario baseline scenario assets is associated with a 1.6-percentage-point Percentage points Percent decline in annual loan growth to the private sector 2.0 0 (World Bank 2023b). Increased debt service costs 1.0 -5 -10 could also crowd out other essential public 0.0 -15 expenditures and increase the rigidity of -1.0 -20 government spending, and also contribute to high -2.0 -3.0 -25 inflation to by encouraging governments to “print -30 2023 2024 2025 2023 2024 2025 money” to erode high levels of debt. 23Q4 24Q2 24Q4 25Q2 25Q4 China World Slowdown in China C. Deviation of global inflation from D. Growth impact of sharper baseline scenario slowdown in China In the baseline forecast, China is expected to Percentage points Percentage points support global activity with output growth of 5.1 0.0 0.4 percent in 2023, 4.4 percent in 2024, and 4.3 0.0 -0.5 percent in 2025. Persistent challenges in China’s -0.4 real estate market, however, present downside risks -1.0 -0.8 to this outlook. Falling housing prices, defaults by -1.2 major developers, and declining lending are 2024 2025 2024 2025 -1.5 already weighing on consumption and investment. 23Q4 24Q2 24Q4 25Q2 25Q4 EMDEs excl. China SAR Sources: Oxford Economics; World Bank. In a scenario with a sharper-than-assumed real Note: EMDEs = emerging market and developing economies; SAR = South Asia. estate sector slowdown, the real estate sector A. Bars show the percentage points deviation between the output growth assuming a sharper slowdown in China and output growth forecasted under the baseline scenario. would increasingly weigh on the Chinese B. Solid line shows the percent deviation of the oil price assuming a sharper slowdown in China and the oil price forecasted under the baseline scenario. economy, possibly reducing growth to 4.9 percent C. Solid line shows the percent deviation between global inflation assuming a sharper slowdown in China and global inflation forecasted under the baseline scenario. in 2023 and 2.0 percent in 2024. In response, the D. Bars show growth revisions between the China slowdown scenario and the baseline scenario. government could introduce a variety of stimulus SAR includes 6 countries. measures, including infrastructure spending and support for credit issuance. These measures could push up growth in 2025 to 5.3 percent. than in other EMDE regions and has risen sharply In this scenario, the slowdown and subsequent over the past decade. As a result, falls in the rebound would have widespread spillovers to other valuation of government debt can result in countries, primarily through external demand and significant erosion of the market value of banks’ commodity prices (figure 1.7). Oil prices would be assets. At 86 percent of GDP in the average South 22 percent lower in 2024, on average, than in the Asian country, government debt is higher than in baseline scenario. Global inflation would move in other EMDE regions. And it is rising as a result of the same direction, and would be 1.4 percentage growing government spending, low (and in some points lower at the end of 2024. On balance, cases declining) domestic revenue, and increasing central banks worldwide could be expected to debt service costs (spotlight). Any loss of begin loosening monetary policy earlier in this credibility resulting from fiscal slippages, delays in scenario than in the baseline. The U.S. Federal debt restructuring negotiations, or the revelation Reserve, for example, could begin cutting its of losses at a major state-owned bank or enterprise, policy rate as early as November 2023, reducing it could drive borrowing costs and debt to by the end of 2024 to 150 basis points below the unsustainable levels. baseline. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 15 The spillovers to South Asia would generally be FIGURE 1.8 Climate risks smaller than to other EMDE regions. South Asian South Asia is the EMDE region most vulnerable to climate risks, with the growth could be 0.5 of a percentage point lower in largest number of people affected by natural disasters in the past decade. 2024, compared to a 1.0 percentage point growth Because of South Asia’s large agricultural sector, extreme weather events can be especially disruptive. Shocks to food markets and prices can slowdown for EMDEs overall (excluding China). exacerbate food insecurity because South Asia is the EMDE region with This is partly because the region is more closed to the highest share of food in consumption baskets. trade than other EMDE regions, and partly B. Number of people affected by because the region is heavily reliant on energy A. Vulnerability to climate risk, 2017- 21 average natural disasters, 2013-22 average imports, whose prices would decline. Index Million people Percent 0.6 EMDE average 70 Total affected 4 Total share of population affected (RHS) Climate change-related disasters 0.5 60 50 3 0.4 South Asia is highly exposed to both the short- 0.3 40 2 and long-term adverse effects of climate change, 0.2 30 20 more so than most other EMDE regions (figure 0.1 10 1 1.8). Much of the region’s population lives in 0.0 0 0 dense river valleys that are increasingly subject to SAR SSA EAP MNA LAC ECA SAR EAP SSA LAC MNA ECA severe floods. A recent example is the floods that C. Share of food in CPI basket and D. Prevalence of undernourishment in submerged one-third of Pakistan last year, causing share of agriculture in value added, the population latest economic losses equivalent to more than 4 percent Percent Percent of GDP (World Bank 2022a). Bangladesh’s losses 60 Food share of CPI 20 Percent 40 Agriculture share of value added (RHS) 2020-21 Avg. SAR World from tropical cyclones alone are estimated to 50 16 average 0.7 percent of GDP per year (World Bank 40 30 12 2022b). According to the Global Climate Risk 30 20 8 Index, Bangladesh ranks seventh, Pakistan eighth, 20 10 and Nepal tenth among the countries most 10 4 severely affected by extreme weather events 0 SAR SSA ECA EAP LAC MNA 0 0 AFG PAK IND BGD LKA NPL globally during 2000–19 (Eckstein, Künzel, and Schäfer 2021). A recent study designated Sources: FAOData Explorer; Ha, Kose, and Ohnsorge (2023); IMF Consumer Price Index Database; International Disaster Database (EM-DAT); Maldives Bureau of Statistics; Notre Dame Global Afghanistan among the countries most at risk Adaptation Initiative; WDI (database); World Bank. Note: AFG = Afghanistan; Avg. = Average; BGD = Bangladesh; BTN = Bhutan; CPI = Consumer from heat waves (Thompson et al. 2023). The Price Index; EAP = East Asia and Pacific; ECA = Europe and Central Asia; EMDEs = emerging countries at highest risk from natural disasters are market and developing economies; IND = India; LAC = Latin America and the Caribbean; LKA = Sri Lanka; MDV = Maldives; MNA = Middle East and North Africa; NPL = Nepal; PAK = Pakistan; SAR = often already poor and ill-equipped to weather South Asia; SSA = Sub-Saharan Africa. A. Regional aggregates computed using 2015 GDP as weights. Values shown are average over 2017 shocks (Rentschler, Salhab, and Jafino 2022). -21. Sample includes 148 EMDEs (22 in EAP, 22 in ECA, 31 in LAC, 18 in MNA, 8 in SAR, and 47 in SSA). B. Bars show the total population affected by natural disasters, and diamond shows the share of total The growing frequency and severity of weather population affected; annual averages over 2013-22. Sample includes 144 EMDEs (22 in EAP, 20 in ECA, 31 in LAC, 18 in MNA, 8 in SAR, and 45 in SSA). disasters pose risks to food production in both C. Regional aggregate computed using 2015 GDP as weights. The share of food and non-alcoholic South Asia and elsewhere. Disruptions to either beverages in CPI basket for the most recent reporting period is used. Sample includes 137 EMDEs (21 in EAP, 18 in ECA, 29 in LAC, 15 in MNA, 8 in SAR, and 46 in SSA). The share of agriculture, local or global food supplies could drive up food forestry, and fishing value added in GDP (2021-22 average). Sample includes 138 EMDEs (18 in EAP, 22 in ECA, 31 in LAC, 15 in MNA, 8 in SAR, and 44 in SSA). prices and households’ living expenses. Food is a D. Aggregates computed using 2015 population as weights, excluding countries with missing values. World sample represents 99 percent of world population. sizable part of households’ consumption baskets in South Asia, with over 45 percent of the region’s CPI basket consisting of food and non-alcoholic beverages, compared with an EMDE average of 2023; Gill and Nagle 2022; Nasir, Kishwar, and about 29 percent. Meyer 2023). High and volatile food prices make it more challenging to maintain a nutritious diet, Rising food prices would hurt the urban poor exacerbating food insecurity, which is widespread most, as more than half of their budgets are in the region. devoted to food and, unlike their rural counterparts, they do not produce food themselves Increasing climate risks also have longer-term (Aksoy and Hoekman 2010; Dovonou and Xie implications, especially since the agriculture sector 16 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 accounts for 17 percent of total value-added and the EMDE average. In Bangladesh and India, this more than 40 percent of employment in South rapid growth has been supported by public Asia. Agricultural production is already being investment growth of around 10 percent per depressed in areas rendered too hot for crops or year—triple the EMDE average. Sustaining this destroyed by flooding. The heat waves, droughts, pace of public investment growth may become and changing weather patterns associated with increasingly challenging as government debt and climate change could make some areas entirely borrowing costs rise. uninhabitable. Coastal areas and the entirety of In all countries in the region, private investment Maldives risk being covered by rising water levels. growth has slowed since the pre-pandemic period Another risk is that coastal areas may become (2015–19), or is forecast to do so. The weakness unsuitable for crop production due to soil and of private investment growth has been mirrored in water salination (Hooijer and Vernimmen 2021). the region’s below-average net inflows of FDI. In Policy challenges 2021–22, FDI in South Asia accounted for around 1.5 percent of GDP, considerably less than In the near term, policy priorities include the EMDE average of 2 percent. preserving financial stability and restoring fiscal Accelerating the pace of catch-up with high- sustainability. In the longer term, a range of income countries will require substantial new structural measures will be needed to accelerate investment and substantial increases in growth and job creation in a sustainable manner. productivity. The productivity gaps are large: These include strengthening private investment productivity levels in South Asia are the second- growth and seizing the opportunities presented by lowest among EMDE regions after Sub-Saharan both the global energy transition and the global Africa (Dieppe 2021). Currently, growth in the effort to diversify supply chains. region is not strong enough for most countries to Strengthening investment reach high-income thresholds within a generation. Potential growth in the region averages about 5 Many factors that supported investment growth percent, but would have to be 8 percent or higher globally in the early 2000s weakened around the in most countries to reach high-income status by time of the global financial crisis. Investment 2050. Weak investment also threatens to delay the growth subsequently trended downward around region’s progress with the energy transition. the world (Stamm and Voristek 2023). This included South Asia, as manufacturing growth Strengthening private investment will depend on weakened amid sluggish global demand, financial many factors. These include the presence of stress, and uncertainties related to government complementary infrastructure, a supportive policies (Kasyanenko et al. 2023). However, the institutional and business environment, a sound decline in South Asian investment growth was financial system, and fewer distortionary policies smaller than in other EMDE regions, although affecting markets. with wide disparities across countries. • Public investment. Effective public investment More recently, investment growth has generally and high-quality infrastructure can crowd in followed two contrasting patterns in South Asia. private investment. Public infrastructure In Bhutan, Pakistan, and Sri Lanka, the average projects, such as the construction of the annual growth rate of investment in the past five Padma Bridge in Bangladesh, and various years has been negative or near zero, with public railways and road projects in India, have the investment particularly weak in Pakistan and potential to spur investment and economic private investment growth particularly weak in activities in the surrounding area (World Bhutan (figure 1.9). Bank 2023c, 2023d). The efficiency of public investment projects in many countries could By contrast, in the second group of countries— be improved through a supportive public Bangladesh, India, Maldives, and Nepal— investment management framework (World investment growth has been robust and well above Bank 2023e, 2023f, 2023g). SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 17 • Supportive institutions. Better public FIGURE 1.9 Investment weakness institutions also tend to attract more private Investment growth in some South Asian countries has been negative or investment and FDI (Ali, Fiess, and anemic in recent years, while in others it has been supported by robust MacDonald 2010; Gwartney, Holcombe, and public investment. Private investment growth in South Asia has slowed from its pre-pandemic averages. Private investment weakness has in part Lawson 2006; Heilbron and Whyte 2019). reflected below-average FDI inflows. Current growth rates are not sufficient For example, spurts of investment climate for most countries to reach high-income thresholds within a generation. reforms, especially in EMDEs, have been A. Real total investment growth B. Real public investment growth associated with an increase in real investment Percent Percent growth of about 6 percentage points per year 9 2018-22 Avg. EMDEs 12 2018-22 Avg. EMDEs (Stamm and Voristek 2023). Surveys of firms 6 9 regularly show that policy and regulatory 3 6 uncertainty, followed by taxation and 0 3 burdensome regulations, are the most critical -3 0 barriers to private sector investment (OECD -6 -3 2015). Public institutions can also provide -9 BGD MDV IND NPL PAK BTN LKA -6 IND BGD BTN PAK NPL critical complementary services to enable C. Real private investment growth D. FDI inflows into EMDE regions functioning markets. For example, in many countries in the region, land records that are Percent 20 Latest annual data 2015-19 Avg. Percent of GDP 4 complete, transparent, and integrated across 15 2018-22 Avg. EMDEs World 10 different parts of government would help 5 3 improve the use of land for business purposes 0 2 -5 (World Bank forthcoming). -10 -15 1 • Business environment. Allowing greater scope -20 -25 0 for competition could unleash private BGD IND NPL BTN PAK LAC ECA SSA EAP SAR MNA investment. For example, in Pakistan, certain E. FDI inflows into South Asian F. GDP growth rate required to reach tax policies discourage investment in the countries income thresholds by 2050 tradable sector, and certain investment laws Percent of GDP Percent of GDP Percent GDP growth forecast, 2023-25 avg. 3 15 discriminate against foreign investors (World 2018-22 Avg. EMDEs 15 GDP growth required to reach HIC level by 2050 GDP growth required to reach UMIC level by 2050 Bank 2023f). In Bhutan, Nepal, and Pakistan, 2 10 12 reducing subsidies or budgetary support to 9 state-owned enterprises could allow for greater 1 5 6 private sector participation while also 0 0 3 increasing fiscal space (World Bank 2022c, AFG (RHS) LKA PAK NPL IND BGD BTN MDV 0 IND BGD BTN NPL PAK LKA 2023f, 2023h). In Pakistan, similarly, state- owned enterprises tend to have low Sources: UN Population Division (database); WDI (database); World Bank (Macro Poverty Outlook). Note: Avg. = Average; AFG = Afghanistan; BGD = Bangladesh; BTN = Bhutan; EAP = East Asia and investment rates, while also consuming Pacific; ECA = Europe and Central Asia; EMDEs = emerging market and developing economies; FDI = Foreign Direct Investment; HIC = High-income country; IND = India; LAC = Latin America and the government resources equivalent to around 23 Caribbean; LKA = Sri Lanka; MDV = Maldives; MNA = Middle East and North Africa; NPL = Nepal; PAK = Pakistan; SAR = South Asia; SSA = Sub-Saharan Africa; UMIC = Upper middle-income percent of the fiscal deficit in FY2023 (World country. Arithmetic annual averages. Aggregates computed using 2015 GDP as weights. Bank 2023i). A. Figure shows the annual growth of real gross fixed investment (in local currency), average of 2018 -22. Sample includes 123 EMDEs. B. Figure shows the annual growth of real public fixed investment (in local currency), average of 2018 • Access to finance. Private investment also -22. Sample includes 93 EMDEs. C. Figure shows annual growth of real private fixed investment (in local currency), average of 2015- depends on access to finance. Adverse 2019. "Latest data” refers to 2023, except for Bhutan and India, which are based on 2020-21 average due to limited data and to even out the deep contractions of 2020 and strong rebounds of 2021. liquidity shocks caused by troubled banks can D. Figure shows net inflow of foreign direct investment as percent of GDP (2018-22 average). hinder investment (Kalemli-Ozcan, Kamil, Sample includes 148 EMDEs (22 in EAP, 23 in ECA, 31 in LAC, 18 in MNA, 8 in SAR, and 46 in SSA). World sample includes 194 countries, representing 99.7% of world economy. and Villegas-Sanchez 2016). In Sri Lanka, E. Figure shows net inflow of foreign direct investment as percent of GDP (2018-22 average). Sample includes 148 EMDEs. more robust deposit insurance, provisions for F. Figure shows the GDP growth rate forecast for 2023-25 and the GDP growth rate required to achieve high-income and upper middle-income status by 2050. Population growth is from United nonperforming loan resolution, and Nation population projections. The threshold for high- and upper middle-income status is assumed to prudential supervision of concentrated loan grow by 1.5 percent, its average growth rate between 1999 and 2019. The income threshold is based on GNI per capita in current U.S. dollars (Atlas method). Since the income threshold is based on GNI, exposures can help strengthen the financial while the forecast is based on GDP, this figure assumes equal growth rates of the two measures. Estimates for the required growth rate from other studies (e.g., Behera et al. 2023) assume an system. Better governance of state-owned inflation rate differential and real appreciation between advanced economies and the country. 18 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.10 Restrictions on trade and foreign exchange enterprises would also improve the allocation transactions of capital (World Bank 2023j). The South Asia generally employs an above-average number of trade and dominance of state-owned banks in South foreign exchange restrictions. In response to recent currency pressures, Asia has been found to limit access to finance South Asian countries raised import and foreign exchange restrictions further to help address balance of payments issues and to stabilize foreign for smaller, newer businesses with strong exchange markets. growth potential but limited collateral or track records, as state-owned banks have tended to A. Non-tariff measures in EMDE B. Cost of trading in EMDE regions, direct credit on the basis of non-economic regions, 2020 2019 Number of measures Percent criteria or have been managed inefficiently 4 200 (IMF 2020; Melecky 2021). In Bangladesh, 3 150 for example, reducing the role of the state in the financial sector, capital market reform, 2 100 and stronger bank governance could help 1 50 improve the allocation of capital (World Bank 2022b). Unclear or difficult-to-enforce 0 0 SAR ECA EAP LAC MNA SSA SSA SAR LAC EAP MNA ECA ownership of assets such as land can limit the collateral available to potentially successful C. Restrictive export measures in SAR D. Restrictive import measures in SAR small companies, which could be an countries and other EMDEs countries and other EMDEs Number of Number of important engine of vigorous private 2022 2022 product groups 12 2015-19 Avg. product groups 50 2015-19 Avg. investment growth (Zhang et al. 2020). Asset EMDEs 2022 10 EMDEs 2022 EMDEs 2015-19 40 EMDEs 2015-19 quality forbearance measures in banks can 8 30 help zombie firms to stay afloat and crowd out 6 20 credit for healthy productive firms (Chari, 4 10 Jain, and Kulkarni 2021). 2 0 AFG BGD BTN IND NPL PAK LKA 0 AFG BGD BTN IND NPL PAK LKA Vigorous private investment is critical not only for growth but also for environmental reasons. Private E. Capital controls, 2019 F. New foreign exchange restrictions, investment in energy-saving technologies will be 2021–22 Index, key if the region is to keep pace with the global Inflows Number of tightened Higher=Stronger Outflows exchange restrictions energy transition (chapter 2). Such investment 1.2 Inflows: EMDE average 20 1.0 Outflows: EMDE average could be encouraged through combinations of 0.8 15 market-based regulations, carbon taxes, cuts in 0.6 10 fossil fuel subsidies, more reliable grid power, and 0.4 5 efforts to increase awareness among businesses of 0.2 the benefits of energy-saving technological 0.0 0 BGD IND PAK LKA BGD BTN IND MDV NPL PAK LKA innovations. Sources: Fernández et al. (2016); IMF (2022); Kose and Ohnsorge (2023); UNCTAD COVID-19 Removing trade and foreign currency Trade Measures Database; World Bank; WTO Trade Monitoring Database. Note: AFG = Afghanistan; BGD = Bangladesh; BTN = Bhutan; EMDEs = emerging market and restrictions developing economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan. A. Bars show the unweighted average number of non-tariff measures (NTM) implemented by EMDEs South Asian countries impose more restrictions on in each region in 2020. B. Data from Kose and Ohnsorge (2023; chart 6.5.B). Bilateral trade costs (as defined in the trade and capital flows than countries in other UNESCAP/World Bank database) measure the costs of a good traded internationally in excess of the EMDE regions (figure 1.10). Applied tariff rates same good traded domestically and are expressed as ad valorem tariff equivalent. Bilateral trade costs are aggregated into individual country measures using 2018 bilateral country exports shares. are higher than in any other EMDE regions: the Bars show unweighted averages, whiskers show interquartile ranges. Sample includes 53 EMDEs (9 in EAP, 12 in ECA, 16 in LAC, 4 in MNA, 2 in SAR, and 10 in SSA). weighted average applied tariff on imports in C.D. Dashed lines mark the EMDE average, weighted by 2015 GDP. Restrictive measures include duties, tariffs, taxes, custom procedures, quantitative restrictions, and others. For export measures, South Asia was 9.4 percent in 2020, compared EMDEs include 62 economies. For import measures, EMDEs include 90 countries. Product groups counted at the 2-digit Harmonized System (HS) level. Method counts number of measure-product with 6.7 percent among other EMDE regions. group pairs, and so a product group affected by two restrictive measures is counted twice. Non-tariff trade barriers such as long custom E. Dashed lines mark the average of 68 EMDEs, weighted by 2015 GDP. Index captures the severity of inflow and outflow capital control restrictions, including restrictions on money market, bonds, clearance times and quantity control measures are equity, mutual funds, financial credits, trade credits, and derivatives. F. Number of tightened measures includes those in 2021 and part of 2022. The cut-off date is June also more numerous than the EMDE average. The 30, 2022 for Bhutan, Maldives, Nepal; July 31, 2022 for India and Sri Lanka; August 31, 2022 for Pakistan; and September 30, 2022 for Bangladesh. cost of trading goods between South Asia and the SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 19 rest of the world was around 140 percent of the 2023b). Such restrictions can be circumvented, for value of the good itself, second highest among example by channeling foreign exchange through EMDE regions. The region also maintains more informal markets such as Hundi and Hawala stringent capital controls than the average EMDE (Biswas 2012; Steinkamp and Westermann 2022). (Fernández et al. 2016). This was true prior to the This circumvention may eventually lead to even pandemic and the use of these restrictions has greater losses of foreign reserves (Gray 2021). expanded further over the past three years. Lowering these barriers to trade and capital flows In recent years, increases in the prices of energy could help the region integrate into the global and other commodities, together with tightening marketplace, with substantial benefits to long- monetary policies in advanced economies, have term productivity. Pakistan, for example, could exacerbated balance-of-payments pressures in boost productivity, diversify its exports, and South Asia. In response, several countries in the increase product sophistication by reforming region have increased the use of restrictive import export subsidy and import duty schemes (World measures, such as quantitative restrictions, tariffs, Bank 2022e). The cost of shipping goods to and and more cumbersome custom procedures. In from Bhutan is elevated, and the country could 2022, the number of product groups affected by unlock greater trade opportunities through various restrictive import measures in Bangladesh investment in physical and digital infrastructure, was seven times the EMDE average, and more combined with an improved and more predictable than 10 times the EMDE average in Pakistan and regulatory environment (World Bank 2020). Sri Lanka. While these restrictions may have Moving toward low and uniform import tariffs in helped reduce pressures in the external sector, they Bangladesh could also spur private investment, have also led to import shortages and depressed increase competitiveness, and promote export economic activity (World Bank 2023h, 2023k). diversification (World Bank 2023c). They also added to fiscal pressures, particularly in countries such as Nepal that are dependent on Improving fiscal positions import duties for government revenues. All countries in South Asia have had persistently Restrictions have also been introduced on the large fiscal deficits. As a result, government debt export side. These included a ban on Indian rice burdens in the region have risen faster than in the exports introduced in July 2023 to slow a rise in average EMDE since 2010 (spotlight). As in other domestic rice prices that had resulted from an regions, government debt in South Asia soared unfavorable monsoon, and to prepare for the during the pandemic as fiscal revenues fell and possibility of agricultural productivity losses in the expenditures on support programs rose. The event of a severe El Niño. region’s government debt has risen even during periods of strong growth, however. Slowing the Some South Asian countries imposed controls on rise in debt burdens will likely require continued foreign exchange transactions. Since 2021, many strong growth, while reforms simultaneously limit countries in the region adopted additional financing costs, increase revenues, and improve restrictions on foreign exchange. These included spending efficiency. higher minimum financing requirements, more restrictive use of letters of credit, higher advance Servicing this debt has become significantly more payments on imports, increased repatriation costly as global and national interest rates rise. To requirements on exports and other proceeds from some extent, borrowing costs can be limited by overseas, and foreign exchange quotas. The most sound and transparent debt management strategies recent restrictive measures include a profit that allows for financing to be provided at longer repatriation requirement for foreign investors in maturities, at fixed (and favorable) interest rates, Bangladesh, foreign exchange quotas in Bhutan, and in local currency. On their own, however, increased cash margin requirements in Nepal, and such steps are unlikely to be sufficient to increased scrutiny of the repatriation of revenues substantially reduce financing requirements. In from exports in Pakistan (World Bank 2022d, Pakistan, for example, interest payments 20 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 1.11 Fiscal challenges accounted for more than half of federal current Revenue collection is below the EMDE average across South Asia, government expenditures in FY23. contributing to persistently large fiscal deficits. Stronger fiscal rules could help contain rising debt; current rules are generally weaker than the EMDE A more direct approach to improving fiscal average. Governments in South Asia face the challenge of bringing positions would be to tackle the region’s unusually government debt and deficits toward their indicative targets. low government revenue collection. All countries A. Government revenues, 2020-22 B. Strength of fiscal rule but one (Bhutan) collect less revenue than the average EMDE average of nearly 30 percent of GDP, with Percent of GDP 35 Index, Higher=Stronger revenues in Bangladesh and Sri Lanka below 10 EMDE average 8 EMDEs, interquartile range SAR 30 percent of GDP (figure 1.11). Bhutan’s revenues 25 6 are bolstered by external grants from India and 20 4 other development partners. 15 2 10 A variety of measures can help increase revenues, 5 0 Expenditure Budget Debt rule such as expanding the tax base, closing loopholes, 0 LKA BGD PAK IND NPL MDV BTN rule balance rule and strengthening collections. These could include reducing special exemptions and concessional rates C. Change in EMDE government debt, D. Share of debt booms associated 2015-22, by strength of fiscal rules with default of existing taxes. For example, targeting VAT Percentage points of GDP Percent exemptions to exclusively the consumption baskets 10 20 Weak Strong of the poorest 40 percent could increase VAT 15 8 revenue by 2 percent in Nepal (World Bank 10 6 2021). Similarly, in Maldives, halving the personal 5 4 income tax threshold to MVR300,000 (equivalent 2 to US$19,480) could raise income tax revenue by 0 Expenditure Budget Debt rule 0 0.5 percent of GDP (World Bank 2022f). Since rule balance rule Total Domestic External low revenue collection is partly a consequence of large informal economies, a broader-based, simpler E. Debt rule target and actual debt F. Budget rule target and actual fiscal deficit tax system that is perceived as fair could raise Percent of GDP Percent of GDP collections by smoothing the movement of 140 Debt Debt rule target 16 Fiscal deficit Budget balance rule target 120 14 businesses into the formal sector. 12 100 80 10 Spending efficiency could be improved by 60 8 6 reducing subsidies, particularly those on fossil 40 4 fuels. Energy subsidies in South Asia amounted to 20 0 2 nearly 2 percent of GDP in 2021. South Asian 0 India Maldives Pakistan Sri Lanka India Maldives Pakistan Sri Lanka governments in part provide such subsidies through fixed prices and support to state-owned Sources: CEIC; IMF Fiscal Rules Dataset, 1985–2021; WDI (database); World Bank (Macro Poverty Outlook). enterprises in the energy sector. These subsidies Note: AFG = Afghanistan; BGD = Bangladesh; BTN = Bhutan; EMDEs = emerging market and developing economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = are generally inefficient, regressive, costly, and Pakistan; SAR = South Asia. Arithmetic annual averages. environmentally damaging (Damania et al. 2023). A. EMDE average computed using 2015 GDP as weights. Bars show 2020–22 averages of government revenue. Reducing these subsidies would improve fiscal B.C. The fiscal rule strength is constructed following Davoodi et al. (2022, appendix III), and is a sum of legal basis, monitoring, enforcement, and flexibility, weighted by the rule coverage (national, sustainability and also speed the energy transition. supranational, or both). SAR sample includes India, Maldives, Pakistan, and Sri Lanka. Sample includes 65 EMDEs (25 for the expenditure rule, 56 for the budget balance rule, and 52 for the debt Some countries in the region would also benefit rule). A higher index means a stronger fiscal rule. B. Values shown are the unweighted average of fiscal rule strength index for countries in the group, from steps to avoid election-related increases in over 2015–2021. public spending (box 1.1). C. Values shown are the unweighted average change (2015–2022) in government debt for countries with above-average strength of fiscal rule (“Strong”) or below-average strength of fiscal rule (“Weak”), over 2015-21. In many countries globally, fiscal rules have been D. Bars show the share of total government debt booms associated with default (of any type), domestic government debt booms associated with domestic default, and external government debt found useful for containing fiscal deficits (Caselli booms associated with external default, up to one year after the end of a boom. E.F. Latest data available. Budget balance rule for Pakistan is on federal budget deficit excluding and Wingender 2018). Four South Asian foreign grants. Budget balance rule target and debt shown for India are for the central government. countries have adopted debt ceilings—60 percent of GDP in India, Maldives, Sri Lanka, and Pakistan—and deficit targets—3 percent of GDP SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 21 BOX 1.1 Fiscal deteriorations around elections Among EMDEs, primary fiscal deficits, primary government expenditures, and government wage bills have tended to rise significantly around election years. While primary spending increases have tended to be partially reversed in the following year, post-election reversals of primary deficit and government wage bill increases have been more variable and at best partial. e consequent ratcheting up of primary deficits around elections in EMDEs can erode fiscal sustainability over the longer term, while the expansion of government wage bills can result in spending rigidities. In South Asia, in particular, fiscal positions have tended to deteriorate around national elections, and, in some cases, there is also evidence of targeted fiscal actions around subnational elections. While this result is true on average for the region, some countries—notably India in its 2023 budget—have avoided the risk of election-induced current spending increases. is points to a way forward for fiscally constrained governments in South Asia. Introduction in advanced economies (de Haan and Klomp 2013; Kyriacou, Okabe, and Roca-Sagales 2022; de Haan and In 2023 and 2024, parliamentary or presidential Gootjes 2023). elections will be held in seven out of the eight South Asian countries. With fiscal positions already fragile in us, the evidence is that political budget cycles are several South Asian economies and government debt common in EMDEs. is box examines political stocks high, spending increases, or revenue decreases budget cycles for South Asia, in particular, to answer around these elections would add to fiscal pressures. the following questions: A well-established literature has documented political • How pronounced are political budget cycles in budget cycles, in both advanced economies and South Asia? EMDEs. ese have been attributed to three factors. • How do political budget cycles in South Asia First, incumbents may adopt an expansionary fiscal compare with those in other EMDEs? policy designed to benefit voters directly, thus maximizing their chances of re-election (Nordhaus • What are the policy implications? 1975; Dubois 2016). Second, incumbents may introduce policies ahead of elections to spur economic is box contributes to the literature in three ways. growth, in the hope of demonstrating the strength of First, it examines fiscal positions around national their governments (Higashijima 2022; Han 2022). elections in South Asia, whereas the existing literature ird, if the expected outcome of an election is on political budget cycles in the region tends to focus unfavorable for the incumbent or is uncertain, the on specific fiscal actions around subnational elections. incumbent may issue debt to constrain their successor’s Second, it documents that one spending category—the room for maneuver (Alesina and Tabellini 1990). e government wage bill—is particularly susceptible to government wage bill, which accounted for 25 percent political budget cycles around national elections in of primary spending in the average EMDE in 2010–20 EMDEs. In contrast, the existing literature focuses on (and 26 percent of paid workers), can be a particularly aggregate spending or fiscal deficits (de Haan and important instrument for influencing elections Klomp 2013; Kyriacou, Okabe, Roca-Sagales 2022; de (Endegnanew, Soto, and Verdier 2017). Haan and Gootje 2023). ird, this box documents that political budget cycles tend to be only partially Several empirical studies have found evidence for reversed after the election whereas the existing literature election effects. A statistically significant—albeit focuses on fiscal aggregates in the election year itself generally small—political budget cycle has been (Brender and Drazen 2005; Strong 2023).b identified in many cross-country studies.a is cycle appears to be more prominent in EMDEs, where is box reports the following findings. income levels and governance are typically weaker than Note: This box was prepared by Jakob de Haan (University of b. An exception is Ebeke and Ölçer (2017), who report for a sample Groningen), Franziska Ohnsorge, and Shu Yu. of low-income countries that governments have tended to raise trade a. See Brender and Drazen (2005); Shi and Svensson (2006); Vergne taxes and cut government investment in the two years after elections, (2009); Klomp and de Haan (2011); Philips (2016); and Strong (2023). with no significant cuts in government consumption. 22 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 BOX 1.1 Fiscal deteriorations around elections (continued) • In South Asia, primary fiscal deficits tended to subnational elections in the region’s two largest widen in or just before national elections, on countries between the 1960s and the mid-2000s. average by 0.5 percentage point of GDP, and only In India, the existing literature covers the 1960s to the half of this deterioration was reversed in the two mid-2000s, restricting itself to frequently held years after the election. For several South Asian subnational elections, and has documented narrowly countries, the literature finds evidence of narrowly targeted fiscal actions around state elections. Significant targeted fiscal actions around subnational elections. spending increases around state elections have been • Among EMDEs more generally, primary fiscal reported: in infrastructure-related social programs deficits, primary government expenditures, and during 1960–2005 (Khemani 2010); in interest government wage bills rose significantly around spending on subnational debt during 1960–2006 (Saez elections, on average by 0.7, 0.5, and 0.1 2016); in capital spending during 1959–2012 (Ferris percentage point of GDP, respectively. South Asia and Dash 2019, Khemani 2004); and in farm debt is among the EMDE regions with particularly waivers in 2001/02 and 2018/19 (Mahambare, pronounced election effects. Dhanaraj, and Mittal 2022). State banks appear to have increased agricultural lending around state elections • On average among EMDEs, primary spending during 1992–1999 (Cole 2009). Similarly, contested increases averaging 0.5 percentage point of GDP constituencies benefited from greater improvements in were virtually fully reversed within a year following power supply around state elections during 1992–2009 the election. However, increases in the government (Baskaran, Min, and Uppal 2015), while commodity wage bill—small (0.1 percentage point of GDP) revenue collections declined and capital spending rose but statistically significant—were not reversed: in around state elections during 1974–1995 (Chaudhuri fact, they continued. ere was wide variation in and Dasgupta 2006). the extent to which primary fiscal deficit increases in election years were reversed but, on average, the In Pakistan, government spending was significantly reversal amounted to less than half of the increase higher in election years and significantly lower after during the election years. e consequent elections during 2000–07 (Nasir, Nazir, and Khawaja ratcheting-up of deficits as well as wage bills around 2022). Fiscal deficits were significantly larger in election elections could erode fiscal sustainability and lock years during 1973–2009 (Sieg and Batool 2012). in spending rigidities over the longer term. In contrast, for Bangladesh, no study has shown clear is box draws on data for 122 EMDEs for 1984– evidence of political budget cycles. e one study 2022. Data on fiscal outcomes and country examining the question fails to find any systematic characteristics are from the IMF’s World Economic impact of political factors on disaster relief during 2010 Outlook and Government Finance Statistics databases, –14 (Karim and Noy 2020). However, monetary policy and the World Bank’s World Development Indicators. appears to have been significantly more accommodative Election dates are from the Database of Political in election years during 1980–2008 (Joarder, Hossein, Institutions until 2020 and assembled from news reports and Ahmed 2016). for 2021–2022. Event study Political budget cycles in South Asia An event study of government spending around national Fiscal positions deteriorated considerably around several elections in South Asia since 1991 suggests the presence elections in South Asian countries. e literature has of political budget cycles in most countries in the also found evidence of more narrowly targeted fiscal region, although of varying intensity. actions around subnational elections in several South Asian countries. Including elections planned for 2023–24, there will have been 53 presidential or parliamentary elections in Literature review the region since 1990: seven in Bangladesh; four in Bhutan; eight in India; eight in Pakistan; four in Nepal; e literature on political budget cycles in South Asia twelve in the Maldives; and ten in Sri Lanka. Elections has identified significant budget cycles around SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 23 BOX 1.1 Fiscal deteriorations around elections (continued) FIGURE B1.1.1 Fiscal positions around elections in South Asia There is a tendency toward a five-year bunching of elections in South Asia. Most elections have been accompanied by a considerable widening of primary fiscal deficits, and sometimes an increase in government spending, either in the election year or in the preceding year. The change in the primary balance tended to be partially reversed in the two years following the elections, although the increase in primary spending has not. A. Elections B. Change in primary fiscal balance C. Change in primary government around elections expenditures around elections Percent of SAR GDP Number of countries Percentage points of GDP Percentage points of GDP 100 Number of countries (RHS) 5 0.4 0.8 Share of SAR GDP 80 4 0.2 0.6 60 3 0.0 40 2 0.4 -0.2 20 1 0.2 -0.4 0 0 1980 1984 1988 1992 1996 2000 2004 2008 2012 2016 2020 2024 -0.6 0.0 Elections Post-elections Elections Post-elections Sources: The Database of Political Institutions 2020; International Monetary Fund; World Bank; de Haan, Ohnsorge, and Yu (forthcoming). Note: Based on 28 parliamentary or executive elections in SAR since 1991. Unweighted averages unless otherwise indicated. “Elections” is defined as the unweighted average of the change in the year before the election and the election year; “Post-elections” is defined as the unweighted average of the change in the year following the election and the subsequent year. in South Asia have tended to be bunched together primary government expenditures, primary fiscal about every five years, as they are again in 2023–24. balances, and the government wage bill (all in percent of GDP). e sample includes up to 122 EMDEs for For the seven South Asian countries in the sample, on 1984–2022. e regression results are shown in annex average, the primary fiscal balance deteriorated either in tables 1.1.1–1.1.3. the election year or in the year preceding the election by 0.5 percentage point of GDP; only half of this EMDE elections were typically accompanied by a fiscal deterioration was unwound over the two years following deterioration (figure B1.1.2). In the average election the election (figure B1.1.1). In some cases, the widening year, the primary deficit widened by 0.7 percentage of primary fiscal deficits in the runup to elections point of GDP, mostly because primary government reflected spending increases. spending rose by 0.5 percentage point of GDP. Government wage bills were higher in election years, on Political budget cycles in South Asia and other average by 0.1 percentage point of GDP. For EMDEs robustness, other components of government spending, Data for the South Asia region are severely limited and including government investment, were tested for provide an inadequate basis for reliable policy lessons to similar systematic changes around elections but none be drawn. To broaden the analysis of political budget was found. cycles, the larger group of EMDEs is examined (de Haan, Ohnsorge, and Yu forthcoming). On the Primary government spending increases around assumption that the typical South Asian country elections were short-lived and typically reversed within a behaves similarly to the typical EMDE, once the main year—with the exception of increases in government country characteristics are controlled for, lessons may be wage bills, which accounted for about 25 percent of inferred for South Asia. primary expenditures in the average EMDE during 2010–20. us, the 0.5-percentage-point of GDP A generalized-method-of-moments regression is increase in primary government spending in the average estimated of fiscal outcomes on elections (annex 1.1.1). election year was virtually entirely unwound in the year e fiscal outcomes that are examined are aggregate after the election. In contrast, the smaller (0.1 24 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 BOX 1.1 Fiscal deteriorations around elections (continued) FIGURE B1.1.2 Political budget cycles in EMDEs In election years and years preceding elections in EMDEs, primary fiscal deficits, primary fiscal spending, and the government wage bill tended to increase significantly. While primary spending increases have tended to be largely reversed in the following year, post-election reversals of primary deficits and government wage bills have been more variable. A. Change in primary fiscal balance B. Change in primary government C. Change in government wage bill around elections expenditures around elections around elections Percentage points of GDP Percentage points of GDP Percentage points of GDP 1.0 1.0 0.3 0.5 0.2 0.5 0.0 0.1 0.0 -0.5 0.0 -1.0 -0.5 -0.1 -1.5 -1.0 -0.2 Preceding Election Subsequent Preceding Election Subsequent Preceding Election Subsequent year year year year year year year year year Sources: The Database of Political Institutions 2020; de Haan, Ohnsorge, and Yu (forthcoming); International Monetary Fund; World Bank. Note: Coefficient estimates from a generalized-method-of-moments (GMM) panel regression of elections on the primary fiscal balance (in percent of GDP), primary government expenditures (in percent of GDP), and compensation of employees (in percent of GDP), controlling for country characteristics. The sample includes up to 122 EMDEs for 1984–2022. Yellow whiskers indicate 90 percent confidence intervals unless otherwise specified. “Preceding year” is the coefficient on a lead of the election variable, and “Subsequent year” is the coefficient on the lagged election variable. percentage point of GDP) but statistically significant B1.1.3). Election-year increases in fiscal deficits, increase in the government wage bill in the election year primary balances, or government wage bills were was not systematically unwound in the post-election statistically significant only in three regions, and South year. With regard to the primary fiscal balance, the Asia was one of the only two regions where all three unwinding was more variable and, on average, smaller, fiscal outcomes increased significantly. e inclusion of than in the case of primary expenditure: in fact, there regional dummy variables left other coefficient estimates was too much heterogeneity in post-election movement broadly unchanged. In South Asia, specifically, election for a statistically significant unwinding of the election- years were associated with 0.6 percentage point of GDP year increase to be identified. higher primary deficits, 0.8 percentage point of GDP higher primary spending, and 0.2 percentage point of Since increases in the primary fiscal deficit and the GDP higher government wage bills than in non- government wage bill around elections are not election years. Two of these three regions (including systematically unwound after elections, they can South Asia) were also the regions with more frequent cumulate to sizable increases over the course of several switches between fully democratic and less democratic elections. Since 1990, for example, the average EMDE political regimes. ese two regions accounted for two- in the regression sample has held a presidential or thirds of all regime switches in EMDEs during 1975– parliamentary election every three years. Assuming that 2022. fiscal deficits are financed by increases in government debt, the regression coefficients from annex table 1.1.1 Policy implications imply that government debt would already be more e empirical analysis suggests that deteriorations in than 10 percentage points of GDP higher and the fiscal positions, stemming particularly from spending government wage bill 0.6 percentage point of GDP increases, have been common in EMDEs at election higher than initially by the time that the fourth election times, regardless of political regimes. Yet the evidence cycle takes place. that such fiscal actions affect election outcomes is Together with three other regions, South Asia has had decidedly mixed. Insignificant, or even adverse, effects particularly pronounced political budget cycles (figure on election outcomes for the incumbent government SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 25 BOX 1.1 Fiscal deteriorations around elections (continued) FIGURE B1.1.3 Political budget cycles in EMDE regions South Asia is among the regions with the most pronounced changes in fiscal positions around elections. A. Change in primary fiscal balances in B. Change in primary spending in C. Change in wage bills in elections elections elections Percentage points of GDP Percentage points of GDP Percentage points of GDP 0.5 0.6 1.5 0.0 0.4 1.0 -0.5 0.2 -1.0 0.5 0.0 -1.5 -2.0 0.0 -0.2 LAC SAR SSA EAP LAC SAR SSA LAC SAR SSA Sources: The Database of Political Institutions 2020; de Haan, Ohnsorge, and Yu (forthcoming); International Monetary Fund; World Bank. Note: Coefficient estimates from a generalized-method-of-moments (GMM) panel regression of elections on the primary fiscal balance (in percent of GDP), primary government expenditures (in percent of GDP), and compensation of employees (in percent of GDP), controlling for country characteristics. The sample includes up to 122 EMDEs for 1984–2022. Yellow whiskers indicate 90 percent confidence intervals unless otherwise specified. LAC stands for “Latin America and the Caribbean”, EAP stands for “East Asia and the Pacific”, SAR stands for “South Asia”, and SSA stands for “Sub-Saharan Africa. have been reported for the United States (Peltzman 2006 a, b; Gootjes and de Haan 2022). 1992), EMDEs in Eastern Europe (Enkelmann and Leibrecht 2013) and Latin America (Kraemer 1997), • Fiscal rules. Fiscal rules, such as the Stability and and a large cross-country sample (Brender and Drazen Growth Pact in Europe and balanced budget 2008). However, more recent cross-country studies requirements in some U.S. states, can constrain have reported that incumbents have benefited from incumbents’ ability to engage in election-motivated fiscal actions in elections (Bojar 2017; Klomp and de fiscal expansions (de Haan and Klomp 2013; Rose Haan 2013), while voters sometimes punished 2006; Alt and Rose 2009; Cioffi, Messina, and incumbents for fiscal consolidation (Mulas-Granados Tommasino 2012; Ebeke and Ölçer 2017; Gootjes, 2004), although at other times they did not (Alesina de Haan, and Jong-A-Pin 2021). A growing 2012). number of countries have adopted such institutional fiscal constraints: in 2015, 92 e lack of reversals of fiscal deficit increases around countries had fiscal rules in place, up from 7 elections raises concerns about an erosion of fiscal countries in 1990 (Lledó et al. 2017). sustainability over the longer term. Similarly, even the small, but statistically significant, ratcheting-up of • Robust governance and control of corruption. Political government wage bills around elections in the average budget cycles have been less pronounced in EMDE will tend to lock in spending rigidities that may countries with stronger checks and balances, become difficult to unwind in times of need. stronger rule of law, and less corruption (Streb, Lema, and Torrens 2009; Shi and Svensson 2006; To help prevent fiscal deteriorations around elections Lee and Min 2021). Checks and balances in the and their longer-term consequences, the establishment political system discourage incumbents from using of more robust fiscal frameworks and institutional policy for re-election purposes. arrangements could be considered, as suggested by the experience of other countries. ere are indications that, at least in some South Asian countries, the 2023–24 election season may break from • Fiscal transparency. ere is empirical evidence that past practice. In India, for example, the latest greater transparency in fiscal policymaking may government budget is on track for fiscal consolidation make election-motivated fiscal policy action less amid upcoming elections (World Bank 2023d). likely by making them more visible (Alt and Lassen 26 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 in India; 3.5 percent in the Maldives and Pakistan; characteristics can raise the reputational costs and 5 percent in Sri Lanka. Yet, most of these to governments of noncompliance and limit countries are among those in South Asia with the their ability to adjust targets around elections. highest government debt-to-GDP ratios. During Transparent implementation of fiscal rules, the pandemic, many countries activated escape alongside regular reporting and monitoring, clauses to suspend fiscal rules. This allowed can help build their credibility and increase governments to provide much-needed support for the probability of compliance (Andersen vulnerable groups, but also added substantial debt. 2013). No countries in South Asia are projected to achieve their budget or debt rule targets in 2023. Managing the energy transition Achieving compliance represents a challenge. As the world presses ahead with the energy The design of fiscal rules in South Asia is less transition, South Asia will need to improve its binding than in other EMDEs, which can energy efficiency to keep pace. In fact, the global diminish their effectiveness. For example, in a energy transition presents an opportunity for sample of EMDEs with fiscal rules over 2015–21, South Asia to upgrade technologies and lift those with weaker designs—measured in terms of productivity, cut pollution, reduce reliance on coverage, legal basis, monitoring, enforcement, energy imports, and create jobs. Currently, the and flexibility—had larger increases in debt energy intensity of the region’s output is twice the between 2015 and 2022. A variety of best global average, and substantially higher than in practices in the design of fiscal rules has been other EMDEs (figure 1.12). identified, and could be adopted by South Asian India and Pakistan already rank among the world’s governments (Caselli and Reynaud 2019). five EMDEs with the largest public investment in renewable energies (chapter 2). But substantial • Medium-term objectives with short-term private investment will be needed for firms to flexibility. A medium-term debt objective can adopt green technologies. Governments can provide the flexibility needed to prevent the support the adoption of energy-saving and low- erosion of political support for the rule during emission technologies. Measures could include adverse events. Various types of escape ensuring the availability of financing, incentivizing mechanisms can ensure that the rule remains a shift toward green energy by removing fossil-fuel applicable even during economic shocks subsidies, introducing carbon taxes, or introducing (Eyraud et al. 2018). In Germany and market-based regulation. Firms sometimes vastly Switzerland, for example, deviations from underestimate the savings from new technologies. target are accumulated over several years. As a result, improving access to information about Once this accumulation exceeds a certain the availability, cost-saving potential, and threshold, adjustments must be made over the competitiveness of green technologies can help next few years to reduce the fiscal deficit boost adoption at limited cost. Access to reliable (German Federal Ministry of Finance 2022; energy grids can encourage firms to phase out OECD 2011). energy inefficient backup energy systems. • Safeguards for priority spending. A fiscal rule The energy transition is likely to create shifts in that excludes capital expenditure can South Asia’s labor markets (chapter 3). In almost encourage public investment, but it needs all countries in the region, pollution-intensive jobs guardrails against creative accounting (IMF outnumber green jobs, and they account for 6–11 2009, 2018). Excluding capital expenditure percent of all jobs in the region. Pollution- from an expenditure rule can help ensure that intensive jobs tend to be concentrated among governments’ ability to raise recurrent lower-skilled and informal workers. Green jobs spending (such as public wage bills) around tend to be filled by higher-skilled and better-paid elections is limited. workers. Experience from economic • Transparency. Effective fiscal rules tend to be transformations in other countries suggests they clear and simple (Eyraud et al. 2018). Such can have significant employment and earnings SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 27 effects, both in the aggregate and for specific FIGURE 1.12 Managing the energy transition groups of workers. The energy transition will require the adoption of greener and more energy- efficient technologies by firms, but such innovation is held back in part by A wide range of policies can facilitate the necessary the tendency of firms to underestimate potential savings. In almost all adjustment in labor markets while protecting countries in South Asia, the share of workers in green jobs is less than that in pollution-intensive jobs. The energy transition will disproportionately vulnerable workers. These include: enhancing improve job prospects for better educated workers in the formal economy access to education and training, finance, and markets; measures to facilitate labor mobility; and A. Energy intensity, 2020 B. Average electricity consumption with old and new technologies strengthening social safety nets. Some countries Toe/thousand $US kWh have already begun to put such policies in place. 0.4 Other EMDEs World 0.10 Clutch motor Servo motor Nepal’s Green Resilient and Inclusive Development 0.3 0.08 Strategic Action Plan (GRID SAP) aims to combine 0.06 climate-conscious growth strategies and improved 0.2 0.04 air quality with equitable job creation. Channeling 0.1 0.02 investments into green initiatives and supporting 0.00 0 small and medium enterprises in areas that have SAR IND PAK BGD LKA Actual consumption Beliefs potential for growth (particularly finance, tourism, C. Shares of workers in green jobs D. Marginal probability of working in a resilient connectivity, renewable energy, forestry, and pollution-intensive jobs green job waste management, and agriculture), as envisioned Percent of workers Percentage points in the GRID, can foster labor participation and 14 Green jobs Pollution-intensive jobs 4 12 economic opportunities in the formal economy 10 2 while reducing job disparities. Similarly, 8 0 Bangladesh’s Climate-Smart Agriculture Investment 6 -2 Plan aims to support and protect the livelihoods of 4 -4 2 farmers, especially women, and improve rural 0 -6 Secondary Tertiary Informal workers’ green skill development, through capacity BGD IND LKA MDV NPL PAK education education building, knowledge sharing, and financial Sources: Annual Survey of Industries, India; European Commission; national statistical offices; OECD resources. Green Growth database; World Bank. Note: BGD = Bangladesh; EMDEs = emerging market and developing economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan; SAR = South Asia; Toe = tons of oil An important collateral benefit of the energy equivalent. A. Energy intensity is defined as energy consumption (in tons of oil equivalent) relative to nominal transition will be reduced air pollution. The GDP (in thousands of U.S. dollars) in 2020 (chapter 2). emission of pollutants that accompany South B. Estimates of mean electricity consumption based on hourly readings of electricity meters installed in one clutch and one servo motor sewing machine in each of 124 intensive treatment firms. Meter Asia’s currently energy-intensive production readings collected for every day in January. Mean baseline beliefs about daily electricity consumption by a clutch motor and servo motor sewing machine in the full sample of firms measured in the processes has been shown to cause material baseline survey. C. Green jobs are those in occupations with a positive share of environmentally friendly tasks. economic and human losses by depressing worker Pollution-intensive jobs are those with above-median pollution intensity. More details can be found in annex 3.1 (chapter 3). Labor force surveys are available for Bangladesh (2015), India (2018), Sri productivity and worsening health and education Lanka (2019), Maldives (2019), Nepal (2017), and Pakistan (2018). outcomes (Behrer, Choudhary, and Sharma 2023; D. Marginal probabilities as estimated in probit regressions of a dummy variable of being employed in a green job, conditional on being in an urban location, having completed secondary or tertiary World Bank 2023l). Nine of the world’s ten most- education, being aged 24–54 or 55 or older, and being informally employed (annex 3.1.3, annex 3.1.4, chapter 3). The regressions control for industry and subnational entity dummies. polluted cities are in South Asia (World Bank 2023l). Some 60 percent of South Asia’s population live in heavily polluted areas. Pollution in the region is often trapped in large airsheds shaped by climatology and geography that span multiple countries (World Bank 2023l). In some cities, the majority of air pollution originates from neighboring states and countries. Airshed-wide air quality management, requiring cooperation across multiple South Asian countries, can benefit both the region’s people and economy. 28 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX 1.1.1. Methodology that buoy government revenues), lagged government debt (to control for budget e panel regression estimate the effect of constraints that preclude any spending increases), elections on three fiscal outcomes. e election and the lagged dependent variable (to control for variable takes the timing of elections into account. path dependence). To mitigate concerns about Specifically, it prorates the months of the year up endogeneity and the inclusion of the lagged to the election date, as in Franzese (2000). e dependent variable, the regression is estimated fiscal outcomes that are examined are aggregate using a generalized method of moments (GMM) primary government expenditures, primary fiscal estimator as in Gootjes, de Haan, and Jong-A-Pin balances, and the government wage bill (all in (2021). e regression results are shown in annex percent of GDP). tables 1.1.1 -1.1.3. e results are robust to excluding insignificant variables such as real GDP e regression controls for the following country per capita. characteristics: per capita real GDP (in logs; which proxies shifts in voter preferences as incomes rise), e sample includes up to 122 EMDEs for 1984- real GDP growth (to capture business cycle-related 2022. ese EMDEs cover all types of political changes to fiscal outcomes), inflation (to control regimes. for inflation-related increases in nominal incomes ANNEX TABLE 1.1.1 Election effects (in percent of GDP) Primary balance Primary expenditures Compensation of employees Election -0.65*** 0.52*** 0.14** (0.18) (0.16) (0.06) GDP growth 0.08*** -0.04 -0.03*** (0.02) (0.04) (0.01) Lagged government debt 0.01*** -0.01*** 0.00 (0.00) (0.00) (0.00) Inflation -0.00*** 0.00*** -0.01 (0.00) (0.00) (0.01) Real GDP per capita (logs) 0.18 0.69 0.43 (0.30) (0.50) (0.37) Lagged primary balance 0.61*** (0.07) Lagged primary expenditure 0.78*** (0.04) Lagged compensation of employees 0.74*** (0.06) Constant -0.65*** 0.52*** 0.14** (0.18) (0.16) (0.06) Obs 3,011 3,011 1,521 Nr of countries 122 122 96 AR(1) p-val 0.00 0.00 0.00 AR(2) p-val 0.21 0.95 0.24 Sargan-Hansen test p-val 0.32 0.68 0.99 Cragg-Donald test p-val 0.00 0.02 0.00 Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions use the GMM estimator and include year dummies. “Election” is a numerical variable that is constructed using the approach detailed in Gootjes, de Haan, and Jong-A-Pin (2021). The sample includes up to 122 EMDEs for the period 1984-2022. See de Haan, Ohnsorge, and Yu (forthcoming) for details. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 29 ANNEX TABLE 1.1.2 Election timing effects in EMDEs (in percent of GDP) Primary balance Primary expenditures Compensation of employees Lead of election -0.31* 0.20 -0.03 (0.17) (0.17) (0.07) Election -0.65*** 0.52*** 0.14** (0.18) (0.16) (0.06) Lag of election 0.22 -0.44** 0.07 (0.17) (0.17) (0.07) Observations 2,892 3,011 3,011 2,892 3,011 3,011 1,521 1,521 1,521 Nr of countries 122 122 122 122 122 122 96 96 96 AR(1) p-val 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.25 AR(2) p-val 0.27 0.21 0.25 0.96 0.95 0.99 0.25 0.24 0.99 Sargan-Hansen test p-val 0.19 0.32 0.19 0.82 0.95 0.95 0.99 0.99 0.58 Cragg-Donald test p-val 0.00 0.00 0.00 0.01 0.02 0.02 0.00 0.00 0.00 Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions use the GMM estimator and include year dummies. All dependent variables are in percent of GDP. “Election” is a numerical variable that is constructed using the approach detailed in Gootjes, de Haan, and Jong-A-Pin (2021). “Lead of election” captures the “Election” in the following year, while “lag of election” captures the “Election” in the previous year. The same set of controls as in annex table 1.1.1 is included but not shown here for brevity. The sample includes up to 122 EMDEs over the period 1984-2022. See de Haan, Ohnsorge, and Yu (forthcoming) for details. ANNEX TABLE 1.1.3 Election effects Primary balance Primary expenditures Compensation of employees Election (EAP region) -0.31 0.43 0.37*** (0.40) (0.51) (0.11) Election (ECA region) -0.19 -0.16 -0.05 (0.35) (0.41) (0.09) Election (LAC region) -1.03*** 0.97*** 0.16* (0.26) (0.30) (0.10) Election (MNA region) 0.95 0.18 0.02 (0.70) (0.64) (0.44) Election (SAR region) -0.61** 0.79*** 0.20** (0.24) (0.18) (0.08) Election (SSA region) -0.90** 0.56* 0.17 (0.36) (0.29) (0.14) Observations 3,011 3,011 1,521 Nr of countries 122 122 96 AR(1) p-val 0.00 0.00 0.00 AR(2) p-val 0.19 0.94 0.22 Sargan-Hansen test p-val 0.10 0.93 0.99 Cragg-Donald test p-val 0.11 0.02 0.00 Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All regressions use the GMM estimator and include year dummies. “Election” is a numerical variable that is constructed using the approach detailed in Gootjes, de Haan, and Jong-A-Pin (2021). The same set of controls as in annex table 1.1.1 is included but not shown here for brevity. The sample includes up to 122 EMDEs for the period 1984-2022. See de Haan, Ohnsorge, and Yu (forthcoming) for details. 30 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 References tional Trade and Economic Development 21 (3): 347– 60. Aksoy, M. A., and B. Hoekman, eds. 2010. Food Bojar, A. 2017. “Do Political Budget Cycles Work? Prices and Rural Poverty. Washington, DC: World A Micro-level Investigation of Pre-electoral Budget- Bank. ing and its Electoral Consequences.” Electoral Studies Alesina, A. 2012. “Fiscal Policy After the Great Re- 45 (February): 29–43. cession.” Atlantic Economic Journal 40: 429–35. Brender, A., and A. Drazen. 2005. “Political Budget Alesina, A., and G. Tabellini. 1990. “A Positive The- Cycles in New versus Established Democracies.” ory of Fiscal Deficits and Government Debt.” The Journal of Monetary Economics 52 (7): 1271–95. Review of Economic Studies 57 (3): 403–14. Brender, A., and A. Drazen 2008. “How do Budget Ali, F. A., N. Fiess, and R. MacDonald. 2010. “Do Deficits and Economic Growth Affect Reelection Institutions Matter for Foreign Direct Investment?” Prospects? Evidence from a Large Panel of Coun- Open Economies Review 21 (2): 201–19. tries.” American Economic Review 98 (5): 2203–20. Alt, J. E., and D. Lassen. 2006a. “Transparency, Caselli, F., and J. Reynaud. 2019. “Do Fiscal Rules Political Polarization, and Political Budget Cycles in Cause Better Fiscal Balances? A New Instrumental OECD Countries.” American Journal of Political Variable Strategy.” IMF Working Paper 2019/049, Science 50 (3): 530–50. International Monetary Fund, Washington, DC. Alt, J. E., and D. Lassen. 2006b. “Fiscal Transparen- Caselli, F., and P. Wingender. 2018. “Bunching at 3 cy, Political Parties, and Debt in OECD Countries.” Percent: The Maastricht Fiscal Criterion and Gov- European Economic Review 50 (6): 1403–39. ernment Deficits.” IMF Working Paper 2018/182, International Monetary Fund, Washington, DC. Alt, J., and S. Rose. 2009. “Context-Conditional Political Budget Cycles.” In The Oxford Handbook of Chari, A., L. Jain, and N. Kulkarni. 2021. “The Un- Comparative Politics, edited by C. Boix and S. C. holy Trinity: Regulatory Forbearance, Stressed Banks Stokes. Oxford: Oxford University Press. and Zombie Firms.” NBER Working Paper 28435, National Bureau of Economic Research, Cambridge Andersen, T. M. 2013. “The Swedish Fiscal Policy MA. Framework and Intermediate Fiscal Policy Targets.” Swiss Journal of Economics and Statistics 149 (2): 231 Chaudhuri, K., and S. Dasgupta. 2006. “The Politi- –48. cal Determinants of Fiscal Policies in the States of Arteta, C., S. Kamin, and F. U. Ruch. 2023. India: An Empirical Investigation.” Journal of Devel- “Financial Spillovers of Rising U.S. Interest Rates.” opment Studies 42 (4): 640–61. In Global Economic Prospects, June, 99–122. Wash- Cioffi, M., G. Messina, and P. Tommasino. 2012. ington, DC: World Bank. “Parties, Institutions and Political Budget Cycles at Baskaran, T., Min, and Y. Uppal. 2015. “Election the Municipal Level.” Bank of Italy Temi di Discus- Cycles and Electricity Provision: Evidence from a sione (Working Paper) 885, Bank of Italy, Rome. Quasi-Experiment with Indian Special Elections.” Cole, S. 2009. “Fixing Market Failures or Fixing Journal of Public Economics 126 (C): 64–73. Elections? Agricultural Credit in India.” American Behera, H., V. Dhanya, K. Priyadarshi, and S. Goel. Economic Journal: Applied Economics 1(1): 219–50. 2023. “India @ 100.” Reserve Bank of India Bulletin Damania, R., E. Balseca, C. de Fontaubert, J. Gill, 77 (7): 89–103. K. Kim, J. Rentschler, J. Russ, and E. Zaveri. 2023. Behrer, A. P., R. Choudhary, and D. Sharma. 2023. Detox Development: Repurposing Environmentally “Air Pollution Reduces Economic Activity: Evidence Harmful Subsidies. Washington, DC: World Bank. from India.” Policy Research Working Papers Davoodi, H., P. Elger, A. Fotiou, D. Garcia-Macia, 10515, World Bank, Washington, DC. X. Han, A. Lagerborg, W. R. Lam, and P. A. Medas. Biswas, A. K. 2012. “Import Tariff Led Export Un- 2022. “Fiscal Rules and Fiscal Councils: Recent der-Invoicing: A Paradox.” The Journal of Interna- Trends and Performance during the COVID-19 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 31 Pandemic.” IMF Working Paper 2022/011, Interna- Fernández, A., M. W. Klein, A. Rebucci, M. tional Monetary Fund, Washington, DC. Schindler, and M. Uribe. 2016. “Capital Control Measures: A New Dataset.” IMF Economic Review 64 de Haan, J., and B. Gootjes. 2023. “Conditional (3): 548–74. National Political Budget Cycles.” In Elgar Encyclo- pedia of Public Choice, edited by C. Bjørnskov and Ferris, J. S., and B. B. Dash. 2019. “Expenditure R.M. Jong-A-Pin. Cheltenham, U.K.: Edward Elgar. Visibility and Voter Memory: A Compositional Ap- proach to the Political Budget Cycle in Indian States, de Haan, J., and J. Klomp. 2013. “Conditional Po- 1959–2012.” Economics of Governance 20 (2): 129– litical Budget Cycles: A Review of Recent Evidence.” 57. Public Choice 157 (3): 387–410. Franzese, R. J. 2000. “Electoral and Partisan Manip- de Haan, J., F. Ohnsorge, and S. Yu. Forthcoming. ulation of Public Debt in Developed Democracies, “Election-induced Fiscal Policy Cycles in Democrat- 1956–90.” In Institutions, Politics and Fiscal Policy, ic and Non-democratic Emerging Market and De- edited by Strauch, R. R. and J. von Hagen, 61–83. veloping Economies.” Mimeo. Boston, MA: Springer US. Dieppe, A. ed. 2021. Global Productivity: Trends, German Federal Ministry of Finance. 2022. Drivers, and Policies. Washington, DC: World Bank. “Germany’s Federal Debt Rule (Debt Brake).” Ger- Dovonou, V. O., and Z. L. Xie. 2023. “The Distri- man Federal Ministry of Finance, Berlin. butional Impact of Inflation in South Asia: An Em- Gill, I., and P. Nagle. 2022. “Inflation Could Wreak pirical Approach.” Policy Research Working Paper Vengeance on the World’s Poor.” Brookings (blog). 10527, World Bank, Washington, DC. March 18, 2022. https://www.brookings.edu/ Dubois, E. 2016. “Political Business Cycles 40 Years articles/inflation-could-wreak-vengeance-on-the- after Nordhaus.” Public Choice 166 (1): 235–59. worlds-poor/. Ebeke C., and D. Ölçer. 2017. “Fiscal Policy over Gootjes, B., and J. de Haan. 2022. “Do Fiscal Rules the Election Cycle in Low-Income Countries.” In Need Budget Transparency to Be Effective?” Europe- Fiscal Politics, edited by V. Gaspar, S. Gupta, and C. an Journal of Political Economy 75 (2): 102210. Mulas-Granados. Washington, DC: International Monetary Fund. Gootjes, B., J. de Haan, and R. Jong-A-Pin. 2021. “Do Fiscal Rules Constrain Political Budget Cycles?” Eckstein, D., V. Künzel, and L. Schäfer. 2021. Public Choice 188 (1): 1–30. “Global Climate Risk Index 2021 | Germanwatch e.V.” January 25, 2021. https:// Gray, S. T. 2021. “Recognizing Reality—Unification www.germanwatch.org/en/19777. of Official and Parallel Market Exchange Rates.” IMF Working Paper 2021/025, International Mone- Endegnanew, Y., M. Soto, and G. Verdier. 2017. tary Fund, Washington, DC. “Do Elections Affect the Wage Bill?” In Fiscal Poli- tics, edited by V. Gaspar, S. Gupta, and C. Mulas- Gwartney, J. D., R. G. Holcombe, and R. A. Law- Granados. Washington, DC: International Monetary son. 2006. “Institutions and the Impact of Invest- Fund. ment on Growth.” Kyklos 59 (2): 255–73. Enkelmann, S., and M. Leibrecht. 2013. “Political Ha, J., M. A. Kose, and F. Ohnsorge. 2023. “One- Expenditure Cycles and Election Outcomes: Evi- Stop Source: A Global Database of Inflation.” Jour- dence from Disaggregation of Public Expenditures nal of International Money and Finance 137. by Economic Functions.” Economics Letters 121 (1): 128–32. Ha, J., M. A. Kose, F. Ohnsorge, and H. Yil- mazkuday. 2019. “Sources of Inflation: Global and Eyraud, L., X. Debrun, A. Hodge, V. Lledo, and C. Domestic Drivers.” In Inflation in Emerging and De- Pattillo. 2018. “Second-Generation Fiscal Rules: veloping Economies: Evolution, Drivers, and Policies, Balancing Simplicity, Flexibility, and Enforceabil- editing by J. Ha, M. A. Kose, and F. Ohnsorge, 143 ity.” Staff Discussion Notes 18/04, International –206. Washington, DC: World Bank. Monetary Fund, Washington, DC. 32 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Han, K. 2022. “Political Budgetary Cycles in Auto- Kasyanenko, S., P. Kenworthy, F. U. Ruch, E. cratic Redistribution.” Comparative Political Studies Vashakmadze, D. Vorisek, and C. Wheeler. 2023. 55 (5): 727–56. “Regional Dimensions of Investment: Moving in the Right Direction?” In Falling Long-Term Growth Pro- Heilbron, A., and R. Whyte. 2019. Institutions for spects: Trends, Expectations, and Policies, edited by M. Investment: Establishing a High-Performing Institu- A. Kose and F. Ohnsorge, 211-85. Washington, tional Framework for Foreign Direct Investment. DC: World Bank. Washington, DC: World Bank. Khemani, S. 2004. “Political Cycles in a Developing Higashijima, M. 2022. The Dictator's Dilemma at the Economy: Effect of Elections in the Indian States.” Ballot Box: Electoral Manipulation, Economic Maneu- Journal of Development Economics 73 (1): 125–54. vering, and Political Order in Autocracies. Ann Arbor: University of Michigan Press. Khemani, S. 2010. “Political Economy of Infrastruc- ture Spending in India.” Policy Research Working Hooijer, A., and R. Vernimmen. 2021. “Global Li- Paper 5423, World Bank, Washington, DC. DAR Land Elevation Data Reveal Greatest Sea-Level Rise Vulnerability in the Tropics.” Nature Commu- Kilic Celik, S., M. A. Kose, F. Ohnsorge, and F. U. nications 12 (1): 3592. Ruch. 2023. “Potential Growth: A Global Data- base.” Policy Research Working Paper 10354, World IMF (International Monetary Fund). 2009. “Fiscal Bank, Washington, DC. Rules—Anchoring Expectations for Sustainable Pub- lic Finances.” International Monetary Fund, Wash- Klomp, J., and J. de Haan. 2011. “Do Political ington, DC. Budget Cycles Really Exist?” Applied Economics 45 (3): 329–41. IMF (International Monetary Fund). 2018. “How to Select Fiscal Rules? A Primer.” FAD How-To Note Klomp, J., and J. de Haan 2013. “Political Budget 9, International Monetary Fund, Washington, DC. Cycles and Election Outcomes.” Public Choice 157 (1/2): 245–67. IMF (International Monetary Fund). 2020. “State- Owned Enterprises: The Other Government.” In Kose, M. A., S. Kurlat, F. Ohnsorge, and N. Fiscal Monitor - April 2020, 47–74. Washington, Sugawara. 2022. “A Cross-Country Database of Fis- DC: International Monetary Fund. cal Space.” Journal of International Money and Fi- nance 128 (November): 102682. IMF (International Monetary Fund). 2022. Annual Report on Exchange Arrangements and Exchange Re- Kose, M. A., and F. Ohnsorge, eds. 2023. Falling strictions 2021. Washington, DC: International Long-Term Growth Prospects: Trends, Expectations, Monetary Fund. and Policies. Washington, DC: World Bank. Joarder, M. A. M., A. N. Hossain, and M. U. Ah- Kraemer, M. 1997. “Electoral Budget Cycles in Lat- med. 2016. “Does the Central Bank Contribute to in America and the Caribbean: Incidence, Causes the Political Monetary Cycles in Bangladesh?” Eco- and Political Futility.” Inter-American Development nomic Change and Restructuring 49 (4): 365–94. Bank Working Paper, Washington, DC. Kalemli-Ozcan, S., H. Kamil, and C. Villegas- Kyriacou, A., T. Okabe, and O. Roca-Sagales. 2022. Sanchez. 2016. “What Hinders Investment in the “Conditional Political Budget Cycles: The Role of Aftermath of Financial Crises: Insolvent Firms or Time Preference.” Economics and Politics 34 (1): 67– Illiquid Banks?” The Review of Economics and Statis- 91. tics 98 (4): 756–69. Laeven, L., and F. Valencia. 2020. “Systemic Bank- Karim, A., and I. Noy. 2020. “Risk, Poverty or Poli- ing Crises Database II.” IMF Economic Review 68 tics? The Determinants of Subnational Public (2): 307–61. Spending Allocation for Adaptive Disaster Risk Re- duction in Bangladesh.” World Development 129 Lee, D., and S. Min. 2021. “Defective Democracy (May): 104901. and the Political Budget Cycle.” Journal of Compara- tive Economics 49 (4): 947–61. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 1 33 Lledó, V., S. Yoon, X. Fang, S. Mbaye, and Y. Kim. Data Evidence from India.” Governance 29 (1): 47– 2017. “Fiscal Rules at a Glance.” Background Paper, 65. International Monetary Fund, Washington, DC. Shi, M., and J. Svensson. 2006. “Political Budget Mahambare, V., S. Dhanaraj, and P. Mittal. 2022. Cycles: Do They Differ across Countries and Why?” “The Political Budget Cycles in the Presence of a Journal of Public Economics 90 (8–9): 1367–89. Fiscal Rule: The Case of Farm Debt Waivers in In- dia.” Journal of Policy Modeling 44 (3): 701–21. Sieg, G., and I. Batool. 2012. “Pakistan, Politics and Political Business Cycles.” The Pakistan Development Melecky, M. 2021. Hidden Debt: Solutions to Avert Review 51 (1): 153–66. the Next Financial Crisis in South Asia. Washington, DC: World Bank. Stamm, K. K., and D. Vorisek. 2023. “The Global Investment Slowdown: Challenges and Policies.” In Mulas-Granados, C. 2004. “Voting against Spending Falling Long-Term Growth Prospects: Trends, Expecta- Cuts – The Electoral Costs of Fiscal Adjustments in tions, and Policies, edited by M. A. Kose and F. Europe.” European Union Politics 5 (4): 467–93. Ohnsorge, 153-210. Washington, DC: World Bank. Nasir, M., S. Kishwar, and M. Meyer. 2023. “Losing Steinkamp, S., and F. Westermann. 2022. to Inflation: Unmasking Inflation Inequality among “Development Aid and Illicit Capital Flight: Evi- Households.” Unpublished paper, World Bank, dence from Nepal.” The World Economy 45 (7): 2305 Washington, DC. –36. Nasir, M., R. Nazir, and I. Khawaja. 2022. “Political Streb, J., D. Lema, and G. Torrens. 2009. “Checks Budget Cycle: A Sub-National Evidence from Paki- and Balances on Political Budget Cycles: Cross- stan.” Journal of Business Cycle Research 18 (3): 343– Country Evidence.” Kyklos 62 (3): 425–46. 67. Strong, C. O. 2023. “Tying One’s Hand: The Effect of Fiscal Rules on the Political Business Cycle in Af- Nordhaus, W. D. 1975. “The Political Business Cy- rica.” Journal of African Economics 32 (4): 438–67. cle.” Review of Economic Studies 42 (2): 169–90. Thompson, V., D. Mitchell, G. C. Hegerl, M. Col- OECD (Organization for Economic Cooperation lins, N. J. Leach, and J. M. Slingo. 2023. “The Most and Development). 2011. Restoring Public Finances. At-Risk Regions in the World for High-Impact Paris: OECD Publishing. Heatwaves.” Nature Communications 14 (1): 2152. OECD (Organization for Economic Cooperation Vergne, C. 2009. “Democracy, Elections and Alloca- and Development). 2015. “Lifting Investment for tion of Public Expenditures in Developing Coun- Higher Sustainable Growth.” In OECD Economic tries.” European Journal of Political Economy 25 (1): Outlook 2015/1. Paris: OECD Publishing. 63–77. Peltzman, S. 1992. “Voters as Fiscal Conservatives.” World Bank. forthcoming. Strategies for India’s Leap Quarterly Journal of Economics 107 (2): 327–61. to High Income Status. India Country Economic Memorandum. Washington, DC: World Bank. Philips, A. 2016. “Seeing the Forest through the Trees: A Meta-Analysis of Political Budget Cycles.” World Bank. 2016. Global Economic Prospects: Spillo- Public Choice 168 (3): 313–41. vers amid Weak Growth. January. Washington, DC: World Bank. Rentschler, J., M. Salhab, and B. A. Jafino. 2022. “Flood Exposure and Poverty in 188 Countries.” World Bank. 2020. “Country Partnership Frame- Nature Communications 13 (1): 3527. work for the Kingdom of Bhutan for the Period FY2021-24.” World Bank, Washington, DC. Rose, S. 2006. “Do Fiscal Rules Dampen the Politi- cal Business Cycle?” Public Choice 128 (3): 407–31. World Bank. 2021. “Fiscal Policy for Sustainable Development: Nepal Public Expenditure Review.” Sáez, L. 2016. “The Political Budget Cycle and Sub- World Bank, Washington, DC. national Debt Expenditures in Federations: Panel 34 CHAPTER 1 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 World Bank. 2022a. Pakistan Country Climate and World Bank. 2023e. “Maldives Development Up- Development Report. Washington, DC: World Bank. date: Navigating a Tight Line.” April. World Bank, Washington, DC. World Bank. 2022b. Bangladesh Country Climate and Development Report. Washington, DC: World World Bank. 2023f. “Pakistan Development Up- Bank. date: Recent Economic Developments, Outlook, and Risks.” April. World Bank, Washington, DC. World Bank. 2022c. “Bhutan Development Up- date.” October. World Bank, Washington, DC. World Bank. 2023g. “Sri Lanka Development Up- date: Time to Reset.” April. World Bank, Washing- World Bank. 2022d. South Asia Economic Focus: ton, DC. Coping with Shocks: Migration and the Road to Resili- ence. Fall. Washington, DC: World Bank. World Bank. 2023h. “Nepal Development Update: Fine-Tuning Policy in a Turbulent Environment.” World Bank. 2022e. From Swimming in Sand to April. World Bank, Washington, DC. High and Sustainable Growth: A Roadmap to Reduce Distortions in the Allocation of Resources and Talent in World Bank. 2023i. “Pakistan Federal Public Ex- the Pakistani Economy. Pakistan Country Economic penditure Review 2023.” World Bank, Washington, Memorandum, Washington, DC: World Bank. DC. World Bank. 2022f. “Maldives Public Expenditure World Bank. 2023j. “Sri Lanka: International Devel- Review: Restoring Fiscal Health.” World Bank, opment Association Program Document for a Pro- Washington, DC. posed Credit on Shorter Term Maturity Loan Terms.” World Bank, Washington, DC. World Bank. 2023a. “Afghanistan Economic Moni- tor.” Issue 8/2023, World Bank, Washington, DC. World Bank. 2023k. “Pakistan Development Up- date: Recent Economic Developments, Outlook, and World Bank. 2023b. South Asia Economic Focus: Risks.” World Bank, Washington, DC. Expanding Opportunities: Toward Inclusive Growth. Spring. Washington, DC: World Bank. World Bank. 2023l. Striving for Clean Air: Air Pollu- tion and Public Health in South Asia. Washington, World Bank. 2023c. “Bangladesh Development Up- DC: World Bank. date: Trade Reform: An Urgent Agenda.” April. World Bank, Washington, DC. Zhang, L., W. Cheng, E. Cheng, and B. Wu. 2020. “Does Land Titling Improve Credit Access? Quasi- World Bank. 2023d. “India Development Update.” Experimental Evidence from Rural China.” Applied Spring. World Bank, Washington, DC. Economics 52 (2): 227–41. SPOTLIGHT An ounce of prevention, a pound of cure: Averting and dealing with sovereign debt default SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 37 Spotlight. An ounce of prevention, a pound of cure: Averting and dealing with sovereign debt default South Asia has the highest average government debt-to-GDP ratio among emerging market and developing economy (EMDE) regions, at 86 percent in 2022, and four of the region’s countries are rated in or near sovereign debt distress. The risk of sovereign defaults in the region is heightened not only by high levels of government debt but also by the increases in global interest rates over the past two years: the vast majority of past defaults occurred around the end of U.S. monetary policy tightening cycles and in countries with above- median government debt-to-GDP ratios. Past experience also shows that more than one-third of defaults failed to lower government debt or borrowing costs in a lasting manner. Defaults that succeeded in lowering debt or borrowing cost were accompanied more frequently than others by above-median debt restructurings, growth accelerations, and fiscal consolidations. South Asia’s above-average economic growth mitigates some of the default risks. Some South Asian countries have reduced their default risk by predominantly borrowing from domestic creditors. However, this strategy comes at a price: high domestic shares of government debt have been associated with higher borrowing costs and lower bank credit to the private sector. With the external environment likely to remain challenging over the next several years, it is all the more important to adopt policies to accelerate sustainable growth and shore up fiscal positions. Introduction • What were the salient features of past defaults? South Asia has the highest average government • What were the characteristics of defaults that debt-to-GDP ratio among EMDE regions, at 86 succeeded in improving fiscal positions? percent in 2022 (figure SL.1). Four South Asian countries are already rated by the Moody’s ratings • What are the policy implications? agency or by the IMF/World Bank Debt The literature on debt defaults is extensive and has Sustainability Analysis as in or near sovereign-debt been growing rapidly. A key focus has been on distress and default. the short-term debt dynamics around debt defaults (Asonuma and Trebesch 2016; Erce, Sovereign debt defaults are costly. A large Mallucci, and Picarelli 2022). A few studies have literature has identified lasting output losses and examined longer-term government debt dynamics borrowing cost increases after external debt after external debt defaults or through debt defaults, especially defaults that were accompanied reduction episodes. However, these studies have by banking crises and were not pre-emptive (box predominantly focused on the role of either fiscal SL.1). Defaults have also resulted in heavy social consolidation or face value reductions.1 This costs, including higher poverty and worse health spotlight extends this analysis to both external and outcomes (Farah-Yacoub et al. 2022). And many domestic debt defaults and to a wider range of debt defaults have left government debt-to-GDP correlates that appear to have helped lower ratios higher after the default than before government debt or interest rates on government (Benjamin and Wright 2009; Bolton, Gulati, and debt after a sovereign default. Panizza 2022). This spotlight reports several findings. In light of the current increased risk of debt default in South Asia, this spotlight reviews past experience and draws policy lessons. Specifically, it explores the following questions: 1 For studies on debt dynamics after external debt default, see Benjamin and Wright (2009); Bolton, Gulati and Panizza (2022); IMF (2023); Reinhart and Trebesch (2016); Sturzenegger and Zettelmeyer (2007). For studies on debt dynamics in debt reduction Note: This spotlight was prepared by Franziska Ohnsorge and episodes, see Baldacci, Gupta, and Mulas-Granados (2012) and Hayley Pallan. IADB (2023). 38 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE SL.1 Government debt government debt-to-GDP ratios or govern- Compared with other EMDE regions, South Asia has the highest average ment borrowing costs over the following five government debt-to-GDP ratio. years. A. Government debt B. South Asia: Government debt • Mitigating factors in South Asia: Above-average Percent of GDP 140 2010 2022 Percent of GDP 140 Other EMDE regions SAR GDP growth and preponderance of domestic 120 120 debt. South Asia’s above-average growth 100 100 mitigates some of the default risks arising 80 80 60 from its above-average government debt-to- 60 40 40 GDP ratios. On balance, however, regression 20 20 0 analysis suggests that the probability of default in the average South Asian EMDE is higher MDV NPL BTN IND BGD LKA PAK 0 SAR EMDEs than in the average EMDE. Debt risks in Sources: IMF (various staff reports); Kose et al. (2022); World Bank. Note: Latest data for government debt are for 2022. All debt data are end-year. some South Asian countries may also be A.B. Unweighted averages (at 2010–19 average prices and market exchange rates). mitigated by the high share of government A. Yellow whiskers indicate minimum-maximum range for seven South Asian economies, and the interquartile range for EMDEs. debt (on average, more than half) that is owed B. Yellow shaded area indicates range of GDP-weighted averages of other EMDE regions. Yellow line indicates GDP-weighted average for South Asia (SAR). to domestic creditors. In the past, predominantly domestically financed debt runups were less likely to end in default than • The how, when, and where of sovereign debt externally financed debt runups. However, defaults. The vast majority of defaults this strategy comes at a price: above-median (including defaults on external debt) since share of domestic debt have been associated 1979 have taken place within a year of the end with higher government borrowing cost, of a U.S. monetary policy tightening cycle, in shorter debt maturities, and a smaller share of countries without fiscal rules, and in countries bank credit allocated to the private sector. with above-median government debt-to-GDP ratios. The average debt default took almost • Policy priorities: Debt restructuring, boosting three years to resolve, mostly because external growth, and putting fiscal positions on a sound debt defaults required multi-year resolutions. footing. Slowing global growth, elevated global Fewer than half of all debt defaults since 1979 borrowing costs, and high government debt included a face value reduction rather than levels increase the probability of sovereign only a maturity extension and/or interest rate debt default, and reduce the likelihood that reduction. sovereign debt defaults, when they occur, will be successful. But the historical record also • Successful versus unsuccessful defaults. Histori- suggests that measures to boost domestic cally, fewer than two-thirds of sovereign debt growth and put fiscal positions on a sounder defaults successfully reduced the government footing will improve the chances of success. debt-to-GDP ratio or the effective interest rate on government debt five years later. Debt For the purposes of this study, a successful debt defaults that failed to lower government debt- default is defined as one that is followed by a to-GDP ratios were twice as likely as reduction in the government debt-to-GDP ratio successful ones to be followed by another or in the effective interest rate on government default within five years. debt, or both, over the subsequent five years. The effective interest rate on government debt is • Features of successful defaults. Sovereign debt defined as net interest payments relative to the defaults accompanied by above-median debt stock of government debt at the end of the restructurings, domestic or global growth previous year. accelerations, fiscal consolidations, and IMF- supported policy programs more frequently This spotlight draws on a dataset of 177 external turned out to be successful in reducing or domestic sovereign debt defaults in 64 EMDEs SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 39 over 1979–2018 (annex table SL.1.1). Data for FIGURE SL.2 The how, when, and where of sovereign external sovereign debt defaults—defaults on debt defaults private external creditors—are from Asonuma and On average in EMDEs, debt defaults in 1970–2018 took almost three years Trebesch (2016), while data for domestic to resolve, although domestic defaults were often resolved much faster. Most external defaults involved face value reductions but most domestic sovereign debt defaults—defaults on debt ones did not. Defaults were bunched around the end of U.S. Federal instruments governed by domestic law—are from Reserve monetary policy tightening cycles, and in countries with high debt Erce, Mallucci, and Picarelli (2022). For a subset -to-GDP ratios and without fiscal rules. South Asia’s above-average growth mitigates some of the default risk arising from its above-average of 88 external debt defaults between 1970 and government debt levels. 2014, estimates of the size of restructured debt, the “haircuts” (reductions in net present value), A. Average size and duration of B. Share of defaults, by duration defaults and face value reductions are available from Percent of GDP Months Percent of defaults 14 40 Cruces and Trebesch (2013). 12 35 50 Domestic External All 40 10 30 This spotlight builds on two quantitative exercises. 8 25 30 20 A panel regression estimation corrected for 6 15 20 4 10 selection bias is used to identify the correlates of 10 2 5 0 more successful defaults (annex SL.1). Separately, 0 0 <6 months 6-12 12-36 > 36 months months months an event study of past government debt booms— Size (LHS) Duration (RHS) encompassing overall, domestic, and external C. Share of face value reductions D. Shares of defaults in the most government debt—is conducted to explore the common circumstances consequences of the composition of government Percent of defaults Percent of defaults debt for debt default (annex SL.2). 80 Total 100 All defaults External defaults 80 60 The how, when, and where 40 60 40 of sovereign debt default 20 20 0 0 End of Fed Without fiscal High debt Among the 64 EMDEs in the sample, defaults Domestic External tightening rule occurred on average almost three times per E. Predicted probability of debt default F. Deviation in predicted probability of country between 1979 and 2018. Only in about debt default from EMDE average, 2024 one-third of the countries in the sample did Percent Percentage points 4 0.3 defaults occur just once in these four decades. 3 0.2 Defaults took a wide variety of forms (Bolton, 2 0.1 Gulati, and Panizza 2022). Three studies have 1 detailed the anatomy of sovereign debt defaults in 0 0.0 Baseline 1 st.dev. more 3.45 ppt higher 2.2 ppt lower 17 ppt higher dom. growth borrowing cost risk aversion debt ratio EMDEs (Asonuma and Trebesch 2016; Cruces -0.1 and Trebesch 2013; Erce, Mallucci, and Picarelli -0.2 MNA LAC EAP ECA SSA SAR 2022). They document the following findings. Sources: Asonuma and Trebesch (2016); Cruces and Trebesch (2013); Erce, Mallucci, and Picarelli (2022); World Bank. • Creditors. Two-thirds of the defaults in the Note: EAP = East Asia and Pacific; ECA = Europe and Central Asia; LAC = Latin America and the sample involved external creditors and most Caribbean; MNA = Middle East and North Africa; SSA = Sub-Saharan Africa. A. Derived as GDP-weighted averages of the unweighted average of defaults on domestic-law and (70 percent) of these external defaults foreign-law obligations in Erce, Mallucci, and Picarelli (2022). B. Derived from shares of defaults on domestic-law and foreign-law obligations in Erce, Mallucci, and occurred in the 1980s. Since 1990, external Picarelli (2022). debt defaults account for fewer than half of all C. For domestic defaults, data from Erce, Mallucci, and Picarelli (2022) for 134 defaults in 1979-2018. For external defaults, data from Cruces and Trebesch (2013) for 180 defaults in 1970-2010. defaults in the sample. D. Share of all defaults that occurred in the year of the end of U.S. Federal Reserve tightening cycle as defined in World Bank (2022) or in the subsequent year. Share of all defaults that occurred in countries without a fiscal rule or in countries with above-median (across the full EMDE sample) government debt at the time of default. All defaults include defaults on domestic and external • Magnitude. On average, EMDEs defaulted on creditors; external defaults refers to defaults on external creditors. Gray line denotes 50 percent. debt instruments equivalent to 12 percent of E.F. Based on coefficient estimates from probit regression, specified in annex SL.1. Baseline probability of debt default is the coefficient estimate of the constant in a probit regression of default on annual GDP (Figure SL.2; Erce, Mallucci, a constant. Sample includes 156 EMDEs between 1970–2022. E. Yellow whiskers indicate 95 percent confidence intervals. and Picarelli 2022). External debt defaults F. Calculation assumes that global risk sentiment and the U.S. federal funds rate remain constant at their 2023 values and that global and regional growth materializes as forecast in the June 2023 Global Economic Prospects report. 40 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 were larger than domestic debt defaults: the • National recessions. Defaults were also more affected debt instruments were equivalent in common during national recessions or in the value to 13 percent of GDP, compared with subsequent year. The sample includes 323 10 percent of GDP in domestic debt defaults. national recessions, defined by negative annual output growth, in the 64 EMDEs that • Resolution period. The average debt default defaulted at some point during 1979–2018. took almost three years to resolve through Hence, national recession years and the first restructuring (Erce, Mallucci, and Picarelli year after each national recession accounted 2022). Domestic debt defaults were resolved for 26 percent of the country-years in the much faster than external ones: more than sample. However, 47 percent of the defaults four-fifths of domestic debt defaults were occurred during these years. On average, there resolved in less than six months, compared were three defaults in every 100 national with just over one-tenth of external defaults. recession country-years, which is three times • Face value reductions. Macroeconomic as many as in every 100 non-recession conditions improved after a default only when country-years. it involved face value reductions (Reinhart Defaults were more common in countries with and Trebesch 2016). Just under half of all weaker fiscal institutions and fiscal positions. Most debt defaults involved reductions in the face defaults (70 percent since 1990) occurred in value of the defaulted debt, while the rest countries that did not have a fiscal rule in place at solely involved maturity extensions and/or the time. Similarly, 67 percent of defaults interest rate reductions. Face value reductions occurred in country-years when government debt were common in external debt defaults, was above the EMDE median. occurring in about two-thirds of cases (Cruces and Trebesch 2013), but rare in domestic An econometric exercise helps establish the main debt defaults, occurring in fewer than one- correlates of the probability of debt default. A fifth of cases (Erce, Mallucci, and Picarelli probit regression is used to model the probability 2022). of a debt default, drawing on the main factors identified in the literature on early warning Most defaults occurred during the 1980s. indicators for financial crises (e.g., Kaminsky and However, whether the 1980s are included in the Reinhart 2000). The probability of default is analysis or not, several regularities were found in modeled as a function of the U.S. federal funds the timing of their occurrence. rate (as a proxy for global borrowing costs), a • End of global policy tightening cycles. Defaults proxy for global investor risk sentiment, changes were bunched around the end of U.S. in global GDP growth, lagged domestic GDP monetary policy tightening cycles. Thus, 63 growth, lagged currency depreciation, and the percent of defaults since 1990 occurred either lagged government debt-to-GDP ratio. For lack of in the year that a U.S. monetary policy a measure with a sufficiently long time series, tightening ended or in the next year (Figure global risk sentiment is proxied by the excess SL.2). number of domestic and external debt defaults globally that cannot be explained by the U.S. • Global recessions. Defaults were also bunched federal funds rate and changes in global growth. around global recessions. Global recessions are rare events: they happened only about once a The general patterns described above also emerge decade in the sample period (1982, 1991, in the econometric estimation of the probability of 2009). Yet 23 percent of defaults occurred default in the more limited sample of countries during these three years or the years and years for which a complete set of data is immediately following. As a result, the average available (annex table SL.1.2; Figure SL.2). global recession was accompanied by almost Defaults occurred in about 2 percent of country- three defaults whereas the average non- year pairs. Defaults were more likely in years when recession year featured only two. global borrowing costs (as proxied by the U.S. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 41 federal funds rate) or investor risk aversion (as FIGURE SL.3 Features of successful debt defaults proxied by global excess defaults) were higher, Since 1990, fewer than two-thirds of sovereign debt defaults were after a domestic growth slowdown, or when successful in lowering government debt-to-GDP ratios or effective interest rates on government debt five years after the default. Defaults on the eve of government debt was high. For example, the global or domestic growth accelerations, with fiscal consolidation, or coefficient estimates suggest that a 3.5-percentage- during IMF-supported policy programs were more often associated with point increase in the U.S. federal funds rate—the success than other defaults. External defaults with above-median haircuts and restructurings were also more often successful. actual change since early 2022—would have increased the probability of default by almost one- A. Share of successful defaults B. Number of defaults half. South Asia’s above-average growth mitigates Percent of defaults Number of defaults some of the default risk arising from its above- 100 Since 1979 Since 1990 40 Successful Unsuccessful average government debt level. On balance, 80 30 however, the regression estimates suggest that the 60 20 probability of default in the average South Asian 40 10 EMDE is higher than in the average EMDE. 0 20 1980s 1990s 2000s 2010s 1980s 1990s 2000s 2010s A large literature has estimated the short- and 0 Debt-reducing Interest-rate reducing Debt-reducing Interest-rate reducing long-term macroeconomic and financial consequences of debt default (box SL.1). It finds C. Share of debt-reducing defaults in D. Share of debt-reducing external that external sovereign debt defaults have been the most common circumstances defaults, by restructuring terms associated with significant and often lasting Percent of defaults Within the indicated group Otherwise Percent of defaults Within the indicated group Otherwise 100 100 output losses, increases in borrowing costs, 80 80 exclusion from international capital markets, and 60 60 40 40 disruptions in trade and financial systems. In the 20 20 0 short term, trade flows plunged temporarily as 0 Above-median Above-median IMF program consolidation acceleration acceleration restructuring Domestic trade credit was disrupted, sovereign bond yields growth growth Global haircut Fiscal rose by as much as 400 basis points, and output declined by 1–6 percent. Since 2000, a return to international capital markets has, on average, E. Share of interest rate-reducing F. Share of interest rate-reducing taken about two years, somewhat less time than in defaults in the most common external defaults, by restructuring circumstances terms the 1980s and 1990s. In the medium term, output Percent of defaults Percent of defaults was still 3–10 percent lower five years after default 100 Within the indicated group Otherwise 100 Within the indicated group Otherwise and, when reductions in the face value of debt 80 80 60 60 were large, sovereign borrowing costs were still 40 40 20 more than 100 basis points higher. 20 0 0 median restructuring Above- haircut acceleration acceleration consolidation IMF program Domestic median Above- growth growth Global Fiscal Features of successful debt defaults Sources: Asonuma and Trebesch (2016); Cruces and Trebesch (2013); International Monetary Fund; World Bank. Fewer than two-thirds of the sovereign debt Note: Default episodes are differentiated between those that featured lower (“Successful”) or higher (“Unsuccessful”) government debt-to-GDP ratios or effective interest rates on government defaults in the sample were successful in the sense debt five years after the default than in the year of default. Based on 177 domestic or external default episodes in 64 EMDEs during 1979–2018. that government debt-to-GDP ratios or effective C.E. Defaults since 1990. Share of successful defaults among defaults under the circumstances indicated on the horizontal axis. “IMF program” indicates that an IMF-supported policy program interest rates on government debt were lower after was in place at the time of default; “Fiscal consolidation” indicates an improvement in the cyclically five years (Figure SL.3). Five years after successful adjusted fiscal balance between the year of default and five years after default; “Global growth acceleration” and “Domestic growth acceleration” indicate a two-year global or domestic growth defaults, government debt was, on average, 24–34 acceleration from the time of default. D.F. “Above-median restructuring” indicates above-median size of restructured debt in percent of percentage points of GDP lower, whereas it was total government debt at time of default, as calculated by Cruces and Trebesch (2013). “Above- median haircut” indicates above-median market haircut at time of default, as calculated by Cruces 5–27 percentage points higher after unsuccessful and Trebesch (2013). Sample includes 88 external debt defaults since 1979, of which 43 occurred from 1990 onwards. defaults. Effective interest rates on government debt were up to 2 percentage points lower five years after successful defaults, whereas they were up to 1 percentage point higher after unsuccessful 42 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE| OCTOBER 2023 BOX SL.1 Literature review: Costs of sovereign debt default Sovereign debt defaults have been associated with significant and sometimes lasting output losses, increases in borrowing costs, exclusion from international capital markets, and trade and financial system disruptions. In the short term, trade flows plunged temporarily as trade credit was disrupted, sovereign bond yields rose by as much as 400 basis points, and output declined by 1–6 percent. In the 2000s, a return to international capital markets took two years on average. Over the medium term, output was still 3–10 percent lower, on average, than before defaults. A large literature has identified the costs of debt default, eight years after default (Esteves, Lennard, and Kenny typically focusing on external debt defaults. Several 2021; Furceri and Zdzienicka 2011; Sturzenegger types of cost have been identified: output losses and 2004). Over the ten years it took, on average, to resolve disruptions to international trade, borrowing cost the 32 external debt defaults among 37 EMDEs during increases and exclusion from capital markets, financial 1970–1998, output was 5 percent below a no-default system disruptions, political strains, and social costs baseline (De Paoli, Hoggarth, and Saporta 2009). In a (Borensztein and Panizza 2009; Farah-Yacoub et al. few cases, output losses compared with a no-default 2022; Gelpern and Panizza 2022; Tomz and Wright baseline may have been as large as 23 percent over the 2013). This box reviews the empirical literature on the course of the resolution of debt default (Jorra 2011). following questions: While these level effects were long-lived, growth effects were short-lived. Levy-Yeyati and Panizza (2011) show • What have been the output and trade losses that trend (Hodrick-Prescott-filtered) output growth associated with debt default? did not change statistically significantly after default • What has been the impact of default on borrowing over the medium term. costs and private credit? Heterogeneity in output losses. Several studies have • How long were defaulting countries excluded from found that long-term output losses after default capital markets? depended on country characteristics and circumstances. Output losses were larger when external defaults Output and trade losses coincided with financial crises (Borensztein and Panizza 2009; Kuvshinov and Zimmermann 2019), when they Since government debt is typically the foundation of were not pre-emptive (Asonuma et al. 2020; Asonuma domestic financial systems, debt default disrupts and Trebesch 2016), when they were more coercive financial flows and economic transactions more (Trebesch and Zabel 2017), and when they were not broadly. This can cause lasting economic damage. accompanied by face value debt reductions (Reinhart Short-term output losses. For external debt defaults, and Trebesch 2016). Output losses associated with many studies have shown significant output declines in domestic debt defaults were larger than those associated the year of default. Most studies have identified output with external defaults. For example, among 40 EMDEs contractions in the year of external debt default of 1–6 during 1950–2010, domestic debt defaults were percent (Borensztein and Panizza 2009; Esteves, Kenny, followed by statistically significant 2.7 percent per and Lennard 2021; Furceri and Zdzienicka 2011; Tomz capita GDP losses after five years whereas external debt and Wright 2007). Causality in this correlation could defaults were not associated with statistically significant go in either direction. In quarterly data for 39 (mostly) lasting losses (Malinen and Ropponen 2019). Larger EMDEs during 1970–2005, Levy-Yeyati and Panizza post-default debt reductions were associated with (2011) show that output declines preceded default and smaller output losses (Forni et al. 2021). that output did not decline significantly further in the quarter of default. Over the longer term, Marchesi and Trade losses. By disrupting financial systems, debt Masi (2018) even find that output rose more after default can reduce trade credit and, hence, trade. Trade defaults in 1975–2013. credit and trade flows were indeed significantly lower in default years (Borensztein and Panizza 2009; Rose Long-term output losses. Several studies have shown 2005). Similarly, defaults were particularly damaging to lasting output losses after external default, with output export-oriented firms; however, the impact was short- levels 3–10 percent below a no-default baseline four to lived (Borensztein and Panizza 2010). SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 43 BOX SL.1 Literature review: Costs of sovereign debt default (continued) Borrowing cost increases Private borrowing cost increases. The increase in sovereign bond yields in the defaulting country tended Sovereign borrowing cost increases. After default, to spill over to private sector borrowing costs and sovereign bond spreads were as much as 400 basis private credit. Corporate borrowing spreads were 30– points higher in a sample of 31 EMDEs during 1997– 200 basis points higher during episodes of sovereign 2004, but the effect faded quickly (Borensztein and stress, although there is some evidence that the Panizza 2010). This may account for the fact that long- correlation between corporate and sovereign bond yields term average returns on sovereign bonds of EMDEs— weakened during sovereign defaults (Bevilaqua, Hale, despite defaults—outperformed benchmark returns by and Tallman 2020). Default led to credit rationing for 3–4 percentage points (Meyer, Reinhart, and Trebesch corporate borrowers that was large and persistent 2022). That said, in a larger sample of defaults during (Esteves and Jalles 2016). 1970–2014, borrowing spreads were still 200 basis points higher five years after default (Catão and Mano Private credit contractions and slowdowns. Defaults 2017). Borrowing cost increases were larger and longer- were associated with reductions in private credit. While lasting when haircuts on debt were larger: The average debt was being renegotiated in defaults during 1981– 40 percent haircut in a sample of 23 countries was 2004, foreign bond issuances and syndicated bank loans associated with 127 basis points higher spreads five declined by 30 percent during debt renegotiations and years later (Cruces and Trebesch 2013). Spreads were another 14 percent in the first year after restructuring also larger for serial defaulters (Catao and Mano 2017). (Arteta and Hale 2008). Domestic private sector credit growth was also significantly lower during defaults in Loss of capital market access. On average since the 81 countries during 1980–2005 (Gennaioli, Martin, 1970s, it took five to six years for partial capital market and Rossi 2014). In defaults during 1998–2012, banks access to be restored after default (Cruces and Trebesch with higher sovereign bond holdings reduced their 2013; Richmond and Dias 2008). The period of lending more than their less exposed peers (Gennaioli, exclusion appears to have shrunk from 4–6 years in the Martin, and Rossi 2018). Even foreign direct 1980s to 2–3 years in the 2000s (Richmond and Dias investment (FDI) flows declined after defaults, 2008; Gelos, Sahay, and Sandleris 2011). Market access especially from creditor countries (Fuentes and Saravia was delayed when haircuts on debt were large (Cruces 2010). and Trebesch 2013), when countries were smaller, or when international risk sentiment was unfavorable (Richmond and Dias 2009). ones. Defaults that succeeded in lowering were common to both the full sample of defaults government debt-to-GDP ratios were more likely and the subsample of external defaults.2 to “stick”: fewer than one-quarter of them relapsed into another default over the next five years, • Domestic growth accelerations. Defaults were compared with about one-half of unsuccessful more frequently successful when they defaults. Government debt defaults in the 1980s occurred at the beginning of a domestic were particularly prone to failing to reduce output growth spurt. Nine-tenths of all government debt relative to GDP. defaults that occurred just before a two-year growth acceleration were successful at Since 1990, there have been several circumstances lowering debt-to-GDP ratios or interest rates, in which defaults were more likely to succeed in reducing interest rates and government debt-to- 2 Defaults in the 1980s shared most of these features, but with GDP ratios over the subsequent five years (Figure global growth accelerations and fiscal consolidations playing smaller SL.3; annex table SL.1.3). These circumstances roles in helping to lower debt-to-GDP ratios in external defaults. 44 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE SL.4 Debt and borrowing costs after successful • IMF-supported policy programs. Two-thirds of and unsuccessful debt defaults fiscal consolidations in the sample were Reductions in government debt ratios in the five years after default were accompanied by IMF-supported policy steeper when domestic output growth was stronger and when global programs. As a result, more than two-thirds of investor risk sentiment was more favorable. Interest rate reductions were steeper when global growth was stronger, global borrowing costs declined defaults in the context of IMF-supported more, and global investor risk sentiment was more benign. programs were successful, compared with fewer than one-half of those that were not A. Predicted change in government B. Predicted change in effective accompanied by IMF-supported programs. debt ratio accompanying a change in interest rate on government debt correlates accompanying change in correlates • Debt restructuring. In the subsample of 88 Percentage points of GDP Percentage points 0 0.0 external debt defaults since 1970 for which -1 debt restructuring estimates are available, greater restructuring was more frequently -2 -0.5 associated with successful defaults (Cruces and -3 Trebesch 2013). Two-thirds of external debt -4 -1.0 defaults with an above-median share of Lower fed. Higher global Better risk Higher domestic growth Better risk sentiment funds rate growth sentiment restructured debt in total government debt were successful, compared with about one-half Source: World Bank. Note: Predicted change in government debt (A) or effective interest rate on government debt (B) is of other defaults. Similarly, almost nine-tenths defined as 1 percentage point increase in the variable indicated on the horizontal axis (0.5 of a of external debt defaults with above-median standard deviation decline in the case of global risk sentiment) during the sample period multiplied by the coefficient estimate from a panel regression estimation, controlling for selection bias, as haircuts turned out to be successful, compared described in annex SL.1. Charts only show calculations for statistically significant coefficient estimates. Yellow whiskers indicate 95 percent confidence intervals. with about one-half of other defaults. An econometric exercise helps establish the most compared with around one-half of defaults robust correlates of debt and borrowing cost that occurred without such growth reductions in the event of default. Correcting for accelerations. the selection of countries defaulting, a panel regression is used to estimate the effects of • Global growth accelerations. Defaults that macroeconomic conditions on the cumulative occurred on the eve of global growth changes in the government debt-to-GDP ratio and accelerations were also more likely to succeed. in the effective interest rate on government debt Since 1990, more than two-thirds of defaults over the five years following default. The that occurred just ahead of global growth correlates of these two dependent variables include accelerations lasting at least two consecutive cumulative changes in global GDP growth, in years were successful, about twice the global borrowing costs (proxied by the U.S. proportion that were successful in other years. federal funds rate), in global risk aversion (proxied This pattern was less pronounced in the by the excess number of defaults), and in domestic 1980s, when the vast majority of defaults were GDP growth over the five years following default. unsuccessful in lowering debt-to-GDP ratios The results are robust to several alternative despite a global growth spurt in the middle of specifications (annex SL.1). Ideally, the regression the decade. would also account for the effects of the features of national defaults, including the terms and • Fiscal consolidation. Defaults that were duration of restructuring agreements, and for the accompanied by fiscal consolidations were magnitude of fiscal consolidation (i.e., changes in considerably more likely to succeed. Almost the cyclically-adjusted or structural fiscal balance). three-quarters of defaults that were followed However, the data needed for such analysis are by an improvement in the cyclically adjusted available for too small a subset of the sample of fiscal balance over the subsequent five years debt defaults to yield meaningful results. were successful, compared with fewer than one-half of other defaults. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 45 The results of the panel regression broadly accord FIGURE SL.5 Government debt composition and with the patterns described above (Figure SL.4; features of debt booms annex SL.1; annex table SL.1.4). In South Asia and EMDEs more broadly, domestic debt accounts for most government debt and for most of the government debt buildup over the past decade. Since 2004, domestic debt booms have been more common • Government debt. The estimation shows that than external debt booms. While domestic and external debt booms have reductions in government debt-to-GDP ratios had similar average amplitudes, speeds, and durations, domestic in the five years after default tended to be government debt booms have been less likely than external debt booms to be associated with debt default. steeper when they were accompanied by stronger domestic growth and more favorable A. Composition of government debt B. South Asia: Share of domestic global investor risk sentiment. Thus, a 1- government debt, 2021 percentage-point increase in domestic growth Percent of GDP 100 External Domestic Percent of government debt 100 EMDEs after default was associated with a decline in 80 80 the government debt-to-GDP ratio that was 60 60 steeper by 2 percentage points of GDP. An 40 40 improvement in global investor risk sentiment 20 by half a standard deviation (approximately 0 20 2010 Latest 2010 Latest the improvement observed between 2020 and 0 SAR EMDEs IND PAK LKA NPL BGD 2021) was also associated with a similarly- sized decline in the government debt-to-GDP C. Share of countries in government D. Share of debt booms associated ratio. debt booms with default Percent Percent 80 Total Domestic External 10 • Government borrowing costs. While domestic 60 8 growth was the main correlate of reductions 6 in government debt-to-GDP ratios after 40 4 default, global growth and global interest rates 20 were the main correlates of reductions in 2 0 government borrowing costs. Specifically, a 1- 2004 2006 2008 2010 2012 2014 2016 2018 2020 0 Total Domestic External percentage-point decline in the U.S. federal funds rate, a one-half standard deviation E. Amplitude of government debt F. Duration of government debt improvement in global investor risk booms booms sentiment, or 1 percentage point faster global Percentage points of GDP 30 Years 10 growth was each associated with 0.3–0.6 of a 20 8 percentage point lower government interest 6 10 rates five years after default. 4 0 2 External domestic boom Domestic external boom Total during Total during 0 Domestic debt: A costly Domestic External mitigating factor Sources: Asonuma and Trebesch (2016); Erce, Mallucci, and Picarelli (2022); Kose et al. (2022); Reinhart and Rogoff (2011); World Bank. South Asia stands out among EMDE regions for Note: BGD = Bangladesh; IND = India; LKA = Sri Lanka; NPL = Nepal; PAK = Pakistan; SAR = South Asia region. governments’ high reliance on domestic A.B. Latest data are for 2021 for measures of debt composition. Weighted averages for South Asia borrowing (Figure SL.5). It accounts for almost (SAR) and EMDEs. SAR includes Bangladesh and India (from Kose et al. 2022) as well as Nepal, Pakistan, and Sri Lanka (from various IMF Article IV staff reports). Domestic government debt for three-quarters of government debt accumulation Nepal, Pakistan, and Sri Lanka is domestic currency-denominated debt; for Bangladesh and India domestic government debt is debt held by domestic residents. between 2010 and 2021 in the average South C.-F. Debt booms are defined as debt accumulations where in at least one year the debt-to-GDP ratio rises above its Hodrick-Prescott-filtered trend by more than one standard deviation. Gaps in the Asian country, compared with less than two-thirds domestic and external government debt series are interpolated. There have been 105 government debt booms, 53 domestic government debt booms, and 34 external government debt booms since in the average EMDE. At end-2021 (the latest 2004. available data), it accounted for 56 percent of C. Lines show the share of countries per year experiencing a total government debt boom, domestic government debt boom, or external government debt boom. The denominator is the total number of government debt in the average South Asian countries with available data per year and per debt type. D. Bars show the share of total government debt booms associated with default (of any type), country and more than half in three of the domestic government debt booms associated with domestic default, and external government debt booms associated with external default, up to one year after the end of a boom. region’s four largest economies. E.F. Bars show the average amplitude and duration of government debt booms. Yellow whiskers correspond to the interquartile range. Amplitude is defined as the change in government debt from the start to the end of a boom. Duration is the number of years the government debt boom lasts. 46 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE SL.6 Costs of domestic debt Greater reliance on domestic debt is associated Above-median shares of domestic debt in government debt have been with a lower probability of debt distress, given that associated with shorter debt maturities, higher interest rates, and larger domestic debt is usually denominated in the shares of domestic credit directed toward the central government. This may have contributed to above-average interest spending in South Asia. domestic currency (and is therefore less vulnerable to exchange rate shocks) and that domestic investors are less prone to loss of market A. Effective interest rate on B. Effective government interest rates, government debt, 2022 by domestic share of government confidence (Grigorian 2023; Panizza 2010). debt, 2010–22 Percent Percent 8 4.5 This conclusion is supported by an event study of the resolution of past government debt booms 6 4.0 (annex SL.2). Domestic and external government 4 debt booms were similar in amplitude and 2 3.5 duration: on average, they lasted about five years and featured cumulative government debt 0 SAR EMDEs 3.0 Above median Below median increases of, on average, 11 percentage points of GDP. However, external debt booms were about C. Average maturity of government D. Average maturity of government three times more likely to result in debt default debt, 2022 debt, by domestic share of either in the last year of the boom or in the government debt subsequent year (Figure SL.5). Years Years 10 13 8 12 This reduction in crisis risk, however, comes with 6 11 costs. EMDEs with above-median shares of 4 domestic debt have an average government debt 10 maturity that is shorter by two years and an 2 9 average effective interest rate that is higher by 0 8 EMDEs SAR Above median Below median nearly 1 percentage point than other EMDEs. Reliance by governments on domestic financing E. Financial system claims on general F. Financial system claims on general could also compound other obstacles to private government, 2021 government, by domestic share of investment by crowding out financing for private government debt Percent of domestic credit Percent of domestic credit investment: the share of total bank credit of credit 50 20 that is allocated to the central government is 6 40 15 percentage points higher in EMDEs with above- 30 median domestic shares of government debt. 10 20 Reflecting the relatively high share of domestic 10 5 debt in South Asia, governments in the region spend almost 3 percentage points of GDP more 0 0 SAR EMDEs Above median Below median on net interest payments, the average maturity of their debt is shorter by four years, and government Sources: IMF (various staff reports); Kose et al. (2022); World Bank. Note: Latest data is for 2022 for interest spending and typically 2021 for average maturity of credit accounts for almost 20 percentage points government debt. more of the banking system’s domestic credit A. Net interest spending is defined as the difference between the primary fiscal balance and the overall fiscal balance. South Asia includes Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, and (Figure SL.6). Sri Lanka. B. Effective government interest rate is defined as net interest spending as share of gross government debt in the previous year. For South Asia, data are only available for India, Sri Lanka, and Pakistan. A high share of domestic debt does not appear to A.C.E. GDP-weighted averages (at 2010–19 average prices and market exchange rates). reduce the cost of debt distress once it occurs. In B.D.F. Unweighted averages for 2010–22 (B) or 2021 or latest available data (D, F) for EMDEs with above-median or below-median share of domestic debt of government debt. fact, output losses associated with domestic debt default have, on average, been higher than those associated with external default. In a study of 40 EMDEs during 1950–2010, domestic debt defaults were followed by statistically significant per capita income losses, averaging 2.7 percent SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 47 after five years, whereas external debt defaults were FIGURE SL.7 Prospects for GDP growth and not associated with statistically significant lasting government revenues per capita income losses (Malinen and Ropponen South Asia’s potential output growth—defined as the highest growth rate 2019). that can be maintained in the long term without igniting inflation—is expected to remain stronger than in other EMDE regions over the remainder of the 2020s but below the rate achieved in the 2010s. Shoring Policy implications up fiscal positions in a lasting manner will be challenging given the region’s exceptionally low revenue ratios. The results suggest that the current, challenging A. Potential growth prospects B. Government revenues, 2022 global economic environment increases the probability of sovereign debt defaults and reduces Percent 6.5 Percent 3.0 Percent of GDP 40 the chance that, when defaults occur, they will 6.0 2.5 30 improve fiscal positions. Global output growth is 5.5 2.0 expected to weaken in 2023, with global financial 20 conditions expected to remain tight. Unfavorable 5.0 1.5 10 global economic conditions are likely to persist 4.5 2011-21 2022-30 2011-21 2022-30 1.0 beyond 2023. Global growth over the remainder SAR (LHS) World (RHS) 0 SAR Other EMDEs of the 2020s is projected at 2.2 percent a year, down from 2.6 percent a year in the 2010s (Kilic Sources: Kasyanenko et al. (2023); Kilic Celik, Kose and Ohnsorge (2023); World Bank. A. GDP-weighted average (at 2010–19 average exchange rates and prices) of potential growth, Celik, Kose, and Ohnsorge 2023). Meanwhile, based on a production function approach, as defined in Kasyanenko et al. (2023) and Kose and global financing conditions are expected to remain Ohnsorge (2023). B. Data for Afghanistan is unavailable. Blue bars are aggregates for South Asia (SAR) and EMDEs tight as advanced-economy central banks maintain and GDP-weighted averages (at 2010–19 average exchange rates and prices). Yellow whiskers indicate the minimum-maximum range for South Asian countries. elevated policy rates to rein in inflation, with bouts of financial stress likely to recur (World Bank 2023). one-fifth to two-thirds of fiscal consolidations Headwinds from the challenging external have been successful at achieving lasting environment heighten the importance of national reductions in debt or deficits (Balasundharam et policies that reduce the probability of default and al. 2023; Gupta et al. 2004). But there is some increase the chances of successful defaults when evidence that faster domestic growth increased the they occur. Building the foundations for stronger probability of a fiscal consolidation episode growth is critical, as is fiscal consolidation that beginning or being sustained (Gupta et al. 2004). returns fiscal positions to a sound footing Conversely, at the high levels of government debt embedded in robust fiscal institutions. prevailing in much of South Asia, fiscal Encouragingly, South Asia’s potential growth over consolidation is unlikely to weigh on growth. the remainder of the 2020s is expected to remain Fiscal multipliers at high levels of debt have been the highest among EMDE regions, at 6.0 percent shown to be near-nil (Huidrom et al. 2020; per year (Figure SL.7; Kasyanenko et al. 2023). Ilzetzki, Mendoza, and Vegh 2013). But South Asian countries may struggle to achieve lasting fiscal consolidations given their above- In addition, fiscal consolidation can be designed in average government debt levels and below-average a manner that is more likely to succeed. For revenue ratios. example, cuts in current spending (especially on transfers and wages), while simultaneously Faced with these external and structural protecting or raising capital spending, have been headwinds, a policy agenda that combines fiscal associated with a higher likelihood of consolidation with growth-boosting reforms has consolidation being sustained (Gupta et al. 2004). the potential to yield particularly high dividends. Fiscal adjustment, when necessary, has been more Chapter 1 details specific policy priorities for likely in countries with less rigid spending, in boosting growth and strengthening fiscal positions particular where wages, pensions, and debt service and management in South Asia. Growth will help accounted for a smaller share of spending (Munoz sustain fiscal consolidation efforts. In general, only and Olaberria 2019). Revenue increases also raised 48 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 the probability of a consolidation being sustained ANNEX SL.1. Regression (Gupta et al. 2004, 2005). Other factors that increased the probability of success included a analysis: Methodology and competitively valued currency and broad-based data political support (Balasundharam et al. 2023). Data Prompt and comprehensive debt restructuring helps both creditors and debtors (Kose et al. This spotlight draws on a dataset of up to 177 2020). Waiting to restructure debt until after a external or domestic sovereign debt defaults in 64 default occurs has been associated with larger EMDEs spanning 1979–2018 (annex table declines in output, investment, private sector SL.1.1). Data for external sovereign debt defaults credit, and capital inflows than preemptive debt are from Asonuma and Trebesch (2016) while restructurings (Asonuma et al. 2020; Asonuma data for domestic sovereign debt defaults are from and Trebesch 2016;). In the past, shallow Erce, Mallucci, and Picarelli (2022). Data for the agreements have been followed by more cyclically adjusted fiscal balance and total, restructurings until a more lasting resolution was external, and domestic government debt are from found (Meyer, Reinhart, and Trebesch 2019). Kose et al. (2022). Data for global growth rates are from the World Bank’s Global Economic Prospects If history is any guide, any sovereign defaults in report. Data for the U.S. federal funds rate (as a South Asia in the foreseeable future would incur proxy of global borrowing costs) are from Haver considerable economic and social costs. Reducing Analytics. Data for real GDP growth, inflation, the likelihood of defaults, and ensuring that any and depreciation are from the IMF World defaults that do occur achieve lasting Economic Outlook database (April 2023 edition). improvements in debt profiles and service costs, requires proactive policy efforts to boost long-term Estimation of correlates of successful growth and put fiscal positions on a sustainable default: Approach and baseline results footing. A selection-bias corrected panel regression is used to estimate the correlates of successful default. The selection bias correction takes into account the specific characteristics and circumstances that tip countries into debt defaults. The implicit assumption of the approach is that sovereigns are forced into default by circumstances rather than strategically choosing default to achieve specific macroeconomic outcomes (Gelpern and Panizza 2022). The approach follows a two-step process to estimate the evolution of fiscal outcomes after default, as in Heckman (1976). First, a probit regression is used to estimate the probability of default depending on global circumstances and country characteristics. In a second step, correcting for the selection into default, a panel regression is used to estimate the effects of macroeconomic conditions on the cumulative changes in government debt-to-GDP ratios and the effective interest rate on government debt in the five years following default. This exercise is SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 49 ANNEX TABLE SL.1.1 Countries and default years Country Year Country Year Country Year Country Year Country Year AGO 2010 CPV 2018 LKA 1996 PAK 1998 TUR 1999 ARG 1982 CRI 1981 MAR 1983 PAK 1999 UKR 1981 ARG 1985 CRI 1984 MAR 1985 PAN 1984 UKR 1998 ARG 1988 CRI 1986 MAR 1989 PAN 1987 UKR 1999 ARG 1989 DMA 2003 MDA 2001 PAN 1988 UKR 2000 ARG 2001 DOM 1981 MDA 2002 PAN 1998 URY 1983 ATG 1998 DOM 1982 MDG 1981 PER 1979 URY 1985 ATG 2008 DOM 1987 MDG 1982 PER 1983 URY 1987 BIH 1992 DOM 1996 MDG 1985 PER 1984 URY 1989 BLZ 2006 DOM 2004 MDG 1987 PER 1985 URY 2002 BLZ 2012 ECU 1982 MDG 2002 PER 1992 URY 2003 BOL 1980 ECU 1983 MEX 1982 PHL 1983 VEN 1983 BOL 1982 ECU 1984 MEX 1984 PHL 1986 VEN 1986 BOL 1984 ECU 1986 MEX 1986 PHL 1988 VEN 1989 BOL 1988 ECU 1997 MEX 1987 PHL 1990 VEN 1995 BRA 1982 ECU 1999 MEX 1988 POL 1981 VEN 1998 BRA 1983 ECU 2008 MKD 1991 POL 1982 VEN 2002 BRA 1984 EGY 1984 MKD 1992 POL 1983 VNM 1982 BRA 1986 GAB 1997 MLI 2011 POL 1986 ZAF 1985 BRA 1989 GAB 2001 MMR 1984 POL 1988 ZAF 1989 BRA 1990 GHA 1982 MMR 1987 POL 1989 ZAF 1992 BRA 1996 GMB 2017 MNE 1991 PRY 1986 BRB 2018 GTM 1989 MNG 1997 PRY 2002 CAF 1992 HND 1981 MNG 2003 ROU 1981 CHL 1983 HND 1990 NGA 1982 ROU 1983 CHL 1984 IDN 1997 NGA 1983 ROU 1986 CHL 1986 IRQ 1986 NGA 1986 RUS 1991 CHL 1990 JAM 1980 NGA 1987 RUS 1998 CIV 1983 JAM 1983 NGA 1988 RUS 1999 CIV 1989 JAM 1984 NGA 1989 RWA 1989 CIV 2000 JAM 1986 NGA 1995 RWA 1994 CIV 2001 JAM 1990 NIC 1981 SDN 2007 CIV 2011 JAM 2010 NIC 1982 SLB 1996 CMR 1985 JAM 2013 NIC 1983 SLV 1981 CMR 1993 JOR 1989 NIC 1985 SLV 2017 CMR 2001 KWT 1990 NIC 1994 SRB 1991 COD 1997 LBR 1980 NIC 1996 SRB 1992 COG 1992 LBR 1989 NIC 2003 SYC 2010 CPV 1998 LBR 2016 NIC 2008 TTO 1988 Sources: Asonuma and Trebesch (2016); Erce, Mallucci, and Picarelli (2022); World Bank. 50 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE SL.1.2 Marginal probability of default (1) (2) (3) Federal funds rate (percent) 0.00256*** 0.00284*** (0.000738) (0.000752) Change in global growth (percentage points) 0.000155 -0.000368 (0.000953) (0.000965) Sentiment (number) 0.00323*** 0.00297*** (0.000766) (0.000773) Lagged government debt (percent of GDP) 8.58e-05** 8.16e-05** (3.79e-05) (3.77e-05) Lagged domestic growth (percent) -0.000702** -0.000733** (0.000349) (0.000346) Lagged depreciation (percent) 0.0322*** 0.00517 (0.0119) (0.00601) Lagged inflation (percent) -7.64e-05 (6.43e-05) Observations 4,338 4,338 4,338 Source: World Bank. Note: Estimated marginal probabilities of a probit regression. The dependent variable is one when a country enters default and zero otherwise. Investor risk sentiment (“Sentiment”) is proxied by the “excess” global number of defaults that cannot be explained by the U.S. federal funds rate and changes in global growth, computed as the residual from a regression of global defaults on the U.S. federal funds rate and the first difference in global output growth. Domestic variables are lagged by one year, including output growth, depreciation, and inflation. The sample includes 145 EMDEs, encompassing 84 defaults in 46 EMDEs, over 1982–2018. The regression sample spans 1982 to 2022. ***, **, and * indicate significance at the 1, 5, and 10 percent levels. intended to assess the correlates of default and, in residual from a regression of the global number of cases where default does occur, the correlates debt defaults on the U.S. federal funds rate and associated with fiscal outcomes. The direction of the change global output growth. Domestic causality remains an open question. factors in the first stage are all lagged by one year, including government debt (in percent of GDP), First stage. The first stage estimates a probit output growth and depreciation.3 The implicit regression: assumption is that external factors and lagged domestic factors are exogenous to the occurrence Pr(defaultit) = 0 + Zt + Cit-1) + it . of debt default. The cumulative normal distribution is denoted by The first stage results are shown in annex table Φ. The dependent variable is a dummy set to 1 if SL.1.2. The table shows three columns: one with a default occurs and zero otherwise in Equation 1. only global factors, a second with only domestic Regressors in the probit model (Zt, Cit-1) include factors, and a third with both global and domestic global and domestic factors, respectively. Global factors. The third column is the most factors included in the first stage are the U.S. comprehensive and is the baseline specification. It federal funds rate, an indicator for investor risk shows that an increase in the U.S. federal funds sentiment, and the change in global output rate is associated with a statistically significant growth. For lack of an alternative measure with a sufficiently long time series reaching back into the 1970s, the sentiment variable is defined as “excess 3 Note that in alternative specifications the first stage regression defaults.” This measure is computed as the includes inflation. However, it is not statistically significant and, as a result, not included in the baseline estimations. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 51 ANNEX TABLE SL.1.3 Share of successful defaults (Percent of defaults in the group) Success = reduction in government debt Success = reduction in effective interest rate Within the Within the Otherwise Otherwise indicated group Indicated group All defaults Defaults since 1990 Domestic growth acceleration 100 54 88 53 Global growth acceleration 70 57 67 55 Fiscal consolidation 75 59 71 53 IMF program 70 46 69 43 Defaults since 1979 Domestic growth acceleration 72 43 88 55 Global growth acceleration 57 46 67 57 Fiscal consolidation 71 53 71 53 IMF program 64 43 69 47 Defaults on external creditors Defaults since 1990 Domestic growth acceleration 100 54 100 67 Global growth acceleration 70 57 83 67 Fiscal consolidation 75 59 100 50 IMF program 70 46 100 25 Above-median haircut 100 59 100 73 Above-median restructuring 100 60 100 75 Defaults since 1979 Domestic growth acceleration 67 41 100 70 Global growth acceleration 38 52 83 71 Fiscal consolidation 67 67 100 50 IMF program 75 42 100 40 Above-median haircut 50 48 100 75 Above-median restructuring 100 45 100 77 Sources: Asonuma and Trebesch (2016); Cruces and Trebesch (2013); Erce, Mallucci, and Picarelli (2022); World Bank. Note: Successful default is defined a default that is followed by a reduction in the government debt-to-GDP ratio or, alternatively, in the effective interest rate on government debt between the year of default and five years later. The effective interest rate on government debt is defined as net interest spending relative to the previous year’s government debt stock. The percentage of successful defaults is calculated among all defaults that occurred under the circumstances indicated in the first column. “Fiscal consolidation” indicates an improvement in the cyclically adjusted fiscal balance between the year of default and five years after default; “IMF program” indicates that an IMF program was in place at the time of default; “Domestic growth acceleration” and “Global growth acceleration” indicate a two-year domestic or global growth acceleration from the time of default. For the subset of 88 external debt defaults since 1979, including 43 that occurred from 1990 onwards, data is available for restructuring terms. “Above-median restructuring” indicates above-median size of restructured debt in percent of total government debt at time of default, as calculated by Cruces and Trebesch (2013). “Above-median haircut” indicates above-median market haircut at time of default, as calculated by Cruces and Trebesch (2013). increase in the probability of debt default. An statistically significant once global factors are increase in global output growth is associated with included. Predicted probabilities from the a somewhat lower probability of default but this estimation in the third column are presented in result is not statistically significant. Among the text and figures. Annex table SL.1.3 shows the domestic factors, both higher government debt share of successful defaults associated with and lower output growth are associated with different circumstances. statistically significantly higher probabilities of default. Exchange rate depreciation is controlled Second stage. The second stage estimates a panel for in the baseline specification, but is not regression: 52 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE SL.1.4 Fiscal outcomes after default Five-year change in Dependent variable Five-year change in government debt effective interest rate Change in domestic growth (percentage points) -2.053*** 0.0578 (0.510) (0.0387) Change in global growth (percentage points) 2.289 -0.410** (2.127) (0.173) Change in federal funds rate (percentage points) -0.394 0.272* (1.722) (0.149) Change in sentiment (number) 3.895*** 0.923*** (1.478) (0.319) Lagged government (percent of GDP) -0.609*** (0.114) Lagged effective interest rate (percent) -0.476*** (0.0721) Constant 181.2*** 8.441** (34.21) (3.876) Inverse Mill’s Ratio -57.48*** -2.36 (14.796) (1.44) Observations 4,333 4,309 Source: World Bank. Note: Results of the outcome (second) stage of a Heckman selection bias regression. The second stage is a panel regression, where the dependent variable is the change in government debt (in percent of GDP) or change in the effective interest rate on government debt (in percent) between default and five years after default. Global investor risk sentiment is defined as in the probit model. The sample includes up to 145 EMDEs, and 59 debt defaults in 37 EMDEs over 1982–2018. The regression sample ranges from 1982 to 2022. ***, **, and * indicate significance at the 1, 5, and 10 percent levels. ∆yitt+H = + 1 gitdomestic,t+H + 2 0 gtglobal,t+H + 3 statistically significant. Investor sentiment is also FFRt t+H + 4 Excesstt+H + 5 IMFit + 6 yit-t + it important: a 0.5 standard deviation improvement in global risk sentiment—comparable to the where the dependent variable is the change in change between 2020 and 2021—is associated either the government debt-to-GDP ratio or the with an about 2-percentage-point steeper decline change in the effective interest rate on government in government debt over the five years following debt between horizon H and the start of default. default. The regressions include the following correlates: the change in domestic output growth gitdomestic,t+H, Global output growth, global interest rates, and the change in global output growth gtglobal,t+H, the global risk sentiment after default are the most change in the U.S. federal funds rate FFRtt+H , the important factors associated with effective interest change in investor risk sentiment Excesstt+H, a rates on government debt after default. All of these dummy variable for an IMF program being in coefficients are statistically significant. The overall place at the time of default, and the lagged level of conclusion is that (i) to reduce debt in a lasting the dependent variable (either government debt or manner, domestic growth is key; and (ii) to lower net interest spending). The results of the second effective interest rates on government debt, “luck” stage focus on five-year changes (H=5) and are in the form of external developments is an shown in annex table SL.1.4. important factor. Overall, the analysis suggests that growth-enhancing domestic policies and good Domestic output growth after default is the most luck can re-enforce each other to achieve “success” important factor associated with a decline in after default. government debt. A 1-percentage-point increase in domestic output growth is associated with an Since the approach used here imposes a clear about 2-percentage-point steeper decline in sequencing—default first, debt or interest rate government debt, and the relationship is reduction second—there is little a priori reason to SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 53 believe that the exogenous external or past ANNEX SL.2. Event study: developments that triggered the default event should affect how debt or interest rates evolve over Methodology the five years following default. Nevertheless, the exclusion restriction is tested for each of the An event study of government debt booms is explanatory variables of the first stage regression. conducted to examine how different types of debt When the second-stage regression is augmented by booms ended. Government debt booms are each of the five first-stage variables, global defined as episodes in which government debt-to- borrowing cost and sentiment are both statistically GDP ratios increased above their Hodrick- insignificant. In the interest rate reduction Prescott-filtered trend by more than 1 standard regressions, changes in global growth, lagged deviation in at least one year. domestic growth, and lagged exchange rates are also statistically insignificant. These results suggest is yields 105 total government debt booms, 53 that the exclusion restriction holds for at least two domestic government debt booms, and 34 external variables in each second-stage regression. government debt booms, among 114 EMDEs since 2004. By this definition, four South Asian The inverse Mills ratio is statistically significant in countries (Bangladesh, Bhutan, Nepal and Sri the debt reduction regressions, and marginally Lanka) were in government debt booms in 2021, significant in the interest rate reduction regressions three (Bangladesh, Nepal and Sri Lanka) in 2022, that include a smaller sample of debt defaults (at and two (Nepal and Sri Lanka) in 2023. Two 15 percent significance). This suggests that a South Asian countries (Bangladesh and India) selection model approach is appropriate in this were in domestic government debt booms in context. 2020, but only one (Bangladesh) in 2021 and 2022. Endogeneity is, of course, a concern when including contemporaneous changes in domestic growth as a correlate of fiscal outcomes in the second-stage regression. However, an appropriate instrumental variable for domestic growth is not immediately apparent. Therefore, the results presented here can only be interpreted as correlations, not as causal effects. Robustness tests A dummy variable for an IMF program being in place at the time of default was not always significant in either the first or second stage regression once other factors were controlled for. Neither was the lagged current account balance or interactions between global growth and trade openness or lagged current account balances and the U.S. federal funds rate. Face value reductions, size of restructured debt, and the haircut to restructuring debt, as well as the magnitude of fiscal consolidation (measured as changed in the cyclically adjusted or structural fiscal balance), are also likely to have important relationships with fiscal outcomes after default. However, data on these measures are too limited to derive meaningful results. 54 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Department, Graduate Institute of International and References Development Studies, Geneva. Arteta, C., and G. Hale. 2008. “Sovereign Debt Borensztein, E. and U. Panizza. 2009. “The Costs of Crises and Credit to the Private Sector.” Journal of Sovereign Default.” IMF Staff Papers 56 (4): 683– International Economics 74 (1): 53–69. 741. Asonuma, M. T., M. M. Chamon, A. Erce, and A. Borensztein, E., and U. Panizza. 2010. “Do Sasahara. 2019. “Costs of Sovereign Defaults: Sovereign Defaults Hurt Exporters?” Open Economies Restructuring Strategies, Bank Distress and the Review 21: 393–412. Capital Inflow-Credit Channel.” IMF Working Paper WP/19/69, International Monetary Fund, Catão, L. A. V., and R.C. Mano. 2017. “Default Washington, DC. Premium.” Journal of International Economics 107 (C): 91–110. Asonuma, T., and C. Trebesch. 2016. “Sovereign Debt Restructurings: Preemptive or Post-Default.” Cruces, J. J., and C. Trebesch. 2013. “Sovereign Journal of the European Economic Association 14 (1): Defaults: The Price of Haircuts.” American Economic 175–214. Journal: Macroeconomics 5 (3): 85–117. Asonuma, T., M. Chamon, A. Erce, and A. Sasahara. De Paoli, B., G. Hoggarth, and V. Saporta. 2009. 2020. “Costs of Sovereign Defaults: Restructuring “Output Costs of Sovereign Crises: Some Empirical Strategies and the Credit-Investment Channel.” Estimates.” Working Paper 362, Bank of England, Working Paper 19/69, International Monetary London. Fund, Washington, DC. Erce, A. and M. Picarelli. 2023. “Financial Stability Balasundharam, V., O. Basdevant, D. Benicio, A. Implications of Domestic-law Debt Defaults.” Ceber, Y. Kim, L. Mazzone, H. Selim, and Y. Yang. Extended abstract. https://debtcon6.princeton.edu/ 2023. "Fiscal Consolidation: Taking Stock of sites/g/files/toruqf3611/files/documents/picarelli.pdf. Success Factors, Impact, and Design." Working Erce, A., E, Mallucci, and M. Picarelli. 2022. “A Paper 23/63, International Monetary Fund, Journey in the History of Sovereign Defaults on Washington, DC. Domestic-Law Public Debt.” International Finance Baldacci, E., and C. Mulas-Granados. 2012. “Debt Discussion Papers 1338, Board of Governors of the Reduction, Fiscal Adjustment, and Growth in Credit Federal Reserve System, Washington, DC. Constraint Economies.” Journal of Applied Economics Esteves, R. and J. Jalles. 2016. “Like Father like 18 (1): 71–98. Sons? The Cost of Sovereign Defaults in Reduced Baldacci, E., S. Gupta, and C. Mulas-Granados. Credit to the Private Sector.” Journal of Money, 2012. “Reassessing the Fiscal Mix for Successful Credit and Banking 48 (7): 1515–45. Debt Reduction.” Economic Policy 27 (71): 365–406. Esteves, R., S. Kenny, and J. Lennard. 2021. “The Benjamin D, and M. Wright. 2009. “Recovery Aftermath of Sovereign Debt Crises: A Narrative Before Redemption? A Theory of Delays in Approach.” CEPR Discussion Paper 16166. Center Sovereign Debt Renegotiations.” Working Paper for Economic and Policy Research, Washington, 2009–15, Centre for Applied Macroeconomic DC. Analysis, Crawford School of Public Policy, The Farah-Yacoub, J. P., C. Graf von Luckner, R. Australian National University. Ramalho, and C. Reinhart. 2022. “The Social Cost Bevilaqua, J., G. B. Hale, and E. Tallman. 2020. of Sovereign Debt Default.” Policy Research “Corporate Yields and Sovereign Yields.” Journal of Working Paper 10157, World Bank, Washington, International Economics 124: 103304. DC. Bolton, P., M. Gulati, and U. Panizza. 2022. Forni, L., G. Palomba, J. Pereira, and C. Richmond. “Sovereign Debt Puzzles,” Working Paper 2021. “Sovereign Debt Restructuring and Growth.” HEIDWP27-2022, International Economics Oxford Economic Papers 73 (2): 671–97. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SPOTLIGHT 55 Fuentes, M., and D. Saravia. 2010. “Sovereign IADB (Inter-America Development Bank). 2023. Defaulters: Do International Capital Markets Punish Dealing with Debt: Less Risk for More Growth in Them?” Journal of Development Economics 91 (2): Latin America and the Caribbean. Washington, DC: 336–47. Inter-America Development Bank. Furceri, D., and A. Zdzienicka. 2012. “How Costly Ilzetzki, E., E. G. Mendoza, and C. A. Végh. 2013. are Debt Crises?” Journal of International Money and “How Big (Small?) Are Fiscal Multipliers?” Journal of Finance 31 (4): 726–42. Monetary Economics 60 (2): 239–54. Gelos, R. G., R. Sahay, and G. Sandleris. 2011. Jorra, M. 2011. “The Heterogeneity of Default “Sovereign Borrowing by Developing Countries: Costs: Evidence from Recent Sovereign Debt What Determines Market Access?” Journal of Crises.” MAGKS Joint Discussion Paper Series in International Economics 83 (2): 243–54. Economics 51. Philipps-University Marburg, Faculty of Business Administration and Economics, Gelpern, A., and U. Panizza. 2022. “Enough Marburg, Germany. Potential Repudiation: Economic and Legal Aspects of Sovereign Debt in the Pandemic Era.” Annual Kaminsky, G. L. and C. M. Reinhart. 2000. “On Review of Economics 14: 545–70. Crises, Contagion, and Confusion.” Journal of International Economics 51(1): 145-68. Gennaioli, N., A. Martin, and S. Rossi. 2014. “Sovereign Default, Domestic Banks, and Financial Kasyanenko, S., P. Kenworthy, S. Kilic Celik, F. U. Institutions.” The Journal of Finance 69 (2): 819–66. Ruch, E. Vashakmadze, and C. Wheeler. 2023. “The Past and Future of Regional Potential Growth: Gennaioli, N., A. Martin, and S. Rossi. 2018. Hopes, Fears, and Realities.” Policy Research “Banks, Government Bonds, and Default: What Do Working Paper 10368, World Bank, Washington, the Data Say?” Journal of Monetary Economics 98 DC. (October): 98–113. Kilic Celik, S., M. A. Kose, and F. Ohnsorge. Grigorian, D. 2023. “Restructuring Domestic 2023. “Potential Growth Prospects: Risks, Rewards Sovereign Debt: An Analytical Illustration.” and Policies.” Working Paper 2023-19, Centre for Working Paper 23/24, International Monetary Applied Macroeconomic Analysis, Crawford School Fund, Washington, DC. of Public Policy, The Australian National University. Gupta, S. B. Clements, E. Baldacci, and C. Mulas- Kose, M. A. and F. Ohnsorge. 2023. Falling Long- Granados. 2004. "The Persistence of Fiscal Term Growth Prospects: Trends, Expectations, and Adjustments in Developing Countries." Applied Policies. Washington, DC: World Bank. Economics Letters 11 (4): 209–12. Kose, M. A., F. Ohnsorge, C. M. Reinhart, and K. S. Gupta, S., E. Baldacci, B. Clements, and E. R. Rogoff. 2022. “Aftermath of Debt Surges.” Annual Tiongson. 2005. "What Sustains Fiscal Review of Economics 14 (August): 637–63. Consolidations in Emerging Market Countries?” Kose, M. A., P. Nagle, F. Ohnsorge, and N. International Journal of Finance and Economics 10 Sugawara. 2020. Global Waves of Debt. Washington, (4): 307–32. DC: World Bank. Heckman, J. 1976. “The Common Structure of Kuvshinov, D. and K. Zimmermann. 2019. Statistical Models of Truncation, Sample Selection “Sovereigns Going Bust: Estimating the Cost of and Limited Dependent Variables and a Simple Default.” European Economic Review 119: 1-21. Estimator for Such Models.” Annals of Economic and Social Measurement 5 (4): 475–92. Levy-Yeyati, E. and U. Panizza. 2011. “The Elusive Costs of Sovereign Defaults.” Journal of Development Huidrom, R., M. A. Kose, J. J. Lim, and F. Economics 94(1): 95-105. Ohnsorge. 2020. “Why Do Fiscal Multipliers Depend on Fiscal Positions?” Journal of Monetary Malinen, T. and O. T. Ropponen. 2019. “To Economics 114 (October): 109–125. Default or Not? The Aftermath of Sovereign Defaults and IMF Austerity Programs During 56 SPOTLIGHT SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Economic Crises” Available at SSRN: http:// European Economic Association 14 (1): 215–51. dx.doi.org/10.2139/ssrn.2785491. Richmond, C., and D. A. Dias. 2008. "Duration of Marchesi, S., and T. Masi. 2018. “Life After Default: Capital Market Exclusion: Stylized Facts and Private vs. Official Sovereign Debt Restructurings.” Determining Factors." Department of Economics, Centro Studi Luca d'Agliano Development Studies University of California, Los Angeles. Working Paper 437, Milan, Italy. Rose, A. K. 2005. “One Reason Countries Pay their Meyer, J., C. M. Reinhart, and C. Trebesch. 2019. Debts: Renegotiation and International Trade.” “Sovereign Bonds since Waterloo.” Working Paper Journal of Development Economics 77 (1): 189–206. 25543, National Bureau of Economic Research, Cambridge, MA. Sturzenegger, F. 2003. “Default Episodes in the 90s: Fact Book, Toolkit and Preliminary Lessons.” Meyer, J., C. M. Reinhart, and C. Trebesch. 2022. mimeograph, International Monetary Fund, "Sovereign Bonds since Waterloo." The Quarterly Washington, DC. Journal of Economics 137 (3): 1615–80. Sturzenegger, F., and J. Zettelmeyer. 2007. Muñoz, E. and E. Olaberria. 2019. "Are Budget “Creditors’ Losses Versus Debt Relief: Results from a Rigidities a Source of Fiscal Distress and a Decade of Sovereign Debt Crises.” Journal of the Constraint for Fiscal Consolidation?" Policy European Economic Association 5 (2–3): 343–51. Research Working Paper 8957, World Bank, Washington, DC. Tomz, M., and M. L. J. Wright. 2007. “Do Countries Default in ‘Bad Times’?” Journal of the Panizza, U. 2010. “Is Domestic Debt the Answer to European Economic Association 5 (2–3): 352–60. Debt Crises?” In Overcoming Developing Country Debt Crises, edited by B. Herman, J. A. Ocampo, Tomz, M., and M. L. J. Wright. 2013. “Empirical and S. Spiegel. Oxford: Oxford University Press. Research on Sovereign Debt and Default.” Annual Review of Economics 5 (August): 247–72. Panizza, U., F. Sturzenegger, and J. Zettelmeyer. 2009. “The Economics and Law of Sovereign Debt Trebesch, C., and M. Zabel. 2017. "The Output and Default.” Journal of Economic Literature 47 (3): Costs of Hard and Soft Sovereign Default." 651–98. European Economic Review 92 (February): 416–32. Reinhart, C. M. and K. S. Rogoff. 2011. “The World Bank. 2021. Debt Transparency in Developing Forgotten History of Domestic Debt.” The Economic Economies. Washington, DC: World Bank. Journal 121(552): 319-50. World Bank. 2023. Global Economic Prospects Report. Reinhart, C. M., and C. Trebesch. 2016. “Sovereign June. Washington, DC: World Bank. Debt Relief and Its Aftermath.” Journal of the CHAPTER 2 Recruiting firms for the energy transition SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 59 Chapter 2. Recruiting firms for the energy transition As the world presses ahead with the energy transition, new energy-saving technologies offer South Asian countries an opportunity to modernize their economies. Currently, the energy intensity of output of South Asian economies is almost twice the global average—despite a decline over the past two decades that was almost entirely driven by firm-level, within-sector cuts in energy intensity. While the region’s firms have been early adopters of basic energy-saving technologies, they have lagged in the adoption of more advanced technologies, with smaller firms lagging particularly far behind. Policies that have been effective at encouraging technology adoption among firms include market-based regulation, dissemination of accurate information on energy savings, and financial support. Introduction together with India account for 87 percent of the region’s GDP. All South Asian countries except Widespread adoption of energy-saving tech- Bangladesh and India witnessed faster emissions nologies is needed as a critical part of South Asia’s growth than other EMDEs between 2010 and energy transition away from fossil fuels. With a 2020. strong and concerted effort, the region can avoid The sources of growth in emissions vary across lagging behind in progress toward its climate goals South Asian countries but, in general, the main and remaining reliant on old, energy-inefficient sources relate to energy use (figure 2.1).1 The technologies that the rest of the world will be power generation sector is a major source of abandoning more quickly. The region’s policy growing emissions, being the largest contributor makers are well aware of this challenge and have to GHG emissions growth between 2010 and begun to take action. This chapter examines 2020 in three countries (Bangladesh, India, and policies that can further accelerate the adoption of Sri Lanka). In addition to grid energy use, the energy-saving technologies. heavy use of fuels for in-house energy generation Compared to its share in the global population, in the manufacturing, mining, and construction South Asia’s contributions to global greenhouse sectors has also been among the three largest gas (GHG) emissions is small, at just under 10 contributors to emissions growth in Bangladesh, percent of global GHG emissions versus 24 Bhutan, India, and Nepal. The expansion of grid percent of the global population. However, per electricity supply from renewable sources, notably unit of output, South Asia’s emissions are almost hydroelectricity in Nepal, may reduce the need twice the global average. The discrepancy in part for such in-house energy sources. Nonetheless, reflects low labor productivity (chapter 3). The improvements in energy efficiency among good news is that this is already changing, and industrial firms will need to be an important part that further improvements are possible even of any strategy to reduce GHG emissions. without large amounts of new investment. A slowdown in energy consumption growth The region is currently contributing through rising energy efficiency will have disproportionately to global greenhouse gas important environmental benefits, in addition to emissions: the region’s share of GHG emissions is lowering GHG emissions. The region’s power more than twice its share in global GDP (figure sector is not only the largest source of the region’s 2.2.1). India accounts for 80 percent of South GHG emissions but also its largest source of Asia’s GHG emissions, but the share of global PM2.5 pollution (figure 2.2). Except for the emissions is also more than twice that of global GDP in Bhutan, Nepal, and Pakistan, which 1 At 20 percent, South Asia has the second-lowest share of renewable electricity generation among EMDE regions (after the Middle East and North Africa), despite a 4-percentage-point increase since 2016. While Bhutan and Nepal rely virtually entirely on Note: This chapter was prepared by Siddharth Sharma, with renewable energy, the share of renewable electricity generation is well contributions from Jonah Matthew Rexer. below the global average in Bangladesh, India, and the Maldives 60 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.1 South Asia’s contribution to global As a result of these PM2.5 emissions, South Asia is emissions home to nine of the world’s ten most polluted SAR’s share of global GHG emissions in recent years has been more than cities (World Bank 2023a). The region also has twice its share of global GDP. Emissions rose faster in most SAR countries the second-highest rate of deaths attributable to air than in other EMDEs, on average, during 2010–20, but only slightly faster in most cases. Energy use, including that in the manufacturing sector, was pollution (after Sub-Saharan Africa). In 2019, the major source of emissions growth. deaths from air pollution in SAR outnumbered deaths from tuberculosis by a factor of five. A. Share of global GHG emissions, B. Growth in GHG emissions, 2010–20 Average life expectancy across Bangladesh, India, 2021 Percent Percent Percent Nepal, and Pakistan would be five years higher if 1.2 GHG emissions GDP 12 80 SAR Other EMDEs pollution concentrations permanently complied 1.0 10 0.8 8 60 with WHO guidelines (Greenstone, Hasenkopf, 0.6 6 and Lee 2022). Pollution-induced sickness, 0.4 4 40 0.2 2 mortality, and learning losses have been shown to 0.0 0 20 reduce incomes materially in the region (Behrer, BGD BTN IND (RHS) SAR (RHS) PAK LKA MDV NPL 0 Choudhary, and Sharma 2023). MDV NPL LKA BTN PAK BGD IND The transition away from fossil fuels will have C. Source of growth in GHG D. Sources of growth in GHG wide-ranging consequences. This chapter focuses emissions, 2010–20 emissions, 2010–20 on policies to encourage the transition among Percentage points Percentage points 8 Energy: Manufacturing 4 Energy: Manufacturing firms. Specifically, it discusses the following Energy: Transportation and other 6 Energy: Electricity/heat Agriculture Energy: Transportation and other Energy: Electricity/heat questions. 3 Industrial Processes Agriculture 4 Industrial Processes 2 • How has energy consumption evolved in 2 South Asia? 1 0 -2 0 • Which factors are associated with shifts by BGD BTN IND LKA MDV NPL PAK Other EMDEs SAR firms toward less polluting technologies? Sources: Climate Watch; EDGARv7.0_GHG database; European Commission; OECD; World Development Indicators. • What are the policy implications? Note: BGD = Bangladesh; BTN = Bhutan; EMDEs = emerging market and developing economies; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan; SAR = South Asia. A. Chart shows South Asia’s share of global GHG emissions compared with global nominal GDP in Contributions to the literature U.S. dollars for 2021. B.-D. Charts show annual average growth in emissions between 2010 and 2020 using Climate Watch data. Latest available data (2020) for GHG emissions sources. “Energy: Electricity/heat” comprises This chapter provides the first region-wide emissions from fossil fuel use in electricity and heat plants, that is, the power sector. “Energy: Transportation and other” category comprises emissions from fossil fuel use in transportation, overview of the policy challenges facing South Asia buildings, agriculture, and fishing. “Energy: Manufacturing” comprises emissions from fossil fuel use in manufacturing, mining, and construction (excluding grid energy use). in reducing energy-related GHG emissions, building on the World Bank’s Country Climate and Development Reports for Bangladesh, Nepal, and Pakistan, and similar work underway in the rest of the region. It makes several additional Maldives, all South Asian countries’ shares of contributions to the literature. global PM2.5 emissions are multiples of their shares of global GDP, ranging from 2 to 15 times First, it examines the contributions of firm- and higher. In Bhutan and Nepal, where the energy sector-specific factors to high energy consumption sector is predominantly based on hydro power, the in South Asia. Previous studies have conducted high PM2.5 and GHG intensity of GDP is driven such analysis only for individual countries by other sources. These sources include brick kilns (Kumar, Mittal, and Pradhan 2023; Martin and other small firms, solid fuel combustion in the 2011). The consistent approach applied here residential sector for cooking and heating, current enables more robust cross-country comparisons. management practices for municipal waste in the region (including the burning of plastics), and Second, using new survey data, this chapter is the transport. first study to compare the diffusion of energy- SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 61 saving technologies among firms in South Asia FIGURE 2.2 Air pollution in South Asia and other EMDEs, and to identify which types of SAR is the EMDE region with the highest air pollution and second-highest firms have been more successful in adopting rate of deaths from air pollution after Sub-Saharan Africa. The energy and transport sectors are important sources of pollution, especially in India, the energy-efficient technologies. Maldives, and Sri Lanka. Third, using the interim findings of a randomized A. PM2.5 pollution, 2018 B. Share of global PM2.5 pollution, control trial for the leather goods industry in 2018 Bangladesh, the chapter illustrates the potential kt Percent Percent 1,000 2.5 PM2.5 GDP 25 energy-efficiency gains that could be achieved by 2.0 20 800 the adoption of low-cost new technology among 1.5 15 firms. The analysis reinforces the results of a large 600 1.0 10 0.5 5 literature that demonstrates the benefits of 400 0.0 0 technology adoption among farmers, non- IND (RHS) SAR(RHS) PAK LKA NPL BGD 200 agricultural firms, and households, but is the first 0 SAR EAP SSA LAC ECA MNA to illustrate how information spillovers among firms can accelerate this process. C. Sources of PM2.5 pollution, 2018 D. Deaths caused by air pollution, 2019 Fourth, this chapter includes the first systematic Percent of PM 2.5 emissions Deaths per 100,000 population Power Transport Other 200 review of the literature on policies that can 100 EMDE average encourage firms to adopt more energy-efficient 80 150 60 technologies. Previous reviews have pertained to 100 40 obstacles that firms face in technology adoption 20 50 more broadly (e.g., Verhoogen 2023), but this 0 IND BGD BTN SAR MDV LKA PAK NPL EMDEs chapter focuses specifically on obstacles to the Other 0 SSA SAR EAP MNA ECA LAC adoption of energy-saving technologies and on experiences with policies to remove such obstacles Sources: EDGARv7.0_GHG database; European Commission; World Development Indicators; World Health Organization. and encourage energy savings by firms. Note: EMDEs = emerging market and developing economies; EAP = East Asia and the Pacific; ECA = Europe and Central Asia; LAC = Latin America and the Caribbean; MNA = Middle East and North Africa; SAR = South Asia; SSA = Sub Saharan Africa. BGD = Bangladesh; BTN = Bhutan; IND = Main findings India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan. kt = kiloton. A. Unweighted cross-country averages. PM2.5 pollution is defined as the amount of small dust or soot particles in the air measuring 2.5 microns or less in width. This chapter offers the following main findings. C. The “other” category includes manufacturing, agriculture, residential sector, municipal waste and soil dust. First, the energy intensity of output in South Asia is almost twice the global average, despite having fallen faster than elsewhere over the past decade. Third, interim results from the midline survey of a The energy intensity of its economic activity has randomized control trial conducted among leather fallen mainly owing to declining energy intensity firms in Bangladesh suggest that firms significantly within sectors, as opposed to shifts of activity and underestimate the potential savings from energy- resources toward less energy-intensive sectors. In efficient new technologies. Firms were, on average, the case of India, the only country with available initially willing to pay only 56 percent of the firm-level panel data, energy intensity has declined purchase price of a new energy-saving product. markedly within firms. Once given the relevant information, 10 percent of firms adopted the new technology within the Second, while firms in South Asia have been earlier short span of three months. This information adopters of simple energy-efficient technologies appears to have spread rapidly to geographically than firms in other EMDE regions, they have close firms, spurring technology adoption. been slower adopters of more advanced technologies. The adoption of such technologies Fourth, a wide range of policies will be needed to has tended to be faster among larger firms, improve economy-wide energy efficiency. Besides including those with better-educated managers well-designed informational interventions and and better management practices. nudges, firms’ technology adoption could be 62 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.3 Energy intensity in South Asia serve such development goals as job creation and India, and especially its industrial sector, is the main source of energy productivity improvements. consumption in South Asia. The region’s two largest economies are more energy intensive than other EMDEs outside South Asia, with the industrial sector’s energy intensity a multiple of that of other sectors. While rapid Evolution of South Asia’s output growth has lifted energy consumption, this was partially offset by improving energy efficiency in all four South Asian countries with available energy consumption data. The sources of the decline in aggregate energy efficiency varied widely across countries. South Asia’s economic activity is unusually energy -intensive and the industrial sector accounts for A. Countries’ shares of South Asia’s B. Sources of energy consumption in energy consumption, 2020 SAR and other EMDEs the bulk of this energy intensity. Improvements in Percent of SAR consumption Percent energy intensity can unlock improvements in air 100 Agriculture Industry quality, production gains, and job creation, while 80 Transport Services also lowering GHG emissions. Other IND 60 PAK BGD 40 South Asia in global comparison LKA 20 The energy intensity of economic activity in South 0 SAR Non-SAR Asia is almost twice the global average (figure 2.3). In contrast to energy use per unit of output— C. Energy intensity of output, 2020 D. Relative sectoral energy intensities which is the focus of this section—South Asia’s Toe/thousand US$ Economy-wide intensity = 1 energy use per capita is lower than the EMDE 0.4 Other EMDEs World 5 Industry Agriculture Services average. The divergence between South Asia’s 0.3 4 above-average energy intensity of economic 3 activity and below-average energy intensity per 0.2 2 capita in large part reflects its low labor 0.1 1 productivity. 0 0 SAR IND PAK BGD LKA SAR IND PAK BGD LKA India and Pakistan, the region’s two largest economies, are one-quarter to one-third more E. Contributions to energy F. Decomposition of within-sector energy-intensive than the average EMDE outside consumption growth, 2010–20 energy consumption growth, 2010–20 South Asia, respectively. In contrast, Bangladesh Percentage points Percentage points 10 Growth Changes in energy intensity Total 0 and Sri Lanka are less energy intensive, by one- 5 -1 half, than the average non-South Asian EMDE. -2 The latest data show that, in 2020, manufacturing 0 -3 Industry accounted for the largest share of South Asia’s -5 -4 Agriculture energy consumption, and a larger share than in -5 Services other EMDE regions. For three of the four South -10 -6 SAR IND PAK BGD LKA SAR IND PAK BGD LKA Asian countries with available data, the Sources: European Commission; OECD Green Growth database; World Development Indicators. manufacturing sector was the single-largest Note: EMDEs = emerging market and developing economies; SAR = South Asia. BGD = Bangladesh; consumer of energy, with India’s manufacturing IND = India; LKA = Sri Lanka; PAK = Pakistan. Data on energy consumption in South Asia are only available for Bangladesh, India, Pakistan, and Sri Lanka. Latest available data are for 2020. sector alone consuming one-third of South Asia’s C. Energy intensity is defined as energy consumption (in tons of oil equivalent, toe) relative to nominal GDP (in thousands of U.S. dollars) in 2020. total energy production. This reflects the fact that, D.E.F. Based on decomposition into three sectors—agriculture, industry, and services—as detailed in in South Asia, energy intensity in manufacturing is annex 2.1.1. D. Energy intensity in each sector relative to economy-wide energy intensity in 2020. a multiple of that in other sectors. Growth versus energy intensity accelerated by a comprehensive package of South Asia’s energy consumption has increased by regulations, carbon taxes, subsidy cuts, and less than 3 percent per year over the past decade— policies to improve access to reliable electricity, about one-half the rate of output growth over the finance, and external markets. Apart from its same period (figure 2.3). In fact, in all four South direct benefits, greater energy efficiency can also Asian countries with available data, energy SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 63 intensity has declined since 2010, although not FIGURE 2.4 Firm-level energy intensity sufficiently to offset the impact of rapidly rising Within-subsector declines have been the main source of overall declines in output growth on energy consumption, especially energy intensity in the manufacturing and services sectors in South Asia since 2001, with shifts of resources between sub-sectors contributing little. in manufacturing (figure 2.3, annex 2.1.1). Within subsectors, mean firm-level energy intensity in South Asia declined by about 4 percent a year between 2006 and 2022. In the two economies with the highest energy intensity, India and Pakistan, the declines in A. Contributions to average annual B. Average annual change in firm- energy intensity were more modest than in changes in energy intensity, 2006–22 level energy intensity, 2006–22 Bangladesh and Sri Lanka. That said, India’s rate Percentage points Percent Between sector component 0 of decline in energy intensity compares favorably 1 Within sector component 0 -2 to the average for Latin America and the -1 -4 Caribbean (World Bank 2022a). The sector that -2 -6 was the main source of declines in energy intensity -3 -8 in the region was manufacturing, which is by far -4 -5 the region’s most energy-intensive sector. In India, Other SAR BGD IND PAK -10 Other SAR BGD IND PAK EMDEs the services sector also contributed to the decline. EMDEs Source: World Bank Enterprise Survey (WBES), World Bank. Sectoral shifts versus firm-level energy Note: EMDEs = emerging market and developing economies; SAR = South Asia; BGD = Bangladesh; IND = India; PAK = Pakistan. Based on two waves of WBES: Bangladesh 2022 and intensity 2006; India 2022 and 2014; Pakistan 2022 and 2006; Other EMDEs 2006 and 2022. India 2006 unavailable. Total sample size is 73,171 firms. Unweighted country average for other EMDEs. The WBES do not cover firms in agriculture, transportation, mining, and construction. Energy intensity The decline in South Asia’s energy intensity defined as ratio of electricity and wage bill. A. The chart shows the annual average growth rate of aggregate energy intensity decomposed into appears to have reflected improvements in energy changes within-subsectors and between subsectors. This is estimated for 15 2-digit subsectors in manufacturing and services. Method detailed in annex 2.1.2. efficiency within subsectors, rather than shifts in B. The chart shows the annual within-subsector growth rate of mean firm-level energy intensity. It is production to less energy-intensive subsectors in based on OLS regressions of log firm-level energy intensity on year, with country-specific subsector fixed effects. Separate regressions estimated for Bangladesh, India, Pakistan and a pooled dataset manufacturing and services. Indeed, the output of of all other EMDEs. The SAR average is the unweighted average of the estimated annual growth rate of South Asian countries. Method detailed in annex 2.1.2. the average firm in South Asia’s manufacturing and services sectors has become significantly less energy-intensive over time. 2006 and 2022 (figure 2.4). The results of a linear regression analysis suggest that, within subsectors, This conclusion is supported by data reported in the average firm’s energy intensity declined by 4 the World Bank Enterprise Surveys, which report percent per year between 2006 and 2022, the electricity expenses and total wage bills of compared with 2 percent per year, on average, in more than 73,000 firms in 15 two-digit other EMDE regions. The fastest rate of decline in manufacturing and services subsectors in 43 average firm-level energy intensity within EMDEs, including Bangladesh (2006 and 2022), subsectors occurred in India. India (2014 and 2022), and Pakistan (2006 and 2022; annexes 2.1.2 and 2.1.3). Firm-level energy Case study: India intensity is proxied by electricity expenses in percent of the wage bill. Qualitatively similar In India, the Annual Survey of Industries (ASI) results are obtained if energy intensity is measured followed a sample of 150,000 firms in 23 as electricity expenses in percent of revenue. manufacturing subsectors during 2001–18. It collected detailed data on firm characteristics, These data suggest that, on average, within- including energy (grid electricity and fuel) subsector declines in energy intensity reduced expenses and the wage bill (annex 2.1.4). ese overall energy intensity in South Asia by 3 percent data make it possible to compare energy intensity per year between 2006 and 2022. Furthermore, in the same firm at different points in time. Firm- shifts in activity between subsectors with differing level energy intensity is defined as energy expenses energy intensities played a negligible role in in percent of the wage bill. Using an alternative reducing overall energy intensity. Other EMDE definition of energy intensity—energy expenses in regions, on average, experienced smaller within- percent of firm revenues—does not alter the key subsector declines in energy intensity between results. 64 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.5 India: Within-firm reductions in energy subsidies. Conversely, the decline in energy intensity intensity could be explained by rising energy From 2001 to 2018, there were significant declines in energy intensity prices and a high elasticity of demand. In practice, within India’s manufacturing firms, on average. But divergences in energy however, the results are robust to controlling for intensity grew as a result of significantly faster declines among larger firms in faster-growing states with lower initial energy intensity. energy prices. Alternatively, the decline in energy intensity might reflect improved electricity grid A. Changes in energy intensity within B. Mean firm-level dispersion in reliability that makes the use of less efficient firms, 2001-18, by firm type energy intensity within sectors, private generators unnecessary. However, 2001-18 Percent electricity intensity and fuel intensity declined, Electricity intensity 0 Low/Small Group High/Large Group 5.5 which suggests that improving grid reliability -20 alone did not drive the observed decline in energy -40 5.0 intensity. -60 -80 4.5 In Indian manufacturing firms, faster cuts in Low versus high GDP Small versus large firms Low versus high energy energy intensity were associated with faster growth rate states intensity sectors 4.0 2001 2010 2018 employment growth. Between 2001 and 2018, employment growth in sectors with above-median Sources: Annual Survey of Industries for India; World Bank. reductions in energy intensity was 1.5 percentage Note: The sample size is 519,849 firm-year observations with 156,927 unique firms. The measure of electricity (or energy) intensity is the ratio of energy expenses to the total wage bill of each firm. points per year higher than in sectors with below- A. Chart is based on firm-level panel regressions of log of energy intensity on year dummies interacted with dummies for above versus below median GDP growth rate states (first pair of bars); median declines in energy intensity (figure 2.6). firm size below 50 workers versus firm size above 50 workers (second pair of bars); and above versus below median energy intensity sectors (third pair of bars). Each pair of bars depict the Within firms, a 1 percent reduction in energy cumulative percentage drop in energy intensity between 2001 and 2018 in the two corresponding groups. Method detailed in annex 2.1.4. intensity was associated with a 0.2 percent increase B. Chart depicts the within-sector ratio of the 75th percentile and 25th percentile energy intensity in employment during this period. While these levels, averaged across sectors in 2001, 2010, and 2018. results do not necessarily indicate a causal effect of reduced energy intensity on employment, they are robust to defining energy intensity as energy On average, within-firm energy intensity in the expenses in percent of revenues, and are indicative Indian manufacturing sector halved between 2001 of energy savings unlocking firm expansions. and 2018 (that is, fell by 4 percent per year). This is broadly in line with the decline in average The roles of firm-level and within-sector declines energy intensity in the broader South Asian in energy intensity among Indian firms are also a sample. It is also consistent with results from feature of advanced economies that have another recent study of energy intensity trends in undergone steep declines in energy intensity. In Indian firms based on a different dataset (Kumar, advanced economies, declines in energy intensity Mittal, and Pradhan 2023). However, the rate of were driven initially by structural change but decline in energy intensity varied with regional subsequently by within-sector, firm-level economic growth, initial sectoral energy intensity, improvements. In the United States, for example, and firm size (figure 2.5). Firms in states with market share reallocation toward less energy- above-median output growth rates, firms in intensive sectors was an important driver of falling sectors with below-median initial energy intensity economy-wide energy intensity during 1960–80 (in 2001), and larger firms reduced their energy but since then, within-industry declines in energy intensity significantly faster than other firms. In intensity have accounted for the larger share of particular, firms with more than 50 employees cut declines in overall energy intensity (Levinson their energy intensity 20 percentage points more 2021; Wing 2008). A similar pattern is observed than smaller firms. As a result, the dispersion of in a survey of studies that examine energy- energy intensity across firms within sectors intensity in a broad range of advanced economies widened over time. (Ang and Zhang 2000). In principle, if the price elasticity of energy Structural change that reallocated market share to demand is low, the decline in energy intensity less energy-intensive sectors may have been limited might reflect declining energy prices or growing in recent years in South Asia for two reasons: SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 65 limited relative price changes and the timing of FIGURE 2.6 India: Energy intensity cuts and major structural reforms. First, energy subsidies employment growth remain high in much of the region (see the policy In 2001-18, deeper declines in energy intensity were associated with faster section of this chapter). In India, ASI data on employment growth, both within sectors and within firms. electricity prices and wages in firms during 2001– 18 indicate that electricity prices have increased A. Average annual employment B. Firm-level employment growth over time but not as rapidly as wages. As a result, growth, by energy intensity decline, associated with 1 percent decline in 2001–18 energy intensity, 2001–18 the price of electricity relative to labor cost has Percent Percent fallen. Second, major structural reforms largely pre 6 1.5 Employment growth associated with 1 percent decline in energy intensity -date the sample used in this exercise. In India, 1.2 Average annual employment growth 4 major trade liberalization and industrial de- 0.9 licensing led to structural change that reallocated 2 0.6 market shares toward less energy-intensive 0 industries (Barrows and Ollivier 2018; Martin Sectors with below- Sectors with above- 0.3 median median 2011). The impacts of these policy changes, which cuts in intensity cuts in intensity 0.0 were enacted in large part in the 1990s, may have begun to diminish in the 2010s. Sources: Annual Survey of Industries for India; World Bank. Note: The measure of electricity (or energy) intensity is the ratio of energy expenses to the total wage bill of each firm. Energy-efficient technology A. Chart depicts the annual employment growth rate between 2001 and 2018 in 23 manufacturing sectors. The sectors are grouped into those with below-median and above-median cuts in sector- level energy intensity. adoption by firms B. The sample size is an unbalanced panel of 519,849 firm-year observations with 156,927 unique firms for the 2001–18 period. Chart depicts the correlation between log employment and energy intensity at the firm level, estimated with a firm level panel regression with firm fixed effects, as detailed in annex 2.1.4. The important role of within-sector and within- firm cuts in energy intensity suggests that the accelerated adoption of new green technologies at surveyed. Similarly, almost two-thirds of South the firm level could make a meaningful Asian firms had adopted Energy Star certified contribution to the energy transition and efforts to appliances, another basic energy-efficient lower GHG and other pollutant emissions. The technology. This adoption rate was in the upper speed of technology adoption, however, depends half of the range of adoption rates among the on firms’ characteristics, the nature of new other EMDEs surveyed. However, fewer than 7 technologies, and the context in which firms percent of firms had installed programmable operate. thermostats, a more advanced energy-efficient technology, a rate in the lower half of the range of Cross-country evidence the other EMDEs. More advanced technologies, such as Internet-of-Things (IoT)-enabled heating, South Asia’s firms have been early adopters of cooling, and ventilation systems, also had a low basic energy-saving technologies but lagging uptake in South Asia, broadly in line with other adopters of more advanced technologies. In 2022, EMDEs (figure 2.7). the World Bank’s Firm Adoption of Technology (FAT) Survey collected data from 10,090 firms in The faster-than-average adoption of basic energy- seven EMDEs, including 1,936 firms in saving technologies in South Asia is driven to a Bangladesh and 1,455 in India, on technology use, large extent by the high rate of adoption among as well as other firm characteristics (annex 2.1.5). Indian firms. This may reflect the effects of energy The surveys are part of a World Bank project on pricing and efficiency policies in India. There is technology adoption in EMDE firms (Cirera, evidence that a high electricity price for industrial Comin, and Cruz 2022). users (compared with households and farmers) has caused firms to shift to less energy-intensive More than three-quarters of South Asian firms had products (Abeberese 2017). Moreover, as adopted energy-efficient lighting, one of the most discussed in more detail in the policy section of basic energy-efficient technologies. This adoption this chapter, India’s high adoption rates of basic rate was higher than in the other EMDEs energy efficiency technologies may be due to its 66 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.7 Energy-efficient technology adoption by flagship energy efficiency programs—notably, a firms: Cross country evidence large-scale LED lights dissemination program Firms in South Asia have been early adopters of basic energy-efficient known as the Unnat Jyoti by Affordable LED for technologies but lag in the adoption of more advanced technologies. All (UJALA) program. Larger firms with better-educated managers who rely on key performance indicators have a higher level of energy-efficient technology adoption. Larger firms, firms with better-educated managers, A. Share of firms adopting energy B. Share of firms adopting Energy and firms that regularly relied on performance efficient light bulbs Star certified appliances indicators tended to be more energy efficient. A Percent Percent linear regression of a firm-level energy-efficient 100 100 technology index, which ranges in value from 0 to 80 80 6, was used to identify firm characteristics 60 60 associated with greater adoption of energy-saving 40 40 technologies in the pooled sample of Bangladesh 20 20 and India (annex 2.1.5). Compared with a 0 0 baseline average of 1.7 index points, the energy- SAR Other EMDEs SAR Other EMDEs efficient technology index in larger firms was 0.2– 0.4 of a point higher than in their smaller peers. C. Share of firms adopting energy- D. Share of firms adopting efficient VAV HVAC systems programmable thermostats Additionally, firms whose manager held a Percent Percent bachelor’s degree had a 0.3-of-a-point higher 100 100 energy-efficient technology index than other firms; 80 80 and firms that used Key Performance Indicators 60 60 (KPIs) as a management practice had a 0.2-of-a- 40 40 point higher energy-efficient technology index 20 20 than those that did not (figure 2.7). 0 0 SAR Other EMDEs SAR Other EMDEs Unreliable electricity grids may lock in energy- intensive technologies such as backup power E. Share of firms adopting IoT-enabled F. Impact on energy efficient generators. Backup power generators are systems technology level inefficient and polluting (Tong and Zhang 2015). Percent Tech. Index 100 0.8 Moreover, generators only partially compensate 80 0.6 firms for productivity losses incurred due to an 60 0.4 unreliable power grid (Allcott, Collard-Wexler, 40 and O’Connell 2016). Firms in Bangladesh and 0.2 India had significantly higher rates of generator 20 0.0 20-99 100+ Manager Uses 1-2 use than the firms surveyed in other EMDE 0 employees employees w/ BA or KPIs SAR Other EMDEs above regions: three-quarters of surveyed firms in India and Bangladesh reported using generators, three Source: Wave 2, Firm-level Adoption of Technology (FAT) Surveys, World Bank. times as many as the one-quarter of firms, on Note: EMDEs = Emerging Markets and Developing Economies; SAR = South Asia Region. Includes data from World Bank’s FAT Surveys of 10,090 firms in seven EMDEs (Brazil, Bangladesh, average, in other EDME regions (figure 2.8). In Cambodia, Chile, Ethiopia, India, and Georgia). India, almost 30 percent of surveyed firms had A.–E. The charts depict the range of country-level averages of percent of firms adopting technologies in South Asia and other EMDEs. For each country, the average percent of firms adopting a experienced a power outage in the month technology is estimated using sampling weights. The technologies are energy-efficient lighting, preceding the FAT survey; and those that had Energy Star-rated equipment, Variable Air Volume HVAC systems, programmable thermostats, and IoT-enabled systems to control temperature, lighting, or refrigeration. experienced outages were significantly more likely F. Charts depict coefficient estimates with 95 percent confidence intervals from OLS regressions of to be using a power generator (figure 2.8; annex Energy Efficient Technology Index on firm attributes, including employment size, sector, and region dummies. KPI = Key Performance Indicator. The Technology Index ranges from 0 to 6 in value. The 2.1.5). sample for the regression is the FAT Survey Wave 2 pooled data for 2,436 firms in Bangladesh, India and five other EMDEs. Method detailed in annex 2.1.5. South Asia’s firms may be early adopters of basic energy-saving technologies, but they are not EMDE leaders in technology adoption more broadly. The World Bank’s FAT survey has SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 67 collected information on the adoption of a wide FIGURE 2.8 Unreliability of grid power and use of range of general-purpose and industry-specific electricity generators technologies, in addition to energy-saving The use of backup generators among firms is exceptionally high in South technologies. e data suggest that, in general, the Asia. Firms that face power outages are more likely to use generators. level of technology adoption among firms in South Asia is at or below the EMDE average A. Establishments using generators B. Impact of power outages on generator use in India (Cirera, Comin, and Cruz 2022). Percent Percent 100 35 Incidence of power outages Case study: Bangladesh 30 Impact of power outages on generator use 80 25 To assess the potential for promoting the adoption 60 20 of energy-efficient new technologies among firms 40 15 10 through enhanced information provision, a 20 5 randomized control trial (RCT) is being 0 0 SAR Other EMDEs conducted among 504 firms engaged in leather goods and footwear manufacturing in Bangladesh Source: Wave 2, Firm-level Adoption of Technology (FAT) Surveys, World Bank. since 2022 (annex 2.1.6; Chaurey et al. 2023).2 Note: EMDEs = Emerging Markets and Developing Economies; SAR = South Asia Region. Includes data from World Bank’s Firm FAT Surveys of 10,090 firms in seven EMDEs (Brazil, Bangladesh, Firms in this industry have traditionally used Cambodia, Chile, Ethiopia, India, and Georgia). Method is detailed in annex 2.1.5. A. Chart depicts the range of country-level averages of percent of firms using generators for South sewing machines fitted with clutch motors, but a Asia and other EMDE groups. For each country, the average percent of firms adopting a technology is estimated using sampling weights. significantly more energy-efficient replacement— B. Chart depicts the percentage of firms that experienced a power outage (left bar) and the estimated the servo motor—is now available. The RCT marginal effect of power outages (with 95 percent confidence intervals) on the probability of owning/ sharing a generator in India. The latter results from a firm-level probit regression of generator use on conducted a baseline survey and then a dummy for whether firm faced a power outage in the past month, controlling for sector, size and firm age. implemented a randomized intervention to examine how the provision of information on the energy efficiency and ease of installation of servo motors could accelerate its adoption. A midline survey was conducted during March–May 2023 to groups: one control group, and three treatment measure the effects of the intervention groups. approximately three months later. A final, endline survey will be conducted to measure the effects The treatment groups differed in the intensity of nine months after the intervention, from the provision of information to them. The first September-November 2023. Pending the endline treatment group (“T1a”) was shown a video survey, the midline survey results already point to explaining the energy-saving benefits of the servo some patterns in firms’ adoption of energy- motor and the ease with which it can replace a efficient technologies (Chaurey et al. 2023). clutch motor in an existing sewing machine. The second group (“T1b”) was shown the same video Technology adoption after information and also had an electricity meter installed in one intervention of their existing clutch motor sewing machines. Firms in the third treatment group (“T2”) were Most of the firms in the RCT are micro, small, or medium firms, with 90 percent of them having 20 not only shown the video and provided with a or fewer workers. In the baseline survey, the RCT meter, but also given the opportunity to try a servo motor themselves. The details of the firms were asked about their beliefs or expectations intervention are discussed in annex 2.1.6. about the new technology, their willingness to pay for it, their pre-existing technology, and other firm Firms initially vastly underestimated the energy characteristics. Following the baseline survey, the savings from the new motors. The readings taken RCT firms were randomly assigned into four from the electricity meters installed in T2 firms confirmed that servo motors are significantly more 2 The RCT is a collaboration between the World Bank and energy-efficient than clutch motors. Specifically, researchers from Columbia University and Johns Hopkins School of mean electricity consumption per hour in a servo Advanced International Studies. motor machine was 83 percent lower than that in 68 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.9 Energy-efficient technology adoption by firms participating in the study had already firms: Randomized control trial replaced a clutch motor with a servo motor (figure Firms tended to underestimate the potential energy savings from a new 2.9). Unexpectedly, the preliminary results technology (servo motors) and were only willing to pay 56 percent, on indicate that firms in the control group (which average, of the actual purchase price. About 10 percent of firms in each of the four RCT groups of firms adopted servo motors after an informational had not been provided with information) also intervention was delivered to the three treatment groups. Their willingness adopted the new motors at a similar rate. to pay increased by 33 percent. Moreover, firms’ average willingness to pay had increased by one-third. The average price that A. Average electricity consumption for B. Willingness to pay for new old and new technologies: Actual technology firms were willing to pay at the midline was close values versus beliefs to the lower end of the price range of servo motor kWh 0.10 Clutch motor Servo motor Bangladeshi Taka brands available in the local market. 6,000 5,000 0.08 Information spillovers 4,000 0.06 3,000 0.04 2,000 Qualitative information collected by informally 0.02 1,000 interviewing firms in the study locations suggested 0 that firms in the control group adopted the new 0.00 Mean willingness to Market price Actual consumption Beliefs pay technology because of information spillovers. Information from the baseline survey also C. Share of firms adopting new D. Change in the willingness to pay suggested that, for firms participating in the RCT, technology for the new technology Percent other firms are a major source of information Percent 60 60 about new technologies. According to data 50 50 collected in the baseline survey, 74 percent of the 40 40 firms that were aware of servo motors at the start 30 30 of the experiment had heard about them from 20 20 another firm in the same industry (figure 2.10). 10 10 0 Control T1a T1b T2 0 Control T1a T1b T2 This suggests that the information about the energy efficiency of the servo motor that was Sources: Chaurey et al. 2023; World Bank. provided in the RCT could have spread from Note: RCT baseline and midline surveys and meter readings, 2022. Sample includes 504 firms in Bangladesh in the leather goods and footwear industry. treatment to control firms. To test for such A. Estimates of mean electricity consumption based on hourly readings of electricity meters installed in one clutch and one servo motor sewing machine in each of 124 T2 firms. Meter readings collected information spillovers, a measure of (indirect) for every day in January. Mean baseline beliefs about daily electricity consumption by a clutch motor and servo motor sewing machine in the full sample of firms measured in the baseline survey. “exposure” to the informational intervention was B. Mean baseline willingness to pay (WTP) for a servo motor in the full sample of firms. WTP elicited constructed using detailed data on the geographic through Becker-De Groot-Marshack procedure. C. Share of firms adopting a servo motor after the intervention, measured in midline survey. locations of the RCT firms (Chaurey et al. 2023). D. Change in willingness to pay for servo motor after the intervention. Methodology as in Chaurey et al. (2023). A probit regression was estimated to examine the adoption of new servo motors in the control group as a function of each firm’s geographic distance a clutch motor machine (figure 2.9). And yet, in from treated firms. the baseline survey, firms on average believed that a servo motor consumes only 30 percent less Geographic proximity to a larger number of electricity per hour than a clutch motor. Perhaps treatment firms significantly increased the as a result, the average firm was willing to pay only probability of adoption among control firms, 56 percent of the actual purchase price of the servo pointing to sizable information spillovers among motor featured in the trial. The degree of firms. The exposure effect was stronger for underestimation was broadly uniform: there were proximity to strong treatment (T2) firms—those no statistically significant differences in which had received a servo motor—than expectations across different types of firms. proximity to any treatment firm. Quantitatively, the results suggest that the spillover effect could By the time of the midline survey—three months account for 70 percent of the adoption rate in the after the intervention—about one-tenth of the control group. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 69 FIGURE 2.10 Information spillovers among firms Policy implications Firms closer to treatment firms that had received information about the new A wide range of policies can promote transition technology (servo motors) were significantly more likely to adopt that technology. away from fossil fuels. Many of these policies go beyond the scope of this chapter, including A. Sources of information about servo B. Informational spillovers and measures to reduce energy use in buildings, the motors adoption in the control group of firms reduction of environmentally harmful subsidies, Number of firms Percentage points measures to promote the switching of fuels used 400 1.2 300 1.0 for electricity and heat generation, and policies to 200 0.8 promote a shift toward battery-powered motor 100 0.6 vehicles (World Bank 2023b, 2023c). For 0 Other Other Other Suppliers Other 0.4 firms in firms in firms or buyers example, policies to promote the use of energy- the sector the same from a 0.2 locality, different efficient cooling technologies in buildings, cold- but locality 0.0 different and Exposure to T2 Exposure to T1a, T1b chains, and transportation can help meet the rising sector sector or T2 demand for adaption to heat stress in a sustainable Sources: Chaurey et al. 2023, World Bank. manner, and are being operationalized in India Note: Sample of 504 firms in the leather goods and footwear industry. A. Chart summarizes the responses to the baseline survey question: “From whom did you learn that through the India Cooling Action Plan servo motors can be replaced by clutch motors”, directed at firms that reported knowing that servo motors can replace clutch motors in sewing machines. The chart depicts the number of firms (Government of India, 2019; World Bank, selecting each possible response (multiple responses possible). 2022b). There is also considerable merit in B. Chart depicts the estimated marginal impact on the probability of servo motor adoption of exposure to treatment through spillovers among control groups firms, based on a probit regression of regional initiatives to reduce emissions of GHG a dummy indicating whether the firm adopted a servo motor after the intervention on an exposure measure and controls for firm size, age, manager education, and management variables. The probit and other pollutants since emissions spread across is estimated on the midline survey sample of 162 control group firms. Method detailed in Chaurey et al. (2023). borders along airsheds (World Bank 2023a). The analysis of this chapter demonstrates the potential benefits of multi-pronged government policies that remove a variety of constraints on firms’ adoption of energy-efficient technologies energy-efficient technologies in a small-scale and (box 2.1). The chapter has focused on the role of incremental manner that is unlikely to cause firms’ access to accurate information about new stranded asset. For example, the energy-efficient technologies and the role of human capital and motor examined in the Bangladesh RCT can management quality and access to reliable power replace inefficient motors in existing sewing grids in the adoption by firms of energy-efficient machines and, hence, its diffusion need not cause technology. The literature has also identified a “stranding” of existing sewing machines (the key regulations addressing market failure (such as the assets of firms in leather goods and garment failure to compensate firms for the positive sectors). externalities from green technology adoption), access to credit, and openness to trade and FDI as The extent to which constraints on technology important factors in the adoption of energy- adoption are binding depends, in part, on the efficient technologies. technology. Energy-efficient technologies range from basic LED lights to highly advanced IoT- South Asia’s above-average energy intensity enabled heating, cooling, and ventilation systems. suggests considerable room for technological There are also many industry-specific energy- catchup among firms, with limited risk of saving technologies, such as servo motor sewing “stranded assets.” Stranded assets could arise if machines in the leather goods and garment firms make major capital investments into industries. In addition, there are a host of technologies that may rapidly become obsolete as a beneficial emission abatement technologies, such result of the energy transition. Since South Asian as industrial air purifiers. These technologies differ firms lag in technology adoption for all but the along many dimensions, including maturity, most basic energy-efficient technologies, there upfront costs, and divergences between private appears to be ample room for firms to adopt and social benefits. 70 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 BOX 2.1 Literature review: Addressing barriers to technology diffusion in firms The literature identifies several constraints on the adoption of new technologies in general, including energy-efficient technologies: lack of finance, information, or scale economies, as well as high costs and perceived difficulty of operating new technology. Advisory support, financial support, and regulation have been shown to accelerate technology adoption. This chapter focuses on the adoption of energy-efficient new technology from own experience, peers, and technologies by firms. An extensive literature has technology extension agents; better schooling; larger established broader patterns in firms’ adoption of new scale economies; access to finance; and conducive technologies in general—not just energy-efficient behavioral factors (Foster and Rosenzweig 2010). technologies. This box distills the main findings of this broader literature on two questions. Interventions to encourage technology adoption • Which factors have been associated with tech- nology adoption? A recent meta-analysis of the evidence from firm-level studies suggests that a range of interventions can be • Which interventions have encouraged technology successful in promoting the adoption of new adoption? technologies in EMDEs (Alfaro-Serrano et al. 2021). The review considers the following types of Constraints on, and enablers of, new interventions: (i) direct funding (including loans, technology adoption subsidies, insurance, and in-kind or cash grants); (ii) indirect financial support (such as loan guarantees and Reviews of the evidence on technology diffusion, mostly policies that reduce input costs); (iii) direct non- from advanced economies, identify several constraints financial support (informational interventions, public on firms’ adoption of new technologies. ese include: technology extension services, and advisory and low or heterogenous benefits of new technologies; high consulting support); and (iv) regulations and standards upfront costs of acquiring the technology or learning (defined as rules, policies, and characteristics of the how to use it effectively; lack of access to accurate environment that affect agents’ incentives, such as information about the technology; uncertainty about business regulation and trade policies). the costs and benefits of the technology; costly access to inputs that are complementary with the technology; Barring one, the studies that meet the systematic review and low returns from early adoption of technologies criteria employed in the meta-analysis are all from that require a large network of users to be viable (Hall EMDEs. Most of the studies examine direct non- 2005; Williams and Bryan 2021). financial support or direct financial support. About half find positive and statistically significant effects of Evidence also suggests that technological diffusion in interventions on technology adoption. For each type of EMDEs is accelerated by exporting to advanced- intervention, some studies find the expected positive economy markets and easier availability of higher- and significant impact. However, there is no quality inputs, and that it is constrained by restricted intervention type for which all, or all but one, studies access to credit, limited managerial ability, and find a significant positive impact. The meta-analysis misaligned incentives between workers and managers suggests that tailoring the intervention to the context (Verhoogen 2023). A review drawing largely on may improve their impact. evidence from agriculture in EMDEs highlights several factors that encourage technology adoption: low cost of new technologies; a simple process of learning about the SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 71 Policy initiatives to improve energy efficiency are FIGURE 2.11 Literature on policies for firm technology already underway in SAR. For example, adoption and energy efficiency gains Bangladesh’s Energy Efficiency and Conservation There is strong evidence that regulation, especially market-based Master Plan includes a series of programs aimed at regulation, is effective in encouraging energy-saving technology adoption. Information campaigns and behavioral nudges have fewer unintended large industrial energy consumers, residential consequences, but their effectiveness is more uncertain. Evidence that consumers, buildings, private companies, and financing policies can promote technology adoption has only recently government agencies (Government of the People’s begun to emerge. Republic of Bangladesh 2015). These include energy auditing and benchmarking; energy A. Studies reporting successful B. Studies reporting successful intervention, by country intervention, by type of policy efficiency-label certification; low-interest loans and Percent of studies Percent of studies subsidies for energy-efficient investment; 100 Weighted by impact factor Unweighted 100 Weighted by impact factor Unweighted preferential taxes that apply to high-efficiency 80 80 60 equipment; and technical capacity building. In 60 40 partnership with the World Bank, Bangladesh is 40 20 0 also facilitating the adoption of green practices Financing Market-based Administrative Information 20 regulation regulation and environmental standards in microenterprises 0 Europe Other China United India through the Sustainable Microenterprise and States Resilient Transformation project (World Bank 2023d). India introduced several flagship Source: World Bank. Note: Results from a review of 45 academic and policy studies on the impact of specific policy programs to improve energy efficiency in interventions (regulation, information/behavioral, and finance) on either firm technology adoption or firms’ energy efficiency. Impact weighting according to the RePEc ranking of the journal or working households and firms after the passage of the paper series in which the study was published. Energy Conservation Act (2001) (Kumar, Mittal, and Pradhan 2023). The innovative UJALA program distributes LED lighting at a studies were for the United States (15 studies), the large scale through bulk procurement and a Euro area (15 studies), China (six studies), and delayed payback schedule. Consumers face a small India (six studies). upfront fee and can pay back the cost of the lights through adjustments in their electricity bill. There are three takeaways from the literature review. Although UJALA is aimed at households, it may have had spillover impacts on firms by reducing • First, regulation and/or environmental the cost of LED lights. The retail price of LED taxation can help close the gap between the lightbulbs fell by 70 percent between 2014 and social and private costs of emissions. They are 2017 (Singh and Bajaj 2018). The “Perform- typically effective in achieving this primary Achieve-Trade” (PAT) program helps firms reduce objective but command-and-control energy intensity through energy audits, technical regulation, especially, can bring considerable advice, and tradable energy savings certificates unintended consequences. In addition, in an (Government of India 2023). environment of weak regulatory capacity, such as in South Asia, regulation may be less To identify the policy options with the best track effective (Duflo et al. 2013; 2018). record of encouraging energy-efficient technology adoption by firms, a comprehensive literature • Second, information campaigns and nudges review was conducted, based on 45 studies can avoid these unintended costs, but published in peer-reviewed journals or policy generally have smaller effects. publications over the past three decades. The studies considered were specifically about energy- • Third, policies can appropriately be used to efficient or pollution-abatement technologies. The subsidize technology adoption, to the extent review included 18 studies of regulation and that environmental benefits are not fully environmental taxes, 16 studies of information captured by adopters, or to ease financing campaigns or behavioral nudges, and 11 studies of constraints that prevent technology adoption. financing constraints (figure 2.11). Most of the Empirical evidence on the effectiveness of 72 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 financing support, although promising, is to unregulated entities, particularly where limited. These policy options are discussed in environmental policy is only partially enforced. An more detail below. example is China’s regulation applying energy use quotas to the “Top 1,000” most energy-intensive Regulation, taxes, and subsidies firms. Regulated firms, typically large conglomerates, used their corporate structures to Regulation is the most common approach to shift energy consumption to unregulated entities, reducing CO₂ emissions and pollution, and undermining energy savings, while still suffering improving energy efficiency. Environmental substantial output losses (Chen et al. 2021). regulations generally fall into two categories: Second, regulation may cause productivity losses. command-and-control policies, such as mandates Water pollution regulations in China reduced on energy efficiency standards and programs that firms’ productivity by 24 percent, implying large set hard quotas on firms’ harmful emissions, and losses from the policy (He, Wang, and Zhang market-based mechanisms, such as cap-and-trade 2020). Such productivity losses are more schemes, that allow firms to exceed quota limits by pronounced in the case of pollution controls, since purchasing permits from those that do not. abatement technology imposes costs on firms Command-and-control regulations have often without any concomitant benefit to them (as proven effective. About two-thirds of the studies opposed to society). In contrast, regulations of the impact of regulations document significant relating to energy efficiency yield benefits to firms impacts. Policies to manage air and water by promoting the adoption of technologies that pollution often involve quotas on emissions and lower energy costs (Allcott and Greenstone 2012) phase-outs of technologies that do not meet and improve productivity (Adhvaryu, Kala, and minimum environmental standards (Greenstone Nyshadham 2020). Third, voluntary compliance and Hanna 2014; Harrison et al. 2015). Such tends to be ineffective, so robust enforcement of policies have been effective in reducing air environmental quotas is essential (Kube et al. pollution in China (He, Wang, and Zhang 2020; 2019). But enforcement is challenging, requiring Bu et al. 2022), India (Harrison et al. 2015), and monitoring and penalties that may be difficult to the United States (Shapiro and Walker 2018). implement in the face of low state capacity (Duflo Harrison et al. (2015) study an Indian supreme et al. 2018), corruption (Duflo et al. 2013), or court ruling that imposed air pollution quotas on complex incentive design (Blundell 2020). firms in 17 cities across the country. Firms Market-based regulations have shown similar adapted to the regulation, not by scaling down benefits to command-and-control regulation, but output, but rather by adopting new pollution with fewer distortions. Thus, almost nine-tenths abatement technologies that allowed them to of the studies of market-based regulations lower overall emissions at given output levels. documented positive effects. Two of the best- Command-and-control environmental regulations known market-based schemes are the 2005 are more common in SAR than market-based European Union Emissions Trading System regulations. For example, India’s 2020 National (ETS) and California’s cap-and-trade program for Clean Air Program has identified 122 cities that carbon and other air pollutants. Both programs do not meet India’s air quality standards for have strong records of success. The EU carbon particulate matter concentrations and requires Air market not only increased the adoption of energy- Quality Action Plans that will include the phasing efficient technologies and reduced carbon -out of certain non-compliant technologies. emissions (Colmer et al. 2023), but also boosted Similar city-focused action plans are being low-carbon technological innovation, as measured implemented in urban areas of Bangladesh and the by patenting activity, by as much as 10 percent Punjab province in Pakistan (World Bank 2023a). (Calel and Dechezleprêtre 2016). This aligns with However, command-and-control policies can also evidence on pollution taxes in OECD countries have costly unintended consequences. First, firms (Brown, Martinsson, and Thomann 2022). In may respond to quotas by shifting their operations California, provisions of the Clean Air Act were SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 73 implemented via a cap-and-trade scheme rather FIGURE 2.12 Subsidies and effective carbon prices than quotas, avoiding the adverse side-effects of Most countries in South Asia spend more than 1.5 percent of their GDP on quota systems experienced in other U.S. states energy subsidies. In some South Asian countries, subsidies are sufficiently (Curtis and Lee 2019). Studies of the EU ETS large to turn carbon prices negative. suggest no adverse effects on competitiveness A. Energy subsidies, 2021 B. Total carbon price, 2021 (Colmer et al. 2023), firms’ relocation (Martin et al. 2014), or carbon leakage (Dechezleprêtre et al. Percent of GDP 3 PPP US$/ton of CO2 60 2022). Sadayuki and Arimura (2021) demonstrate 30 positive technology adoption and emissions 2 0 spillovers from a carbon market in Japan, as firms -30 -60 transferred green technologies to their unregulated 1 -90 plants. ese approaches are, of course, not -120 entirely immune to the risk of emissions shifting, 0 PAK BGD SAR IND LKA BGD PAK LKA IND NPL SAR Other EMDEs particularly if regulations are unevenly applied across integrated markets (Bartram, Hou, and Sources: Agnolucci et al. (2023); International Energy Association; World Bank. Note: EMDEs = emerging market and developing economies; SAR = South Asia; BGD = Kim 2022). Another important caveat is that these Bangladesh; IND = India; LKA = Sri Lanka; NPL = Nepal; PAK = Pakistan. A. Unweighted averages for South Asia (SAR) and EMDE commodity importers. regimes generally have imposed low carbon prices; B. The Total Carbon Price combines a comprehensive set of direct carbon pricing policies with higher-priced regimes could generate more adverse indirect interventions on carbon-containing energy source to measure the aggregate carbon price signal faced by agents in the economy. The Direct Carbon Price component of the Total Carbon consequences. Price includes all carbon taxes and emission trading systems, adjusted for the share of the country’s carbon emissions covered by such direct carbon taxes. The Indirect Carbon Price component includes fuel excise taxes, fuel subsidies, and value-added tax (VAT) deviations (arising if VAT rates Market-based approaches are already being piloted on fuels are below the standard VAT rate). For each fuel and sector, the Indirect Carbon Price is estimated as the deviation between the retail price and the supply cost, adjusting for the upstream in South Asia, with encouraging results. A recent carbon price. A negative Total Carbon Price is to be interpreted as a net subsidy on carbon, while a positive Total Carbon Price is to be interpreted as a tax. Further details are available in Agnolucci et randomized experiment with emissions trading for al. (2023). SAR and EMDE averages are emissions-weighted. particulate matter in the Indian state of Gujarat reduced emissions by up to 30 percent, and it did so at a low cost (Greenstone et al. 2023). India’s and increasing adoption of carbon taxes, their PAT program incorporates a market-based effects on firm technology choices and energy regulatory mechanism to help participating firms intensity remain under-studied. achieve energy efficiency improvements at a low cost (Government of India 2023). In this e elimination of subsidies on fossil fuels can be program, highly energy-intensive firms are a powerful mechanism to discourage their use. assigned an energy-intensity target and can trade Most countries in SAR spend more than 1.5 excess energy savings certificates on an active percent of their GDP on energy subsidies. India exchange market. ranks among the world’s five largest providers of fossil-fuel subsidies, especially for coal and Carbon taxes are another policy option, one that liquified petroleum gas (Damania et al. 2023). In retains the desirable properties of emissions Bangladesh, Pakistan, and Sri Lanka, fossil fuels trading but also generates government revenue and fossil fuel-generated electricity are so heavily and reduces monitoring costs. One benefit is that subsidized compared with world prices that the revenue from carbon taxes can be used to offset effective price of carbon is negative (figure 2.12). the negative welfare impacts of emissions In contrast, Nepal does not subsidize petroleum restrictions on the poor. Another oft-cited products, and even imposes a pollution tax on justification for carbon taxes is that they encourage them (World Bank 2021). e removal of energy firms to innovate and adopt low-carbon subsidies would encourage more efficient use of technologies (Timilsina 2022). Empirical support fossil fuels and shifts toward green energy sources for this hypothesis comes from experience with and less fossil-fuel-intensive activities—provided pollution taxes, which have been shown to alternatives are readily available and affordable increase research and development expenditures (Damania et al. 2023). Lasting subsidy reduction on clean technology (Brown, Martinsson, and tends to be politically challenging; it has often omann 2022). Notwithstanding the promise been more successful when implemented at times 74 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 of falling energy prices and in combination with of laboratory estimates because households do not social benefit reform (World Bank 2020a). use and maintain their stoves properly (Hanna, Duflo, and Greenstone 2016). A related literature Information and behavioral nudges on agricultural technology adoption generally finds that differences in expected returns are e empirical evidence presented in this chapter reflected in farmers’ adoption of new technologies suggests that well-targeted, low-cost interventions (Jack 2013; Suri 2011). e literature also to address informational constraints can help suggests that in some cases, the low adoption of boost the adoption of energy-efficient technologies seemingly profitable energy-saving technologies— by firms provided the benefits to them outweigh dubbed the “energy efficiency paradox”—reflects the costs. Such programs can exploit informational rational, well-informed decisions on the part of spillovers to economize on program costs. For firms or households: that is, the technology in example, the government could focus question is not profitable in real world conditions interventions on early or “lead” adopters in (Allcott and Greenstone 2012; Anderson and manufacturing clusters, with the expectation that Newell 2004; Gerarden, Newell, and Stavins other firms in the cluster will be influenced by 2017; Gillingham, Keyes, and Palmer 2018). these early adopters. Technology diffusion programs in the agriculture sector already employ Second, lack of information is usually only one of such low-cost designs leveraging spillovers, but many constraints, and it may not be the one that evidence to support similar programs in non- is binding. It is often for this reason that policies agriculture sectors has been limited so far.3 that provide firms information on their energy use Interventions to address informational constraints and strategies to raise its efficiency—so-called include both information campaigns and energy audits—have produced mixed results. In behavioral nudges to encourage firms to adopt the Industrial Assessment Centers program in the energy-saving technologies. United States, firms adopted recommendations from energy audits only 50 percent of the time Firms often report a lack of awareness about (Anderson and Newell 2004). Similarly, a energy-efficient technologies in surveys, and these randomized evaluation of an intensive energy information shortfalls correlate negatively with consulting program among large Indian technology adoption (De Groot, Verhoef, and manufacturing firms showed only modest effects Nijkamp 2001; Hochman and Timilsina 2017). on energy efficiency and technology adoption (Ryan 2018). In fact, increased efficiency enabled However, for a number of reasons, there are often firms to expand, leading to increases in overall only limited benefits from information campaigns energy usage despite a reduction in energy alone. intensity, an example of the so-called “rebound First, especially among smaller firms and effect” (Gillingham, Rapson, and Wagner 2016). households, new technologies do not always work ird, there are complementarities between as well in real-world conditions as they do in information acquisition and broader business laboratory settings. For example, the actual practices that may need to be exploited for more benefits from a cleaner cookstove may fall far short efficient technology use to result. Energy audits have been found to be more effective when paired 3 Positive spillovers between firms, learning externalities, or with monitoring (Yajima and Arimura 2022), externalities from lower pollution might justify subsidizing the adoption of green technologies. That said, this chapter does not financing (Bodas-Freitas and Corrocher 2019; examine the case for large-scale “industrial policies,” typically Kalantzis and Revoltella 2019) and top understood to be sectoral or place-based, and “explicitly target(ing) management participation (Blass et al. 2014). the transformation of the structure of economic activity in pursuit of some public goal” (Juhasz, Lane, and Rodrik 2023, page 4). The Business training programs have been shown to be policies that this chapter discusses do not have a sector or location effective in diffusing better business practices focus because there is scope for energy-saving technologies in most along with more efficient technologies (Bloom et sectors. The typical case for industrial policy would be more relevant for issues related to the growth of the renewable energy production al. 2013; Cirera and Maloney 2017; Piza et al. sector. 2016). Among both U.K. and U.S. firms, better SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 75 management practices have been associated with presents more objective evidence. Even temporary reduced energy intensity (Bloom et al. 2010; Boyd boosts to U.S. firms’ cashflows reduced firms’ and Curtis 2014; Martin et al. 2012). emissions (Xu and Kim 2022). Similarly, Management has been shown to matter for contractions in bank credit supply have been pollution abatement, while credit constraints also shown to reduce green technology adoption matter for investment in new “green” machinery (Accetturo et al. 2022; De Haas et al. 2023) and (De Haas et al. 2023). Good management has also increase emissions (De Haas et al. 2023). Capital limited the costs of regulatory compliance by structure also matters: a 2006 tax reform in reducing the productivity losses associated with Belgium that lowered the cost of equity relative to pollution abatement technology (Hottenrott, debt reduced firms’ emissions intensity (De Haas Rexhäuser, and Veugelers 2016). Similarly, and Popov 2023). But while there is strong reductions in coal usage among Chinese firms in evidence that household and consumer subsidies response to higher carbon pricing has been greater encourage the adoption of energy-saving with better management (Yong et al. 2021). technologies (Gillingham, Keyes, and Palmer 2018), there is limited evidence of the “Nudges” that address behavioral biases can be a effectiveness of policies that ease firms’ credit cost-effective complement to information constraints. provision. Interventions such as reminders, defaults, and peer comparisons have been found to Better access to grid electricity may reduce reduce household energy consumption at a cost of reliance on inefficient, polluting substitutes. As only 2.8c per kWh saved—just half of the social suggested by the findings from the FAT surveys marginal cost of electricity generation discussed in this chapter, a more reliable grid may (Gillingham, Keyes, and Palmer 2018). Social reduce the use of portable generators. It may also pressure has been shown to increase the adoption improve firm revenues and productivity (Allcott, of environmentally friendly propane fuel among Collard-Wexler, and O’Connell 2016). Measures Mexican brickmakers (Blackman and Bannister to improve grid reliability may be complemented 1998). Default settings also matter: automatic bill- by interventions to promote the diffusion of pay increased energy use by reducing the cleaner electricity backups that use renewable awareness of energy consumption for both energy sources. households and firms (Sexton 2015). While this evidence is generally from studies of households, it Policy reforms that make economies more open may be particularly relevant in South Asia where internationally, including by improving access to more than 90 percent of firms are small and export markets and foreign investment may, informal, and the distinction between firms and among their many potential benefits, induce faster households is often not meaningful (Bussolo and adoption of energy-efficient technologies. First, Sharma 2022). local firms must compete more with imports, incentivizing investment in energy-saving Access to finance, markets, and public technology (Gutiérrez and Teshima 2018). services Second, firms have greater access to imported inputs, which may be more energy efficient Financing constraints are an oft-cited obstacle to (Martin 2011). Third, external demand shocks firms’ technology adoption. A number of studies may lead firms to upgrade technology, reducing document correlations between firms’ reported emissions intensity (Barrows and Ollivier 2018). credit constraints and technology adoption (Caporale, Donati, and Spagnolo 2023; Fleiter, The effects of FDI are in some respects similar to Schleich, and Ravivanpong 2012), energy those of international trade. Since multinational intensity (Biscione et al. 2023; Zhang, Li, and Ji subsidiaries usually have access to better 2020), and emissions (Andersen 2017). While technology, they are likely to be able to reduce the these studies typically rely on self-reported energy intensity of their activity more easily constraints in survey data, some recent work (Eskeland and Harrison 2003). This can lead to some unintended consequences, however: for 76 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 2.13 Historical comparison example, a policy-driven multinational divestment Compared with their elevated levels in ECA economies in the 1990s, the in the Nigerian oil sector increased pollution emissions and energy intensities of GDP in SAR economies is significantly because local firms were heavier polluters (Rexer lower. However, the pollution intensity of several SAR economies is significantly higher. Several SAR countries also have considerably higher 2022). One concern about increased openness is pollution intensities than China had in 2012 when it embarked on reforms the “pollution haven hypothesis”—that economic aimed at lowering air pollution. Despite rapid economic growth, ECA countries, and to a lesser extent, China, reduced their emissions and integration with uneven regulation will lead firms energy consumption in the years following 1995 and 2012. to relocate production to less regulated countries. While Eskeland and Harrison (2003) find no A. GHG emissions intensity of GDP, B. PM2.5 emissions intensity of GDP, support for this hypothesis, more recent work has 2021 2018 Mt CO2 eqiv./US$ billion of GDP Kt/US$ billion of GDP found some evidence in its favor (Kellenberg 3.0 ECA (1995) China (2012) 8 ECA (1995) China (2012) 2009; Levinson and Taylor 2008; Wagner and 2.5 6 Timmins 2009). Nonetheless, environmental 2.0 regulation is just one of many factors driving FDI 1.5 4 and trade flows. 1.0 2 0.5 Besides reforms to enhance openness to foreign 0.0 0 PAK NPL BTN IND BGD LKA MDV PAK NPL BTN IND BGD LKA MDV trade and capital flows, policies to support national innovation systems and orient them C. Energy intensity of GDP, 2020 D. GHG emission commitments in toward green technologies may help facilitate the historical comparison, 2021–30 Toe/US$ billion of GDP development of energy-efficient technologies Percent GHG emissions GDP 0.8 ECA (1995) China (2012) 9 suited to local conditions. Such policies would 6 include measures to strengthen intellectual 0.6 3 property rights protection, enhance research and 0 0.4 -3 development capabilities, facilitate collaboration 0.2 -6 between public and private actors, and incentivize -9 technology transfer (Cirera and Maloney 2017). 0.0 IND PAK BGD LKA MDV ECA IND PAK LKA BGD 95-05 It can be done; it has been done E. ECA 1995–2005: Growth in emis- F. China since 2012: Growth in sions, energy consumption, and real emissions, energy consumption, and GDP real GDP Given the unintended consequences of, and Percent per year Percent per year interactions among, individual policy 6 Emissions, consumption Output 9 Emissions, consumption Output interventions, comprehensive policy packages are 6 4 more likely to be successful than narrowly targeted 3 2 0 ones. This has been the experience both of a group 0 -3 of nine countries in Europe and Central Asia -2 -6 GHG PM2.5 Energy (ECA) between 1995 (when their GHG emissions -4 GHG PM2.5 Energy emissions (2012-21) emissions (2012-18) consumption growth and energy intensity peaked) and 2005, and of emissions emissions consumption (2012-20) China between 2012 (when its PM2.5 emissions Sources: Crippa et al. (2022); EDGAR database; European Commission; World Development peaked) and 2018. In both cases, emissions, indicators; World Bank. energy consumption, and pollution were cut Note: EMDEs = emerging market and developing economies; SAR = South Asia; BGD = Bangladesh; BTN = Bhutan; IND = India; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan. despite rapid economic growth. A. GHG emissions (in mt of CO2 equivalent) relative real GDP (at 2021 prices and exchange rates). Data for SAR for 2021. Data for ECA in 1995 are the GDP-weighted average of 9 ECA countries. ECA region, 1995–2005. Despite 4 percent B. Emissions in kt per real GDP in billions of U.S. dollars (in 2021 prices and exchange rates). 2018 data for SAR countries . Data for ECA in 1995 are the GDP-weighted average of nine ECA countries. average annual output growth between 1995 and C. Energy intensity of GDP is defined as energy consumption in tons of oil equivalent relative to real 2005, PM2.5 pollution and GHG emissions GDP in billions of U.S. dollars (in 2021 prices and exchange rates). Data for ECA economies in 1995 are the GDP-weighted average of nine ECA countries. Data for South Asian economies are for 2020, declined by a cumulative 37 and 8 percent, the latest available data. respectively, in nine countries of the ECA region D. For South Asia, average annual percent change in GHG emissions consistent with emissions commitments in latest national Nationally Determined Contributions and real GDP growth consistent (Bulgaria, Belarus, Georgia, Kazakhstan, Moldova, with potential growth through 2030. For ECA region, annual average aggregate GHG emissions and Poland, Romania, Russia, and Ukraine; figure real GDP growth during 1995–2005. The ECA region includes nine countries (Bulgaria, Belarus, Georgia, Kazakhstan, Moldova, Poland, Romania, Russian Federation, and Ukraine). 2.13). These declines reflected a plunge in energy E.F. Average annual percent change in GHG emissions, energy consumption, and real GDP. intensity, which was accompanied by a 2.6 percent E. Real GDP-weighted average for nine ECA economies. F. Latest data: GHG emissions 2021; PM2.5 emissions 2018; for energy consumption 2020. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 77 decline in energy consumption. This change ANNEX 2.1. Methodology occurred as overindustrialized former Soviet Union economies restructured away from heavy 2.1.1 Sectoral decomposition of energy industry. In addition, the share of coal in consumption growth electricity and non-electric heat production Energy consumption growth is decomposed into declined, reducing the emissions intensity of within-sector changes in energy intensity and power generation (Breitenfellner et al. 2021). This between-sector changes in the shares of each sector transformation was achieved through a in real GDP. Matching data from the OECD’s comprehensive structural transformation that Green Growth Database for energy use with involved: privatization; enterprise restructuring; sectoral national accounts data from Haver the liberalization of prices, foreign exchange, and Analytics allows for a decomposition into three trade; competition policy; financial sector reform; sectors: agriculture, industry, and services. Since and legal reform (Sachs 1996; Sachs and Woo different classifications are used for energy use and 1994). real GDP, industry in the national accounts is China, 2012–18. Despite 7 percent average interpreted as industry including construction. annual output growth, PM2.5 air pollution was Energy use in industry is interpreted as including cut by one-sixth in these six years, again through a energy consumption in “other” sectors, which is comprehensive policy shift. Monitoring and largely building-related. Services in the national reporting of air pollution was enhanced. accounts are interpreted as services including Tightened emission standards for motor vehicles transport services, and energy consumption in and power plants were robustly enforced. Straw services is interpreted as including transport burning was prohibited. New air pollution control services. For Sri Lanka, energy consumption in the standards were put in place and enforced. agriculture sector is unavailable. Performance evaluations and promotion of Energy intensity is defined as energy consumption government officials were tied to emission (in tons of oil equivalent) relative to real GDP (in reductions (Greenstone et al. 2021; Lu et al. constant local currency terms). This allows the 2020; Zeng et al. 2019). These efforts also helped following sectoral decomposition: reduce growth of emissions and energy i i i i consumption to one-quarter to one-half the pace Et 3 ( et − e( t −1) ) (Yt + Y( t −1) ) − 1 =  i =1 + of output growth. E( t −1) eti−1 Yt i−1 i i 3 (Yt i − Yt i−1 ) (et + e( t −1) )  i =1 Y i e(it −1) , t −1 where Et is aggregate energy consumption, Yt is aggregate real GDP, Yit is real value added of sector i, Eti is energy consumption in sector i, i E and e = Y is energy intensity of sector i. e first i t t t i term on the right of the equation denotes the contribution of sectoral changes in energy intensity to energy consumption growth; the second term denotes the contribution of sectoral output growth to energy consumption growth. 2.1.2 Firm-level approaches to measuring energy intensity Definition. At the firm level, energy intensity is defined as energy expenses as a percentage of the wage bill. 78 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Conceptual framework. At the firm level, the Hence, the energy intensity of a firm, which is the physical definition of energy intensity, that is, the ratio of the share of expenditures on energy and unit of energy consumed per unit output, cannot labor, depends on the elasticity of substitution, the be applied for lack of data: firm-level datasets relative efficiency of energy use, and the relative typically do not report units of energy price of the two factors. e literature on the consistently. Instead, they report expenditure on elasticity of energy demand typically finds that energy (or electricity). Moreover, because physical σ < 1 , which signifies a complementary units of output are not comparable across firms relationship between labor and energy (Martin from different sectors, the denominator in the 2010). Assuming that σ < 1, energy intensity falls expression for energy intensity would necessarily with higher relative energy efficiency AE /AL and consist of output in value terms and could reflect lower relative energy price PE /PL. differences in output prices across firms. Hence, given relative prices, lower energy intensity Studies of firm-level energy intensity typically use can be interpreted as higher relative energy expenditure-based definitions, such as energy efficiency. expenses in percent of costs (Shapiro and Walker 2018). e interpretation of this definition can be A change in the relative energy price can also lead clarified by illustrating the firm’s optimality to changes in energy intensity. For example, if conditions with a general production function. σ < 1, a decline in the energy price faced by the Consider a firm whose production involves using firm would lower energy intensity. e firm-level labor and electricity as imperfect substitutes. e regression analysis of energy intensity undertaken output is given by: in this chapter includes controls for relative prices of energy and labor in robustness checks when σ −1 σ −1 σ −1 data availability permits. e results reported are σ σ σ Y = A (( AL L ) + ( AE E ) ) robust to controlling for relative prices. Here, AE and AL represent the efficiency associated An alternative definition of energy intensity, with energy use and labor use, respectively, and σ energy expenditure in percent of total costs, is also is the elasticity of substitution between the inputs used in robustness checks, and the results reported (and the elasticity of input demand). A is a general are robust to this alternative definition. (that is, not input-specific) efficiency term. 2.1.3 Firm-level approaches: World Bank e total cost incurred by the firm is given by: Enterprise Surveys Sample. e World Bank Enterprise Surveys E xpenditure = PE E + PL L (WBES) are firm-level surveys covering a representative sample of the formal private sector. Here, PE and PL represent the price of a single e surveys are stratified by firm size, sector and unit of energy and labor, respectively. location. e repeated cross-sectional survey data Solving the firm’s cost minimization problem consist of two waves of the surveys: Wave 1 yields the following expression: (between 2006 and 2016) and Wave 2 (between 2017 and 2022). Overall, these repeated World PE E A P Bank Enterprise Surveys have gathered data on = ( E * L ) (σ −1) electricity expenses and the total wage bill among PL L AL PE 73,171 firms in 15 2-digit manufacturing and services subsectors in 43 EMDEs, including Note that Bangladesh (2006 and 2022), India (2014 and 2022), and Pakistan (2006 and 2022; annex table PE E Energy Expenditure 2.1.3.1). = = Energy Intensity PL L Wage Bill SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 79 ANNEX TABLE 2.1.3.1 World Bank Enterprise Surveys coverage of South Asia and Other EMDEs Country Year Number of observations India 2014 9,281 India 2022 9,376 Bangladesh 2007 1,504 Bangladesh 2022 998 Pakistan 2007 935 Pakistan 2022 1,300 Albania 2007 304 Albania 2019 377 Argentina 2006 1,063 Argentina 2017 991 Armenia 2009 374 Armenia 2020 546 Azerbaijan 2009 380 Azerbaijan 2019 225 Belarus 2008 273 Belarus 2018 600 Bolivia 2006 613 Bolivia 2017 364 Bosnia and Herzegovina 2009 361 Bosnia and Herzegovina 2019 362 Ecuador 2006 658 Ecuador 2017 361 Egypt, Arab Rep. 2013 2,897 Egypt, Arab Rep. 2020 3,075 Georgia 2008 373 Georgia 2019 581 Guatemala 2006 522 Guatemala 2017 345 Hungary 2009 291 Hungary 2019 805 Jordan 2013 573 Jordan 2019 601 Kazakhstan 2009 544 Kazakhstan 2019 1,446 Kenya 2007 657 Kenya 2018 1,001 Kyrgyz Republic 2009 235 Kyrgyz Republic 2019 360 Lebanon 2013 561 Lebanon 2019 532 Liberia 2009 150 Liberia 2017 151 Madagascar 2009 445 Madagascar 2022 402 Malaysia 2015 1,000 Malaysia 2019 1,221 80 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 2.1.3.1 continued Country Year Number of observations Moldova 2009 363 Moldova 2019 360 Montenegro 2009 116 Montenegro 2019 150 Morocco 2013 407 Morocco 2019 1,096 Mozambique 2007 479 Mozambique 2018 601 North Macedonia 2009 366 North Macedonia 2019 360 Paraguay 2006 613 Paraguay 2017 364 Romania 2009 541 Romania 2019 814 Russian Federation 2009 1,004 Russian Federation 2019 1,323 Serbia 2009 388 Serbia 2019 361 Sierra Leone 2009 150 Sierra Leone 2017 152 South Africa 2007 937 South Africa 2020 1,097 Tajikistan 2008 360 Tajikistan 2019 352 Timor-Leste 2009 150 Timor-Leste 2021 238 Tunisia 2013 592 Tunisia 2020 615 Turkey 2008 1,152 Turkey 2019 1,663 Ukraine 2008 851 Ukraine 2019 1,337 Uruguay 2006 621 Uruguay 2017 347 Uzbekistan 2008 366 Uzbekistan 2019 1,239 West Bank and Gaza 2013 434 West Bank and Gaza 2019 365 Zambia 2007 484 Zambia 2019 601 Source: World Bank Enterprise Surveys . SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 81 Note that the World Bank Enterprise Surveys only sum of the changes in the sector weights across the report total electricity expenses (including waves, each weighted by the mean of the sector expenses on fuel for portable electricity energy intensities in the waves. generators). ey do not collect data on the consumption of other fuels in firms. Hence, in (es2 + e1 2 1 s )( ws − ws ) what follows, energy intensity refers to electricity Between1,2 =  s 2 intensity. Within- and between-sector energy intensity By construction, the sum of these within and changes. In order to analyze the within- and between sector components equals the change in between-sector decomposition of energy intensity aggregate economy-wide energy intensity between growth, all the firms were first categorized into one the two waves. of 15 possible sectors. Countries with fewer than Changes in mean firm-level energy intensity. e 10 sectors in any of the two waves were dropped. within-sector change in mean firm-level energy To decompose aggregate energy intensity into that intensity is computed through firm-level attributed to changes in energy intensity within regressions in a dataset that stacks data from Wave sectors and that attributed to changes in the shares 1 and Wave 2 for each South Asian country. Note of sectors in the economy, energy intensity was that the World Bank Enterprise Surveys are not a first computed at sector level by dividing the panel at the firm level. For each country, an aggregate electricity cost of all firms in the sector ordinary least squares regression of log firm-level by their aggregate wage bill, across all countries energy intensity on year t with sector fixed effects and years. Survey sampling weights were used is estimated. when aggregating firm-level costs to their sector- level aggregates. Log ( Energy Intensity ) ist = α + β Yeart + λ s + ε ist t Let ws be the share of sector s in the total Here, i indexes firm, s indexes sector and t indexes t the year of the survey. λs are sector fixed effect. economy wage bill in Wave t and es be the Hence, β captures the annual percent change in energy intensity (sector energy to sector wage cost mean firm-level energy intensity within sectors. ratio) of sector s in Wave t. When estimating the regression for other EMDEs, e economy’s aggregate energy intensity in Wave data from other EMDEs are pooled together and t is country-sector specific fixed effects are employed. e regression results are shown in annex table t etotal =  wst est 2.1.3.2. s In a robustness analysis, energy intensity is e within-sector component of the change in measured as electricity expenditures as percent of energy intensity between Wave 1 and Wave 2 is total revenue. e regression results of this defined as alternative specification are qualitatively similar to those obtained in the main specification. ( ws2 + w1 2 1 s )( es − e s ) Within1,2 =  2.1.4 Firm-level approaches: India’s Annual s 2 Survey of Industries at is, it is the sum of the changes in energy Sample. India’s Annual Survey of Industries (ASI) intensity in each sector between Waves 1 and 2, includes 519,849 observations covering 156,927 each weighted by the mean of the sector weights unique firms in 22 manufacturing subsectors across the two waves. during 2001–18. e ASI is a plant (factory)-level dataset that is representative of the formal Similarly, the between sector component is the manufacturing sector at the state and sector level. e ASI panel is an unbalanced panel; that is, the 82 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 2.1.3.2 World Bank Enterprise Surveys: Change in mean firm level energy intensity within sectors Dependent variable: Log Energy Intensity Variable India Pakistan Bangladesh Other EMDEs Year -0.0856*** -0.00968** -0.0173*** -0.0192*** (0.00245) (0.00397) (0.00381) (0.00155) Constant -0.811*** -1.548*** -2.031* -2.962** (0.0517) (0.190) (1.201) (1.209) Observations 17,831 2,204 2,440 47,120 R-squared 0.112 0.061 0.095 0.165 Country specific sector fixed effects Yes Yes Yes Yes Source: World Bank staff estimates based on World Bank Enterprise Surveys. Note: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. OLS regressions include fixed effects for fifteen 2-digit manufacturing and services subsectors. These sector fixed effects are country specific in the regression shown in column (4), which is estimated on pooled survey data from all EMDEs excluding South Asian countries. number of firms in the panel in each year varies where XLowit and XHighit represent a set of firms’ due to entry, exit and the survey design. e ASI characteristics. ree characteristics were panel is a rotating panel, designed such that except considered: firm size, state-level manufacturing for firms with more than 100 workers, not every growth rate and sector-level initial energy firm is tracked every year. On average, each firm intensity. When considering firm-size, XLowit is a appears once every three years. e ASI covers all dummy variable for firms with 1–49 employees firms with more than 100 workers each year, while (“Small firms”) and XHighit is a dummy variable smaller firms are on average revisited once every for firms with more than 50 employees (“Large three years. firms”). Similarly, for state-level manufacturing growth rate, XLowit is a dummy variable for being Trends in energy intensity. Panel regressions are in states with below-median manufacturing value- estimated to identify: (i) the overall change in added growth during 1999–2014 (“Slow-growing energy intensity over time; and (ii) the firms’ states”), and XHighit is a dummy variable for characteristics associated with higher energy being in states with above-median manufacturing intensity change over time. To identify the change value-added growth (“Fast-growing states”). in energy intensity over time, a regression with the Finally, for sector-level initial energy intensity, following specification is estimated: XLowit is a dummy variable for being from a sector with low initial energy intensity in 2001 L og ( E nergy Intensity ) it = β t o Yeart + γ i + ε it and XHighit is a dummy variable for being from a sector with high initial energy intensity in 2001. where Yeart is a dummy variable for year t and γi Annex table 2.1.4.1 shows the results of these represents firm-level fixed effects to account for regressions. firm-specific trends. To identify firms’ characteristics associated with higher energy Annex table 2.1.4.2 shows the results of similar intensity change over time, regressions with the panel regressions as shown above, but with following specification are run: controls for relative price of electricity and labor included. e empirical specifications are as Log (energy intensity )it =  β lt Yeart * XLowit + follows:  β Year * XHigh t h t it + γ i + ε it SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 83 ANNEX TABLE 2.1.4.1 India: Within-firm trends in energy intensity Log Energy Log Energy Log Energy Log Energy Intensity Intensity Intensity Intensity Year = 2018 -0.708*** (0.00931) Slow growing state * Year = 2018 -0.598*** (0.0163) Fast growing state * Year = 2018 -0.771*** (0.0112) Firm size < 50 employees* Year = 2018 -0.585*** (0.0150) Firm size >= 50 employees* Year = 2018 -1.036*** (0.0145) Low energy intensity * Year = 2018 -0.732*** (0.0125) High energy intensity * Year = 2018 -0.685*** (0.0139) Observations 519,849 519,849 519,849 519,849 Firm fixed effects Yes Yes Yes Yes Source: World Bank staff estimates based on Annual Survey of Industries for India. Note: Standard errors are clustered at the firm-level and included in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The OLS regressions include dummies for all years between 2001 and 2018. Columns 2-4 also include a full set of year dummy interactions with specific firm characteristics. These coefficients are not displayed for concision. The omitted (baseline) year dummy is 2000. Log (energy intensity)it =  βltYeart * XLowit + Log ( Energy Expense ) it =  β ht Yeart + γ i + ε it Penergy Total Sales  β Year * XHigh t h t it + β p * Log ( Plabor ) + γ i + ε it Energy Expense Log ( )it =  β ltYeart * XLowit Total Sales Note that the estimated coefficient of 0.382 on relative price in the table above implies that a 1 +  β ht Yeart * XHighit + γ i + ε it percent increase in relative energy price increases Energy intensity and employment. To examine relative energy expenditure by 0.38 percent (that the relationship between energy intensity and is, less than 1 percent). is implies a price employment in the cross-section of firms, the elasticity of energy consumption of -0.62. following OLS regression is estimated on 2018 Annex table 2.1.4.3 shows the results of similar ASI cross-sectional data: regressions where energy intensity is instead measured as the share of energy expense in total Log ( employment ) i = Log ( Energy Intensity ) i sales. e empirical specifications are as follows: + λi + ε i Here, λi are sector dummies. 84 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 2.1.4.2 India: Within-firm trends in energy intensity, controlling for relative prices Log Energy Log Energy Log Energy Log Energy Intensity Intensity Intensity Intensity Year = 2018 -0.368*** (0.0104) Slow growing state * Year = 2018 -0.283*** (0.0167) Fast growing state * Year = 2018 -0.416*** (0.0120) Firm size < 50 employees* Year = 2018 -0.243*** (0.0151) Firm size >= 50 employees * Year = 2018 -0.744*** (0.0146) Low energy intensity * Year = 2018 -0.383*** (0.0130) High energy intensity * Year = 2018 -0.353*** (0.0146) Log (energy price/wage) 0.382*** 0.381*** 0.397*** 0.382*** (0.00557) (0.00558) (0.00556) (0.00557) Observations 502,023 502,023 502,023 502,023 Firm fixed effects Yes Yes Yes Yes Source: World Bank staff estimates based on Annual Survey of Industries for India. Note: Standard errors are clustered at the firm-level and included in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The OLS regressions include dummies for all years between 2001 and 2018. Columns 2-4 also include a full set of year dummy interactions with specific firm characteristics. These coefficients are not displayed for concision. The omitted (baseline) year dummy is 2000. To examine the relationship between changes in apparel, food processing, pharmaceuticals, leather energy intensity and employment within firms goods, bricks, iron and steel, cement, a residual over time, the following firm fixed effects OLS category comprising all other manufacturing regression is estimated on the ASI panel: subsectors, wholesale and retail, land transport, and health. e surveys are stratified by location Log (employment )it = Log ( Energy Intensity ) i (cities in the case of Bangladesh), firm size and +γ i + ε i subsector. Estimates of average rates of technology adoption use survey sampling weights. e results are shown in annex table 2.1.4.4. Regression specification. e following OLS regression is estimated to examine the association 2.1.5 World Bank Firm-level Adoption of between energy-efficient technology adoption and Technology Surveys firms’ characteristics. Sample. In 2022, the World Bank’s Firm Adoption of Technology (FAT) Surveys Wave 2 surveyed Energy Technology Indexi = α +  β1s ⋅ Sizeis + 10,090 firms in seven EMDEs, including 1,936 β s 2 ⋅ Ageia +  β 3s ⋅ Managementtm + firms in Bangladesh and 1.455 firms in India, about their technology use, as well as other firm β 3 ⋅ Multinational + β 5 ⋅ Finance + characteristics. e FAT surveys in South Asia are β 6 ⋅ Subsidy + β 7 ⋅ GBF + µi + π i + ε i representative of formal sector firms in select manufacturing and services subsectors including Here, Energy Technology Indexi is an index SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 85 ANNEX TABLE 2.1.4.3 India: Within-firm trends in energy intensity, measuring energy intensity in percent of total sales Log Energy Log Energy Log Energy Log Energy VARIABLES Expenditure/ Sales Expenditure/ Sales Expenditure/ Sales Expenditure/ Sales Year = 2018 -0.313*** (0.0110) Slow growing state * Year = 2018 -0.277*** (0.0184) Fast growing state * Year = 2018 -0.331*** (0.0138) Firm size < 50 employees * Year = 2018 -0.319*** (0.0174) Firm size >= 50 employees* Year = 2018 -0.449*** (0.0171) Low energy intensity * Year = 2018 -0.327*** (0.0155) High energy intensity * Year = 2018 -0.298*** (0.0158) Observations 520,336 520,336 520,336 520,336 Firm fixed effects Yes Yes Yes Yes Source: World Bank staff estimates based on Annual Survey of Industries for India. Note: Standard errors are clustered at the firm-level and included in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The OLS regressions include dummies for all years between 2001 and 2018. Columns 2-4 also include a full set of year dummy interactions with specific firm characteristics. These coefficients are not displayed for concision. The omitted (baseline) year dummy is 2000. ANNEX TABLE 2.1.4.4 India: Employment and energy intensity (1) (2) (3) Log Employment Log Employment Log Employment 2018 2018 2001-18 Log energy intensity -0.169*** -0.164*** -0.215*** (0.00594) (0.00676) (0.00324) Observations 36,006 36,006 519,849 Sector fixed effects No Yes No Firm fixed effects No No Yes Source: World Bank staff estimates based on Annual Survey of Industries for India. Note: The first two columns present results for cross-sectional OLS regressions for firms in 2018 while the last column presents the results from a firm-level OLS panel regression of firms from 2001-18. Robust standard errors are included in parentheses for the first two columns while standard errors clustered at the firm level are included within parentheses in the last column. *** p<0.01, ** p<0.05, * p<0.1. 86 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 2.1.5.1 Average energy efficient technology usage rates: Country details (Percent of firms) Energy Star EE lighting VAV HVAC system Programmable thermostats IoT enabled systems Bangladesh 59.5 76.3 6.2 1.4 1.1 Brazil 43.4 90.7 69.3 35.1 15.8 Cambodia 2.4 31.7 0.5 1.1 1 Chile 39.7 50.3 23.4 14.1 4.9 Ethiopia 14.3 20 1 0 0 Georgia 10.2 45.8 37.1 10.9 4 India 70 90.9 39 6.7 5 Source: World Bank Firm Level Adoption of Technology Surveys. Note: Energy Star refers to a U.S. government-backed program for measuring energy efficiency. "VAV HVAC" refers to variable air volume heating, ventilation, and air conditioning systems. "IoT" refers to Internet of Things-enabled systems to control premises temperature, lighting system, and/or refrigeration units. measuring the level of energy- efficient technology P (Yi = 1 | o utage i , X i ) = Φ (α + β . outa gei + γ . X i ) in firm i, constructed as the sum of dummies for each energy-efficient technology listed in annex Here, Yi is a dummy for whether the firm uses a s table 2.1.5.1. Sizei includes the set of firm size generator, outage is a dummy indicating whether a groups (20–99 and 100+ employees), Agei is the the firm experienced a power outage in the past set of firm age groups (6–10, 11–15, and 16+), month. Xi is a set of firm attributes: sector, size M anagem ent i represents the set of management and firm age. e regression is estimated on the related variables (Managers with BA or above, India FAT sample and on a pooled sample Managers with experience in multinational firms, containing all seven EMDE countries. It cannot Having formal incentive for workers, following 1– be estimated separately for Bangladesh owing to 2 KPI, and following 3+ KPI), Multinational is an the absence of variation in power outages in that indicator for firms having business with survey (that is, all firms in Bangladesh sample multinational firms. Finance is an indicator for report facing a power outage in the past month). firms that needed to borrow money but could not Annex table 2.1.5.3 reports the marginal estimated borrow, Subsidy is an indicator for firms that probabilities. benefited from government program/subsidy, and is an index for the adoption of general business 2.1.6 Bangladesh Randomized Control Trial function technology. e specifications also Study sample. A randomized control trial (RCT) include industry (μi) and region (πi) fixed effects. is being conducted among 504 small and medium e estimation results are presented in annex table firms in leather goods and footwear manufacturing 2.1.5.2. e main specification, which is the in Bangladesh since 2022 (Chaurey et al. 2023). foundation of the discussion in the chapter, is in e RCT measures the impact on informational column 5. It shows that there is a statistically interventions on the adoption of servo motors, a significant association between energy-efficient new energy-efficient alternative to the traditional technology adoption and firm size, manager clutch motors used in sewing machines in leather education and KPI monitoring. Having business goods, footwear and garment firms. e baseline links with multinational firms and having survey was conducted in 2022, and a midline benefited from government programs are not survey conducted in March–May 2023. A final, significantly associated with technology adoption. endline survey will be conducted to measure the Use of generators. e association between power effects nine months after the intervention, from outages and the use of generators is estimated with September-November 2023. e report is based the following probit regression: on the interim analysis of the baseline and midline SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 87 ANNEX TABLE 2.1.5.2 Correlates of energy efficient technology adoption index Technology Technology Technology Technology Technology adoption index adoption index adoption index adoption index adoption index Employment 20-99 0.127** 0.133** 0.163** 0.166** 0.218*** (0.0643) (0.0637) (0.0688) (0.0686) (0.0700) Employment 100+ 0.0159 0.0636 0.0795 0.0994 0.387*** (0.0667) (0.0659) (0.0801) (0.0789) (0.0966) Age 6-10 0.101 0.0191 0.00713 -0.0472 (0.113) (0.108) (0.107) (0.100) Age 11-15 0.167 0.0175 -0.000759 -0.0353 (0.109) (0.106) (0.105) (0.0990) Age 16+ 0.403*** 0.131 0.109 0.0463 (0.105) (0.102) (0.102) (0.0959) Manager with BA degree or above 0.358*** 0.360*** 0.281*** (0.0628) (0.0630) (0.0665) Business with multinational firm 0.0133 -0.00476 -0.0211 (0.0882) (0.0904) (0.0884) CEO/manager has experience in -0.0221 -0.0158 -0.00179 multinational firm (0.0780) (0.0793) (0.0765) Firm needed to borrow money but 0.0885 0.0876 0.0521 could not (0.0659) (0.0665) (0.0670) Formal incentives for workers 0.280*** 0.263*** 0.0193 (0.0786) (0.0792) (0.0836) Uses 1-2 key performance indicators 0.443*** 0.419*** 0.174** (0.0641) (0.0633) (0.0730) Uses 3+ key performance indicators 0.434*** 0.412*** 0.134 (0.0852) (0.0847) (0.0908) Benefited from government program -0.109 -0.0943 -0.0712 or subsidy (0.0841) (0.0833) (0.0836) Exporter -0.0819 (0.0735) Sector, region fixed effects No No No No Yes Dependent variable mean 1.777 Observations 3,391 3,384 2,436 2,434 2,436 R-squared 0.003 0.023 0.157 0.161 0.214 Source: World Bank staff estimates based on World Bank Firm Level Adoption of Technology (FAT) Surveys Wave 2 for Bangladesh and India. Note: Robust standard errors are included in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Charts depict coefficient estimates with 95% confidence intervals from OLS regressions of Energy Efficient Technology Index on firm attributes including sector and region dummies. The index (range 0-6) is the sum of dummies for whether the firm uses energy efficient lighting, Energy Star rated equipment, Variable Air Volume HVAC systems, programmable thermostats, and Internet of Things enabled systems to control temperature, lighting, or refrigeration. The sample for the regression is the FAT Survey Wave 2 pooled dataset for Bangladesh and India. N = 2436. “Size 20-99” and “Size 100+” are dummies for firm employment size; the omitted size dummy is “Employment below 20 workers.” “Manager with BA degree or above” is a dummy for manager education. “Formal incentives for workers” is a dummy indicating the use of formal incentive schemes for workers by the firm. “Uses 1-2 key performance indicators” and “Uses 3+ key performance indicators” are dummies indicating the number of Key Perfor- mance Indicator (KPIs) monitored by the firm; the omitted KPI dummy is “Firm does not monitor any KPI”. 88 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 2.1.5.3 Power outages and generator use Generator Use Generator Use Power outage (=1) 0.161*** 0.155*** (0.0368) (0.0238) Size and age controls Y Y Sector dummies Y Y Country fixed effects Y Sample India All EMDEs Observations 1444 9243 Source: World Bank staff estimates based on World Bank Firm Level Adoption of Technology Surveys. Note: Robust standard errors are included in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The table shows the estimated marginal effect of power outages on the probability of owning/ sharing a generator, resulting from a firm-level Probit regression of generator use on a dummy for whether firm faced a power outage in the past month, controlling for sector, size and firm age. Column (1) is estimated on India survey data. Column (2) is estimated on Bangladesh, Brazil, Cambodia, Chile, Ethiopia, Georgia, and India, and includes country fixed effects in addition to the other controls. surveys discussed in Chaurey et al. (2023). e • Control (labelled C): right-skewed distribu- details of the RCT intervention are discussed tion, no video below. • Video only (labelled T1a): right-skewed e intervention. e RCT has provided distribution, plus detailed information about information about the servo motors in varying the servo motor using a video on a tablet intensities to the managers of the study firms. It device has elicited managers’ beliefs about cost savings from the servo motors through a belief elicitation • Video + meter (labelled T1b): right-skewed procedure, and elicited managers’ willingness to distribution plus video from T1a plus smart pay for the servo motors using a Becker-DeGroot- electricity meter on one machine with a clutch Marschak (BDM) procedure. In the BDM motor procedure, the firm’s manager is asked to state the • Motor + video + motor + 2 meters (labelled maximum price at which they would be willing to T2): left-skewed distribution plus video from purchase the servo motor. ey have the T1a plus two smart electricity meters, one on opportunity to purchase the motor at a randomly a clutch motor machine and one on the drawn price if that price is lower than their stated machine with the new servo motor. maximum willingness to pay for the motor. ere are three treatment arms and one control arm. In Approach. A number of methodologies was all four arms, the RCT is tracking the evolution of applied. Heterogeneity in firms’ willingness to pay managers’ beliefs about cost savings from the servo at baseline is examined in a linear regression of motors and their willingness to pay for them. e firms’ reported willingness to pay for servo motors arms differ in the extent of information provided on firm characteristics. e change in firms’ about the servo motors and in the distribution willingness to pay after the intervention is derived from which prices are drawn in the BDM from a linear regression of the willingness to pay procedure. All three treatment arms are shown a on a dummy variable for the midline variable and video explaining the energy saving benefits of firm dummies. Spillovers are derived from a probit servo motors. For the control and first two regression of technology adoption among control treatment arms, the RCT draw prices from a right group firms on the number of treated firms within -skewed distribution; almost no firms purchase the a limited geographic distance (Chaurey et al. servo motor at this price. For the third treatment 2023). arm, the RCT draws prices from a left-skewed distribution; almost all firms purchase the servo motor at this (very low) price. e arms are the following: SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 89 Environmental Economics and Management 88 References (March): 134–58. Abeberese, A. B. 2017. “Electricity Cost and Firm Bartram, S. M., K. Hou, and S. Kim. 2022. “Real Performance: Evidence from India.” The Review of Effects of Climate Policy: Financial Constraints and Economics and Statistics 99 (5): 839–52. Spillovers.” Journal of Financial Economics 143 (2): Accetturo, A., G. Barboni, M. Cascarano, E. Garcia- 668–96. Appendini, and M. Tomasi. 2022. “Credit Supply Behrer, A. P., R. Choudhary, and D. Sharma. 2023. and Green Investments.” SSRN Scholarly Paper. “Air Pollution Reduces Economic Activity: Evidence Available at SSRN: https://ssrn.com/ from India.” Policy Research Working Paper 10515. abstract=4093925. World Bank, Washington, DC. Adhvaryu, A., N. Kala, and A. Nyshadham. 2020. Biscione, A., D. Boccanfuso, A. De Felice, and F. “The Light and the Heat: Productivity Co-Benefits Porcelli. 2023. “Barriers to Firms’ Energy Efficiency of Energy-Saving Technology.” The Review of in Transition Countries.” Applied Economics 55 (36): Economics and Statistics 102 (4): 779–92. 4258–72. Agnolucci, P., C. Fischer, D. Heine, M. Montes De Blackman, A., and G. J. Bannister. 1998. Oca Leon, J. D. C. Pryor, and S. Hallegatte. 2023. “Community Pressure and Clean Technology in the “Measuring Total Carbon Pricing.” Policy Research Informal Sector: An Econometric Analysis of the Working Paper 10486, World Bank, Washington Adoption of Propane by Traditional Mexican DC. Brickmakers.” Journal of Environmental Economics Alfaro-Serrano, D., T. Balantrapu, R. Chaurey, A. and Management 35 (1): 1–21. Goicoechea, and E. Verhoogen. 2021. “Interventions Blass, V., C. J. Corbett, M. A. Delmas, and S. to Promote Technology Adoption in Firms: A Muthulingam. 2014. “Top Management and the Systematic Review.” Campbell Systematic Reviews 17 Adoption of Energy Efficiency Practices: Evidence (4): e1181. from Small and Medium-Sized Manufacturing Firms Allcott, H., A. Collard-Wexler, and S. D. in the US.” Energy 65 (February): 560–71. O’Connell. 2016. “How Do Electricity Shortages Bloom, N., B. Eifert, A. Mahajan, D. McKenzie, and Affect Industry? Evidence from India.” American J. Roberts. 2013. “Does Management Matter? Economic Review 106 (3): 587–624. Evidence from India.” The Quarterly Journal of Allcott, H., and M. Greenstone. 2012. “Is There an Economics 128 (1): 1–51. Energy Efficiency Gap?” Journal of Economic Bloom, N., C. Genakos, R. Martin, and R. Sadun. Perspectives 26 (1): 3–28. 2010. “Modern Management: Good for the Andersen, D. C. 2017. “Do Credit Constraints Environment or Just Hot Air?” The Economic Journal Favor Dirty Production? Theory and Plant-Level 120 (544): 551–72. Evidence.” Journal of Environmental Economics and Blundell, W. 2020. “When Threats Become Management 84 (July): 189–208. Credible: A Natural Experiment of Environmental Anderson, S. T., and R. G. Newell. 2004. Enforcement from Florida.” Journal of Environmental “Information Programs for Technology Adoption: Economics and Management 101 (May): 102288. The Case of Energy-Efficiency Audits.” Resource and Bodas-Freitas, I.-M., and N. Corrocher. 2019. “The Energy Economics 26 (1): 27–50. Use of External Support and the Benefits of the Ang, B. W., and F. Q. Zhang. 2000. “A Survey of Adoption of Resource Efficiency Practices: An Index Decomposition Analysis in Energy and Empirical Analysis of European SMEs.” Energy Policy Environmental Studies.” Energy 25 (12): 1149–76. 132 (September): 75–82. Barrows, G., and H. Ollivier. 2018. “Cleaner Firms Boyd, G. A., and E. M. Curtis. 2014. “Evidence of or Cleaner Products? How Product Mix Shapes an ‘Energy-Management Gap’ in U.S. Emission Intensity from Manufacturing.” Journal of Manufacturing: Spillovers from Firm Management Practices to Energy Efficiency.” Journal of 90 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Environmental Economics and Management 68 (3): Colmer, J., R. Martin, M. Muûls, and U. J. Wagner. 463–79. 2023. “Does Pricing Carbon Mitigate Climate Change? Firm-Level Evidence from the European Breitenfellner, A., M. Lahnsteiner, T. Reininger, and Union Emissions Trading Scheme.” Discussion J. Schriefl. 2021. “Green Transition: What Have Paper 232. Available at SSRN: https://ssrn.com/ CESEE EU Member States Achieved So Far?” Focus abstract=3725482. on European Economic Integration Q4/21: 61–76. Crippa, M., D. Guizzardi, M. Banja, E. Solazzo, M. Brown, J. R., G. Martinsson, and C. Thomann. Muntean, E. Schaaf, F. Pagani, et al. 2022. CO2 2022. “Can Environmental Policy Encourage Emissions of All World Countries: JRC/IEA/PBL 2022 Technical Change? Emissions Taxes and R&D Report. Luxembourg: Publications Office of the Investment in Polluting Firms.” The Review of European Union. Financial Studies 35 (10): 4518–60. Curtis, E. M., and J. M. Lee. 2019. “When Do Bu, C., K. Zhang, D. Shi, and S. Wang. 2022. Environmental Regulations Backfire? Onsite “Does Environmental Information Disclosure Industrial Electricity Generation, Energy Efficiency Improve Energy Efficiency?” Energy Policy 164 and Policy Instruments.” Journal of Environmental (May): 112919. Economics and Management 96 (July): 174–94. Bussolo, M., and S. Sharma, eds. 2022. Hidden Damania, R., E. Balseca, C. de Fontaubert, J. Gill, Potential: Rethinking Informality in South Asia. K. Kim, J. Rentschler, J. Russ, and E. Zaveri. 2023. Washington, DC: World Bank. Detox Development: Repurposing Environmentally Calel, R., and A. Dechezleprêtre. 2016. Harmful Subsidies. Washington, DC: World Bank. “Environmental Policy and Directed Technological Dechezleprêtre, A., C. Gennaioli, R. Martin, M. Change: Evidence from the European Carbon Muûls, and T. Stoerk. 2022. “Searching for Carbon Market.” Review of Economics and Statistics 98 (1): Leaks in Multinational Companies.” Journal of 173–91. Environmental Economics and Management 112 Caporale, G. M., C. Donati, and N. Spagnolo. (March): 102601. 2023. “Small and Medium Sized European Firms De Groot, H. L. F., E. T. Verhoef, and P. Nijkamp. and Energy Saving Measures: The Role of 2001. “Energy Saving by Firms: Decision-Making, Financing.” Energy Policy 179 (August): 113613. Barriers and Policies.” Energy Economics 23 (6): 717– Chaurey, R., G. Nayyar, S. Sharma, and E. 40. Verhoogen. 2023. “How Do Managers’ Beliefs De Haas, R., and A. Popov. 2023. “Finance and about New Technologies Evolve? Informational Green Growth.” The Economic Journal 133 (650): Interventions and the Adoption of Energy-Efficient 637–68. Stitching Motors in Bangladesh.” Unpublished manuscript. De Haas, R., R. Martin, M. Muûls, and H. Shweiger. 2023. “Managerial and Financial Barriers Chen, Q., Z. Chen, Z. Liu, J. C. Suárez Serrato, and during the Green Transition.” CentER Discussion D. Xu. 2021. “Regulating Conglomerates in China: Paper Nr. 2021-008, European Banking Center Nr. Evidence from an Energy Conservation Program.” 2021-002, Available at SSRN: https://ssrn.com/ NBER Working Paper 29066, National Bureau of abstract=3796924. Economic Research, Cambridge, MA. Duflo, E., M. Greenstone, R. Pande, and N. Ryan. Cirera, X., D. Comin, and M. Cruz. 2022. Bridging 2013. “Truth-Telling by Third-Party Auditors and the Technological Divide: Technology Adoption by the Response of Polluting Firms: Experimental Firms in Developing Countries. Washington, DC: Evidence from India.” The Quarterly Journal of World Bank. Economics 128 (4): 1499–1545. Cirera, X., and W. F. Maloney. 2017. The Duflo, E., M. Greenstone, R. Pande, and N. Ryan. Innovation Paradox: Developing-Country Capabilities 2018. “The Value of Regulatory Discretion: and the Unrealized Promise of Technological Catch- Estimates from Environmental Inspections in India.” Up. Washington, DC: World Bank. Econometrica 86 (6): 2123–60. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 91 Eskeland, G. S., and A. E. Harrison. 2003. “Moving Greenstone, M., R. Pande, N. Ryan, and A. to Greener Pastures? Multinationals and the Sudarshan. 2023. “Can Pollution Markets Work in Pollution Haven Hypothesis.” Journal of Developing Countries? Experimental Evidence from Development Economics 70 (1): 1–23. India.” Warwick Economics Research Paper 1453, University of Warwick, Coventry, U.K. Fleiter, T., J. Schleich, and P. Ravivanpong. 2012. “Adoption of Energy-Efficiency Measures in Gutiérrez, E., and K. Teshima. 2018. “Abatement SMEs—An Empirical Analysis Based on Energy Expenditures, Technology Choice, and Audit Data from Germany.” Energy Policy 51 Environmental Performance: Evidence from Firm (December): 863–75. Responses to Import Competition in Mexico.” Journal of Development Economics 133 (July): 264– Foster, A. D., and M. R. Rosenzweig. 2010. 74. “Microeconomics of Technology Adoption.” Annual Review of Economics 2 (1): 395–424. Hall, B. 2005. “Innovation and Diffusion.” In The Oxford Handbook of Innovation. Oxford, U.K.: Gerarden, T. D., R. G. Newell, and R. N. Stavins. Oxford University Press. 2017. “Assessing the Energy-Efficiency Gap.” Journal of Economic Literature 55 (4): 1486–1525. Hanna, R., E. Duflo, and M. Greenstone. 2016. “Up in Smoke: The Influence of Household Gillingham, K., A. Keyes, and K. Palmer. 2018. Behavior on the Long-Run Impact of Improved “Advances in Evaluating Energy Efficiency Policies Cooking Stoves.” American Economic Journal: and Programs.” Annual Review of Resource Economics Economic Policy 8 (1): 80–114. 10 (1): 511–32. Harrison, A., B. Hyman, L. Martin, and S. Nataraj. Gillingham, K., D. Rapson, and G. Wagner. 2016. 2015. “When Do Firms Go Green? Comparing Price “The Rebound Effect and Energy Efficiency Policy.” Incentives with Command and Control Regulations Review of Environmental Economics and Policy 10 (1): in India.” NBER Working Paper 21763, National 68–88. Bureau of Economic Research, Cambridge, MA. Government of India, 2019. “India Cooling Action He, G., S. Wang, and B. Zhang. 2020. “Watering Plan.” Government of India, New Delhi. Down Environmental Regulation in China.” The Government of India. 2023. “Impact of Energy Quarterly Journal of Economics 135 (4): 2135–85. Efficiency Measures for the Year 2021-22.” Hochman, G., and G. R. Timilsina. 2017. “Energy Government of India, New Delhi. Efficiency Barriers in Commercial and Industrial Government of the People’s Republic of Bangladesh. Firms in Ukraine: An Empirical Analysis.” Energy 2015. “Energy Efficiency and Conservation Master Economics 63 (March): 22–30. Plan up to 2030.” Government of the People’s Hottenrott, H., S. Rexhäuser, and R. Veugelers. Republic of Bangladesh, Dhaka, Bangladesh. 2016. “Organisational Change and the Productivity Greenstone, M., and R. Hanna. 2014. Effects of Green Technology Adoption.” Resource “Environmental Regulations, Air and Water and Energy Economics 43 (February): 172–94. Pollution, and Infant Mortality in India.” American Jack, B. K. 2013. “Market Inefficiencies and the Economic Review 104 (10): 3038–72. Adoption of Agricultural Technologies in Greenstone, M., C. Hasenkopf, and K. Lee. 2022. Developing Countries.” Agricultural Technology “Air Quality Life Index: Annual Update: June Adoption Initiative. J-PAL (MIT) and CEGA (UC 2022.” Energy Policy Institute, The University of Berkeley). Chicago. Juhasz, R., N. Lane, and D. Rodrik. 2023. “The Greenstone, M., G. He, S. Li, and E.Y. Zou. 2021. New Economics of Industrial Policy.” NBER “China’s War on Pollution: Evidence from the First Working Paper 31538, National Bureau of 5 Years.” Review of Environmental Economics and Economic Research, Cambridge, MA. Policy 15 (2): 281–99. Kalantzis, F., and D. Revoltella. 2019. “Do Energy Audits Help SMEs to Realize Energy-Efficiency 92 CHAPTER 2 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Opportunities?” Energy Economics 83 (September): Piza, C., T. A. Cravo, L. Taylor, L. Gonzalez, I. 229–39. Musse, I. Furtado, A. C. Sierra, and S. Abdelnour. 2016. “The Impact of Business Support Services for Kellenberg, D. K. 2009. “An Empirical Investigation Small and Medium Enterprises on Firm Performance of the Pollution Haven Effect with Strategic in Low- and Middle-Income Countries: A Systematic Environment and Trade Policy.” Journal of Review.” Campbell Systematic Reviews 12 (1): 1–167. International Economics 78 (2): 242–55. Rexer, J. 2022. “The Local Advantage: Corruption, Kube, R., K. Von Graevenitz, A. Löschel, and P. Organized Crime, and Indigenization in the Massier. 2019. “Do Voluntary Environmental Nigerian Oil Sector.” Available at SSRN: https:// Programs Reduce Emissions? EMAS in the German ssrn.com/abstract=3894597. Manufacturing Sector.” Energy Economics 84 (October): 104558. Ryan, N. 2018. “Energy Productivity and Energy Demand: Experimental Evidence from Indian Kumar, A., A. Mittal, and A. K. Pradhan. 2023. Manufacturing Plants.” NBER Working Paper “Magnitude and Determinants of Energy Intensity: 24619, National Bureau of Economic Research, Evidence from Indian Firms.” Environmental Science Cambridge, MA. and Pollution Research 30 (2): 3270–81. Sachs, J.D., and W. T. Woo. 1994. “Structural Levinson, A. 2021. “Energy Intensity: Factors in the Economic Reforms of China, Eastern Deindustrialization, Composition, Prices, and Europe, and the Former Soviet Union.” Economic Policies in the United States.” Resource and Energy Policy 9 (18): 101–45. Economics 65 (August): 101243. Sachs, J. D. 1996. “The Transition at Mid Decade.” Levinson, A., and M. S. Taylor. 2008. “Unmasking American Economic Review 86 (2): 128–33. the Pollution Haven Effect.” International Economic Review 49 (1): 223–54. Sadayuki, T., and T. H. Arimura. 2021. “Do Regional Emission Trading Schemes Lead to Carbon Lu, X., S. Zhang, J. Xing, Y. Wang, W. Chen, D. Leakage within Firms? Evidence from Japan.” Energy Ding, Y. Wu, S. Wang, L. Duan, and J. Hao. 2020. Economics 104 (December): 105664. “Progress of Air Pollution Control in China and Its Challenges and Opportunities in the Ecological Sexton, S. 2015. “Automatic Bill Payment and Civilization Era.” Engineering 6 (12): 1423–31. Salience Effects: Evidence from Electricity Consumption.” Review of Economics and Statistics 97 Martin, L. 2011. “Energy Efficiency Gains from (2): 229–41. Trade: Greenhouse Gas Emissions and India’s Manufacturing Sector.” Unpublished manuscript. Shapiro, J. S., and R. Walker. 2018. “Why Is Pollution from US Manufacturing Declining? The Martin, R. 2010. “Why Is the USA so Energy Roles of Environmental Regulation, Productivity, Intensive? Evidence for US Multinationals in the and Trade.” American Economic Review 108 (12): UK.” GRI Working Papers 15, Grantham Research 3814–54. Institute on Climate Change and the Environment, London School of Economics and Political Science, Singh, S., and K. Bajaj. 2018. “Market-making for London. Low-carbon Energy Technologies: The UJALA Scheme in India.” Bonn: Deutsche Gesellschaft für Martin, R., M. Muûls, L.B. de Preux, and U. J. Internationale Zusammenarbeit (GIZ) GmbH. Wagner. 2014. “Industry Compensation under Relocation Risk: A Firm-Level Analysis of the EU Suri, T. 2011. “Selection and Comparative Emissions Trading Scheme.” American Economic Advantage in Technology Adoption.” Econometrica Review 104 (8): 2482–2508. 79 (1): 159–209. Martin, R., M. Muûls, L.B. de Preux, and U. J. Timilsina, G. R. 2022. “Carbon Taxes.” Journal of Wagner. 2012. “Anatomy of a Paradox: Economic Literature 60 (4): 1456–1502. Management Practices, Organizational Structure and Energy Efficiency.” Journal of Environmental Tong, Z., and K. M. Zhang. 2015. “The Near- Economics and Management 63 (2): 208–23. Source Impacts of Diesel Backup Generators in SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 2 93 Urban Environments.” Atmospheric Environment 109 World Bank. 2023b. South Asia Economic Focus: (May): 262–71. Expanding Opportunities: Toward Inclusive Growth. Spring. Washington, DC: World Bank. Verhoogen, E. 2023. “Firm-Level Upgrading in Developing Countries.” Journal of Economic World Bank. 2023c. Bangladesh Development Literature 61 (4). Update: Trade Reform: An Urgent Agenda. April Wagner, U. J., and C. D. Timmins. 2009. 2023. World Bank, Washington, DC. “Agglomeration Effects in Foreign Direct Investment World Bank. 2023d. Project Information Document: and the Pollution Haven Hypothesis.” Sustainable Microenterprise and Resilient Environmental and Resource Economics 43 (2): 231– Transformation (SMART). World Bank, 56. Washington, DC. Williams, H., and K. Bryan. 2021. “Innovation: Market Failures and Public Policies.” In Handbook of Xu, Q., and T. Kim. 2022. “Financial Constraints Industrial Organization 5, edited by K. Ho, A. and Corporate Environmental Policies.” The Review Hortaçsu, and A. Lizzeri, 281-388. Amsterdam: of Financial Studies 35 (2): 576–635. North-Holland. Yajima, N., and T. H. Arimura. 2022. “Promoting Wing, I.S. 2008. “Explaining the Declining Energy Energy Efficiency in Japanese Manufacturing Intensity of the U.S. Economy.” Resource and Energy Industry through Energy Audits: Role of Economics 30 (1): 21–49. Information Provision, Disclosure, Target Setting, Inspection, Reward, and Organizational Structure.” World Bank. 2020. Global Economic Prospects. June. Energy Economics 114 (October): 106253. Washington, DC: World Bank. World Bank, 2021. Fiscal Policy for Sustainable Yong, S. K., U. J. Wagner, P. Shen, L. De Preux, M. Development: Nepal Public Expenditure Review. Muûls, R. Martin, and J. Cao. 2021. “Management Washington, DC: World Bank. Practices and Climate Policy in China.” Available at SSRN: https://ssrn.com/abstract=3797466. World Bank. 2022a. Consolidating the Recovery: Seizing Green Growth Opportunities. Semiannual Zeng, Y., Y. Cao, X. Qiao, B.C. Seyler, and Y. Tang. Report for Latin America and the Caribbean. 2019. “Air Pollution Reduction in China: Recent Washington, DC: World Bank. Success but Great Challenge for the Future.” Science of the Total Environment 663: 329–37. World Bank. 2022b. Climate Investment Opportunities in India’s Cooling Sector. Washington, Zhang, D., J. Li, and Q. Ji. 2020. “Does Better DC: World Bank. Access to Credit Help Reduce Energy Intensity in World Bank. 2023a. Striving for Clean Air: Air China? Evidence from Manufacturing Firms.” Energy Pollution and Public Health in South Asia. Policy 145 (October): 111710. Washington, DC: World Bank. CHAPTER 3 Stranded jobs? The energy transition in South Asia’s labor markets SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 97 Chapter 3. Stranded jobs? The energy transition in South Asia’s labor markets The transition away from fossil fuels in South Asia will have significant labor market impacts, which could leave many workers stranded in lower-wage jobs in declining industries. In all South Asian countries except India, pollution-intensive jobs outnumber green jobs and account for 6–11 percent of all jobs; only in India do green jobs outnumber—and then only slightly—pollution-intensive jobs, which account for 9 percent of all jobs. Pollution-intensive jobs are concentrated among lower-skilled and informal workers, whereas green jobs tend to be held by higher-skilled, better-paid, and formal-sector workers. Experience from past major economic transformations, especially in resource sectors, suggests that the transition from fossil fuels will have large effects on the structure of employment and earnings, with lasting losses for some workers, and will cause significant internal worker migration. A wide range of policies will be needed to facilitate the necessary adjustment in labor markets while protecting vulnerable workers. These include: the provision of better access to high-quality education and training, finance, and markets; measures to facilitate labor mobility; and strengthening social safety nets. Introduction capacity is being enlarged, allowing both countries to export electricity to India (Asian Development The widespread adoption of more energy-efficient Bank 2013; IEA 2021a; IRENA 2022a, 2022b; and green technologies, and cutbacks in highly World Bank Group 2022a). polluting activities, are likely to have significant The region has strong potential for solar and wind effects in South Asia’s labor markets. Workers power generation (figure 3.2). Nearly 60 percent could be stranded in lower-wage jobs in declining of its area has solar exposure of at least 3.5 kWh/ industries, or when assets like land and capital are kWp with low seasonality, and about half of its no longer productive and labor is not mobile. This area has wind speeds of greater than 5 meters per chapter examines the important role of labor second (World Bank et al. 2023a). Solar-powered market policies in supporting South Asia’s energy electricity generation in the region has grown transition, while also promoting the region’s rapidly: between 2015 and 2022, it grew by 20 broader development objectives. percent in Bangladesh, 40 percent in the South Asia is a major source of GHG emissions, Maldives, and almost 70 percent in Nepal and Sri accounting for almost 10 percent of global Lanka (figure 3.3; Ember 2022). emissions in 2021—more than twice its 4 percent Policy makers in South Asia have started to gear share of global GDP (figure 3.1). As part of the policies explicitly to supporting the green 2015 Paris Agreement, several South Asian transition. For example, India’s 2023–24 budget countries have committed to cuts in emissions or has been dubbed a “Green Budget” for including emissions intensity. green growth as one of its seven goals, with sizable A significant focus of the region’s efforts under the investments in renewable energy and energy- Paris Agreement is to increase the share of saving technologies. Sri Lanka’s government has renewables in electricity generation. India is on prepared a Natural Adaptation Plan and National track to produce 50 percent of its energy from Environment Action Plan and is formulating a renewables by 2030 (Birol and Kant 2022). In Climate Change Act and a new Environment Act. Bhutan and Nepal, hydroelectric energy already Bangladesh’s National Solar Energy Road Map accounts for almost all power generation, and 2021–2044 and Pakistan’s Alternative and Renewable Energy (ARE) Policy 2019 outlined renewable energy targets to be achieved by increasing solar and wind capacity (World Bank Note: This chapter was prepared by Margaret Triyana. Group 2022b; 2022c). 98 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 3.1 South Asia: GHG emissions and policy in the context of a favorable outlook for commitments investment in the region as a whole: over the next South Asia’s recent share of global GHG emissions has been more than 10 years, consensus forecasters expect South Asia twice its share in global GDP. The commitments under the Paris to be the EMDE region with the strongest Agreement of several South Asian countries imply a slowdown in their emissions growth between 2021 and 2030. investment growth. A. Share of global GHG emissions and B. GHG emissions growth implied by South Asia’s energy transition will have significant global GDP, 2021 commitments consequences for its labor markets. Already the Percent 1.2 GHG emissions GDP Percent 12 Percent 8 2021-30 2015-21 number of jobs in the region related to the 0.9 9 4 production of renewable energy is almost as high 0.6 6 as in the United States, although far lower than in 0 0.3 3 China. Most of the region’s renewable energy jobs -4 0.0 0 are in India’s and Pakistan’s hydropower sectors, SAR (RHS) MDV NPL BGD BTN IND (RHS) PAK LKA -8 but there is significant potential for employment -12 IND PAK BGD LKA MDV growth in the region’s wind and solar energy sectors. Indeed, Bangladesh and India are already Sources: Climatewatch.org; EDGARv7.0_GHG database; European Commission; World among the world's top five countries in terms of Development Indicators; World Bank. Note: BGD = Bangladesh; IND = India ; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; employment in solar energy, and India is among PAK = Pakistan, SAR = South Asia Region. A. Chart shows South Asian countries’ shares of global GHG emissions compared with their shares the world’s top 10 countries in terms of in global nominal GDP in U.S. dollars in 2021. employment in wind energy (IRENA 2022c). B. Since different countries have defined their commitments in different ways, the chart shows estimated emissions growth based on the latest available nationally determined contributions plans as recorded by Climatewatch.org. For India, which has a target for the decline in emissions intensity but not for emissions growth, the emissions growth shown assumes that the targeted cut in Policies relating to labor markets can help smooth emissions intensity is achieved. the energy transition as it gains traction, by facilitating the mobility of workers, improving FIGURE 3.2 Renewable energy potential in South Asia worker skills, providing social safety nets that South Asia has high potential for solar power generation, and parts of the reduce the long-term scarring of job losses, and region also have high potential for wind power generation. developing well-coordinated regional economic programs. India, for example, launched the Skill A. Solar power potential B. Wind power potential Council for Green Jobs in 2015 to prepare its labor market for the green transition. In addition to reducing GHG emissions, the region’s energy transition has the potential to raise labor productivity and incomes significantly. Currently, South Asia, along with Sub-Saharan Africa, has the lowest labor productivity among EMDE regions, and it also has the highest share of Source: ESMAP. informal employment, in part due to a sizable Note: Charts show South Asia’s solar and wind energy potential. Solar radiation is based on the median radiation for the district. Solar suitability is based on a cutoff of 3.5 kWh/kWp. Wind speed is agricultural sector in employment and economic based on the 90th percentile wind speed in the district. Wind suitability is based on a cutoff of 5 m/s. A darker color indicates greater solar (A) or wind (B) power potential. activity (chapter 1, figure 3.4). Agricultural workers are at particular risk of being stranded in low-productivity jobs since their assets and skills Large-scale investment in green technologies is tend to be tied to land. e major investments in already underway in South Asia. India and sectors like power, industrial, and services sectors Pakistan already count among the five EMDEs to achieve the transition away from fossil with the largest public investment in renewable fuels may raise growth, as well as the level of labor energy (IRENA 2023). e region is also investing productivity, by encouraging more firms heavily in electricity networks and battery storage and workers to operate in the formal sector (IEA 2021b). ese investments are taking place (chapter 2). SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 99 While the energy transition from fossil fuels will FIGURE 3.3 Economic activity have wider-ranging consequences, this chapter Solar electricity generation is increasing rapidly in South Asia. India and focuses on the implications for workers. It discusses Pakistan are among the five EMDEs with the largest public investments in renewable energy. the following questions. A. Growth in solar electricity B. Investment in renewable energy: • What are the labor market implications of a generation, 2015–22 international finance flows, 2011–20 shift from pollution-intensive industries? Percent US$, billions 70 35 • What was the labor market impact of past 60 30 major structural transformations around the 50 25 40 20 world? 30 15 20 10 • What are the policy implications? 10 5 0 0 Contributions to the literature NPL LKA MDV IND BGD CHN U.S EU Brazil India Pakistan Nigeria U.K Individual aspects of the challenges arising from C. Green investment, 2018–19 D. Number of renewable energy jobs, the green transition have been explored in detail in US$, billions US$, billions Thousands of jobs Thousands of jobs World Bank Country Climate and Development Electricity networks and battery storage 1,000 Hydropower 6,000 Reports for Bangladesh, Nepal, and Pakistan. is 400 Renewable energy power generation 80 800 Solar Wind energy 5,000 chapter, for the first time, examines together the 300 60 600 Other 4,000 3,000 associated labor market policy challenges faced by 200 40 400 2,000 South Asia’s economies. It makes several 100 20 200 1,000 additional contributions to the literature. 0 China EAP U.S ECA LAC AFR India 0 0 0 excl. (RHS) (RHS) (RHS) (RHS) U.S IND BGD PAK CHN China (RHS) First, this chapter provides a first assessment of the share of South Asia’s workers that are most likely Sources: Ember; ESMAP; IEA; ILO; IRENA. Note: AFR = Africa; BGD = Bangladesh; CHN = China; EAP = East Asia and Pacific; ECA = to be affected by the energy transition—favorably Europe and Central Asia; EU = European Union; IND = India; LAC = Latin America and the Caribbean; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan; UK= United or unfavorably. Research has shown that shifts Kingdom; U.S. = United States. from pollution-intensive to green jobs have been A. Change in solar energy production between 2015 and 2022 for Bangladesh, India, the Maldives, Nepal, Sri Lanka, and selected comparators. Energy production measured in terawatt net job-increasing in the EU (Bali Swain, Karimu, hours. B. Renewable energy finance flows by technology, donor, financial instrument and financial and Gråd 2022), Japan (Kuriyama and Abe 2021), institution/agency in the five largest EMDE recipients between 2011 and 2020. C. Investment in green technologies, 2018–19. Investment in India is average annual investments Switzerland (Füllemann et al. 2020), the United over the period 2016–20. Investment for the power sector includes refurbishments, upgrades, new States (Garrett-Peltier 2017), and a sample of 11 builds and replacements for all fuels, and technologies for on-grid, mini-grid, and off-grid generation. Additionally, they include investment in transmission, distribution, and battery storage. advanced economies, one advanced-economy D. Number of direct and indirect renewable energy jobs, by energy type. Direct jobs are those created by core activities within the renewable energy industry. Indirect jobs supply or support the region, and the Middle East and North Africa renewable energy industry, such as the provision of materials or positions in government ministries (IRENA 2012). Data are primarily from 2021. Other renewable category includes (Meyer and Sommer 2014). Another study for the biogas, concentrating solar-thermal power, geothermal, liquid biofuels, municipal and industrial waste, solid biomass, tide, wave, and ocean energy. EMDE average includes Angola; Argentina; United States has shown that tighter Brazil; China; Egypt, Arab Rep.; Indonesia; Iran, Islamic Rep.; Mexico; Nigeria; Poland; the environmental regulation has not been associated Russian Federation; South Africa, and Thailand. with changes in aggregate employment but with higher demand for green skills, especially technical intensive jobs in South Asia. e existing literature and engineering ones (Vona et al. 2015, 2018). is sparse and based on data for advanced Data limitations mean that comprehensive economies. U.S. workers in green jobs appear, on exercises such as these cannot be conducted for average, to be better-educated and more South Asia. at said, this chapter informs the experienced, and to have higher cognitive and debate by quantifying for the first time the shares social skills than other U.S. workers (Consoli et al. of workers employed in green and pollution- 2016). In a broad sample of advanced economies intensive jobs in South Asia in a manner that and EMDEs, workers in green jobs also appear to allows comparisons across countries. be higher-skilled, more urban, and better-paid (IMF 2022). In the United Kingdom and the Second, the chapter is the first study to describe the United States, some green jobs have been characteristics of workers in green and pollution- associated with greater engineering skills (Curtis 100 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 3.4 Labor productivity and informal employment Second, workers in pollution-intensive jobs were South Asia, alongside Sub-Saharan Africa, is the EMDE region with the concentrated in the manufacturing and lowest labor productivity; it also has the highest share of informal construction sectors. Almost 50 percent of workers employment. in pollution-intensive jobs were in low-skilled occupations in textile and garment-related A. Labor productivity relative to B. Share of informal employment, advanced economies, 2021 2018 processing, food processing, and painting and Percent of advanced-economy Percent of total employment wood treatment. On average, they were labor productivity 20 80 significantly less educated and more often 60 informally employed than other workers. Their 15 informal employment arrangements and lack of 40 10 education were reflected in lower average wages. 5 20 0 Third, green jobs were more dispersed, both across 0 ECA EAP LAC MNA SAR SSA SAR SSA LAC MNA EAP ECA sectors of production and geographically. Sources: Haver Analytics; World Bank; Yu and Ohnsorge (2021). Although manufacturing was the largest employer Note: EAP = East Asia and Pacific; ECA = Europe and Central Asia; LAC = Latin America and the of workers in green jobs, it accounted for only one Caribbean; MNA = Middle East and North Africa; SAR = South Asia; SSA = Sub-Saharan Africa. A. Working-age population-weighted averages. Labor productivity is defined as nominal U.S. dollar -third of green jobs, compared with one-half of GDP relative to working-age population (aged 15–64 years). B. Working-age population-weighted averages. Share of informal employment is proxied by the share pollution-intensive jobs. Across the region, there of self-employment in total employment. were only four states where green jobs accounted for more than 10 percent of employment—half and Marinescu 2022; Sofroniou and Anderson the number of states where pollution-intensive 2021). This chapter confirms some of these results jobs accounted for more than 10 percent of in the South Asian context. employment. Workers in green jobs were significantly better educated and less frequently Third, this chapter draws upon historical informally employed than other workers and, on experience with sector-specific booms and busts, as average, their wages were 7 percent higher even well as major economic transformations in China after controlling for their higher education and and India. It synthesizes the findings of a large formal employment arrangements. literature on resource booms in the first meta Fourth, the experience of higher-income countries regression analysis of this literature. This analysis with energy sector booms and busts, and the summarizes quantitatively 43 studies that estimate major economic transformations in China and the labor market effects of major resource booms India in the 1990s, suggest that the green and busts in more than 50 countries from 2004. transition may have labor market effects that go The chapter also conducts the first review of well beyond changes in the shares of green and research on the labor market impacts of China’s pollution-intensive jobs. Some regions may be able and India’s economic transformations in the to leverage the green transition into employment 1990s. and earnings increases in both green and non- green jobs, with overall increases in employment Main findings and income. Other regions may suffer lasting This chapter sets out the following main findings. negative labor market effects, including employment and earnings losses that leave some First, in all South Asian countries except India, workers stranded in low-productivity activities, pollution-intensive jobs outnumbered green jobs, which may even carry over into future generations. with pollution-intensive jobs accounting for 6–11 percent of all jobs. In India, pollution-intensive Fifth, policies that can smooth the labor market jobs accounted for 9 percent of all jobs, below the transition and create jobs for all will be crucial to 11 percent accounted for by green jobs. The vast managing these long-term risks. Policies that have majority of jobs are classified as “pollution- proven useful elsewhere include measures that neutral”—neither green nor pollution-intensive. facilitate labor mobility—both geographically and SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 101 across occupations—into more productive FIGURE 3.5 Green and pollution-intensive jobs in South employment, support for the upgrading of worker Asia skills, well-designed social safety nets, and well- In most South Asian countries, pollution-intensive jobs outnumber green coordinated region-specific economic and labor jobs. About one-half of pollution-intensive jobs are in manufacturing whereas green jobs are more widely dispersed across sectors. market programs. A. Share of green jobs and pollution- B. Distribution of green jobs and Labor market effects of the intensive jobs pollution-intensive jobs Percent of workers Percent of jobs energy transition 14 Green jobs Pollution-intensive jobs Agriculture Construction Manufacturing Transportation Retail Electricity 12 100 80 10 Labor force surveys provide a window into the 8 60 40 nature of the jobs most likely to be affected by the 6 20 0 energy transition—green and pollution-intensive Green Jobs Poll. int. Poll. int. Poll. Int. Poll. int. Poll. int. Poll. int. Poll. int. Green Green Green Green Green Green 4 2 jobs—and into the characteristics of the workers 0 in those jobs. BGD IND LKA MDV NPL PAK SAR BGD IND LKA MDV NPL PAK Sources: National statistical offices; World Bank. Green and pollution-intensive jobs Note: BGD = Bangladesh; IND = India ; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan. Green jobs are those in occupations that have some environmentally friendly tasks, such as recycling. Pollution-intensive jobs are those that are most common in industries with above- Green and pollution-intensive jobs are defined on median pollution intensity, such as machinery mechanic, as defined in annex 3.1. Labor force surveys are available for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), the basis of occupational classifications. The vast Pakistan (2018), and Sri Lanka (2019). majority of jobs are “pollution-neutral”—neither B. All other sectors include: mining and quarrying; financial services and insurance; health and social work, education, public administration and defense; compulsory social security; other community, particularly green nor particularly pollution- social, and personal services. intensive. “Green” jobs are defined as occupations that include at least some environmentally friendly tasks, such as those related to renewable energy or half of them, with about one-sixth in environmental protection or activities such as construction. Green jobs were more dispersed: repair or recycling (Granata and Posadas 2022). only one-third of workers in green jobs were in “Pollution-intensive” jobs are defined as manufacturing, the single largest green-job sector, occupations that are most common in the most with construction and retail trade accounting for a polluting industries and include occupations such little more than one-tenth each. Despite the large as textile and garment trades workers and share of employment in agriculture, only about 4 machinery mechanics (Vona et al. 2018). Annex percent of green jobs and 1 percent of pollution- 3.1 defines green and pollution-intensive jobs in intensive jobs are in agriculture. detail. Data are available from labor force surveys for Bangladesh (2015), India (2018), the Maldives Pollution-intensive jobs were also more (2019), Nepal (2017), Pakistan (2018), and Sri geographically concentrated than green jobs. Lanka (2019). Pollution-intensive jobs accounted for 10 percent or more of employment in two states or union In the six South Asian economies with available territories in India, which also had above-average data, 2–11 percent of workers were employed in poverty; three provinces in Pakistan; and three green jobs (figure 3.5; annex table 3.1.2). The provinces in Sri Lanka (figure 3.6). Green jobs range was narrower for pollution-intensive jobs, at were more dispersed: across the whole region, 6–11 percent. In all countries except India, there were only four states with a share of green pollution-intensive jobs outnumbered green jobs. jobs above 10 percent, all of them being in India, In India, which accounts for 70 percent of the with two also heavily reliant on pollution- region’s labor force, workers in pollution-intensive intensive jobs. jobs accounted for about 9 percent of all jobs, and workers in green jobs for about 11 percent. Characteristics of workers in green jobs Pollution-intensive jobs were highly concentrated Workers in green jobs differed systematically from in manufacturing, which accounted for about one- other workers. A probit regression is used to 102 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 FIGURE 3.6 Regional distribution of jobs in South Asia drawn from labor force surveys for Bangladesh Workers in pollution-intensive jobs are concentrated in a few states and (2015), India (2018), the Maldives (2019), Nepal districts, while workers in green jobs are more geographically dispersed. (2017), Pakistan (2018), and Sri Lanka (2019). A. Share of green jobs B. Share of pollution-intensive jobs In the region as a whole, highly-educated workers were more likely to be employed in green jobs (figure 3.7). In India and Sri Lanka, workers who had completed secondary or tertiary schooling were more likely to be employed in green jobs; but in Bangladesh and Pakistan, the opposite was the case. The difference appears to reflect, in part, differences in the skill requirements for the most common green jobs in Bangladesh and Pakistan Sources: National statistical offices; World Bank. compared with other countries: more than three- Note: Green jobs are those in occupations that have some environmentally friendly tasks. Pollution- quarters of green jobs in Bangladesh and Pakistan intensive jobs are those that are most common in industries with above-median pollution intensity as defined in annex 3.1. Labor force surveys are available for Bangladesh (2015), India (2018), the were mid-skilled, compared with fewer than two- Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). A darker color indicates a higher share of local workers employed in green (A) or pollution-intensive (B) jobs. thirds of green jobs in the other countries in the region and fewer than half of green jobs in India. On average in the region, workers in informal FIGURE 3.7 Green and pollution-intensive jobs in South employment were less likely than workers in Asia: Worker characteristics formal employment to be employed in green jobs. Across the region, workers in pollution-intensive jobs are systematically However, this was not the case in the Maldives or less educated and more often informally employed. Workers in green jobs are systematically better-educated and less often informally employed. Sri Lanka, the two countries with the smallest informal sectors in the region, or Nepal. A. Marginal probability of working in a B. Marginal probability of working in a green job relative to the average job pollution-intensive job relative to the In all countries other than the Maldives, wages average job were higher for workers in green jobs than other Percentage points Percentage points 4 4 workers. In South Asia as a whole, the average 2 2 wage in green jobs was 31 percent higher than in 0 non-green jobs (figure 3.8; annex tables 3.1.5 and 0 -2 3.1.6). Much of this wage differential reflected -4 -2 worker characteristics. But even after controlling -6 -4 for worker, industry, and location characteristics, Secondary Tertiary Informal Secondary Tertiary Informal education education education education workers in green jobs received 7 percent higher wages than their peers in non-green jobs. This is Sources: National statistical offices; World Bank. Note: Green jobs are those in occupations with a positive share of environmentally friendly tasks, similar to evidence from a study of mostly such as recycling. Pollution-intensive jobs are those that are most common in industries with above- median pollution intensity, such as machinery mechanic as defined in annex 3.1. Labor force surveys advanced economies where a skill premium of are available for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan almost 7 percent was found for workers in green (2018), and Sri Lanka (2019). A.B. Marginal probabilities as estimated in probit regressions of a dummy variable of being employed jobs (IMF 2022). in a green job (A) or polluting job (B), conditional on being in an urban location, having completed secondary or tertiary education, being aged 24–54 or 55 or older, and being informally employed (annex tables 3.1.3 and 3.1.4). The regressions control for industry and subnational entity dummies. Pakistan and the Maldives were exceptions, where workers in green jobs did not receive significantly higher wages than other workers with similar estimate the probability of being in a green job characteristics. This, in part, reflected differences conditional on education level (completion of in the most common green jobs, in particular a primary, secondary, and tertiary schooling), age, greater prevalence of mid-skilled green jobs in urban location, and informal employment, Pakistan and of green jobs in low-wage controlling for subnational and industry fixed construction in the Maldives. In Pakistan, effects (annex table 3.1.3). The data are again considerably more green jobs were mid-skilled— SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 103 about 80 percent of them—than in South Asia as FIGURE 3.8 Green and pollution-intensive jobs in South a whole, where the proportion was 43 percent. In Asia: Wage premiums and discounts the Maldives, the share of green jobs in typically On average, workers in green jobs receive significantly higher wages than lower-wage construction was double the share of other workers whereas those in pollution-intensive jobs receive lower wages. In part, this reflects worker characteristics, but even controlling for green jobs in typically higher-wage manufacturing, these, in most South Asian countries, workers in green jobs earn a wage whereas the reverse was true in South Asia overall. premium. Characteristics of workers in pollution- A. Raw wage differential between B. Wage premium for green and workers in green jobs and other pollution-intensive jobs intensive jobs workers Percent Percent Workers in pollution-intensive jobs also differed 35 10 systematically from their peers in non-pollution- 30 8 6 intensive jobs. Workers with secondary or tertiary 25 4 20 education were less likely, and informal workers 15 2 more likely, to be in pollution-intensive jobs 10 0 -2 (annex table 3.1.4). 5 -4 0 Green Pollution-intensive SAR IND BGD LKA NPL PAK MDV job job Workers in pollution-intensive jobs, on average, received almost 10 percent lower wages than the C. Wage premium in green jobs D. Wage premium in pollution- intensive jobs average worker in the region. In India, this raw Percent Percent wage differential reflected worker, industry, and 30 5 location characteristics. In most other South Asian 20 0 countries, wages in pollution-intensive jobs were 2 10 -5 –11 percent lower than average wages even after 0 -10 controlling for worker characteristics, industry, -10 -15 and location. In Bangladesh, however, workers in -20 -20 pollution-intensive jobs received higher wages SAR LKA NPL IND BGD PAK MDV SAR MDV NPL PAK LKA IND BGD than their peers with similar characteristics, in part Sources: National statistical offices; World Bank. because the share of high-skilled workers in Note: SAR = South Asia Region; BGD = Bangladesh; IND = India ; LKA = Sri Lanka; MDV = Maldives; NPL = Nepal; PAK = Pakistan. pollution-intensive jobs is about twice the region’s Green jobs are those in occupations with a positive share of environmentally friendly tasks, such as average. recycling. Pollution-intensive jobs are those that are most common in industries with above-median pollution intensity, such as machinery mechanic as defined in annex 3.1. Labor force surveys are available for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Lessons from past A. Wage differential is the difference in the average wages between workers in green jobs and other workers. B.C.D. A Mincer regression estimates the rate of returns to being in a green and pollution-intensive structural transformations job, dubbed “wage premium.” The analysis entails regressing log earnings on dummy variables for being employed in a green job or pollution-intensive jobs conditional on being in an urban location, having completed secondary or tertiary education, and being aged 24– 54 or 55 or older, potential experience, and being informally employed (annex table 3.1.5). The regressions control for industry The energy transition is likely to involve and subnational entity dummies. considerable structural transformations of the region’s economies. Such structural transformations have occurred in other countries and practices and exit pollution-intensive and at other times. A sizable literature has activities. Structural shifts in agriculture and examined these past experiences, and there are resource sectors have caused comparable policy lessons to be drawn. transformations in the past. While agriculture remains an important sector in the region, its Resource booms and busts share of green and pollution-intensive jobs is relatively small. The growth in renewable energy As the energy transition proceeds, different parts may be more similar to the experience of the of South Asia will undergo employment resource sectors, which indicates that the labor expansions or contractions, depending on the market consequences of the energy transition may speed with which they adopt green technologies reach well beyond the directly affected sectors. 104 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 The labor market effects of resource booms and subnational resource booms and busts, the meta busts have been most extensively studied for five regression analysis is based on data from 23 studies advanced economies (Australia, Canada, Spain, on employment (53 estimates), 27 studies on the United Kingdom, and the United States), earnings (70 estimates), and 13 studies on three EMDEs in Latin America (Brazil, Chile, and spillovers to non-resource sectors (65 estimates). Peru), two EMDEs in Europe and Central Asia (Kazakhstan and the Russia Federation), and 29 A meta regression analysis is conducted on these EMDEs in Sub-Saharan Africa. The effects of studies to estimate the average effect of resource these resource booms and busts have parallels to booms and busts on the local labor market. The the likely effects on labor markets of the green meta regression combines and contrasts multiple transition. studies on resource booms and busts to capture the methodological diversity in such studies. The That said, there are also important differences. In main outcomes of interest are the percent changes particular, the green transition is likely to be in total employment and earnings. Earnings associated with a shift from lower-skilled to higher measures vary widely in these studies and include -skilled jobs, whereas past resource booms have earnings per worker, family earnings, wage and often been associated with shifts in the opposite salary income, GDP per capita, total wages, direction. Therefore, the following review of the annual pay, household income, median income, evidence on resource booms and busts is median earnings, and per capita expenditure. The complemented with a review of the literature on meta regression analysis assumes that all these two broader structural transformations: those measures are suitable proxies for average earnings. following India’s liberalization of international The analysis also examines employment spillovers trade and domestic regulations that began in to non-resource sectors. While the results are 1991, and China’s market-based reforms of the indicative of indirect effects, the meta regression 1990s. analysis cannot distinguish the channels behind the observed spillovers. Literature sources Subnational regions with rapid growth in To identify the labor market effects of resource green jobs: Past resource booms booms and busts, a comprehensive literature review is conducted. The studies summarized in e parts of South Asia with rapid growth in Aragón, Chuhan-Pole, and Land (2015) and green jobs are likely to witness similar benefits and Marchand and Weber (2018) form a starting challenges to those experienced by countries or point for the analysis, supplemented by more subnational regions that experienced resource recent studies. The analysis began with a total of booms in the past. eir experience suggests 43 studies published in peer-reviewed journals, significant employment and earnings gains, with working papers, and policy publications (annex favorable spillovers to non-resource sectors. tables 3.2.1 and 3.2.2). Most of these studies (28 of the 43 studies) examine developments in the In almost all studies of resource booms around the United States, with others analyzing developments world, total employment rose and, in most cases, in Canada (one study), and economies in East Asia average earnings rose too. Brazil’s offshore oil and the Pacific (four studies), Europe and Central boom in the 1980s was an exception because it Asia (four studies), Latin America (eight studies), provided limited employment for residents in the Middle East and North Africa (three studies), regions closest to the oil fields. On average, South Asia (three studies), and Sub-Saharan Africa subnational resource booms were associated with a (30 studies). Together, these studies cover more statistically significant 5 percent increase in than 50 countries from 2004 onward. Some of employment in the local labor market directly these studies examined resource booms at the affected by the booms and a 0.5 percent increase country level, others at the subnational level. After in earnings (figure 3.9). standardizing the results and restricting them to SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 105 e employment gains were virtually universal FIGURE 3.9 Meta regression of the effects of resource across all studies. For earnings gains, the majority booms and busts on employment and earnings of estimates showed increases. About 30 percent of Employment and earnings systematically increased during resource the estimates in studies that reported earnings booms and fell during resource busts declines typically referred to areas with high resource dependence and a rapid expansion of low A. Average effect of resource booms B. Average effect of resource busts -skilled jobs. ese studies generally present Percent Percent evidence consistent with “Dutch disease”—a 8 0 decline in non-resource sectors that accompanies a 6 resource boom in a resource-reliant economy. -1 4 Since green jobs tend to be higher-skilled, the -2 experience of this subset of studies may be of 2 limited relevance to the energy transition. 0 -3 Employment Earnings Employment Earnings Studies of resource booms have often shown that additional jobs in the resource sector created Sources: Calculations based on Marchand and Weber (2018), Aragón, Chuhan-Pole, and Land (2015), and related studies in annex 3.2. additional non-resource-sector jobs, which Note: Estimates are based on random effects meta regressions. Each study’s effect size is standardized to percentage changes. Log changes are interpreted as percentage changes. attracted workers from regions not directly Employment measure is total employment in the labor market considered by each study. Earnings include earnings per worker, family earnings, wage and salary income, GDP per capita, total wages, affected by the boom. With the exception of annual pay, household income, median income, median earnings, and per capita expenditure. Brazil’s offshore oil boom, resource booms were accompanied by increases in the working-age population. e meta regression analysis suggests 1980s and 2010s (Hanson 2023); and the 2012– that the average subnational resource boom was 15 coal bust in China (Zhang 2023).1 After such associated with a small, but statistically significant, resource busts, household finances deteriorated increase in non-resource-sector employment, with significantly not only for directly affected about 0.2 additional jobs created during the boom households but also more broadly in the affected (annex tables 3.2.2 and 3.2.3). All sectors (in both regions (Blonz, Tran, and Troland 2023). goods and services production) experienced statistically significant increases in employment. On average, subnational resource busts were associated with a 1 percent, statistically significant, Subnational regions reliant on pollution- decline in total employment. The decrease in intensive employment: Past resource busts employment was not reversed (Krause 2022). On average, resource busts were associated with 2 Local economic busts, such as those that may percent earnings losses—well in excess of the 0.5 occur in some subnational regions that currently percent earnings gain during resource booms. If rely heavily on pollution-intensive jobs, have been the energy transition generates similarly outsized shown to cause lasting scarring. The literature on earnings losses for workers in pollution-intensive resource busts suggests significant aggregate jobs, it will tend to increase inequality in several employment and earnings declines. However, South Asian countries where these workers’ there is wide heterogeneity in spillovers to the non earnings are already below-average. -resource sector. The spillovers to non-resource sectors varied In many past resource busts, total employment widely because they had different degrees of and average earnings declined in the affected regions and remained depressed for many years. This was the case for regions affected by the 1 In Canada, by contrast, there was no significant decline in decline of U.S. manufacturing due to globalization employment around oil busts in the 1980s (Marchand 2012). In (Autor, Dorn, and Hanson 2021); the U.S. oil Spain, large plant closures led to less-than-proportionate declines in and gas bust in the 1980s (Jacobsen and Parker local employment (Jofre-Monseny, Sánchez-Vidal, and Viladecans- Marsal 2018); and a less-than-proportionate decline was also seen in 2016); the decline of the U.S. coal industry in the the U.K. regions affected by coal mine closures (Beatty, Fothergill, and Powell 2007). 106 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 complementarity with the resource sector. In some on aggregate employment remains an open cases, subnational resource busts were associated question. Major state-owned enterprise with job losses of up to 10 percent in all other restructuring was associated with either sectors. insignificant aggregate employment changes or employment gains, at least among firms that Affected regions may see an exodus of their most underwent outright ownership changes (Boubakri productive workers (Marchand 2012). Significant and Cosset 1998; Gong, Maioli, and Görg 2007; out-migration, especially of younger workers, was Sun and Tong 2003). Privatization, specifically, seen in the U.S. and U.K. regions most affected by was associated with higher employment (Cosset the coal industry decline of the 1980s (Beatty, and Boubakri 2002; Sun and Tong 2003) but Fothergill, and Powell 2007; Hanson 2023), lower employment growth in privatized firms, although not in the Chinese regions most affected pointing to slower job transitions (Gong, Maioli, by the coal industry bust of 2012–15 (Zhang and Görg 2007). 2023) or in the Spanish regions affected by large plant closures (Jofre-Monseny, Sánchez-Vidal, and India. India’s economic liberalization included Viladecans-Marsal 2018). reductions in import tariffs, the deregulation of markets, and reductions in taxes. There were no Many displaced workers may never find another statistically significant employment contractions in job, never migrate, and never return to their pre- the import-competing industries most affected by boom earnings. This was the case for many these reforms, but there was a clear decline in workers affected by the decline of U.S. wages (Banga 2005; Epifani 2003; Topalova manufacturing in the 1990s and early 2000s 2010). The presence of FDI in an industry was (Autor, Dorn, and Hanson 2013; Autor et al. associated with higher wages (Banga 2005) while 2014); the decline of the U.S. coal industry in the higher rates of import protection were associated 1980s (Hanson 2023); and the U.S. oil bust with lower wages (Kumar and Mishra 2008). The (Jacobsen, Parker, and Winikoff 2023). Various wage-elasticity of labor demand increased after the welfare payments rose significantly in resource reforms (Hasan, Mitra, and Ramaswamy 2007), busts and remained statistically significantly larger pointing to faster job transitions for workers. long afterwards (Black, McKinnish, and Sanders Separately, delicensing was associated with an 2003; Hanson 2023). increase in the number of factories, promising increased economic activity and job creation Large-scale economic transformations (Aghion et al. 2008). The experience of major economic transformation, such as the market-based Policy implications economic reforms in China, including state- owned enterprise reforms, in the 1990s, and the The experience of past sectoral transformations trade liberalization and deregulation in India that suggests that regions with rapid green jobs growth started in 1991, points to possible consequences of are likely to benefit from employment and income the green transition for employment and earnings gains, while regions with declining pollution- in South Asia. intensive activities may suffer lasting employment and income losses. The characteristics of workers China. China’s reforms in the 1990s relaxed price in pollution-intensive jobs suggest that job and controls and allowed the operation of private income losses may be concentrated among lower- enterprises, including through the privatization of paid, lower-educated workers. state-owned enterprises. In general, China’s state- owned enterprises have been characterized by Efforts to accelerate overall job creation and to lower labor productivity and greater job stability help the most affected workers and regions adjust than private enterprises (Cheng and Ng 2023; Li are, therefore, a priority. Measures to improved and Yamada 2015). The privatizations changed access to finance and high-quality education and the composition of employment but their impact training, to facilitate the mobility of workers, as SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 107 well as a robust social safety net, can support changing circumstances, have large coverage, and workers in their transition out of declining be efficiently administered (World Bank 2022a). pollution-intensive industries into neutral jobs. Social protection programs implemented before Regional policies can help shore up economic negative shocks occur can serve as automatic activity in the most adversely affected regions. stabilizers of the economy and, during crises, they Each of these policies needs to be carefully can be modified to meet specific objectives. Such designed and coordinated across different levels of programs can aim for income protection or job government to avoid fiscally costly interventions protection, the choice partly depending on the that do not improve aggregate employment or structure of the labor market. In South Asia, productivity and incomes. where few workers are permanently employed full- time in the formal sector, income-based, means- Job creation. Broader job creation, fueled by tested social protection schemes may be needed to stronger growth, could help absorb the workers in protect vulnerable workers. A robust social safety pollution-intensive jobs at risk of being stranded net can also allow policy makers to ease labor during the energy transition (chapter 1). The regulations, which may allow workers to move to employment opportunities created by the energy productive firms and more productive transition itself may not offer promising prospects employment. Broad income support programs are to workers in pollution-intensive jobs because of likely to be less distortionary and more effective at the high-skill bias of green jobs. On average, protecting vulnerable groups than job protection workers in green jobs are higher skilled: 30 percent schemes during the labor market shifts that are of green workers are high-skilled, while only 15 likely to accompany the green transition. percent of non-green workers and 4 percent of pollution-intensive workers are high-skilled. At Facilitating internal worker migration. least initially, the green transition is likely to Connecting workers with jobs, including through constitute a skill-biased shift in the composition of mobile job platforms, may help encourage labor jobs. Faster output growth and, with it, more mobility and improve skill matching and vigorous job creation is needed to absorb workers employment prospects (Lobao et al. 2021; who are currently in pollution-intensive jobs, Ruppert Bulmer et al. 2021). South Asia already especially into pollution-neutral jobs. Indeed, accounts for 9 percent of the world’s international evidence for India suggest that, over time, gains in migration but, considering the size of the region’s energy efficiency have been accompanied by job population, cross-border migration within South creation (chapter 2). Asia is relatively limited (World Bank 2023b). Migration, both internal and cross-border, Skills upgrading. Retraining programs for workers including agriculture workers who may be can facilitate their redeployment. However, stranded due to land, toward higher-productivity evidence on the success of such active labor market employment, could be encouraged by lowering the programs is mixed, both for advanced economies cost of migration, broadening the portability of (Crépon and Van Den Berg 2016) and EMDEs social security benefits (such as in India’s One- (McKenzie 2017). Any retraining or reskilling Nation-One-Ration-Card scheme), and expanding program should be designed and implemented in access to remittance services (World Bank 2022b). close consultation with employers (Christiaensen and Ferré 2020). Subnational policies. The energy transition will have different effects across subnational labor Robust social safety net. A robust social safety net markets and may therefore require changes in can prevent the lasting scarring that often occurs fiscal transfers to ensure continued and well- after adverse economic shocks. When well- targeted public service delivery. In most South designed, it can also encourage the risk-taking by Asian countries (except Bangladesh and Sri workers often needed for gains in productivity. Lanka), subnational government spending, relative Although challenging to implement, the social to GDP, is above the average level in other benefits system needs to be adaptable to rapidly EMDEs. In India, targeted intragovernmental 108 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 transfers for specific purposes have been found to ANNEX 3.1. Methodology: improve service delivery and subnational outcomes (World Bank 2019). This suggests that well- Quantifying green and coordinated programs, including programs pollution-Intensive jobs financed by fiscal transfers, to stimulate economic activity in specific districts and provinces may help Sample. e sample includes 332,128 workers for them adjust to the structural transformation away Bangladesh (2015), India (2018), the Maldives from fossil fuels. The most effective subnational (2019), Nepal (2017), Pakistan (2018), and Sri policies have been found to be those that were well Lanka (2019). District level data is available for -coordinated between different levels of India (648 districts), the Maldives (18 districts), government (Lobao et al. 2021; Oei, Brauers, and Nepal (74 districts), and Sri Lanka (25 districts). Herpich 2020). To facilitate the adjustment Division level data is available for Bangladesh (7 process in the hardest-hit districts and provinces, divisions) and province level data for Pakistan (4 subnational policies could aim to develop the provinces). e data are restricted to employed institutional capacity to catalyze private male workers between the ages 15 and 64. Both investment and expand infrastructure to connect pollution-intensive and green jobs employ few to markets. women and the selection into labor force participation for women is beyond the scope of • Infrastructure. Subnational policies that this chapter. Variables and data sources are shown promote specific investments to improve the in annex table 3.1.1. region’s infrastructure and productive capacity can promote economic activity. For the U.S. Definitions. Green jobs are those in occupations regions that suffered from the decline of the with a positive share of environmentally friendly coal industry, for example, investment in tasks, as defined in Granata and Posadas (2022). A transport infrastructure was associated with green job indicator takes the value one if the permanent increases in employment (Hanson occupation has at least one environmentally 2023). In declining German coal mining friendly task. Environmentally friendly tasks regions, too, investment in infrastructure and include repairs, environmental management, education reduced adjustment costs (Oei, recycling and forest management. Examples of Brauers, and Herpich 2020). A case study green occupations include: bicycle repairer; from Canada’s Just Transition Fund also forestry workers; refuse sorters; farming, forestry, found that a combination of training, income and fisheries advisers; civil engineers; and support, and investments in infrastructure environmental and occupational health inspectors. helped absorb coal workers in Alberta in the Examples of green occupations in agriculture short term (Parkland Institute 2019). include aquaculture workers and agricultural and forestry production managers. A pollution- • Institutional capacity. Strong local institutional intensive occupation is defined as in the most capacity can secure resources needed for the common occupations (at the 6-digit SOC level) in transition. Evidence from the closure of U.S. the five percent of 4-digit NAICS industries that coal mines suggests that strong institutional have the highest emissions of pollutants per capacity was key to a successful transition to a worker, as defined in appendix C of Vona et al. post-coal economy. This allowed local city (2018). Polluting industries are a set of 62 4-digit councils to access grants, turn nearby North American Industry Classification System universities into incubators, develop (NAICS) industries that are in the 95th percentile commuter or tourism economies near of pollution intensity for at least three pollutants population centers, and design effective among CO, VOC, NOx, SO2, PM10, PM2.5, reskilling and job placement programs lead, and CO2. Pollution-intensive occupations (Ruppert Bulmer et al. 2021). are then defined as those 6-digit Standard Occupational Classification (SOC) occupations with a 7 times higher probability of working in SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 109 ANNEX TABLE 3.1.1 Sample description Variable Definition/description Occupations follow the ILO's ISCO-08 4-digit coding for Bangladesh, Sri Lanka, the Maldives, Nepal, and Pakistan and Occupation ILO’s ISCO-88 3-digit coding for India (International Labour Organization 2008). Green job and pollution-intensive job indicators are created based on occupation. Industry includes 1. agriculture (reference group); 2. manufacturing; 3. mining and quarrying; 4. electricity, gas, steam and air condition supply; 5. construction; 6. wholesale and retail trade; 7. transportation and storage; 8. financial services and Industry insurance; 9. health and social work, education, public administration and defense; compulsory social security; other com- munity, social and personal services. Data for Nepal includes an adjustment to agriculture. Education is categorized into less than primary, primary, secondary (which combines lower- and upper-secondary) and Education tertiary (which includes post-secondary). Potential experience Inferred from age and level of education minus 6, the assumed start of primary education. India includes urban location of work. For all other countries, urban residence is assumed to be urban location of work. Urban Data for the Maldives assume Malé Atoll, which includes the capital, Malé, to be urban. Informal indicator is created based on self-employment, casual work pattern, informal salary payments, social security, and Informality business registration. The informality indicator is missing for 2 to 10 percent of the sample. Monthly cash earnings of an employee in main job. Main refers to the work that an individual spent most of the time on in Monthly earnings the past week or month. If same number of hours used in more than one work, this considers the one where they earn the most money. PPP conversion factors for GDP taken for the corresponding country and survey year. Raw monthly earnings divided by PPP adjustment PPP conversion factor. Note : Labor force surveys for Bangladesh (2015), India (2018), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). The Maldives (2019) Household Income and Expenditure Survey is used since there is no separate labor force survey. District-level data are available for India (648 districts), the Maldives (18 districts), Nepal (74 districts), and Sri Lanka (25 districts). Division -level data are available for Bangladesh (seven divisions) and province level data for Pakistan (four provinces). The data are restricted to employed male workers between the ages 15 and 64. The definition of agricultural work was adjusted between the preceding labor force survey and the 2018 labor force survey. To account for this change, the definition of agricultural work was adjusted to count all those individuals who spent at least 4 hours a month on subsistence agriculture activities, following the methodology employed in the 2020 Nepal Jobs Diagnostic. polluting sectors than in any other job.2 ese where Xi is the occupation at the 6-digit level, wi is occupations are collapsed to the 3-digit SOC the share of employment of 6-digit occupation i in occupations using the share of employment in the 3-digit occupation r, and Xr is the occupation 3-digit occupation as follows: variable at the 3-digit level. e 3-digit SOC n occupations are then mapped to the ILO’s 3-digit Xr =  w ×xi =1 i i ISCO occupations. e mapping to the ISCO n  w i occupations effectively collapses the SOC i =1 occupation-level data, which generates a measure 2 Vona et al. (2018) used this threshold to ensure the inclusion of of pollution intensity between 0 and 1. In this certain occupations that are associated with high emissions. analysis, for comparability to the binary nature of 110 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 the green job definition, a pollution-intensive job annex table 3.1.4 shows the estimated marginal indicator takes the value one if the occupation’s probability of being employed in a pollution- pollution intensity is above the median. Annex intensive job. table 3.1.2 shows the share of workers in green and in pollution-intensive jobs in the main sectors. Regression: wage premium. To estimate the wage Examples of pollution-intensive jobs include: premium, a variant of the same regression is machinery mechanics and fitters; tailors, estimated in a linear regression of log earnings: dressmakers, furriers and hatters; painters, log earningsi,j = building structure cleaners; food processing and related trades workers; and textile, garment and δg Greenj + δp PollutionIntensivej related trades workers. In agriculture, such jobs β1 Ejsecondary β2 Ejtertiary β3 Agej24-54 include butchers, fishmongers and related food preparers and food processing and related trades β4 Agej55+ β5 urbanj workers. β6 informalj β6 expj β7 expj2 Regression: worker characteristics. Probit where expj is the potential experience of worker j, regressions are estimated for the probability of a i.e., their age minus the typical age at which their worker being employed in a green job or in a education level is achieved, and the quadratic term pollution-intensive job, conditional on worker of potential experience. is chapter assumes characteristics: formal education begins at age 6. Industry and pi,j = Φ(β1 Ejsecondary β2 Ejtertiary β3 Agej24-54 subnational regional fixed effects are included. β4 Agej55+ β5 urbanj β6 informalj) Annex table 3.1.5 shows the coefficient estimates. Annex table 3.1.6 shows the Blinder-Oaxaca where pi,j is a dummy variable if worker j is decomposition, into explained and unexplained employed in a job of type i (either green or non- components, of the difference between average green); Ejsecondary and Ejtertiary are dummy variables wages in a green or pollution-intensive job and the for worker j having completed secondary or country- or region-wide average wage (Blinder tertiary schooling, respectively; Agej24-54 and 1973; Oaxaca 1973). Agej55+ are dummy variables for worker j being aged 24-54 years or 55 and more years, Both regressions—for the probability of being respectively; urbanj is residence in an urban employed in a green or pollution-intensive job and location by worker j3, and informalj, a dummy for the wage premium—are estimated both as a variable for worker j being informally employed. pooled regression of all labor force surveys with Industry and subnational regional fixed effects are subnational regional fixed effects and separately for included. A similar probit regression is estimated each country. Survey weights are used in all where pi,j is a dummy variable if worker j is estimations. Pooled regression reweights each employed in a pollution-intensive job or non- country’s sample by the share of the labor force in pollution-intensive job. Worker characteristics of the region.4 green jobs are based on the set of jobs that are green at the moment. However, given the early stages of the green transition, there is a possibility that the characteristics of workers who hold green jobs may be systematically different in the future. Annex table 3.1.3 shows the estimated marginal probability of being employed in a green job and 3 Almost all surveys include urban residence instead of location of 4 Using the original survey weight in the pooled regression yields employment (annex table 3.1.1). qualitatively similar results. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 111 ANNEX TABLE 3.1.2 Share of workers in green and pollution-intensive jobs in the main sectors Electricity, gas, Agriculture; Wholesale and Transportation Manufacturing steam and air Construction Other forestry & fishing retail trade and storage condition supply 4 34 3 15 13 4 26 South Asia Green Pollution- 1 52 1 21 14 3 9 intensive 26 14 3 10 33 1 14 Bangladesh Green Pollution- 3 41 1 12 26 0 17 intensive 4 34 3 15 13 4 26 India Green Pollution- 1 51 1 25 11 3 9 intensive 9 2 20 5 14 10 40 Maldives Green Pollution- 0 24 17 25 15 10 8 intensive 9 8 5 17 40 0 20 Nepal Green Pollution- 1 57 1 19 18 0 2 intensive 5 10 6 4 66 1 10 Pakistan Green Pollution- 0 67 2 8 18 1 4 intensive 20 12 2 4 30 2 30 Sri Lanka Green Pollution- 1 45 1 17 28 1 7 intensive Sources: Labor force surveys for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Note: The data are restricted to employed male workers between the ages 15 and 64. Other includes mining and quarrying; financial services and insurance; health and social work, education, public administration and defense; compulsory social security; other community, social and personal services. ANNEX TABLE 3.1.3. The marginal probability of being employed in a green job Region Bangladesh India Maldives Nepal Pakistan Sri Lanka Level of education: -0.003 0.000 -0.003 0.020*** 0.001 -0.004** 0.01 below primary (0.006) (0.003) (0.006) (0.008) (0.005) (0.001) (0.012) Level of education: 0.014*** 0.000 0.015*** -0.007 0.002 -0.009*** 0.002 secondary (0.004) (0.002) (0.005) (0.005) (0.004) (0.002) (0.004) Level of education: 0.028*** -0.005*** 0.030*** -0.005 0.010** -0.008*** 0.024*** tertiary (0.007) (0.002) (0.007) (0.005) (0.005) (0.003) (0.009) Age: 25–54 -0.014** -0.007*** -0.014** 0.003 -0.007 -0.014*** 0.000 (0.006) (0.001) (0.006) (0.005) (0.004) (0.002) (0.004) Age: 55–64 -0.023*** -0.016*** -0.023*** -0.020*** -0.011* -0.020*** -0.013*** (0.008) (0.004) (0.009) (0.007) (0.006) (0.003) (0.005) Urban 0.041*** 0.003** 0.044*** 0.000 0.004 0.005** 0.010*** (0.005) (0.002) (0.006) (0.001) (0.003) (0.002) (0.003) Informality -0.034*** -0.011*** -0.037*** -0.007 -0.006 -0.020*** 0.002 (0.006) (0.001) (0.006) (0.007) (0.005) (0.003) (0.004) No. of observations 227,434 89,932 65,269 5,591 10,218 40,096 15,590 Share of pollution- 0.06 0.03 0.17 0.04 0.02 0.03 0.04 intensive jobs Sources: Labor force surveys for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Note: The data are restricted to employed male workers between the ages 15 and 64. Industry and subnational region fixed effects included, standard errors in parentheses, clustered at the subnational region level. * p<0.10, ** p<0.05, *** p<0.01. 112 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 3.1.4 The marginal probability of being employed in a pollution-intensive job Region Bangladesh India Maldives Nepal Pakistan Sri Lanka Level of education: below primary -0.006 -0.022*** -0.005 -0.009 -0.002 -0.026*** -0.044** (0.004) (0.005) (0.005) (0.010) (0.003) (0.005) (0.018) Level of education: secondary -0.006** -0.009*** -0.006** 0.005 -0.005 -0.009 0.007 (0.003) (0.002) (0.003) (0.006) (0.003) (0.009) (0.006) Level of education: tertiary -0.021*** -0.067*** -0.020*** -0.013*** -0.005 -0.089*** -0.096*** (0.006) (0.002) (0.006) (0.005) (0.005) (0.016) (0.022) Age: 25–54 -0.016*** -0.005** -0.016*** 0.011** 0.001 -0.017*** -0.007 (0.003) (0.002) (0.003) (0.005) (0.003) (0.003) (0.005) Age: 55–64 -0.029*** -0.019*** -0.029*** 0.008 -0.008* -0.032*** -0.025*** (0.005) (0.003) (0.006) (0.011) (0.005) (0.009) (0.009) Urban 0.021*** 0.003 0.021*** 0.001 0.005* 0.031*** 0.011 (0.003) (0.005) (0.003) (0.002) (0.003) (0.002) (0.007) Informality 0.010** 0.015*** 0.008* -0.002 0.004** 0.048*** 0.016*** (0.004) (0.006) (0.005) (0.003) (0.002) (0.005) (0.005) No. of observations 226,755 90,197 62,161 5,551 12,725 40,096 15,632 Share of pollution-intensive jobs 0.09 0.09 0.12 0.06 0.06 0.15 0.11 Sources: Labor force surveys for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Note: The data are restricted to employed male workers between the ages 15 and 64. Industry and subnational region fixed effects included, standard errors in parentheses, clustered at the subnational region level. * p<0.10, ** p<0.05, *** p<0.01. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 113 ANNEX TABLE 3.1.5 Earnings in green and pollution-intensive jobs Region Bangladesh India Maldives Nepal Pakistan Sri Lanka Green job 0.066*** 0.037** 0.065*** -0.053 0.154** 0.017 0.197*** (0.013) (0.012) (0.013) (0.050) (0.060) (0.016) (0.062) Above-median pollution -0.006 0.017* -0.005 -0.110* -0.033 -0.029** -0.026 intensity (0.012) (0.009) (0.013) (0.057) (0.035) (0.005) (0.019) Informality -0.418*** -0.119*** -0.422*** -0.264*** 0.073*** -0.458*** -0.128*** (0.022) (0.006) (0.023) (0.029) (0.025) (0.060) (0.027) Level of education: -0.104*** -0.060** -0.104*** -0.168 -0.086*** -0.056*** 0.035 below primary (0.012) (0.017) (0.012) (0.122) (0.031) (0.002) (0.045) Level of education: 0.186*** 0.136*** 0.186*** 0.221*** 0.281*** 0.178*** 0.110*** secondary (0.009) (0.011) (0.009) (0.026) (0.029) (0.024) (0.018) Level of education: tertiary 0.671*** 0.598*** 0.673*** 0.623*** 0.627*** 0.618*** 0.953*** (0.022) (0.072) (0.023) (0.043) (0.031) (0.038) (0.039) Age: 25–54 -0.033** -0.068*** -0.032** 0.054* -0.019 0.009 0.102*** (0.015) (0.008) (0.015) (0.026) (0.036) (0.011) (0.024) Age: 55–64 0.044* -0.017 0.047** -0.026 0.023 0.085** 0.143** (0.023) (0.024) (0.024) (0.042) (0.062) (0.024) (0.067) Urban residence 0.136*** 0.062*** 0.136*** 0.000 0.049** 0.178*** 0.041 (0.014) (0.007) (0.014) (.) (0.024) (0.014) (0.035) Potential experience 0.040*** 0.024*** 0.040*** 0.038*** 0.038*** 0.035*** 0.030*** (0.002) (0.002) (0.002) (0.004) (0.004) (0.003) (0.004) Potential experience -0.001*** -0.000*** -0.001*** -0.001*** -0.001*** -0.000*** -0.001*** (squared) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) No. of observations 141,512 42,476 66,594 4,695 4,738 14,508 8,501 Monthly earnings in cash 647.47 658.12 638.32 1226.30 621.72 552.17 689.03 (PPP mean) Sources: Labor force surveys for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Note: The data are restricted to employed male workers between the ages 15 and 64. Industry and subnational region fixed effects included, standard errors in parentheses, clustered at the subnational region level. * p<0.10, ** p<0.05, *** p<0.01. 114 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 3.1.6 Decomposition of earnings differential between workers in green jobs and the average worker Region Bangladesh India Maldives Nepal Pakistan Sri Lanka Overall Green workers 6.361*** 6.544*** 6.361*** 7.058*** 6.439*** 6.166*** 6.446*** (0.032) (0.028) (0.025) (0.174) (0.059) (0.072) (0.050) Non-green workers 6.047*** 6.327*** 6.044*** 7.014*** 6.328*** 6.095*** 6.265*** (0.017) (0.037) (0.018) (0.145) (0.019) (0.040) (0.050) Difference 0.314*** 0.217*** 0.317*** 0.044 0.112* 0.071** 0.181*** (0.026) (0.032) (0.020) (0.039) (0.060) (0.036) (0.049) Explained 0.207*** 0.130*** 0.213*** 0.082 0.037 0.010 0.005 (0.025) (0.028) (0.020) (0.058) (0.065) (0.044) (0.034) Unexplained 0.107*** 0.088*** 0.104*** -0.039 0.074 0.061*** 0.176*** (0.028) (0.013) (0.019) (0.043) (0.059) (0.021) (0.048) No. of observations 142,853 43,347 66,671 4,956 4,797 14,508 8,574 Sources: Labor force surveys for Bangladesh (2015), India (2018), the Maldives (2019), Nepal (2017), Pakistan (2018), and Sri Lanka (2019). Note: The data are restricted to employed male workers between the ages 15 and 64. Explained component comes from education, age, urban residence, potential experience, and informality. Industry and subnational region fixed effects included, standard errors in parentheses, clustered at the subnational region level. * p<0.10, ** p<0.05, *** p<0.01. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 115 ANNEX 3.2. Methodology: Meta regression. e meta-analysis regression uses a random effects model, weighted by the inverse of Meta regressions the variance so that more weight will be given to results that are more precisely estimated (Fabregas, Sample. e initial screening included 43 studies Kremer, and Schilbach 2019; Fabregas et al. summarized in Aragón, Chuhan-Pole, Cust and 2019). For each experiment, the observed Poelhekke (2015), Land (2015), and Marchand treatment effect is given by: and Weber (2018) supplemented with recent studies. e systematic reviews are updated by Tˆ =θ +e , j j j including studies conducted between 2015 to the where θj is the true effect for study j and ej is the present, including working papers. Studies are included in the analysis if they estimate a resource within study error, where e j ~ N (0, σ j ) , and σj is boom or bust on the labor market, focusing on the sampling variation in estimating θj. e employment, earnings, and spillovers to non- technique assumes that θ j = µ + δ j and resource sectors. Employment is defined as total δ j ~ N (0, τ 2 ) , where τ2 is the between-study employment in the labor market considered by variance, estimated by the DerSimonan and Laird each study. Earnings are measured in various ways, method (DerSimonian and Laird 1986). e including earnings per worker, family earnings, estimated µ is given by: wage and salary income, GDP per capita, total wages, annual pay, household income, median ˆ= µ w T j j income, median earnings, and per capita w j expenditure. Spillovers to non-resource sectors are measured in terms of employment in all other where wj is study-specific weight given by the sectors, manufacturing, transport, construction, inverse of the variance, which is given by: education and health, government, retail, accommodation, other services, agriculture, and all 1 wj = , services. Some sectors may contract and expand 2 (τˆ + σˆ2 j) along with the resource sector while the converse may be true for others. If the article includes several estimated effects, the Studies must have standard errors or confidence intent-to-treat parameter is the preferred estimate intervals to be included in the meta regression (Croke et al. 2016). e intent to treat parameter analysis. If the estimated effect is not reported as a is the estimated effect of a resource boom or bust percent change, for consistent comparison, the on the labor market outcomes of interest. Ideally, estimates are standardized as the percent change in the parameters in the meta-regression would come the outcome relative to the mean of the from randomized trials, which are considered the comparison group or sample mean. Estimates in gold standard for program evaluation, but such logarithms are interpreted as percent changes. studies do not exist in this context, hence quasi- Studies for spillovers during booms are experimental studies are included in this analysis. standardized to the number of jobs created, while Meta regression results showing the average effect those during busts are standardized to the percent size for each outcome of interest during resource change of employment or earnings in non- booms and busts are shown in annex table 3.2.3. resource sectors. Estimates that cannot be standardized are excluded from the analysis. is approach results in 33 studies whose estimates are included in the regression analysis. e description of the studies is shown in annex table 3.2.1 and the list of studies is shown in annex table 3.2.2. 116 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 3.2.1 Sample description Panel A. Description of screened and analyzed studies Number of studies Number of studies screened Included in analysis Total 43 33 Region/Country Canada 1 1 United States 28 24 Latin America and the Caribbean 6 2 Europe and Central Asia 3 1 East Asia and the Pacific 4 2 South Asia 2 1 Sub-Saharan Africa 4 1 Outcomes (not mutually exclusive) Employment 27 23 Earnings 29 27 Non-resource sector spillover 14 13 Method (not mutually exclusive) Difference-in-differences 16 14 Instrumental variable 12 10 Fixed effects 6 4 Other 10 8 Panel B. Description of analyzed estimates Number of studies Number of estimates Outcomes (not mutually exclusive) Employment 23 53 Earnings 27 70 Spillover 13 65 Sources: Aragón, Chuhan-Pole, and Land (2015; Marchand and Weber (2018), and related studies. Note: Other quantitative methods include propensity score matching and time series analysis. Employment is total employment in the labor market considered by each study. Earnings include earnings per worker, family earnings, wage and salary income, GDP per capita, total wages, annual pay, household income, median income, median earnings, and per capita expenditure. Spillovers to non-resource sectors include all other sectors, manufacturing, transport, construction, education and health, government, retail, accommodation, other services, agriculture, and all services. Some papers analyzed more than one country or region, therefore, the counts of papers by region are not mutually exclusive. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 117 ANNEX TABLE 3.2.2 Studies of the effects of natural resource booms and busts on the labor market Reference Region/Country Boom or Bust Outcome(s) Agerton et al. (2017) United States Boom Employment Allcott and Keniston (2014) United States Boom Earnings, employment, spillover Latin America and the Aragón and Rud (2013) Caribbean Boom Employment, spillover Aragón, Rud, and Toews (2018) Europe and Central Asia Bust Earnings Asher and Novosad (2014) South Asia Boom Employment, spillover Latin America and the Atienza, Lufin, and Soto (2021) Boom Sector linkages Caribbean Bartik et al. (2019) United States Boom Earnings, employment Baum and Benshaul-Tolonen (2019) Sub-Saharan Africa Boom Employment Bazillier and Girard (2020) Sub-Saharan Africa Boom Earnings Betz et al. (2015) United States Both Earnings, spillover Black, McKinnish, and Sanders (2005) United States Both Employment, spillover Brown (2014) United States Boom Earnings, employment Brown (2015) United States Boom Employment Latin America and the Caselli and Michaels (2013) Boom Earnings Caribbean Douglas and Walker (2017) United States Boom Earnings Fetzer (2014) United States Boom Earnings, employment, spillover Feyrer, Mansur, and Sacerdote (2017) United States Boom Earnings, employment Fleming and Measham (2015) East Asia and Pacific Boom Earnings, employment Latin America and the Gradstein and Klemp (2020) Boom Earnings Caribbean Haggerty et al. (2014) United States Boom Earnings Jacobsen (2018) United States Boom Earnings, employment Jacobsen and Parker (2016) United States Both Earnings, employment James and Aadland (2011) United States Boom Earnings James (2015) United States Boom Earnings James and James (2011) United States Boom None (education) Komarek (2016) United States Boom Earnings, employment, spillover Kotsadam and Tolonen (2016) Sub-Saharan Africa Both Employment Krause (2023) United States Bust Earnings, employment, spillover Lee (2015) United States Boom Earnings, employment Latin America and the Loayza, Mier y Teran, and Rigolini (2013) Boom Earnings Caribbean Maniloff and Mastromonaco (2017) United States Boom Earnings, employment Marchand (2012) Canada Both Earnings, employment Markandya et al. (2016) Europe and Central Asia Boom Employment Michaels (2011) United States Boom Earnings Munasib and Rickman (2015) United States Boom Employment, spillover Papyrakis and Gerlagh (2004) United States Boom Earnings 118 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 ANNEX TABLE 3.2.2 continued Reference Region/Country Boom or Bust Outcome(s) Paredes, Komarek, and Loveridge (2015) United States Boom Earnings, employment Pelzl and Poelhekke (2021) East Asia and Pacific Boom Earnings, employment Europe and Central Asia, Latin America and the Caribbean, von der Goltz and Barnwal (2019) Middle East and North Africa, Boom Wealth South Asia, and Sub-Saharan Africa Weber (2012) United States Boom Earnings Weber (2014) United States Boom Earnings, employment Weinstein (2014) United States Boom Earnings, employment Wrenn, Kelsey, and Jaenicke (2015) United States Boom Employment Zuo and Jack (2016) East Asia and Pacific countries Boom Earnings Note: Some papers cover multiple countries and multiple regions. The East Asia and Pacific countries included in the papers screened are: Australia, Cambodia, China, Indonesia, and the Philippines. The Latin America and Caribbean countries included are: Bolivia, Brazil, Chile, Colombia, Dominican Republic, Guyana, Haiti, and Peru. The South Asian countries included are: Bangladesh, India, and Nepal. The Sub-Saharan Africa countries included are: Angola; Benin; Burkina Faso; Burundi; Cameroon; Central African Republic; Congo, Dem. Rep.; Cote d'Ivoire; Ethiopia; Ghana; Guinea; Kenya; Lesotho; Liberia; Madagascar; Malawi; Mali; Mozambique; Namibia; Niger; Nigeria; Rwanda; Senegal; Sierra Leone; Swaziland; Tanzania; Togo; Uganda; Zambia; and Zimbabwe. The Middle East and North African countries are: Egypt, Arab Rep.; Jordan; and Morocco. The Europe and Central Asia countries/regions included are Albania, the European Union, Moldova, and the United Kingdom. ANNEX TABLE 3.2.3 Meta regression results Boom: Boom: Bust: Boom: spillovers to spillovers to spillovers to Boom: Boom: spillovers to Bust: Bust: manufacturing, all non- all non- employment earnings retail and employment earnings transport, and resource resource services construction sectors sectors Overall effect size 5.194 0.480 0.200 0.207 0.166 -1.325 -2.151 0.003 (0.692) (0.105) (0.079) (0.092) (0.046) (0.349) (0.508) (0.007) No. of estimates 31 51 18 16 51 14 18 25 Note: Estimated effects are based on random effects meta regressions for each outcome of interest. Standard errors in parentheses. Each study’s effect size is standardized to percentage changes. Log changes are interpreted as percentage changes. Employment is total employment in the labor market considered by each study (columns 1, 6). Earnings (columns 2, 7) include earnings per worker, family earnings, wage and salary income, GDP per capita, total wages, annual pay, household income, median income, median earnings, and per capita expenditure. Spillovers to non-resource sectors include all other sectors, manufacturing, transport, construction, education and health, government, retail, accommodation, other services, agriculture, and all services. Number of estimates refers to the number of individual results included in the meta regression to compute the overall effect size. Spillovers in columns 3-5 are the number of jobs created in non-resource sectors, spillovers in column 8 are standardized to the percentage of jobs created in non-resource sectors. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 119 References Working Paper 29401, National Bureau of Economic Research, Cambridge, MA. Agerton, M., P. R. Hartley, K. B. Medlock, and T. Autor, D. H., D. Dorn, G. H. Hanson, and J. Song. Temzelides. 2017. “Employment Impacts of 2014. “Trade Adjustment: Worker-Level Evidence.” Upstream Oil and Gas Investment in the United The Quarterly Journal of Economics 129 (4): 1799– States.” Energy Economics 62 (February): 171–80. 860. Aghion, P., R. Burgess, S. J. Redding, and F. Bali Swain, R., A. Karimu, and E. Gråd. 2022. Zilibotti. 2008. “The Unequal Effects of “Sustainable Development, Renewable Energy Liberalization: Evidence from Dismantling the Transformation and Employment Impact in the License Raj in India.” American Economic Review 98 EU.” International Journal of Sustainable (4): 1397–1412. Development & World Ecology 29 (8): 695–708. Allcott, H., and D. Keniston. 2014. “Dutch Disease Banga, R. 2005. “Impact of Liberalisation on Wages or Agglomeration? The Local Economic Effects of and Employment in Indian Manufacturing Natural Resource Booms in Modern America.” Industries.” Working Paper 153, Indian Council for NBER Working Paper 20508, National Bureau of Research on International Economic Relations. Economic Research, Cambridge, MA. Bartik, A. W., J. Currie, M. Greenstone, and C.R. Aragón, F. M., P. Chuhan-Pole, and B. C. Land. Knittel. 2019. “The Local Economic and Welfare 2015. “The Local Economic Impacts of Resource Consequences of Hydraulic Fracturing.” American Abundance: What Have We Learned?” Policy Economic Journal: Applied Economics 11 (4): 105–55. Research Working Paper 7263, World Bank, Baum, S., and A. Benshaul-Tolonen. 2019. Washington, DC. “Structural Transformation, Extractive Industries Aragón, F. M., and J. P. Rud. 2013. “Natural and Gender Equality.” OxCarre Working Paper 211, Resources and Local Communities: Evidence from a Oxford Centre for the Analysis of Resource Rich Peruvian Gold Mine.” American Economic Journal: Economies, University of Oxford, U.K. Economic Policy 5 (2): 1–25. Bazillier, R., and V. Girard. 2020. “The Gold Digger Aragón, F. M., J. P. Rud, and G. Toews. 2018. and the Machine. Evidence on the Distributive “Resource Shocks, Employment, and Gender: Effect of the Artisanal and Industrial Gold Rushes in Evidence from the Collapse of the UK Coal Burkina Faso.” Journal of Development Economics 143 Industry.” Labour Economics 52 (C): 54–67. (March): 102411. Asher, S., and P. Novosad. 2014. “Digging for Beatty, C., S. Fothergill, and R. Powell. 2007. Development: Mining Booms and Local Economic “Twenty Years on: Has the Economy of the UK Development in India.” In Technical Report. Coalfields Recovered?” Environment and Planning A Working Paper, Oxford University. 39 (7): 1654–75. Asian Development Bank. 2013. Asian Development Betz, M. R., M. D. Partridge, M. Farren, and L. Outlook 2013: Asia’s Energy Challenge. Mandaluyong Lobao. 2015. “Coal Mining, Economic City, Philippines: Asian Development Bank. Development, and the Natural Resources Curse.” Energy Economics 50: 105–16. Atienza, M., M. Lufin, and J. Soto. 2021. “Mining Birol, F., and A. Kant. 2022. “India’s Clean Energy Linkages in the Chilean Copper Supply Network Transition Is Rapidly Underway, Benefiting the and Regional Economic Development.” Resources Entire World – Analysis.” International Energy Policy 70 (March): 101154. Agency, Paris. Autor, D. H., D. Dorn, and G. H. Hanson. 2013. Black, D. A., T. G. McKinnish, and S. G. Sanders. “The China Syndrome: Local Labor Market Effects 2003. “Does the Availability of High-Wage Jobs for of Import Competition in the United States.” Low-Skilled Men Affect Welfare Expenditures? American Economic Review 103 (6): 2121–68. Evidence from Shocks to the Steel and Coal Autor, D., D. Dorn, and G. H. Hanson. 2021. “On Industries.” Journal of Public Economics 87 (9–10): the Persistence of the China Shock.” NBER 1921–42. 120 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Black, D. A., T. G. McKinnish, and S.G. Sanders. Croke, K., J. H. Hicks, E. Hsu, M. Kremer, and E. 2005. “Tight Labor Markets and the Demand for Miguel. 2016. “Does Mass Deworming Affect Child Education: Evidence from the Coal Boom and Nutrition? Meta-Analysis, Cost-Effectiveness, and Bust.” ILR Review 59 (1): 3–16. Statistical Power.” NBER Working Paper 22382, National Bureau of Economic Research, Cambridge, Blinder, A. S. 1973. “Wage Discrimination: MA. Reduced Form and Structural Estimates.” Journal of Human Resources 8 (4): 436–55. Curtis, E. M., and I. Marinescu. 2022. “Green Energy Jobs in the US: What Are They and Where Blonz, J., B. R. Tran, and E. E. Troland. 2023. “The Are They?” NBER Working Paper 30332, National Canary in the Coal Decline: Appalachian Household Bureau of Economic Research, Cambridge, MA. Finance and the Transition from Fossil Fuels.” NBER Working Paper 31072, National Bureau of Cust, J., and S. Poelhekke. 2015. “The Local Economic Research, Cambridge, MA. Economic Impacts of Natural Resource Extraction.” Annual Review of Resource Economics 7 (1): 251–68. Boubakri, N., and J. C. Cosset. 1998. “The Financial and Operating Performance of Newly DerSimonian, R., and N. Laird. 1986. “Meta- Privatized Firms: Evidence from Developing Analysis in Clinical Trials.” Controlled Clinical Trials Countries.” The Journal of Finance 53 (3): 1081– 7 (3): 177–88. 110. Douglas, S., and A. Walker. 2017. “Coal Mining Boubakri, and J. C. Cosset. 2002. “The Financial and the Resource Curse in the Eastern United and Operating Performance of Newly Privatized States.” Journal of Regional Science 57 (4): 568–90. Firms: Evidence from Developing Countries.” The Journal of Finance 53 (3): 1081-110. Ember. 2022. “Electricity Generation by Source.” Ember, London, U.K. Brown, J. 2014. “Production of Natural Gas from Shale in Local Economies: A Resource Blessing or Epifani, P. 2003. “Trade Liberalization, Firm Curse?” Economic Review QI: 1–29. Performance, and Labor Market Outcomes in the Developing World: What Can We Learn from Brown, J. 2015. “The Response of Employment to Micro-Level Data?” Policy Research Working Paper Changes in Oil and Gas Exploration and Drilling.” 3063, World Bank, Washington, DC. Economic Review QII: 57–81. Fabregas, R., M. Kremer, M. Lowes, R. On, and G. Caselli, F., and G. Michaels. 2013. “Do Oil Zane. 2019. “SMS-Extension and Farmer Behavior: Windfalls Improve Living Standards? Evidence from Lessons from Six RCTs in East Africa.” Working Brazil.” American Economic Journal: Applied Paper, Agricultural Technology Adoption Initiative Economics 5 (1): 208–38. (ATAI) and University of Texas at Austin. Cheng, Q., and A. Ng. 2023. “Achieving Stability Fabregas, R., M. Kremer, and F. Schilbach. 2019. and Prosperity: The Chinese Way.” Humanities and “Realizing the Potential of Digital Development: Social Sciences Communications 10 (1): 1–15. The Case of Agricultural Advice.” Science 366 (6471). Christiaensen, L., and C. Ferré. 2020. “Just Coal Transition in Western Macedonia, Greece: Insights Fetzer, T. 2014. “Fracking Growth.” CEPDP1278. from the Labor Market.” Jobs Working Paper 54, Centre for Economic Performance, London School World Bank, Washington, DC. of Economics, London. Consoli, D., G. Marin, A. Marzucchi, and F. Vona. Feyrer, J., E. T. Mansur, and B. Sacerdote. 2017. 2016. “Do Green Jobs Differ from Non-Green Jobs “Geographic Dispersion of Economic Shocks: in Terms of Skills and Human Capital?” Research Evidence from the Fracking Revolution.” American Policy 45 (5): 1046–60. Economic Review 107 (4): 1313–34. Crépon, B., and G. J. van den Berg. 2016. “Active Fleming, D. A., and T. G. Measham. 2015. “Local Labor Market Policies.” Annual Review of Economics Economic Impacts of an Unconventional Energy 8 (October): 521–46. Boom: The Coal Seam Gas Industry in Australia.” SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 121 Australian Journal of Agricultural and Resource IMF (International Monetary Fund). 2022. World Economics 59 (1): 78–94. Economic Outlook: War Sets Back the Global Recovery. Washington, DC: International Monetary Fund. Füllemann, Y., V. Moreau, M. Vielle, and F. Vuille. 2020. “Hire Fast, Fire Slow: The Employment IRENA (International Renewable Energy Agency). Benefits of Energy Transitions.” Economic Systems 2022a. “Bhutan - Energy Profile.” International Research 32 (2): 202–20. Renewable Energy Agency, Masdar City, United Arab Emirates. Garrett-Peltier, H. 2017. “Green versus Brown: Comparing the Employment Impacts of Energy IRENA (International Renewable Energy Agency). Efficiency, Renewable Energy, and Fossil Fuels 2022b. “Nepal - Energy Profile.” International Using an Input-Output Model.” Economic Modelling Renewable Energy Agency, Masdar City, United 61 (February): 439–47. Arab Emirates. Gong, Y., S. Maioli, and H. Görg. 2007. IRENA (International Renewable Energy Agency). “Employment Effects of Privatisation and Foreign 2022c. “Renewable Energy and Jobs: Annual Review Acquisition of Chinese State-Owned Enterprises.” 2022.” International Renewable Energy Agency, International Journal of the Economics of Business 14 Masdar City, United Arab Emirates. (2): 197–214. IRENA (International Renewable Energy Agency). Gradstein, M., and M. Klemp. 2020. “Natural 2023. “Global Landscape of Renewable Energy Resource Access and Local Economic Growth.” Finance 2023.” International Renewable Energy European Economic Review 127 (August): 103441. Agency, Masdar City, United Arab Emirates. Jacobsen, G. D. 2019. “Who Wins in an Energy Granata, J., and J. Posadas. 2022. “Which Jobs Are Boom? Evidence from Wage Rates and Housing.” Green?” World Bank. Economic Inquiry 57 (1): 9–32. Haggerty, J., P. H. Gude, M. Delorey, and R. Jacobsen, G. D., and D. P. Parker. 2016. “The Rasker. 2014. “Long-Term Effects of Income Economic Aftermath of Resource Booms: Evidence Specialization in Oil and Gas Extraction: The U.S. from Boomtowns in the American West.” The West, 1980–2011.” Energy Economics 45: 186–95. Economic Journal 126 (593): 1092–128. Hanson, G. H. 2023. “Local Labor Market Impacts Jacobsen, G. D., D. P. Parker, and J. B. Winikoff. of the Energy Transition: Prospects and Policies.” 2023. “Are Resource Booms a Blessing or a Curse? NBER Working Paper 30871, National Bureau of Evidence from People (Not Places).” Journal of Economic Research, Cambridge, MA. Human Resources 58 (2): 393–420. Hasan, R., D. Mitra, and K. V. Ramaswamy. 2007. James, A. 2015. “Is Education Really Underfunded “Trade Reforms, Labor Regulations, And Labor- in Resource-Rich Economies? Evidence from a Panel Demand Elasticities: Empirical Evidence from of U.S. States.” Working Paper 2015–01, India.” The Review of Economics and Statistics 83 (3): Department of Economics, University of Alaska, 466–81. Anchorage. IEA (International Energy Agency). 2021a. James, A., and D. Aadland. 2011. “The Curse of Hydropower Special Market Report: Analysis and Natural Resources: An Empirical Investigation of US Forecast to 2030. OECD Publishing. Counties.” Resource and Energy Economics 33 (2): 440–53. IEA (International Energy Agency). 2021b. “Investment Data Explorer – Data Tools.” IEA. James, A. G., and R. G. James. 2011. “Do Resource 2021. Dependent Regions Grow Slower than They Should?” Economics Letters 111 (3): 194–96. ILO (International Labour Organization). 2008. “ISCO - International Standard Classification of Jofre-Monseny, J., M. Sánchez-Vidal, and E. Occupations.” International Labour Organization, Viladecans-Marsal. 2018. “Big Plant Closures and Geneva. Switzerland. Local Employment.” Journal of Economic Geography 18 (1): 163–86. 122 CHAPTER 3 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Komarek, T. M. 2016. “Labor Market Dynamics Markandya, A., I. Arto, M. González-Eguino, and and the Unconventional Natural Gas Boom: M.V. Román. 2016. “Towards a Green Energy Evidence from the Marcellus Region.” Resource and Economy? Tracking the Employment Effects of Low Energy Economics 45 (August): 1–17. -Carbon Technologies in the European Union.” Applied Energy 179 (October): 1342–50. Kotsadam, A., and A. Tolonen. 2016. “African Mining, Gender, and Local Employment.” World McKenzie, D. 2017. “How Effective Are Active Development 83 (July): 325–39. Labor Market Policies in Developing Countries? A Critical Review of Recent Evidence.” The World Krause, E. 2022. “Job Loss and Selective Migration Bank Research Observer 32 (2): 127–54. in Distressed Communities: Evidence from Appalachia’s Coal Country.” Working paper, Meyer, I., and M. Sommer. 2014. “Employment Harvard University, Cambridge, MA. Effects of Renewable Energy Supply–A Meta Analysis. WWWforEurope Policy Paper 12,” WIFO Kumar, U., and P. Mishra. 2008. “Trade Studies, Austrian Institute of Economic Research, Liberalization and Wage Inequality: Evidence from India.” Review of Development Economics 12 (2): 291 Vienna, Austria. –311. Michaels, G. 2011. “The Long Term Consequences Kuriyama, A., and N. Abe. 2021. “Decarbonisation of Resource-Based Specialisation.” The Economic of the Power Sector to Engender a ‘Just Transition’ Journal 121 (551): 31–57. in Japan: Quantifying Local Employment Impacts.” Munasib, A., and D. S. Rickman. 2015. “Regional Renewable and Sustainable Energy Reviews 137 (1): Economic Impacts of the Shale Gas and Tight Oil 110610. Boom: A Synthetic Control Analysis.” Regional Lee, J. 2015. “The Regional Economic Impact of Oil Science and Urban Economics 50 (January) 1-17. and Gas Extraction in Texas.” Energy Policy 87 Oaxaca, R. 1973. “Male-Female Wage Differentials (December): 60-71. in Urban Labor Markets.” International Economic Li, Z., and T. Yamada. 2015. “Political and Review 14 (3): 693–709. Economic Incentives of Government in Partial Oei, P.-Y., H. Brauers, and P. Herpich. 2020. Privatization.” Journal of Corporate Finance 32 “Lessons from Germany’s Hard Coal Mining Phase- (June): 169–89. out: Policies and Transition from 1950 to 2018.” Loayza, N., A. Mier y Teran, and J. Rigolini. 2013. Climate Policy 20 (8): 963–79. “Poverty, Inequality, and the Local Natural Resource Ohnsorge, F., and S. Yu, eds. 2021. The Long Curse.” Policy Research Working Paper 6366, Shadow of Informality: Challenges and Policies. World Bank, Washington, Dc. Washington, DC: World Bank. Lobao, L., M. Partridge, O. Hean, P. Kelly, S. Chung, and E. N. Ruppert Bulmer. 2021. Papyrakis, E., and R. Gerlagh. 2004. “Resource- “Socioeconomic Transition in the Appalachia Coal Abundance and Economic Growth in the U.S.” Region: Some Factors of Success.” World Bank, Working Papers 2004.62, Fondazione Eni Enrico Washington, DC. Mattei, Milan, Italy. Maniloff, P., and R. Mastromonaco. 2017. “The Paredes, D., T. Komarek, and S. Loveridge. 2015. Local Employment Impacts of Fracking: A National “Income and Employment Effects of Shale Gas Study.” Resource and Energy Economics 49 (August): Extraction Windfalls: Evidence from the Marcellus 62–85. Region.” Energy Economics 47 (January): 112–20. Marchand, J. 2012. “Local Labor Market Impacts of Parkland Institute. 2019. “Alberta’s Coal Phase-Out: Energy Boom-Bust-Boom in Western Canada.” A Just Transition?” Parkland Institute, University of Journal of Urban Economics 71 (1): 165–74. Alberta, Edmonton, Canada. Marchand, J., and J. Weber. 2018. “Local Labor Pelzl, P., and S. Poelhekke. 2021. “Good Mine, Bad Markets and Natural Resources: A Synthesis of the Mine: Natural Resource Heterogeneity and Dutch Literature.” Journal of Economic Surveys 32 (2): 469– Disease in Indonesia.” Journal of International 90. Economics 131 (July): 103457. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 CHAPTER 3 123 Ruppert Bulmer, E., K. Pela, A. Eberhard-Ruiz, and World Bank. 2019. South Asia Economic Focus. J. Montoya. 2021. “Global Perspective on Coal Jobs Making (De) Centralization Work. Washington, DC: and Managing Labor Transition out of Coal.” World World Bank. Bank, Washington, DC. World Bank. 2022a. South Asia Economic Focus. Sofroniou, N., and P. Anderson. 2021. “The Green Coping with Shocks: Migration and the Road to Factor: Unpacking Green Job Growth.” Resilience. Washington, DC: World Bank. International Labour Review 160 (1): 21–41. World Bank. 2022b. Nepal Country Climate and Sun, Q., and W. H. S. Tong. 2003. “China Share Development Report. Washington, DC: World Bank. Issue Privatization: The Extent of Its Success.” Journal of Financial Economics 70 (2): 183–222. World Bank. 2022c. Bangladesh Country Climate and Development Report. Washington, DC: World Bank. Topalova, P. 2010. “Factor Immobility and Regional Impacts of Trade Liberalization: Evidence on World Bank. 2022d. Pakistan Country Climate and Poverty from India.” American Economic Journal: Development Report. Washington, DC: World Bank. Applied Economics 2 (4): 1–41. World Bank. 2023a. “Global Wind Atlas.” Technical von der Goltz, J., and P. Barnwal. 2019. “Mines: University of Denmark (DTU Wind Energy), The Local Wealth and Health Effects of Mineral International Finance Corporation, and World Mining in Developing Countries.” Journal of Bank. https://globalwindatlas.info/en. Development Economics 139 (June): 1–16. World Bank. 2023b. World Development Report Vona, F., G. Marin, D. Consoli, and D. Popp. 2023: Migrants, Refugees, and Societies. Washington, 2015. “Green Skills.” NBER Working Paper 21116, DC: World Bank. National Bureau of Economic Research, Cambridge, Wrenn, D. H., T. W. Kelsey, and E. C. Jaenicke. MA. 2015. “Resident vs. Nonresident Employment Vona, F., G. Marin, D. Consoli, and D. Popp. Associated with Marcellus Shale Development.” 2018. “Environmental Regulation and Green Skills: Agricultural and Resource Economics Review 44 (2): 1– An Empirical Exploration.” Journal of the Association 19. of Environmental and Resource Economists 5 (4): 713– Zhang, H. 2023. “Coal Busts and Urban Recovery: 53. Evidence from China.” Regional Science and Urban Weber, J. G. 2012. “The Effects of a Natural Gas Economics 101 (July): 103921. Boom on Employment and Income in Colorado, Zuo, N., and S. Jack. 2016. “Are Resources Cursed? Texas, and Wyoming.” Energy Economics 34 (5): An Investigation of Chinese Provinces.” Agricultural 1580–88. and Applied Economics Association Annual Weber, J. G. 2014. “A Decade of Natural Gas Meeting, Boston, MA. Development: The Makings of a Resource Curse?” Resource and Energy Economics 37 (August): 168–83. Weinstein, A. L. 2014. “Local Labor Market Restructuring in the Shale Boom.” Journal of Regional Analysis and Policy 44 (1): 71-92. SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SELECTED TOPICS 125 South Asia Development Update: Selected Topics, 2015-23 Growth Fiscal space and disaster resilience Spring 2023, Box 2.3 Rising interest-growth differentials and what it means for developing economies Fall 2022, Box 2.1 Financial markets post-lending support measures Spring 2022, Box 1.3 Shifting gears: Digitization and services-led development Fall 2021, Chapter 3 Digital technologies can also aid agricultural production Fall 2021, Box 3.4 What does a model based on macro trends predict about remittance growth in 2020, and what does it Spring 2021, Box 1.2 miss? How have South Asian women fared during the crisis? Spring 2021, Box 1.3 Without immediate action, learning losses and the resulting economic losses in South Asia could be cata- Spring 2021, Box 2.4 strophic Learning and related income losses due to school closures in South Asia are huge Fall 2020, Box 1.2 Tourism in South Asia has been shattered but there are opportunities Fall 2020, Box 1.3 Assessing India’s economic activity with daily electricity consumption Fall 2020, Box 1.4 Worrying fiscal implications of shuttered tourism in Maldives Fall 2020, Box 1.5 Green and resilient recovery in South Asia Fall 2020, Box 2.2 Early insights from Bangladesh–Informal workers and women are losing livelihoods, and considerable Fall 2020, Box 3.2 uncertainty remains Food price increases need to be addressed with decisive measures Spring 2020, Box 1.2 South Asia Economic Focus forecasting performance Fall 2019, Box 3 Growth expectations from within the region Fall 2019, Box 4 Private cities: Outstanding examples from developing countries and their implications for urban policy Urban Development Series, May 2023 Inequality Distributional impact of high food and energy inflation in South Asia Spring 2023, Box 1.1 Expanding opportunities: A map for equitable growth in South Asia Spring 2023, Chapter 3 Measuring inequality, inequality of opportunity and intergenerational mobility in South Asia Spring 2023, Box 3.1 In South Asia, opportunity gaps in primary education have been shrinking but not at the same pace for Spring 2023, Box 3.2 all countries Are opportunity gaps closing? A stylized version of the opportunity growth incidence curve Spring 2023, Box 3.3 Affirmative action policies in South Asia Spring 2023, Box 3.4 Remittances and the effects on poverty and inequality Fall 2021, Box 1.3 Environment Recruiting firms for the energy transition Fall 2023, Chapter 2 Stranded jobs? The energy transition in South Asia’s labor markets Fall 2023, Chapter 3 Weather extremes and price stability Spring 2023, Box 2.1 Fiscal space and disaster resilience Spring 2023, Box 2.3 The turning point–fossil fuel subsidy reform in South Asia Spring 2023, Box 2.4 The green transition: How will it affect households in South Asia? Fall 2022, Box 2.4 Migration and climate change in South Asia Fall 2022, Box 3.5 126 SELECTED TOPICS SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Environment (continued) How prepared are South Asia's energy firms and workers for the green transition? Spring 2022, Box 2.2 Healthy fiscal balance for a swift recovery: Lessons from natural disasters Fall 2021, Box 2.2 Toward a low carbon future in South Asia Fall 2021, Box 2.3 The “double jeopardy” of fiscal and climate-related risks Spring 2021, Box 2.3 Green and resilient recovery in South Asia Fall 2020, Box 2.2 Striving for clean air: Air pollution and public health in South Asia South Asia Development Matters, July 2023 Glaciers of the Himalayas: Climate change, black carbon, and regional resilience South Asia Development Forum, June 2021 Monetary policy and inflation Distributional impact of high food and energy inflation in South Asia Spring 2023, Box 1.1 Recent changes in exchange rate policy in Bangladesh Spring 2023, Box 1.2 Weather extremes and price stability Spring 2023, Box 2.1 Estimating the spillovers from US monetary policy Spring 2023, Box 2.2 Pass-through of global commodity prices in South Asia Fall 2022, Box 1.1 The dollar is whose problem: Impact of the US dollar dynamics on bilateral trade Fall 2022, Box 1.2 How effective is monetary policy in South Asia? Fall 2022, Box 1.3 Financial markets post-lending support measures Spring 2022, Box 1.3 Food price increases need to be addressed with decisive measures Spring 2020, Box 1.2 The drivers of food price inflation in South Asia Fall 2019, Box 1 Consumer price inflation and food inflation in South Asia Spring 2019, Box 2 Fiscal policy and debt An ounce of prevention, a pound of cure: Averting, and dealing with, debt default Fall 2023, Spotlight Fiscal deteriorations around elections Fall 2023, Box 1.1 The sovereign-bank sector nexus in South Asia Spring 2023, Box 1.3 Fiscal space and disaster resilience Spring 2023, Box 2.3 The turning point–fossil fuel subsidy reform in South Asia Spring 2023, Box 2.4 Crisis in Sri Lanka: Lessons from the Asian financial crisis Fall 2022, Spotlight Rising interest-growth differentials and what it means for developing economies Fall 2022, Box 2.1 Healthy fiscal balance for a swift recovery: Lessons from natural disasters Fall 2021, Box 2.2 Toward a low carbon future in South Asia Fall 2021, Box 2.3 How can South Asia avoid getting caught in a wave of debt? Spring 2021, Box 2.1 What does the economic literature tell us about government spending multipliers in developing Spring 2021, Box 2.2 countries? The “double jeopardy” of fiscal and climate-related risks Spring 2021, Box 2.3 Worrying fiscal implications of shuttered tourism in Maldives Fall 2020, Box 1.5 Fiscal policy should turn countercyclical during this crisis Spring 2020, Box 2.3 Government borrowing crowds out the private sector across the region Spring 2020, Box 3.4 Reducing government ownership has had positive effects in other countries Spring 2020, Box 3.5 SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 SELECTED TOPICS 127 Fiscal policy and debt (continued) Research on oil prices, J-curves, and twin deficits in South Asia Spring 2019, Box 8 Hidden debt: Solutions to avert the next financial crisis in South Asia South Asia Development Matters, June 2021 Trade Pass-through of global commodity prices in South Asia Fall 2022, Box 1.1 The dollar is whose problem: Impact of the US dollar dynamics on bilateral trade Fall 2022, Box 1.2 Where do South Asia's exports stand in 2022? Spring 2022, Box 1.2 An update on trade policy changes affecting South Asia Spring 2019, Box 1 Exports wanted Spring 2019, Chapter 3 Analyzing the current account balance with Vector Autoregressive (VAR) Models Spring 2019, Box 5 A Gravity model to estimate South Asia’s export gaps Spring 2019, Box 6 Constraints to export competitiveness in Pakistan Spring 2019, Box 7 Research on oil prices, J-curves, and twin deficits in South Asia Spring 2019, Box 8 Financial flows The informal foreign exchange market and capital controls: A South Asian tale Spring 2023, Spotlight The sovereign-bank sector nexus in South Asia Spring 2023, Box 1.3 Estimating the spillovers from US monetary policy Spring 2023, Box 2.2 Fintech credits: From competition to collaboration Fall 2022, Box 2.2 Financial markets post-lending support measures Spring 2022, Box 1.3 Central bank digital currency Spring 2022, Box 1.4 What determines domestic market yields? Spring 2022, Box 2.1 Remittances and the effects on poverty and inequality Fall 2021, Box 1.3 What does a model based on macro trends predict about remittance growth in 2020, and what does it miss? Spring 2021, Box 1.2 Public banks: A cursed blessing Spring 2020, Chapter 3 Have public banks hindered subsequent financial development? Spring 2020, Box 3.1 Does the broad public branch network translate into more credit for development targets in Bangladesh? Spring 2020, Box 3.2 In Asia, more public banks are associated with lower interest rate margins Spring 2020, Box 3.3 Reducing government ownership has had positive effects in other countries Spring 2020, Box 3.5 Measurement and significance of remittances Spring 2019, Box 4 Hidden debt: Solutions to avert the next financial crisis in South Asia South Asia Development Matters, June 2021 Labor markets The informal foreign exchange market and capital controls: A South Asian tale Spring 2023, Spotlight Affirmative action policies in South Asia Spring 2023, Box 3.4 (Mis)Measuring migration Fall 2022, Box 3.1 Intraregional migration in South Asia Fall 2022, Box 3.2 128 SELECTED TOPICS SOUTH ASIA DEVELOPMENT UPDATE | OCTOBER 2023 Labor markets (continued) Determinants of economic migration: A framework Fall 2022, Box 3.3 Migration and climate change in South Asia Fall 2022, Box 3.5 Reshaping social norms about gender: A new way forward Spring 2022, Chapter 3 Female labor force participation rates may be affected by a country’s economic structure and by the Spring 2022, Box 3.1 prevalence of norms over women’s employment in specific sectors How have South Asian women fared during the crisis? Spring 2021, Box 1.3 Early insights from Bangladesh–Informal workers and women are losing livelihoods, and considerable Fall 2020, Box 3.2 uncertainty remains Private cities: Outstanding examples from developing countries and their implications for urban policy Urban Development Series, May 2023 Hidden potential: Rethinking informality in South Asia South Asia Development Forum, November 2022 COVID-19 pandemic How is the labor market recovering from the pandemic? Fall 2022, Box 2.3 COVID and migration in South Asia Fall 2022, Chapter 3 Labor market impacts of COVID-19 on the displaced Rohingya population in Cox’s Bazar, Bangladesh Fall 2022, Box 3.4 COVID-19 vaccination and economic activity in South Asia Spring 2022, Box 1.1 Alternative measures of COVID-19 deaths Fall 2021, Box 1.1 Impact of COVID-19 among refugees in South Asian countries Fall 2021, Box 1.2 Rethinking tourism: Seizing on services links post-COVID Fall 2021, Box 3.2 The pandemic has exacerbated the difficulties in measuring GDP in South Asia Spring 2021, Box 1.1 South Asia vaccinates Spring 2021, Chapter 3 How can countries address COVID vaccine hesitancy and increase take-up? Spring 2021, Box 3.1 Methodology for modeling impact of COVID-19 by population groups Spring 2021, Box 3.2 Both the spread of COVID-19 and related containment measures contributed to GDP losses Fall 2020, Box 1.1 The silver lining: Can global value chains thrive in South Asia post-COVID? Fall 2020, Box 2.1 Forecasting COVID caseloads and estimating services activity using the Google mobility index Fall 2020, Box A2.1 The impact of COVID-19 on the informal sector Fall 2020, Chapter 3 How to simulate the impact of the COVID-19 crisis Fall 2020, Box 3.1 Unpacking India’s COVID-19 social assistance package Fall 2020, Box 3.3 Predicting the spread of COVID-19 in South Asia through migration corridors Spring 2020, Box 1.1 Migrant remittances in South Asia may decline during the time of COVID-19 Spring 2020, Box 1.3 Distributional impact of COVID-19. Whose health is affected? Spring 2020, Box 1.4 Identifying the people working in sectors most affected by the COVID-19 crisis Spring 2020, Box 2.2 Note: The South Asia Development Update was called South Asia Economic Focus through April 2023. S outh Asia is expected to grow faster than any other emerging market and developing economy (EMDE) region in 2024–25. However, for all countries, this will represent a slowdown from pre-pandemic averages. Several potential adverse events could derail this outlook, including risks related to fragile fiscal positions. Government debt in South Asia averaged 86 percent of GDP in 2022, above that of any other EMDE region. In some countries, outright defaults have short-circuited growth, while in others, increasing domestic borrowing by governments has driven up interest rates and diverted credit away from the private sector. Elections could add to spending pressures. An urgent policy priority for the region is, therefore, to manage and reduce fiscal risks. Over the longer term, the policy priority is to accelerate growth and job creation in a sustainable manner. The energy transition, away from fossil fuels toward sustainable sources of energy, presents an opportunity for the region to lift productivity, cut pollution, reduce its reliance on fuel imports, and create jobs. South Asia’s energy intensity of output is twice the global average and the region lags in the adoption of advanced energy-efficient technologies. Even fiscally constrained governments can take action to support the energy transition with market-based regulations, information campaigns, broader access to finance, and reliable public power grids. With about 9 percent of the region’s workers employed in pollution-intensive activities, and these workers less educated and more often informally employed than the average worker, the energy transition will create challenging labor market shifts. This calls for measures to boost job creation and facilitate worker mobility, geographically and across sectors.