Official Use Only Report No: ACS8954 SOUTH ASIA The Competitive Advantage of South Asia: New Perspectives on Productivity Volume I June 29, 2016 Trade & Competitiveness Global Practice South Asia Region THE WORLD BANK GROUP Standard Disclaimer: This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank 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. Copyright Statement: The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Acknowledgements This report was prepared by a team led by Gladys Lopez-Acevedo (SARCE), Denis Medvedev (GTCDR), and Vincent Palmade (GTCDR) under the guidance of Esperanza Lasagabaster (Practice Manager, GTCDR) and Martin Rama (Chief Economist, SARCE). Annette Dixon (Vice President, SARVP), Cecile Fruman (Director, GTCDR), and Anabel Gonzalez (Senior Director, GTCDR) linked the team to the World Bank Group’s overall strategy and steered them in that direction. Part 1 of this Volume was authored by Gonzalo Varela and Antonio Martuscelli, with substantial inputs from Apoorva Gupta. Part 2 was authored by Filipe Lage de Sousa and Deeksha Kokas (agglomeration), Michael Ferrantino and Gaurav Nayyar (global value chains), and Xavier Cirera and Ana Cusolito (innovation and firm capabilities, with substantial inputs from Filipe Lage de Sousa). Part 3 was authored by Gladys Lopez-Acevedo, Vincent Palmade, and Dominique van der Mensbrugghe. Bill Shaw streamlined and organized the narrative. Major contributions were received from Alvaro Gonzalez, Atisha Kumar, and Siddharth Sharma, and the team benefitted from the advice by Sebastian Saez. Guoliang Feng, Apoorva Gupta, and Deeksha Kokas provided invaluable research assistance. Isaac Lawson and Tanya Cubbins provided production and logistical support. Sections of the report (in particular Part 3) draw extensively on a recently prepared “Stiches to Riches” report (World Bank, 2016). The peer reviewers were Uri Dadush (Carnegie Endowment for International Peace), Navin Girishankar (GTCDR), Pravin Krishna (Johns Hopkins University), Bill Maloney (GGEVP), and Shubham Chaudhuri (GMFDR). The team is grateful to South Asia country directors and, most importantly, country authorities for their support throughout this endeavor. Contents Acknowledgements ....................................................................................................................................... ii Overview ........................................................................................................................................................ i 1. South Asia’s competitiveness challenge and opportunity ..................................................................... 1 1.1 The region’s competitiveness potential remains largely unrealized ............................................. 1 1.1.1 Multiple pockets of excellence evidence vast untapped potential ........................................ 1 1.1.2 Difficulties in attracting investment and penetrating global markets ................................... 2 1.1.3 Little progress in diversifying the merchandise export basket .............................................. 6 1.1.4 Elusive sophistication and low quality of exports................................................................. 7 1.1.5 Low scores on most competitiveness benchmarks.............................................................. 12 1.2 Improving competitiveness is about raising productivity rather than keeping costs low ........... 14 1.2.1 Macro challenge: contribution of TFP to growth is low and declining............................... 15 1.2.2 Sectoral challenge: slow pace of structural transformation ................................................ 17 1.2.3 Firm challenge: firm growth is low and resources are trapped in small firms .................... 21 1.3 Annex .......................................................................................................................................... 28 2. Productivity performance: firms and linkages .................................................................................... 31 2.1 Business environment challenges continue to weigh on firm performance ................................ 32 2.2 Productivity-boosting agglomeration economies are under-leveraged ....................................... 35 2.2.1 Economic activity in South Asia is highly concentrated..................................................... 35 2.2.2 Agglomeration economies raise firm productivity.............................................................. 37 2.2.3 Resources do not flow easily to more productive firms ...................................................... 41 2.3 Limited success in linking to global value chains ....................................................................... 42 2.3.1 GVC participation supports productivity ............................................................................ 43 2.3.2 South Asia’s success in global and regional GVCs is limited to apparel ........................... 45 2.3.3 Most policy determinants of GVC participation are lacking .............................................. 52 2.4 Firm capabilities are constrained ................................................................................................ 55 2.4.1 Firms lack in managerial quality and do not use resources efficiently ............................... 55 2.4.2 Knowledge and technology adoption are low ..................................................................... 57 2.4.3 Innovation is widespread but novelty is limited ................................................................. 62 2.4.4 Returns to innovation are high ............................................................................................ 67 2.5 Annex .......................................................................................................................................... 70 3. The way forward ................................................................................................................................. 75 3.1 Potential for increased growth through policy reforms ............................................................... 75 3.1.1 Macro benefits: faster exports growth through higher productivity.................................... 75 3.1.2 Sectoral benefits: more jobs, higher earnings, greater inclusion......................................... 77 3.1.3 Firm benefits: greater density of successful firms .............................................................. 81 3.2 Need for greater emphasis on trade policies, spatial policies and firm capabilities .................... 83 3.2.1 Policies to better connect to Global Value Chains .............................................................. 83 3.2.2 Policies to maximize agglomeration benefits ..................................................................... 86 3.2.3 Policies to support innovation and productivity.................................................................. 88 3.3 Annex .......................................................................................................................................... 90 4. References ........................................................................................................................................... 97 Figures Figure 1.1 FDI to GDP ratios ........................................................................................................................ 2 Figure 1.2 Trade to GDP ratios ..................................................................................................................... 3 Figure 1.3 Changes in Export Market Shares by Country – 2000/04 – 2010/14 .......................................... 5 Figure 1.4 Export growth decomposition in South Asia ............................................................................... 7 Figure 1.5 Export sophistication and weighted average income of importers, 2000-2014 ........................... 8 Figure 1.6 Relative Unit Values of Select Export Products by Country and Year – Quality Ladders ......... 9 Figure 1.7 Ease of Doing Business rankings in South Asia ........................................................................ 13 Figure 1.8 Real GDP and GDP per Capita Growth in South Asia .............................................................. 16 Figure 1.9: Growth Accounting – Decomposition of GDP Growth 1990-2014 ......................................... 17 Figure 1.10 Structural transformation in South Asia .................................................................................. 18 Figure 1.11 Employment and value added shares by sector for Bangladesh, India, Pakistan, and Sri Lanka .................................................................................................................................................................... 19 Figure 1.12 Value added shares by sector in Afghanistan, Bhutan, Maldives, and Nepal ......................... 20 Figure 1.13 Size distribution of firms in India, 1990-2010......................................................................... 23 Figure 1.14 Firm size distribution in South Asia and comparator countries............................................... 25 Figure 1.15: Firm employment and export shares by age ........................................................................... 27 Figure 1.1.16: Scatter plot of quartile 1 of employment distribution of firms and income per capita ........ 29 Figure 1.1.17: Scatter plot of quartile 2 (median) of employment distribution of firms and income per capita ........................................................................................................................................................... 29 Figure 1.18: Scatter plot of quartile 3 of employment distribution of firms and income per capita ........... 29 Figure 1.19: Export Orientation Index – Ranking SAR countries and comparators................................... 29 Figure 1.20: Import Orientation Index – Ranking SAR countries and comparators................................... 29 Figure 2.1 Decomposition of misallocation into between- and within-district components....................... 42 Figure 2.3 Structure of the global value chain for apparel .......................................................................... 43 Figure 2.4 GVC participation and firm TFP ............................................................................................... 44 Figure 2.5 GVC participation across regions .............................................................................................. 46 Figure 2.6 South Asia GVC exports by country (share of total exports) .................................................... 47 Figure 2.7 Unit values of India’s auto and electronics exports ................................................................... 49 Figure 2.8 Foreign value added in exports, India and comparators ............................................................ 49 Figure 2.9 Product and market sophistication of South Asia’s textile and apparel exports ........................ 50 Figure 2.10 Regional GVCs in apparel in South Asia ................................................................................ 51 Figure 2.9 Intermediate imports and exports to and from East Asia .......................................................... 52 Figure 2.12 GVC capability gaps vs. ASEAN and SACU ......................................................................... 54 Figure 2.14 Weak performance on management capabilities ..................................................................... 56 Figure 2.15 ICT adoption across countries in South Asia .......................................................................... 58 Figure 2.16 Types of internet use by firms across countries....................................................................... 60 Figure 2.17 Innovation practices in South Asia and comparator countries ................................................ 65 Figure 2.18 In-house vs collaborative innovation ....................................................................................... 66 Figure 3.2 South Asia’s global market share, 2011 and 2030 (under various scenarios) ........................... 76 Figure 3.4 Employment effects of 10 percent rise in China apparel prices, by destination and gender ..... 80 Figure A1 South Asia’s sources of growth, baseline, 2011-2030 ............................................................... 93 Tables Table 1.1 Market Shares in Merchandise Exports ........................................................................................ 4 Table 1.2 Market Shares in Services Exports ............................................................................................... 5 Table 1.3 Apparel competitiveness of South Asian countries, 2012 .......................................................... 14 Table 1.4 Labor productivity and structural transformation ....................................................................... 21 Table 1.5 TFP dispersion (coefficients of variation) by sector and country ............................................... 23 Table 1.6: Distribution of firms by size – South Asia and comparator countries ( percent) ....................... 24 Table 1.7 Sales & Value Added per Worker, India and China, By Firm Size ............................................ 26 Table 2.1 Investment climate constraints in South Asia and comparator countries ................................... 32 Table 2.2 Impact of investment climate on firm performance .................................................................... 33 Table 2.3 Employment in top five districts across South Asia (percent of total employment) ................... 36 Table 2.4 Evolution of spatial concentration of manufacturing in South Asia ........................................... 37 Table 2.5 Estimates of agglomeration economies in India and Bangladesh ............................................... 39 Table 2.6 Estimates of agglomeration economies in Sri Lanka .................................................................. 40 Table 2.7 Agglomeration economies at different quantiles of firm distribution ......................................... 40 Table 2.8 Per capita GVC exports from South Asia (USD) ....................................................................... 47 Table 2.9 GVC capabilities and endowments in South Asia (2010) ........................................................... 53 Table 2.10 Knowledge capital intensity ...................................................................................................... 63 Table 2.11 Types of innovation undertaken (percent of all firms) ............................................................. 66 Table 2.12 Concentration of economic and innovative activities (Herfindahl index) ................................ 67 Table 2.13 Impact of innovation practices on firm productivity................................................................. 69 Table 2.14 Output and factor misallocation in South Asia ......................................................................... 70 Table 2.15 ICT index by country, sector, and size...................................................................................... 71 Table 2.16 Determinants of overall ICT adoption ...................................................................................... 72 Table 2.17 Determinants of R&D adoption and ICT intensity ................................................................... 73 Table 2.18 Determinants of innovation....................................................................................................... 74 Table 3.1 Elasticity of substitution for US and EU apparel imports ........................................................... 78 Table 3.2 Marginal effects of female labor participation with respect to the expected wage ..................... 79 Table 3.3 Apparel demand function estimates ............................................................................................ 94 Table 3.4 Labor demand in textiles and apparel ......................................................................................... 95 Boxes Box 1.1 Barriers to competition and productivity dispersion ..................................................................... 22 Box 2.1 Efforts to improve the investment climate in South Asia .............................................................. 34 Box 2.2 Agglomeration and productivity ................................................................................................... 38 Box 2.4 Measuring GVC participation ....................................................................................................... 46 Box 2.6 Data on ICT adoption in South Asia ............................................................................................. 58 Box 2.7 Determinants of ICT adoption....................................................................................................... 61 Box 2.8 Innovation activities and outputs in the Enterprise Survey ........................................................... 63 Box 2.8 The CDM model............................................................................................................................ 67 Box 3.1 Why focus on women? .................................................................................................................. 79 Overview Which region will become the next global factory? As the work force ages and labor costs rise in China and other East Asian countries, many eyes turn to South Asia. It is a region that is still largely rural, where agriculture accounts for a large share of employment and a substantial fraction of GDP, and it has not been particularly successful in integrating within itself and with the rest of the world. Yet, education levels are on the rise, and more than one million young workers enter the labor market each year—by 2030, 26 percent of the world’s working adults will live in South Asia. This is the region’s greatest opportunity and greatest challenge. What will determine South Asia’s ability to take advantage of the demographic transition to unlock its potential and accelerate growth, create jobs, reduce poverty, and boost shared prosperity? One of the answers lies in improving the region’s competitiveness. While different authors have proposed different definitions for the concept, this report chooses a simple approach by defining competitiveness as productivity (a la Porter, 1990). Despite relatively rapid economic growth, the contribution of total factor productivity (TFP) to the region’s growth has been low and factors subject to diminishing returns – quantity rather than quality of labor and non-ICT capital – have been the main drivers of growth. This calls for greater focus on improving productivity. As the global environment deteriorates, South Asia’s integration remains low The global environment is becoming tougher. The demand for developing countries’ exports is limited by the slow recovery in industrial economies and the impact of declines in commodity prices on resource-rich economies, while the benefits to many commodity importers have been eroded by declining remittances. New mega-regional trade agreements (e.g., the Trans-Pacific Partnership and the Trans-Atlantic Trade and Investment Partnership) may lead to trade and investment diversion away from non-members. Against this background, it has become even more urgent for countries in South Asia to make overdue investments in raising productivity to avoid falling further behind comparator countries in the global marketplace. South Asia’s intra-regional and global ties are relatively weak. Both trade-to-GDP and FDI-to-GDP ratios in the region are well below competitors’ ratios. From1990-2014, the region received, on average, between 2.2 and 2.8 percentage points of GDP less FDI inflows than countries in East Asia. Lack of investment integration in South Asia is particularly evident when considering intra-regional flows. Trade integration is also low. Over 1990-2014, South Asia’s average ratio of exports to GDP varied between 17 and 21 percentage points below East Asia, and the average ratio of imports to GDP was 21-22 percentage points lower. The region has made little progress in diversifying its exports and moving up the value chain. Although South Asia has had some success in penetrating new markets, almost 80 percent of the region’s export growth from 2001 to 2013 came from the sale of the same goods to the same destinations, and the remaining 20 percent came from selling the same products to new markets. Exports remain highly concentrated in textiles and apparel in Bangladesh, Afghanistan, Nepal, Pakistan and Sri Lanka, in i minerals in Bhutan, and in animal and vegetable products in Afghanistan and Maldives. Overall, the region’s export basket does not reflect substantial transformation of production structures or innovative activities. While the sophistication of exports has increased in India, it has remained low in the rest of South Asia and quality (as measured by the prices its products fetch in international markets) has generally remained low and has declined for some countries. Productivity growth has been slow but considerable potential exists for an acceleration GDP in South Asia more than quadrupled from 1990 to 2014, and most countries enjoyed rapid growth in output and per capita income. However, the contribution of total factor productivity to GDP growth has been mixed.1 In India and Pakistan, which achieved high rates of GDP growth, increases in TFP made an important contribution to growth, although this contribution appears to have declined substantially since 2011 and 2006, respectively. In Sri Lanka, while the contribution of TFP picked up in 2014, it has declined from the high level before 2009. In Bangladesh, TFP has played a negligible role in GDP growth during the entire period of analysis. Most of the growth in the region is generated by increases in labor (rather than higher-quality labor) and non-ICT investment, both factors subject to diminishing returns. An important driver of productivity growth in developing countries is the movement of resources from agriculture to higher-productivity manufacturing and services. Between 1960 and 2013, the share of agriculture in South Asia’s GDP decreased from 44 to 19 percent, while the share of industry rose from 18 to 29 percent. Similar to patterns observed in OECD economies, labor productivity differentials between agriculture and more modern activities play an important role in explaining the movement of labor across sectors – with the sensitivity of labor movement to productivity much higher in South Asia than in high income countries. However, the movement of labor from agriculture to industry and services in South Asia has not been rapid enough to substantially reduce the large differences in productivity across sectors. In other words, the region has a significant untapped potential (e.g., compared to OECD economies) to reap productivity gains by further reallocation of labor from lower to higher productivity activities. The dispersion of productivity levels across firms in South Asia is high, indicating large potential for improved efficiency. Productivity gains of as much as 60 percent could be realized in India by reducing distortions in resource allocation from current levels to those of the United States. The size distribution of firms in the region is particularly biased towards small firms, which struggle to grow and tend to be less productive than large firms. Firms aged 25 years or more in Bangladesh, Bhutan, India, and Sri Lanka are 50 – 90 percent larger than new firms, while in China, Indonesia, and Vietnam similar firms are larger by a factor of 2-5. The predominance of small firms accounts for a large share of the productivity gap between South Asia and high-income economies. 1 The analysis is limited to the four countries with the data required to decompose GDP growth into its components, including changes in the quantity of labor, the quality of labor, ICT capital, non-ICT capital, and TFP ii Barriers to the growth of firms can likely be found in policies. Across the region, licensing and size restrictions (which have declined in importance but still exist), labor regulations that increase the cost of hiring and firing, financial sector regulations that favor small enterprises, and inadequate bankruptcy laws may have limited the ability of efficient plants to grow and enabled inefficient plants to survive. Taxes or labor costs that affect larger firms more than smaller firms may reduce the return on investment in large firms. Impediments to reaching foreign markets, both from trade policy and high costs of logistics, can also impede expansion. The business environment in South Asia is poor On average, countries in South Asia score poorly on major indices used globally to capture key aspects of competitiveness, such as the Global Competitiveness Index (GCI) published by the World Economic Forum and the World Bank’s Doing Business report. In the most recent (2015-2016) GCI rankings, India is the only South Asian country in the top half of nearly 140 countries. In the World Bank’s 2016 Doing Business report, all the South Asian economies, with the exception of Bhutan, are ranked in the bottom half. In a 2006 Investment Climate Assessment of the region, the World Bank argued that South Asian countries under-perform comparators on many investment climate dimensions, including infrastructure and electricity supply, access to finance, employee skills, and corruption. Similar results emerge from the most recent round of Enterprise Surveys, where an average firm in South Asia consistently ranks each investment climate constraint as more binding than does an average firm in China or Vietnam. While performance varies substantially across countries and indicators – pointing to significant potential for improvement by leveraging best practices from within the region – the overall gap puts South Asia’s firms at a clear disadvantage vis-à-vis select comparators in other parts of the world. Some evidence suggests that investment climate deficiencies are particularly binding on firms that would grow more rapidly and create more jobs in the absence of distortions South Asian firms face limits in benefiting from agglomeration Economic activity in South Asia is highly concentrated. Locational Gini coefficients for manufacturing are relatively high, and in most countries a small number of districts account for a large share of economic activity (in India, the five largest districts account for 18 percent of total employment). However, the degree of geographic concentration of manufacturing activities in South Asia has not changed substantially in the last two decades. For example, the share of the top five districts in total employment has changed little over time, although which districts are in the top five has changed. This indicates that more productive locations have generally not been successful in attracting additional resources at the expense of less productive locations, although congestion in the major economic centers has not reached sufficiently high levels to push a substantial share of economic activity out to the periphery. In the three South Asian countries with adequate data, district or state borders tend to be “thick” in the sense that impediments to efficient allocation of resources between districts are stronger than distortions within districts. Agglomeration economies – the benefits that accrue to firms and workers from locating close together in cities or clusters – matter for firm productivity: a 10 percent increase in district employment leads to a iii 0.2-0.9 percent increase in TFP of the district’s firms. The effect operates primarily through two channels: localization (i.e., firms in the same industry locating close together to benefit from, for example, a specialized labor pool) and urbanization (i.e., firms from different industries locating close together to benefit from a diverse supplier network, common infrastructure, or a large number of workers). Unlike in high income countries, firms in South Asia benefit more from urbanization than localization economies. These results, robust to controls for geographical location, industrial diversity, and the degree of competition, suggest that cities, catering to wide-ranging industries, may currently be more effective vehicles of supporting firm productivity in South Asia than clusters, which cater to a specific sector – although the two are not mutually exclusive. Nevertheless, regional firms struggle to take advantage of the benefits of urbanization. When large-scale solutions are difficult or costly, improved infrastructure could be delivered through industrial zones or clusters. While a number of traditional approaches to industrial zones in South Asia have not delivered the expected benefits, there are encouraging examples of new approaches from within and outside the region, such as India’s SITP model and China’s plug-and-play industrial zones. Location of clusters often makes all the difference, and countries in the region could make further efforts to identify and develop industrial areas close to ports, resolve pending issues in existing industrial zones, and ensure provisions for worker housing. Providing access to R&D and testing facilities, waste dumping, and recycling facilities would make these zones more attractive to SMEs. The evidence given above on the degree of misallocation of labor and capital, and the importance of agglomeration economies for firm productivity, should focus policy makers’ attention on removing constraints on the growth of firms. The removal of policy-induced distortions that limit the flexibility of labor and capital markets could enable more productive firms to grow. In particular, policies to increase the flexibility of labor markets, especially for women – who face particularly high discrimination in South Asia’s labor markets – are likely to substantially reduce misallocation of labor and improve productivity. Policies directed at improving urban governance and bridging the region’s infrastructure gap will ensure that firms and workers will be matched more easily. Achieving this will require tackling congestion issues head-on. In particular, investments in roads and public transit, provision of quality affordable housing and other basic infrastructure services, and reducing the negative social impact of agglomeration (e.g., crime) should be high on the policymakers’ agenda. There is considerable scope for increasing productivity through participation in GVCs South Asia has the second-highest level of GVC exports in total exports among developing regions, but South Asian countries other than India have negligible exports of GVC products other than apparel. Bangladesh has one of the highest GVC participation rates in the world, although this reflects the fact that Bangladesh exports little else besides garments. India’s participation in GVCs is low, because it has a more diversified export basket. There is also substantial variation in terms of where most firms in each country locate along the value chains: firms in Pakistan tend to be more upstream, while Sri Lankan and Bangladeshi firms are much further downstream. iv Value chains tend to cluster on a regional basis in light of transport and other transactions costs, as well as the need for timely delivery. Bangladesh and Sri Lanka have the highest share of final apparel goods (86 and 44 percent of apparel exports) in the region, and source many apparel inputs from Pakistan and India who focus relatively less on final products (18 and 6 percent of apparel exports). In 2013, two- thirds of India’s exports of knit and crochet fabric were destined for Sri Lanka and Bangladesh, while nearly half of Pakistan’s exports of woven cotton denim were destined for Bangladesh and Sri Lanka There is also an emerging “East Asia/South Asia” regional value chain, especially in intermediate apparel: 70 percent of South Asia’s imported apparel inputs come from East Asia and, together with intra-South Asian trade, account for 94 percent of the regions apparel inputs. South Asian GVC activity is more integrated with East Asia than any other region in the world, except East Asia itself. Final apparel producers in Sri Lanka and Pakistan have been more successful at penetrating higher- income markets than firms in Bangladesh and India, while firms in Pakistan and Bangladesh have shown greater ability to penetrate high-income markets in intermediate apparel. Overall market sophistication, however, declined between 2000 and 2010 in all four countries, indicating either increased sales to middle-income markets or more intense competition in high-income markets, or both. Participation in GVCs, and exposure to international markets more generally, is associated with higher levels of firm productivity in South Asia. Access to foreign markets – either through trade or licensing foreign technology – is associated with stronger outcomes in terms of ICT adoption and innovation, and these in turn have a robust positive relationship with firm-level productivity While the availability of qualified local suppliers is crucial for the expansion of GVCs into new geographic territories, greater exposure to international trade makes firms more viable participants in GVCs, which in turn can further enhance productivity in a virtuous cycle. While more productive firms may self-select to join GVCs, evidence suggests that GVC participation and deeper global integration more generally have positive productivity impacts on firms. South Asia lags on capabilities that matter for GVC participation. Countries in South Asia are, on average, more wage competitive and closer to markets than ASEAN and SACU countries, but compare unfavorably with regard to policy variables such as human capital, institutions, logistics and trade barriers on imports of intermediate inputs. Some countries in the region still have relatively high tariffs on GVC (intermediate and final) goods, and in addition impose high para-tariffs and non-tariff trade barriers. There is also substantial room for improvement with regard to trade facilitation, while the ability to access imported inputs in a timely manner are particularly important for GVC sectors. Therefore, policy actions such as facilitating imports for exporters (e.g., through better functioning duty drawback schemes), reducing average rates of protection and harmonizing tariff schedules across intermediate and final goods, and improving trade logistics to reduce customs clearance and transit times – all areas where the region falls short of its Southeast Asian competitors and global benchmarks – could help raise productivity in South Asia. Surveys of global apparel buyers and interviews underline the key determinants of the ability to participate in GVCs. Cost and quality ranked the highest in all buyer surveys reviewed over the last decade. Lead time is increasing in importance, given the shift toward lean retailing and just-in-time v delivery. Access to and availability of high-quality fabric inputs locally, or at least regionally, is also important. Offering accompanying services, such as input/material sourcing and financing, and apparel product development (referred to as full package services), has become more important as buyers attempt to reduce the complexity of their supply chains. Finally, buyers take into account social and (to a lesser extent) environmental compliance, and political stability and predictability in their sourcing decisions. Social compliance has increased in importance in response to pressure from corporate social responsibility (CSR) campaigns by NGOs, compliance-conscious consumers and, more recently, the increase of disasters in apparel factories. Many South Asian firms are not very efficient The region’s firms over-employ relatively scarce capital and under-employ South Asia’s abundant labor. Estimates with the most recent round of Enterprise Surveys for Bangladesh, India, Nepal, Pakistan, and Sri Lanka indicate that firms tend to hire fewer workers than would be rational at the prevailing wage rates and marginal products of labor: the optimum level of labor in Indian and Sri Lankan firms is 1.7 and 2.2 times current employment levels, respectively, while estimates for Nepal and Pakistan suggest under-utilization on the order of 14-16 times the existing workforce. Bangladeshi firms, on the other hand, appear to over-utilize labor: firms hire approximately 18 percent more workers than would be optimal at the going wage rate. Thus, most large firms in South Asia do not operate close to what would be considered optimum efficiency levels given the prevailing factor prices, bringing down aggregate productivity. Potential reasons for this less-than-rational behavior include limited managerial capacity, labor market rigidities (particularly with regard to firing workers), and spatial distortions which prevent firms locating close to a ready supply of workers (or vice versa). The resources allocated to increasing efficiency are limited in many South Asian firms Many firms under invest in knowledge, and such investment varies widely across firms. Overall (public and private) investment in R&D in South Asia is low and is increasingly falling behind Latin America and particularly East Asia. Larger firms are more likely to engage in R&D activities, and having a license to use foreign technology increases R&D in all countries but Bangladesh. Exporters in India and older firms in Pakistan are also more likely to engage in R&D activities than non-exporters and young firms, respectively. Financial constraints are associated with lower investments in R&D activities for all countries except Bangladesh. Market structure appears to affect R&D only through informal sector competition in India, while other variables related to market structure are not significant– perhaps because less than 9 percent of the sample firms compete in an oligopolistic or monopolistic market. R&D investment varies by country: innovation leaders (Bangladesh and India) have a larger percentage of firms conducting R&D than the average of the Eastern Europe & Central Asia (ECA) and Africa, while laggards (Nepal and Pakistan) display a lower percentage. The incidence of R&D in India is the highest in the region, although the concentration rate is low. Small, non-exporting, national and very young firms are more R&D intensive in India, while in Bangladesh, large, exporting, foreign and old firms are significantly more R&D intensive. In Pakistan, there is a very large concentration of R&D activity in a very small number of firms. vi ICT adoption rates also vary across the region. Indian firms score very high on multiple dimensions of technology use, Pakistan is in line with global peers, but ICT adoption in Bangladesh and Nepal is very low. However, despite widespread internet use, the adoption of e-commerce and other online business tools is relatively low, with the difference particularly stark in the case of India. Firm size, export status, and to a lesser extent import status are important determinants of ICT adoption at the firm level. However, we do not find evidence that young firms or foreign owned firms are more likely to adopt ICT practices. Complementary factors – technology and skills – are important determinants of ICT adoption. Lastly, access to finance and to financial institutions is critical in facilitating the adoption of e-commerce. The region’s moderate achievements on many of these dimensions may explain the limited penetration of some technologies and hint at missed opportunities to improve productivity performance. These findings suggest that in countries like Nepal and Bangladesh, the focus should be on further efforts to foster the general adoption of internet and computers. This requires overcoming infrastructure challenges as well improving the provision of complementary skills – such as technology and human capital. In the case of India, where the use of ICT is already highly mainstreamed, the focus should be on further improvements of ICT practices, in particular e-commerce and other online business tools. Given the large extent of software development and the relative high availability of IT engineers, access to finance and establishment of broad base financial transactions platforms online could be critical in broadening the use of internet for commercialization. Innovation is limited in some countries, and generally confined to imitation rather than new products Beyond enhancing firm efficiency directly, ICT is also an important enabler of firm-level innovation. Firms in Bangladesh and India tend to demonstrate higher rates of innovation activities, while firms in Nepal and Pakistan generally lag behind. Patterns across firms within countries are similar: the acquisition of knowledge capital (e.g., R&D, investments in equipment, and training) is highly concentrated in a few firms, and mature, exporting, and foreign-owned firms tend to be the most innovative. There is some evidence of agglomeration benefits in innovation, as cities with more than 1 million inhabitants have a larger share of innovative firms. Turning to innovation outputs, the region’s leaders exhibit innovation rates around 80 percent, well above the average of the ECA and Africa regions, while Pakistan and Nepal have innovation rates of 15 percent and 21 percent. Process innovation is more important in Bangladesh and India, while product innovation is more important in Nepal and Pakistan. Even among the leaders, most innovation reflects the imitation of existing products and/or processes. Few firms engage in disruptive innovative activities such as introducing new products to the country or to the world. Most of the firms in the region tend to innovate for upgrading the quality of their products, although the introduction of new products is slightly more frequent in India. And most innovation is done in-house (more so than in Africa or ECA), which may contribute to limiting the potential for new products. Innovation improves firm productivity in South Asia. The impact of innovation on productivity in Nepal and Bangladesh is positive, statistically significant, and larger than in OECD countries. In India, the large number of observations allows for separate estimation of product and process innovation, with both vii coefficients positive and statistically significant. The degree of novelty does not introduce any additional effect on productivity. The findings suggest different approaches to innovation policy for countries that are innovation leaders versus laggards. For leaders, the critical challenge is how to generate novel, and if possible radical, innovations. Here, a focus on enhancing complementary factors - managerial capabilities, worker skills and finance – but more importantly breaking the nature of inward innovation development by supporting cooperation with other firms and institutions is warranted. On the other hand, for laggards policy should focus on increasing the number of firms engaged in incremental innovation. Throughout the region, employee education is a significant determinant of ICT adoption and an important enabler of GVC participation, which are associated with improvements in productivity. Thus, investment in skills, including modernizing training institutions and expanding access to on-the-job training can lead to higher efficiency and lower costs. Moreover, efforts to improve business and management practices in South Asia – where the region lags behind comparators – could boost productivity, profits, survival rates, and sales growth. Lead firms in the region have managed to rise to standards of global excellence. The experience of leading firms, which managed to rise to standards of global excellence, demonstrates that world class levels of operational performance, efficiency, and innovation can be achieved with the right management, scale/technology and worker training. For example, a number of firms in the automotive sector learned by becoming domestic suppliers to multinationals entering the region, and then leveraged that experience to access international markets on their own. They fostered innovation by locating close to customers to enable their engineers to work together with those of the client, and gradually built up their capacities from simpler to more complex components. They also invested in skilling and training their workforce, recognizing the linkages between skills, productivity, and innovation. Replicating good performance on a broader scale would boost the region’s global market share South Asia has tremendous potential to increase incomes through policies that enhance productivity and gain market share in exports. A forward-looking model with a particular focus on the region’s economies generates an optimistic baseline scenario through 2030, where productivity growth contributes as much as 2 percentage points per year to increases in regional GDP, the region’s performance in skill-intensive sectors improves, and it becomes the world’s fastest-growing region in terms of exports. However, absolute gains vs. major competitors would still be limited: despite significantly outpacing China in terms of growth rates, India’s 2030 exports of motor vehicles would only just approach China's current levels, and its 2030 electronics exports would remain an order of magnitude below what China exports today. If productivity growth were instead to follow performance in the current decade, exports growth would slow by more than a percentage point per year and performance vis-à-vis China and the rest of East Asia would suffer further. These scenarios are compared with several alternative simulations that give a boost to South Asia’s export performance by improving the region’s domestic and international logistics performance. The viii scenarios involve: (i) a reduction of logistic difficulties due to weak port infrastructure, burdensome customs regulations or inefficient behind-the border services (e.g., warehousing, transportation); (ii) a more rapid implementation of ongoing improvements in the port-to-port trade and transportation costs; and (iii) a reduction in the domestic cost of trade. These simulations lead to further increases in trade and additional gains in global market share. However, the expected impacts of these reforms are not sufficient for the region to catch up with its competitors in Southeast Asia in terms of export volumes. Partially, this is due to the fact that, in our dataset, these margins are already relatively low. Partially, this illustrates that improvements in logistics alone cannot offset the longer list of competitiveness challenges faced by the region. And partially, these results are limited by the assumptions and restrictions of the model which include, perhaps most importantly, an exogenous labor force participation rate driven purely by demographic trends. In a region where youth employment remains a challenge and female labor force participation rates are among the lowest in the world, one could envision – and hope – that more competitive factor and product markets could pull additional workers into the labor force. What is the potential for faster productivity growth in the region to create jobs? To illustrate, this report considers the intensely competitive apparel market. Productivity-enhancing measures which could result in a 10 percent cost advantage vis-à-vis Chinese apparel may lead to a 13–25 percent (depending on country) rise in South Asian countries’ apparel exports to the United States. In turn, increased output could have significant implications for job creation, particularly for women. Given the high labor intensity of apparel manufacturing and the large sensitivity of South Asia’s labor supply, particularly for women, to higher wages, a 10 percent price advantage over China in the US market could translate into employment gains of by 8.9 percent in Pakistan – by far the biggest winner – followed by Bangladesh (4.2 percent) and India (3.3 percent). These would be well-paying jobs; the wage premium of the apparel sector over agriculture ranges from 8 to 27 percent, depending on the country, and is even higher for women. Moreover, jobs created in textiles and apparel are likely to particularly attract low-skilled women. These results point to the critical importance of implementing productivity-enhancing measures in the apparel sector – but also caution that inaction may lead to a decline in market share as competitors that have pursued more aggressive apparel-friendly policies (such as Vietnam and Cambodia) can stand to gain much more than the South Asian countries in terms of market access. ix 1. South Asia’s competitiveness challenge and opportunity 1.1 The region’s competitiveness potential remains largely unrealized Which region will become the next global factory? As the work force ages and labor costs rise in China and other East Asian countries, many eyes turn to South Asia. It is a region that is still largely rural, where agriculture accounts for a large share of employment and a substantial fraction of GDP, and it has not been particularly successful in integrating within itself and with the rest of the world. Yet, more than one million young workers enter the labor market each year and by 2030, 26 percent of the world’s working adults will live in South Asia. This is the region’s greatest opportunity and greatest challenge. In the meantime, the global environment is becoming tougher. The commodity boom is over, putting brakes on demand and tightening fiscal belts in resource-rich countries. While commodity importers benefit from improved terms of trade, many also receive reduced remittances, limiting the benefits to the current account. Slowing global growth and even more pronounced slowdown in global trade make it more challenging for firms to enter and expand in export markets. New mega-regional trade agreements (e.g., TPP and TTIP) promise welfare gains to members but may lead to trade and investment diversion away from non-members. Against this background, it has become even more urgent for countries in South Asia to make overdue investments in boosting competitiveness and raising productivity to avoid falling further behind comparator countries in the global marketplace. There are plenty of examples of pockets of excellence in the region. Exports of goods and services are higher than in China, relative to GDP. At the sectoral level, the software industry in India, the garment sector in Bangladesh and Sri Lanka, and the Sialkot cluster in Pakistan are global success stories. And at the firm level there is also no shortage of global champions (e.g., US Apparel, Orient Craft, Pacific Jeans, and MAS in apparel, Tata Motors, Bharat Forge, Hi-Tech Gear, and HTGL in automotive, Fauji Foundation, Dilmah, and KRBL in agribusiness, and Dixon Technologies and Micromax in electronics). Yet, so far the region as a whole has made relatively little progress in integrating within itself and with the rest of the world, diversifying and increasing the sophistication of its export bundle, moving up the quality ladder, and improving its ranking on many competitiveness benchmarks. It has yet to realize the substantial benefits of economic integration and achieve its full potential – both relative to its endowments and global competitors – while the window of opportunity may not remain open for long. The following discussion develops this observation in more detail. 1.1.1 Multiple pockets of excellence evidence vast untapped potential South Asia, and India in particular, is already well known for having achieved excellence and pre- eminence in the ICT industry. Less well known, is the fact that South Asian countries/locations/firms are becoming major players in important manufacturing industries. These successes include:  The apparel industries in Sri Lanka and Bangladesh which are as big (on a per capita basis) as the ones in China and Vietnam.  The light manufacturing cluster in Sialkot (Pakistan), which despite all odds, have achieved dominant global market shares in products such as soccer balls and surgical instruments. i  Indian auto-part firms which are becoming global players through exports to and acquisition in leading markets such as Germany.  Global electronics and auto part firms establishing their global R&D centers in India  Leading agribusiness firms developing, in partnership with governments, new varieties for the domestic and international markets (e.g. tea in Sri Lanka as well as rice and mint oil in India). These cases were selected for in-depth case studies as part of Volume II of this report so as to better understand the drivers of competitiveness as well as the constraints limiting more often occurrences of such cases to more firms, locations, industries and countries at a time where South Asia is uniquely positioned to take advantage of rising production cost in East Asia. 1.1.2 Difficulties in attracting investment and penetrating global markets Integration in trade and investment can increase productivity. Opening up local markets to foreign trade and investment increases competition, which encourages labor and capital to move from less productive to more productive firms (Melitz, 2003). Further, increased competition may induce firms to improve their efficiency (Helpman and Krugman, 1985), to focus on their core competencies (Bernard et al, 2006) and reduce managerial slack (Hicks, 1935), or to invest in new technology (Aghion et al, 2005). Finally, openness facilitates access to better inputs and technologies, which is particularly important for those developing countries where import substitution policies previously reduced firms’ ability to purchase imported inputs. Despite these well-known benefits of economic integration, South Asia’s intra-regional and global ties are relatively weak. Foreign investment, for example, is low (Figure 1.1). Over the period 1990-2014, the region received, on average, between 2.2 and 2.8 percentage points of GDP less FDI inflows than countries in East Asia (see Annex). Particularly in the later part of this period (2011-2014), FDI inflows in all countries (except Maldives) were substantially below the average of countries at similar levels of development. Figure 1.1 FDI to GDP ratios 1990-1994 2011-2014 ii 20 15 15 FDI to GDP (%), 2011-2014 MDV 10 10 5 5 LKA PAK BGD IND NPL 0 IND LKA BGD PAK NPL AFG -5 0 -10 -5 6 7 8 9 10 11 Log of GDP per capita (PPP, av. 1990-1995) 7 8 9 10 11 Log of GDP per capita (PPP, av. 2011-2014) Source: Authors’ calculations based on WDI Lack of investment integration in South Asia is particularly evident when considering intra-regional flows. Globally, South-South cross border investment has increased, and a recent market survey showed that multinationals in other regions tend to allocate a significant share of outward investment within their region. By contrast, countries in South Asia receive little FDI from within the region. In particular, despite the lower transactions cost of investing in nearby, familiar markets, Indian multinationals tend to invest outside the region (Gomez-Mera et al, 2014). Some of the blame for low FDI inflows can be traced to burdensome regulations governing FDI. Nepal, for example, has failed to attract substantial FDI because of the complicated processes required to repatriate profits, high entry barriers (with a long negative list) and insufficient guarantees of investor protection. The lack of readily-available land with adequate access to infrastructure services has constrained foreign investment, particularly in Bangladesh. For example, in 2011 Samsung’s large intended investment in electronics in Bangladesh fell through because adequate land was not available in an export processing zone. By contrast, one element of the competition among Indian states for investment by major original equipment manufacturers (OEMs) is through offering land in conjunction with tax incentives, although there is some risk that such competition will lead to sub-optimal investment locations and industry fragmentation. Trade integration is also low. Over 1990-2014, South Asia’s average ratio of exports to GDP varied between 17 and 21 percentage points below East Asia, and the average ratio of imports to GDP was 21- 22 percentage points lower than in EAP countries (Maldives, with a highly developed tourism export sector, is again the exception). 2 Countries in the region have become more integrated in the global 2 A slightly different picture emerges if, when measuring the export and import orientation of countries in the SAR region, we control for some non-policy determinants of openness. For example, larger countries tend to trade less with the rest of the world, because they face more domestic trade opportunities than small countries. Similarly, landlocked or island states face higher transportation costs and hence tend to trade less. Once size and whether the country is landlocked or an island state are taken into account, most SAR countries remain less integrated into the global marketplace, both in terms of export and import orientation, than the average. However, India, the largest iii marketplace; the region today accounts for 60 percent more of merchandise trade than in 2000-2004 (from 1.15 dollars of every 100 traded globally, to 1.82 dollars). Nevertheless, growth in the region has been more inward-oriented than in East Asia. Figure 1.2 Trade to GDP ratios 1990-1995 2011-2014 250 200 MDV 200 Trade to GDP (%), 2011-2014 150 150 100 100 LKA LKA IND 50 AFG NPL BGD 50 NPL PAK PAK BGD IND 0 0 6 7 8 9 10 11 7 8 9 10 11 Log of GDP per capita (PPP, av. 1990-1995) Log of GDP per capita (PPP, av. 2011-2014) Source: Authors’ calculations based on WDI South Asian countries can be divided into two groups by their performance in merchandise exports. While exports increased at double-digit rates over 2000-2013 in India, Bhutan, Bangladesh and Afghanistan (14, 16, 13 and 15 percent per annum), growth was slower in Pakistan, Sri Lanka and Maldives (8, 5 and 2 percent per annum) and exports fell in Nepal at 1 percent per annum (Table 1.1). However, regardless of whether countries belong to slow- or fast-growing groups, their share of global merchandise export markets remains small. For example, India, with an 80 percent increase in its market share in the past decade and a half, has just reached 1.5 percent of the global exports market. Bangladesh, with somewhat slower growth of 50 percent, has passed Pakistan to become the second largest merchandise exporter in South Asia with nearly 0.2 percent of the global merchandise market. Afghanistan and Bhutan more than doubled their global market share in the past decade and a half, but still account for less than 0.01 percent of the world market, combined. Table 1.1 Market Shares in Merchandise Exports Afghanistan Bangladesh Bhutan India Sri Lanka Maldives Nepal Pakistan South Asia 2000-2004 0.002 0.107 0.001 0.814 0.074 0.003 0.010 0.139 1.149 2010-2014 0.004 0.162 0.001 1.463 0.057 0.001 0.004 0.135 1.827 Absolute 0.002 0.055 0.001 0.649 -0.017 -0.002 -0.005 -0.005 0.678 growth Percent 112 121 percent 52 percent 80 percent -23 percent -61 percent -54 percent -3 percent 59 percent Growth percent economy in the region, appears to be more integrated than the average (see plots of these export and import orientation indices in Figure 1.19 and Figure 1.20 in the Annex). iv Note: Cell values indicate the share of each country in global merchandise exports, in percentage points Source: Authors’ calculations based on UN Comtrade The textiles and apparel sector is one exception to this general trend (Figure 1.3). South Asia’s share of global exports in garments rose from 7.4 percent in 2000-2004 to 11.6 percent in 2010-2014. More than half of that increase is accounted for Bangladesh (due in particular to effective import facilities for exporters), about 40 percent by India, and the remainder by Pakistan. Gains in market shares in other sectors are almost all below 1 percentage point, and almost fully explained by increased exports from India. Figure 1.3 Changes in Export Market Shares by Country – 2000/04 – 2010/14 5 Change in Export Market Share by 4.5 4 3.5 (percentage points) 3 2.5 Country 2 1.5 1 0.5 0 -0.5 AFG BGD BTN IND LKA MDV NPL PAK Source: Authors’ calculations based on UN Comtrade In services, the region has recorded much better performance. Overall, South Asia’s share of global services exports rose from 0.9 percent in the early 1990s to 3.6 percent in the early 2010s. Every country in the region except Nepal increased its share of global services exports; India, and at a substantially lower scale, Maldives, more than doubled their shares of services exports (Table 1.2) v Table 1.2 Market Shares in Services Exports Afghanistan Bangladesh India Sri Lanka Maldives Nepal Pakistan South Asia 1990-1994 0.05 0.54 0.06 0.01 0.03 0.17 0.87 2000-2004 0.05 1.27 0.07 0.02 0.02 0.12 1.56 2010-2013 0.08 0.06 3.16 0.08 0.05 0.02 0.13 3.58 Absolute 0.01 1.89 0.01 0.02 0.00 0.01 2.02 growth Percent 107 -13 129 26 percent 149 percent 8 percent 10 percent Growth percent percent percent Note: Cell values indicate the share of each country in global services exports, in percentage points Source: Authors’ calculations based on UN Comtrade Limited trade integration reflects elevated trade costs. For example, Nepal charges high tariffs on yarn, a key input into most of its apparel exports. In India, high tariffs on inputs particularly affects the electronics sector, while tariffs on man-made fiber (combined with the problems with duty drawback schemes for exporters) essentially limit exporters to garments made of domestic cotton, which are concentrated during the summer season, thus reducing capacity utilization. On top of high tariff rates, South Asian countries impose high para-tariffs, an additional tax on imported products that is typically complicated, subject to arbitrary enforcement, and is applied irrespective of trade preferences. The average import tax rates in Bangladesh and Sri Lanka are more than double the customs duty average if para-tariffs are included (Kathuria et al. 2015). The decline in tariffs and increase in para-tariffs in Bangladesh have made the latter the more important constraint on imports (World Bank, 2014). By contrast, Sri Lanka’s garment exporters have benefited greatly from zero tariffs on textile imports. In services trade, Bangladesh, India, Nepal and Sri Lanka are substantially more restrictive than high income economies, and even China, according to the World Bank’s Services Trade Restrictiveness Index, although Pakistan’s service restrictiveness is relatively low. Earlier steps to reduce barriers to trade and investment in the region have paid substantial productivity dividends and hint at potential future benefits from further policy efforts to improve integration. In India, for example, the reduction in tariffs on auto parts and electronics greatly increased competition in the domestic market, raising standards among firms and enabling them to increase productivity further by working with demanding clients. More broadly, the sharp fall in the level and dispersion of tariffs in response to the 1991 balance of payments crisis induced firms to improve their efficiency and improved their access to imported inputs (Topalova and Khandelwal, 2011).3 The productivity benefits of reform were smallest in sectors where burdensome regulations limited firms’ ability to adopt new technologies, 4 greater for domestic than foreign firms (probably because foreign firms had already been exposed to competition), and largest in industries that also experienced the most deregulation and FDI liberalization. About a third of the rise in firms’ product diversification was due to increased access to better quality and higher variety of imported intermediate inputs (Goldberg et al 2010). Improved services policies after 1991 also boosted Indian firms’ productivity. Policy changes that facilitated the operations of foreign services firms – particularly in banking, telecommunications, insurance and 3 A similar result was previously reported by Amiti & Konings (2007) for the case of Indonesia. 4 This is consistent with cross-country evidence presented by Bolaky and Freund (2004) that the growth effect of trade depends on a country’s business regulations. vi transport – increased the productivity of foreign and local manufacturing firms that used those services (Arnold et al. 2015). 1.1.3 Little progress in diversifying the merchandise export basket The composition of exports in South Asian countries has changed little in 15 years, which points to limited product innovation. Exports remain highly concentrated in textiles and apparel in Bangladesh, Afghanistan, Nepal, Pakistan and Sri Lanka, in minerals in Bhutan, and in animal and vegetable products in Afghanistan and Maldives. Almost 80 percent of the region’s export growth from 2001 to 2013 came from the sale of the same goods to the same destinations (Figure 1.4), and the remaining 20 percent came from selling the same products to new markets. While the number of product varieties exported increased in almost all countries (with the exception of Nepal), diversification into new products (either in old or new markets) accounted, on average, for only 0.07 percent of export growth. In some countries, most exports go to only a few destinations. For example, the top 5 export destinations account for 97 percent of export revenues in Bhutan and 70 percent of in Maldives. In contrast, India’s top 5 markets purchase only 36 percent of the country’s exports (Figure 1.4). vii Figure 1.4 Export growth decomposition in South Asia Export Growth Decomposition 2001-2013 Number of Destinations Reached 2000-2014 90% 160 80% Number of Export Markets 70% 140 60% 120 50% 40% 100 Reached 30% 20% 80 10% 60 0% 40 20 0 Source: Authors’ calculations based on UN Comtrade Source: Authors’ calculations based on UN Comtrade Number of Products Exported 2000-2014 Share of Export Revenues Accounted for by Top 5 Destinations 5000 100% Number of Product Varieties 4500 90% 4000 80% 3500 70% Exported 3000 60% 2500 50% 2000 40% 1500 30% 1000 20% 500 10% 0 0% Source: Authors’ calculations based on UN Comtrade Source: Authors’ calculations based on UN Comtrade 1.1.4 Elusive sophistication and low quality of exports On average, the sophistication of merchandise exports in South Asia (as measured by the EXPY indicator) is higher than expected given the region’s income level.5 However, with the exception of India, countries 5 In the EXPY indicator, each export good is assigned the value of the average per capita income of other countries exporting that good; the country’s EXPY is the average of these values, weighted by the good’s share of total exports (Hausmann et al. 2006). The EXPY indicator has to be interpreted with caution, because it reflects the product exported rather than the stage of production (the actual task). In today’s increasingly fragmented production viii in the region have not been successful in further increasing export sophistication. While India leapfrogged both Vietnam and Indonesia on this metric between 2000 and 2014, sophistication did not increase in Bangladesh and Sri Lanka, rose steadily but from a low level in Pakistan, and declined in Bhutan (Figure 1.5, top left panel). Even in India, one measure of export quality and sophistication (PRODY),6 has remained low. A recent IMF study finds that the average level of sophistication for India’s manufacturing exports is lower than for the rest of Asia, in sharp contrast to India’s performance in the services sector.7 The average sophistication of the countries that purchase exports from South Asia, as measured by the weighted average of the buyers’ per capita incomes, has declined over time (Figure 1.5, top right panel). This is both because South Asia is moving toward relatively poorer trading partners (Figure 1.5, bottom right panel), and because the existing trading partners are growing less rapidly than average (Figure 1.5, bottom left panel). Figure 1.5 Export sophistication and weighted average income of importers, 2000-2014 Export Sophistication Weighted Average of Income of Importers 25000 1.4 Relative to Median Importer CHN2014 BTN2000 1.2 20000 BTN1990 CHN2000 IND2014 VNM2014 Income of Importers 1 IDN2000 CHN1990 BTN2014 IDN2014 0.8 IDN1990 15000 IND2000 VNM2000 0.6 IND1990 NPL2014 VNM1990 MDV2014 0.4 NPL2000 AFG2014 PAK2014 LKA2000 LKA2014 0.2 PAK2000 10000 BGD2000 BGD2014 BGD1990 LKA1990 PAK1990 0 NPL1990 5000 0 0 5000 10000 15000 20000 25000 GDP Per Capita (PPP, Constant) 2000-2004 2010-2014 Source: Authors’ calculations based on UN Comtrade Source: Authors’ calculations based on UN Comtrade processes, a country may export a sophisticated product (for example, a computer) but only carry out low-skilled assembly activities while the high-tech activity of making the parts is done elsewhere. 6 PRODY is a weighted average of the per capita GDP of countries producing auto goods (including auto parts), with weights derived from Revealed Comparative Advantage calculations. 7 IMF Working Paper 2015. ix Change due to trading partners’ incomes (trade Change due to trade shares (partner incomes weights fixed at 2000-04 level) fixed at 2000-04 level) 2 1.4 Income of Importers Relative Income of Importers Relative 1.2 1.5 1 to Median Importer to Median Importer 1 0.8 0.6 0.5 0.4 0.2 0 0 2000-2004 2010-2014 2000-2004 2010-2014 Source: Authors’ calculations based on UN Comtrade Source: Authors’ calculations based on UN Comtrade Quality, as measured by the price that exporters of a particular product fetch in international markets compared with other producers, is generally low and has declined for some countries. For example, Sri Lankan apparel (e.g., brassieres) was at the higher end of the price spectrum at the turn of the century, but now fetches prices in the lower fifth of the distribution. In tea, where the country has a built-in brand name, Sri Lankan exporters secure prices just above the median. Nepali carpets are in the middle of the distribution of prices, while mineral water exports are in the lowest tenth. Pakistani cotton increased from the bottom 30th percentile to almost the mid-point of the distribution of prices. Trousers, however, are sold at the very bottom of the distribution of prices. Key Bangladesh exports have lost ground along the quality ladder. In cotton t-shirts, for example, the country’s exports fell from the middle of the price distribution in 2000 to the 10th percentile in 2013. Indian cars and cell phones also secure low prices relative to competitors. Maldives is an exception, as some exports of fish are priced at the high end of the market – tilapia exporters, for example, receive 50 percent more than the price received by the median exporter of the same product (Figure 1.6). Figure 1.6 Relative Unit Values of Select Export Products by Country and Year – Quality Ladders LKA Bras (2000) LKA Bras (2013) EST 2.5 2.5 DEU DNK FIN 2 EST 2 CHE CAN BIH ALB Relative Unit Value Relative Unit Value AUS 1.5 1.5 BGR AUT CAN CHE MAR HRV FRA SWE SVN LAO PRT SWE MAR LKA JPN URY UKR SER TUN LVA UNS MMR HUN EGY ITA ITA SVN BRA RUS CZE 1 CRI IRL 1 IND GBR ARG BLR ESP HUN VNM DEU VNM ARG CZE THA ROM ROM GEO HND LTU BGR LVA MAC TUN GRC POL GRC COL SVK COL ESP BLR THA OAS IND SER UNS GTM HRV DOM ISR SVK DNK UKR HKG DOM KHM ISR OAS PHL MEX LKA MDG JOR LUX SGP MYS USA BRA JPN SLV TUR NLD HKG KOR BGD MMR ALB USA IDN .5 CRI URY MUS POL CHN .5 CHN MAC MYS IDN CHL SGP BGD EGY TUR PHL FJI IRL ZAF HND LTU SYR MDV MEX 0 0 0 20 40 60 0 20 40 60 80 Rank Rank x Relative Unit Value Relative Unit Value Relative Unit Value Relative Unit Value .6 .8 1.2 1.4 .5 1.5 2.5 .5 1.5 2.5 .5 1.5 2.5 1 0 1 2 1 2 1 2 0 0 0 0 SAU FRE OAS MUS BEL GEO SER ROM HUN BFA BGR IND ARG MDA SDN MAR IRL UKR PRT ECU PAK NPL LTU AFG PNG SWE ARG BHR NGA EST CHN IRN IDN KWT TZA RUS HUN ESP ZAF BRA SGP GTM OMN VNM NLD THA SAU 20 UZB THA HKG EGY DNK MWI BIH AUS LVA MAR TZA LKA LBN MAC ARE 10 10 ROM CHN ZAF 20 TKM COL IND MKD HRV NPL ZWE USA SVN PRY VNM ITA VNM FRA CRI ABW DEU DOM FRA GBR GBR TUN GRC TUR CHE SYR TUR ZMB UNS IDN IDN SWE HKG 40 POL VCT PRY Rank Rank Rank LTU KEN Rank HKG IRL SYR POL AUT MYS TKM CHN IND ARG ISR BRA CHL TUN USA PER SLV PHL LKA CIV PAK LKA Tea (2000) SGP ZWE KOR 20 PAK 20 PAK Cotton (2000) BEL DEU SGP TTO NPL Carpets (2000) 40 NLD LKA TUR OAS GBR SVN PRT AUS VEN NPL Mineral_water (2006) ESP MAC AUT DEU MAR JAM 60 FIN AZE IRN ITA USA MYS ITA GIN BOL GRC CHE NZL RUS ZAF TUR MEX CAN SGP Relative Unit ValueSLV BRA Relative Unit Value Relative Unit Value Relative IND Unit Value AUT GRC CZE PAN CHE PER NPL ECU MEX NLD JPN KOR DNK ISR SPE EGY LUX CHN SVK MLT HKG FRA JPN GBR ECU 30 30 CHE RUS RUS 80 DEU 60 NLD ROM CAN .5 1.5 2.5 .5 1.5 2.5 .5 1.5 .5 1.5 2.5 1 2 1 2 0 1 2 0 1 2 0 0 0 0 xi MUS LTU CHL IRN IRN PAK SGP BGD LSO TZA ARG ROM IDN MYS MKD GEO ZAF HUN NPL OAS SER AFG PNG UKR THA EST TUR AUS BGR IND ZAF NGA MDA MOZ NIC PER BOL SVN UNS HND HRV MAR 20 ETH ARG PAK VNM RUS TKM IND ZWE BFA LKA IRN CHN 10 IND OAS MWI TUR AUT KAZ CYP IDN UZB HKG SVK USA VNM ZAF ECU ARM TZA NGA UGA KAZ 10 20 POL POL VNM TUN COL TZA PAN UNS PAK GRC NLD LVA RUS MAC 40 ZWE TGO MAR SGP PRK HKG MYS GBR GTM NLD MOZ ESP SYR FRA KEN ARE ARG THA DEU RWA BRB IDN SLV 20 BEN BEL UKR BRA TTO Rank Rank Rank SGP Rank UNS VCT LKA ARE KOR NPL TUR CRI ZAF ITA MLI CZE PAN IRL BLR VNM BHR 60 ESP PSE DNK IND HND FRE LKA Tea (2013) CHN KWT CHN SGP DZA TUR SAU CYP PAK Cotton (2013) MEX NPL Carpets (2013) PHL 20 BRA HKG 40 SVK PER JPN EGY GBR DOM PRT BEL NPL Mineral_water (2013) GRC FIN NLD MEX CHE CHE CHN BEL PRT 30 AUT POL HUN VEN SLV IDN GBR USA DEU PRT LUX 80 LBN THA OAS DEU USA PER SWE DEU NZL SWE KOR AUS JAM NPL AUT BTN ECU JPN CHE NLD CHL SVN GEO UNS USA GRC BRA MLT DNK CZE NOR ITA ITA ISR LVA TUN ISR 30 40 100 BEL 60 Relative Unit Value Relative Unit Value Relative Unit Value Relative Unit Value .5 1.5 .5 1.5 .5 1.5 0 2 0 1 2 1 0 1 2 0 0 0 0 TKM LKA SGP HTI CUB GUY NIC MNG SYR LSO PAN PER FRE SAU HKG PRT DOM NPL IRN KWT KEN KEN GRD MAR MRT LCA MDV ARE ZAF ZAF OMN BWA HKG MDG TZA MEX USA CHN NIC DNK CAN PAK HND SYR CHN MMR LSO USA BGD 20 NAM FJI 5 ESP UGA PAK PAK LBN CIV TTO MWI KOR 20 VNM PRY ITA KAZ EGY BIH ITA BRA BHR SLV LAO FRO IND AFG TTO KHM BEL SPE CHN GTM LTU RUS SYR ZAF THA PRT IRL MAC AUT IND BLR 20 HND KOR VNM COL ISR IRN TUR LVA LUX BRA SER 40 RUS ALB FIN IDN MKD 10 GBR LVA BGD LVA TGO KHM BOL ECU ARE MDA UNS ARG LTU PHL THA MMR RUS 40 NLD MYS ZWE IND OAS EGY AUT AUT DNK OAS MUS MAC SWE HUN SVN UKR Rank Rank TUR Rank CZE Rank GRC OMN TUN CYP GBR MAR BGR USA PRT IDN GRC SVK 60 EGY SGP CRI IRL UNS VNM MOZ MYS POL TUR 15 NLD DEU MDG TTO NOR NPL POL ROM COL AFG Figs (2000) NZL SVK MUS DEU HUN BGR ROM MDV Tilapia (2006) CHL PAK Trousers (2000) HRV TUN 40 YEM ARG HRV BRA 60 MEX FRA CHE SWE PAK SWE MLT LUX OMN BGD Cotton_tshirts (2000) TUN JAM MKD JPN ZAF FRA LUX NLD PRT GRC 80 CHN SEN FRA QAT THA LAO GBR PER EST MDV DEU DEU JOR 20 BEL Relative Unit Value JPN GBR ITA Relative Unit Value SGP Relative Unit Value HND Relative Unit Value POL COL PER SVN LBN SAU IDN CZE MEX ISL ESP BEL ESP MUS PHL NLD BRA ECU HTI GRC ITA DOM UNS SLV GRL MRT 80 FRA FJI ZMB LKA AUT GTM ESP JPN NZL ISR 100 EST LKA IRL USA BRN DNK SYC USP CHL CHL URY CAN NZL CHE 25 PER 60 PHL .5 1.5 2.5 .5 1.5 .5 1.5 .5 1.5 2.5 1 2 1 2 0 1 2 1 2 0 0 0 0 TKM xii VNM HTI HND DOM BGD ARE OAS RUS MMR DZA NIC CHN CRI ARE ESP CHN HND PAK TZA BGD KHM MMR PRY THA SGP EGY SLV LKA KEN MEX MDG SLV GTM IND AUT IRL NPL LTU BEL LUX PRY LAO TUR UGA THA GRL LVA IDN COL HTI 20 ZAF IND IRL ITA PAK JOR FRO CHN NIC BIH 20 TUR KEN VNM SVK NAM KHM TGO MAC ZWE MMR GRC SLV PRT ECU VNM ARM LKA ETH SWE ZAF GRC EGY MAR ARG IDN EGY MDA 20 NLD JOR MEX MYS 10 GBR NZL MAR BRA COL ROM ETH MUS DEU NLD PHL COL BIH ESP 40 HND BRA LVA ISR HUN USA CRI TUR JAM SER DNK ZAF SVK IRN FRE SER PER TTO UKR MEX GBR 40 GRD MKD NLD USA LBN GRC IND LAO SVN DNK AUS Rank BEL Rank FRA NOR BOL PRT Rank IDN SVN DEU Rank PAN THA DOM CHE TUR HRV ITA PHL BGR MDG FRA SYR GBR TUN 60 SVK POL KAZ POL FIN UNS DEU SWE DNK ESP AUT DEU UNS LTU FRA FJI AFG Figs (2013) SGP AUT MYS MDV Tilapia (2013) PAK Trousers (2013) ECU GTM 40 ESP LTU OAS POL ROM NLD 20 JPN SEN BGR NZL DNK 60 MKD MAC CHL BGD Cotton_tshirts (2013) PAK LUX EST ISL PAN ALB RUS CZE UNS CHN HUN RUS BEL ARG CZE CAN TKL CZE HKG 80 BRA PER ARG FRA USA PAN BRA HRV KOR AFG SWE ROM FIN MUS BUN FJI USA LVA ITA ARG GRC PRT CYP FRE AND MUS HKG QAT CHL MYS KOR IRL SGP OAS ISR EST ARE 80 LKA TUN KWT MEX MDV NZL JPN AUT GBR 100 60 ZAF GTM UKR PHL 30 PER Source: Authors’ calculations based on UN Comtrade 1.1.5 Low scores on most competitiveness benchmarks On average, countries in South Asia score poorly on two major indices used globally to capture key aspects of competitiveness: the Global Competitiveness Index (GCI) published by the World Economic Forum and the World Bank’s Doing Business report. In the most recent (2015-2016) GCI rankings, India is the only South Asian country in the top half of nearly 140 countries, featured in the 55 position but lagging well behind China at 28. India is followed by Sri Lanka at 68, Nepal at 100, Bhutan at 105, Bangladesh at 107, and Pakistan at 126. Although all South Asian economies (with the exception of Bhutan) improved since last year, many have yet to regain the ground they lost since 2007: for example, Pakistan has lost 34 places and India, despite making major advances, still ranks 7 positions lower than it did in 2007. When it comes to the components of the overall ranking, the most challenging areas in Bangladesh, Bhutan, Nepal and Pakistan include inadequate supply of infrastructure and corruption, while in Sri Lanka the most problematic factors are inefficient government bureaucracy and access to finance. While India has recently achieved better GCI scores on macro stability, institutions, and infrastructure, its score remains hampered by inadequate electricity supply and poor technology readiness of its businesses. Nevertheless, the importance for firms of these overall indicators of the business climate can be hard to evaluate. For example, India scores poorly on the provision of electricity on the GCI and electricity is found to reduce productivity (see Annex to Volume II), but firms completing the World Bank Enterprise Survey found that electricity provision was only a moderate or minor constraint on their activities. On the other hand, electricity provision was an important constraint on productivity for agribusiness firms in Pakistan. In the World Bank’s 2016 Doing Business report, all the South Asian economies, with the exception of Bhutan, are ranked in the bottom half, with an average ranking on the Ease of Doing Business of 128. By contrast, many of South Asia’s competitors, such as Thailand (49), China (84) and Vietnam (90), are all in the top half of the ranking. Bhutan has the region’s highest rank, at 71, followed by Nepal (99), Sri Lanka (107), India (130), Pakistan (138), Bangladesh (174) and Afghanistan (177). Compared to the 2015 ranking, only India and Sri Lanka improved – going from the position 134 to 130 and 113 to 107 respectively – while all the other countries experienced a setback. xiii Figure 1.7 Ease of Doing Business rankings in South Asia Note: The distance to frontier score benchmarks economies with respect to regulatory practice, showing the absolute distance to the best performance in each Doing Business indicator. An economy’s distance to frontier score is indicated on a scale from 0 to 100, where 0 represents the worst performance and 100 the frontier. Source: Doing Business database. On average, the South Asian economies rank highest in Protecting Minority Investors, with India and Pakistan ranked in the 8th and 25th positions globally. The region’s next best performing category is Starting a Business, but rankings here are already well below comparators: India ranked 155 out of 189 economies, and while Sri Lanka ranked better (98), it nevertheless cost 18.7% of income per capita to set up a firm, compared to 6.4% in Thailand and 4.9% in Vietnam. The areas with the most opportunity for improvement are Enforcing Contracts (where the region’s average ranking is 143), Registering Property (136), and Resolving Insolvency (129). On average, resolving a commercial dispute through the courts takes 1,077 days in South Asia—almost twice the global average of 630 days. The region also performs poorly in logistics rankings. According to the World Bank’s Logistics Performance Index, in 2014 South Asia had the lowest logistics performance among all developing regions due to poor quality of trade and transport related infrastructure, time-consuming clearance processes, low quality of logistics services, and lack of timeliness of shipments in comparison to economies like China, Vietnam and Thailand. Similar to GCI rankings, none of the South Asian economies are featured in the top 50 of the LPI, while China with is ranked at 30, Thailand at 31 and Vietnam at 48. Between 2007 and 2014, only Nepal and Sri Lanka have been able to improve their logistics performance by gaining 25 places and 3 places in the overall LPI rankings – but much ground still remains to be covered. xiv Poor logistics can sharply reduce productivity. Lengthy and unpredictable delays in customs clearance can force firms to hold higher inventories (regional firms in auto components, textiles, electronics, and heavy engineering report maintaining 27 percent higher inventories on average to deal with uncertain delivery times), and can impose delays in production and increased turn-around times. Delays imposed due to poor road infrastructure and lengthy inter-state clearance processes have similar affects. For example, in India, crossing two state borders between origin and destination can add as much as a week to the uncertainty in delivery schedules. 1.2 Improving competitiveness is about raising productivity rather than keeping costs low Competitiveness occupies a central position in government and industry agendas in South Asia.8 However, current performance – whether measured as participation in global markets or common competitiveness benchmarks – has been subdued. For example, the region possesses several key advantages in the apparel sector: it has an abundant supply of workers, labor costs are one-half to one- quarter of China’s, and South Asia is a top cotton producer in an industry where textiles make up close to 70 percent of production costs. Some leading apparel firms in South Asia have achieved world class operational performance by investing in training and technology, reaping economies of scale, and in the case of India and Pakistan, by integrating vertically to avoid barriers to sourcing high-quality inputs on the global market. Nevertheless, although South Asia increased its share of the global apparel market from 7.5 to 12.3 percent from 2000 to 2012, it continues to lag well behind China which accounts for 41 percent of the market (Table 1.3). Despite higher labor costs, China is able to attract buyers by offering a wide range of apparel at short lead times, while high productivity limits total costs despite relatively high wages. No country in South Asia has thus far succeeded in offering a comparable package of goods and services. Table 1.3 Apparel competitiveness of South Asian countries, 2012 Apparel exports as a Rank in top share of world Apparel exports as a Average apparel 15 apparel apparel exports share of country monthly earnings Country exporters (percent) exports (percent) (US$/per hour) Bangladesh 2 6.4 82.8 0.51 India 7 3.5 5.2 1.06 Pakistan 13 1.2 19.0 0.58 Sri Lanka 14 1.2 44.8 0.55 China 1 41 7.1 2.60 Source: World Bank staff calculations based on COMTRADE (exports) and household surveys (earnings) This example illustrates the difficulty in defining national competitiveness, because different countries have become “competitive“ with a different mix of endowments, factor prices, and policies (Porter, 1990). If competitiveness is defined as purely gains in global market share, countries can remain 8 The Indian government’s “Make in India” initiative has industry competitiveness at its core. The Sri Lankan government has set up a National Productivity Secretariat to help “enhance Sri Lankan productivity by energizing the sector to face international competition” (Ministry of Productivity Promotion, Sri Lanka). xv competitive in the short term by keeping production costs low through controlled exchange rates, rigid factor markets, and similar policies. However, such policies and the larger focus on gaining a bigger slice of the pie are quite likely to be a zero-sum game (Krugman, 1994). A better strategy for improved competitiveness is to reduce the transaction costs for firms to compete domestically and globally by providing efficient infrastructure services, a smoother business environment and more effective public services. Still, reducing these costs has its limits as well. On the other hand, investing in productivity-enhancing measures can pay continuous dividends over the long term. Porter (1990) ties competitiveness to the efficiency with which firms combine factors of production (total factor productivity, TFP) and argues that the “only meaningful concept of competitiveness at the national level is productivity.” This is also the perspective adopted by this report: productivity is what drives competitiveness in the long run, and boosting productivity leads to rising living standards through higher wages and returns on investment. Productivity in South Asia has been less studied than various cost factors, so the following discussion focuses on three major challenges to productivity growth in the region. At the macro level, the contribution of total factor productivity (TFP) to growth in South Asia has declined in recent years, and factors subject to diminishing returns –quantity rather than quality of labor and non-ICT investment – have been the main drivers of growth. This calls for greater focus on improving productivity to sustain and accelerate growth, create jobs, reduce poverty, and boost shared prosperity. The main forces that can increase productivity growth are increased integration with the global economy, the movement of resources from agriculture to higher productivity manufacturing and services, and the movement of capital and labor from less productive to more productive firms within narrowly-defined economic activities. At the sectoral level, the movement of labor from agriculture to industry and services (structural transformation) has not been rapid enough in South Asia to markedly reduce the large differences in productivity across sectors. While countries in the region are at different points along this transformation, South Asia is overall in its early stage, leaving significant untapped potential to reap productivity gains by further reallocation of labor from lower to higher productivity activities. At the level of the firm, large productivity differences exist between South Asian firms, and much of the region’s resources are locked away in small, low-productivity firms that neither grow nor exit, indicating the existence of barriers to market entry and exit (Cabral, 2007; Li and Rama, 2015; Tybout, 1996). The consequent “misallocation” of resources accounts for a large share of the difference in productivity between South Asia and high-income economies (Hsieh and Klenow, 2009; Hsieh and Olken, 2014; Pages, 2010).9 The following discussion develops these observations in more detail. 9 Hsieh and Klenow (2009), Pages (2010), and others document that productivity dispersion between top and bottom firms is particularly large in developing countries. The positive (albeit weaker in developing countries) correlation between firm size and productivity, and the fact that medium and large firms are under-represented in South Asia, suggests the existence of major opportunities for improving productivity (i.e., the preponderance of small firms drags down aggregate productivity). xvi 1.2.1 Macro challenge: contribution of TFP to growth is low and declining GDP in South Asia more than quadrupled from 1990 to 2014, and most countries enjoyed rapid growth in output and per capita income (figure 1.7). However, the contribution of total factor productivity to GDP growth has been mixed, as indicated by the four countries with the data (from the Conference Board) required to decompose GDP growth into its components, including changes in the quantity of labor, the quality of labor, ICT capital, non-ICT capital, and TFP (figure 1.7).10 Figure 1.8 Real GDP and GDP per Capita Growth in South Asia 9% 8% 7% 6% 5% 4% 3% 2% 1% 0% Afghanistan Bangladesh Bhutan India Maldives Nepal Pakistan Sri Lanka 1990-2000 1990-2000 2001-2014 2001-2014 1990-2014 1990-2014 Source: Authors’ calculations based on WDI Note: Afghanistan’s growth both for GDP and GDP per capita is calculated over 2002-2014 due to data availability. Three messages emerge from this analysis:  In India and Pakistan, which achieved high rates of GDP growth, increases in TFP made an important contribution to growth.  The contribution of TFP gains to economic growth in the region has declined across the four countries. In India and Pakistan, the contribution of TFP to GDP growth has declined dramatically since 2011 and 2006, respectively. In Sri Lanka, while the contribution of TFP picked up in 2014, it has declined from the high level before 2009. In Bangladesh TFP has played a negligible role in GDP growth during the entire period of analysis.11  Non-ICT investment and increases in the number of workers have been the leading forces behind growth in all four countries. Although investment in ICT has increased its contribution to growth in India, and substantially so in Sri Lanka, most of the growth is still accounted for by more labor 10 Three important critiques have been leveled at this growth accounting framework. First, TFP is measured as a residual, providing an imperfect measure of shifts in the production function, which can reflect many determinants (e.g.: technical change, but also sustained political turmoil, external shocks, institutional changes or measurement errors). Second, the data are calculated by assuming a sufficient degree of competition in factor markets so that factor earnings are proportional to factor productivities. Third, growth accounting cannot measure the fundamental causes of growth (policies, institutions, and history), but simply examines the proximate causes. While these critiques have merit, the framework provides a simple and internally consistent way to organize data, and is useful in generating insights into the process of economic growth. 11 Note that the analysis does not distinguish between a structural decline in TFP growth and a decline from idle or poorly-allocated factors of production during cyclical periods of economic slowdown. xvii (rather than higher-quality labor) and non-ICT investment – both factors subject to diminishing returns. Figure 1.9: Growth Accounting – Decomposition of GDP Growth 1990-2014 Bangladesh India 8 10 6 4 5 2 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 0 -2 -5 TFP Growth Contribution Non-ICT Capital TFP Growth Contribution Non-ICT Capital Contribution ICT Capital Contribution Labor Quantity Contribution ICT Capital Contribution Labor Quantity Contribution Labor Quality GDP Growth Contribution Labor Quality GDP Growth Pakistan Sri Lanka 10 11 8 6 6 4 2 1 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 -2 -4 -4 TFP Growth Contribution Non-ICT Capital TFP Growth Contribution Non-ICT Capital Contribution ICT Capital Contribution Labor Quantity Contribution ICT Capital Contribution Labor Quantity Contribution Labor Quality GDP Growth Contribution Labor Quality GDP Growth Source: Conference Board 1.2.2 Sectoral challenge: slow pace of structural transformation As in other countries at similar stages of development, resources in South Asia are moving from agriculture to manufacturing and services (Figure 1.10). This shift in economic activities from lower- productivity, traditional sectors to more modern and productive ones is known as structural transformation. Between 1960 and 2013, the average share of agriculture in GDP in South Asia fell from 44 percent to 19 percent, while the share of industry increased from 18 to 29 percent. Over the same period, real GDP per capita of the region nearly quintupled. xviii Figure 1.10 Structural transformation in South Asia Source: World Development Indicators database Individual countries are at different points along this transformation, but the overall trend is quite consistent (for convenience, the eight countries are divided between the four larger--Figure 1.11, and the four smaller--Figure 1.12). An increase in real GDP per capita has been associated with a decline in the share of agriculture in employment and value added, and concomitant increases in the shares of services and industry.12 Maldives, a small island economy, has traditionally had a large services sector owing to tourism, and has not experienced much structural change in the last decade. The share of value added of industry, however, appears to plateau at certain levels of income (at least in the larger countries), consistent with the views of structural transformation in the literature (Herrendorf et al., 2013). The share of employment in agriculture has been consistently higher than the share of value added at all levels of GDP per capita, for the four countries with adequate data. This suggests that agricultural productivity has not improved substantially over the decades. The share of services in employment and value added in this sector is high in South Asia, particularly considering the relatively low income levels. The share of services in GDP appears to rise relatively early in the development process, at below US$700 per person (in 2005 prices). This pattern differs from the experience of mature industrialized countries such as United Kingdom, France, and the United States, where the share of the services sector reached high levels only at high levels of GDP per capita (e.g., see Verma, 2012). 12 Sectoral employment shares data for Afghanistan, Bhutan, Maldives, and Nepal are available for very few years, and has not been used in this analysis. xix Figure 1.11 Employment and value added shares by sector for Bangladesh, India, Pakistan, and Sri Lanka Agriculture Industry Services xx Figure 1.12 Value added shares by sector in Afghanistan, Bhutan, Maldives, and Nepal Agriculture Industry Services Workers in South Asia are moving from agriculture to the higher-productivity manufacturing and services sectors (figure 1.10), a transition which has been associated with increases in aggregate productivity.13 However, the share of agriculture in total employment in the region remains high at over 50 percent, despite the fact that labor productivity in industry and services is several times that of agriculture. For example, in India average labor productivity in industry from 2004-2013 was approximately 5 times, and services 6.5 times, the level of productivity in agriculture and services, illustrating the potential for major productivity gains in the region from accelerating the process of structural transformation. 13 The economic development literature has long recognized the role of structural transformation in boosting aggregate productivity. Baumol (1967), Dekle and Vandenbroucke (2012), Ngai and Pissarides (2004), Acemoglu and Guerrieri (2008), and others have shown that differential productivity growth across sectors will attract resources to more productive parts of the economy. xxi Differences in labor productivity play an important role in pulling labor to more productive sectors, above and beyond the “natural” rate of structural transformation (Table 1.4).14 In other words, movement from agriculture to industry and services is significantly faster in periods when the difference in productivity between the two sectors is larger. In this way, the process of structural transformation in South Asia is similar to the experience of OECD countries, although there are differences due either to the two groups of countries being at a different stage of the same path, or the paths being a bit different. In either case, the response of employment to differences in sectoral productivity levels is much higher in South Asia than in OECD economies, indicating that boosting productivity growth in manufacturing and services in South Asia carries a much greater potential for accelerating structural transformation, increasing non-farm employment, and raising income growth. Table 1.4 Labor productivity and structural transformation from agriculture to industry from agriculture to services OECD countries India, Pakistan, and Sri Lanka Standard errors in parentheses, country fixed effects included but not shown. * p<0.05, ** p<0.01, *** p<0.001 1.2.3 Firm challenge: firm growth is low and resources are trapped in small firms In addition to resources moving from less to more productive sectors (i.e., structural transformation), productivity growth can be driven by movement in resources from less-productive to more-productive firms within narrowly defined economic activities. When this mechanism does not function as effectively as it could – for example, due to barriers to competition – the economy suffers from misallocation of resources (Box 1.1). 14 A simple reduced form model is used to quantify the role of productivity in driving the process of structural transformation. The movement of labor from one sector to another is a function of productivity differentials across sectors (see Annex for the description of the model). The natural rate of structural transformation is measured by the coefficient on the share of employment lagged one period. The coefficient on the lagged employment share is always positive, significant, and significantly less than one, indicating a “natural” downward trend in the employment share of agriculture xxii Box 1.1 Barriers to competition and productivity dispersion Consider a fictional example of an economy with two firms. One operates at low productivity, but manages to survive in the market because it has political connections through which it secures subsidized credit. The other firm has no political connections, borrows at the market rate, and therefore faces higher costs. However, the second firm has higher productivity, which enables it to compete with the first firm.15 If labor and capital were to move from the firm with low productivity to the firm with high productivity, aggregate output would be higher, and the difference between productivity levels in the two firms would be lower. Thus, the misallocation of capital results in low output per worker, on average across the two firms. Barriers to competition in the real world also can create and perpetuate substantial misallocation of resources. For example, informal, low productivity retailers in Brazil can secure a large share of the market because they are subject to less stringent labor market regulations, and thus lower labor costs, than higher-productivity supermarkets (Mc Kinsey Global Institute, 1998). Subsidized loans and differential tax-code treatment in Japan are used “to keep mom-and-pop retailers from going out of business” (Lewis 2004). And, prior to reforms, severe restrictions on FDI in retail prevented investment by global best practice retailers in India.16 Evidence shows that the costs of misallocation, in terms of aggregate productivity, can be very high. In a seminal study, Hsieh and Klenow (2009) measure resource misallocation in China and India by comparing productivity dispersion among firms in these countries with the US market.17 They find that firms in China and India produce the same products with vastly different levels of productivity, with a range that is much wider than in the United States. Reducing these productivity gaps to the level of efficiency observed in the United States would increase TFP by 40-60 percent in India (and by 30-50 percent in China), and output would increase by twice as much if investment increased in response to higher productivity. Conversely, a more rapid expansion by less efficient firms than by more efficient firms in the early 1990s reduced TFP growth in Indian manufacturing by 2 percent over the 1987-94 period (Hsieh and Klenow, 2009). While the detailed data required to replicate the Hsieh and Klenow (2009) analysis for other countries in the region are not available, indicators of productivity differences among firms compared to India can be calculated for a few sectors in Bangladesh, Sri Lanka and Nepal. The results show that substantial scope exists for improving productivity by shifting labor and capital to higher-productivity firms in the Nepalese food beverages sector (firms in the lowest ten percent of the TFP distribution are more than 5 times less productive than those at the highest ten percent) and other manufacturing, and less so 15 This fictional example is taken from Hsieh and Klenow (2009). 16 Both examples from Lewis (2004), pp 14-15 and OECD FDI Restrictiveness Indicators. 17 In theory, with no distortions firms producing similar products would have the same level of productivity. If that were not the case, then resources would move from the low productivity firm to higher returns in the high productivity firm, driving productivity of the former upward, and of the latter downward. Given that factors omitted in the model may be responsible for productivity differences, the benchmark for comparison is not ‘zero’ productivity dispersion, but that of the US – a relatively undistorted market. xxiii (relative to India) in Sri Lanka and Bangladesh (Table 1.5).18 The results also hint at the importance of competition: in Bangladesh and Sri Lanka, firms in the apparel sector, which is significantly exposed to competition through exports, shows less productivity dispersion than in India. Table 1.5 TFP dispersion (coefficients of variation) by sector and country (Relative to India) Bangladesh India Sri Lanka Food and Beverages 0.64 1.00 1.56 Textiles 1.79 1.00 Apparel 0.65 1.00 0.98 Basic Metals 2.75 1.00 Other Manufacturing 1.49 1.00 2.95 Source: David Francis based on World Bank Enterprise Surveys Some authors have explained high productivity dispersion – and, consequently, lower overall productivity – by a disproportionately high number of small, unproductive firms which neither grow nor exit, releasing resources into the economy (Li and Rama, 2015). For example, the share of manufacturing firms with less than 10 employees in India is “almost visually indistinguishable from 100 percent” and the most common observation in the sample is a firm with a single employee (Hsieh and Olken 2014). By contrast, in the United States, the most common observation in the sample is a firm with 45 employees. Moreover, the dominance of small firms in India appears to be the same, or perhaps even greater, as it was more than twenty years ago (Figure 1.13). Evidence from the World Bank Enterprise Surveys indicates that the importance of small firms in South Asia is greater than in comparator East Asian countries. 19 Countries with higher levels of GDP per capita tend to have a smaller share of firms with only a few employees (see Figure 1.1.16, Figure 1.1.17, and Figure 1.18 in the Annex to this chapter). 18 As argued by Li and Rama (2015), because micro and small firms are generally underrepresented in datasets like Enterprise Surveys, these indicators of dispersion of productivity likely underestimate true dispersion. 19 The WB enterprise surveys collect information about firms in different countries around the world in a harmonized way providing comparable cross-country information. The dataset includes only formal firms with at least 5 employees. The reader should keep in mind this truncation when interpreting results. xxiv Figure 1.13 Size distribution of firms in India, 1990-2010 xxv Source: Authors’ calculations using combined ASI/NSS data Table 1.6: Distribution of firms by size – South Asia and comparator countries ( percent) Small (5-19) Medium (20-99) Large (100 and more) Afghanistan2014 69.37 25.51 5.12 Bangladesh2013 37.29 35.98 26.74 Bhutan2015 70.88 24.74 4.39 China2012 55.05 32.4 12.55 India2014 42.87 43.75 13.37 Indonesia2009 87.99 9.76 2.25 Nepal2013 82.25 15.58 2.17 Pakistan2013 44.32 39.58 16.1 Philippines2009 52.39 34.63 12.98 SriLanka2011 75.89 18.18 5.93 Vietnam2009 45.22 36.43 18.34 Source: Authors’ calcuations using Enterprise Surveys xxvi Figure 1.14 Firm size distribution in South Asia and comparator countries There are many reasons why small firms may be less productive than larger ones: economies of scale, access to finance, better employees, and stronger business practices. Larger firms tend to innovate xxvii more, particularly in terms of process and organization, because they can more easily secure finance for risky projects and because of the potential for economies of scale in research and development investments (see Del Mel et al., 2008; Cohen and Klepper, 1996b and Ayyagari et al 2007). Larger firms also tend to invest more in administration and adopt better management and overall business practices, which are highly correlated with firm performance (Bloom and Van Reenen 2007; Bloom et al., 2011).20 Regardless of the channel, productivity does appear to be lower in small than in large firms in Asia. For example, in both India and China, sales and value added per worker in small firms are much lower than in large firms (Table 1.7).21,22 This pattern is also documented by Hsieh and Klenow (2009), who argue that the relationship between productivity and size is stronger in China and India than in the United States due to distortions which prevent firms from achieving optimal size (and consistent with Banerjee and Duflo’s (2005) contention that Indian policies constrain its most efficient producers and coddle its least efficient ones). Moreover, India’s productivity growth between 1993 and 2007 was associated with productivity gains within large manufacturing plants (200 or more workers) rather than with gains within small firms or with reallocation between plants (Bollard et al., 2011). Table 1.7 Sales & Value Added per Worker, India and China, By Firm Size India (2014) China (2012) India (2014) China (2012) Sales per Sales per VA per VA per N N N N worker worker worker worker All $ 15,706 6686 $ 36,659 1620 $ 5,855 4774 $ 19,724 1349 Large(100 and over) $ 18,993 1564 $ 37,090 699 $ 6,707 1144 $ 20,279 579 Medium(20-99) $ 14,960 3113 $ 38,880 705 $ 5,611 2378 $ 20,434 587 Small(<20) $ 15,553 2009 $ 33,826 216 $ 5,912 1252 $ 18,475 183 Source: Francis (2015) Another symptom of resource misallocation in the region is that firms face difficulties in growing. In India, manufacturing plants that are 40 years old are only 40 percent larger than young manufacturing plants (under 5 years of age), while in the United States older plants are more than seven times larger than the younger ones (Hsieh and Klenow 2014). The same conclusion holds more broadly across the region: firms aged 25 years or more in India, Sri Lanka, Bangladesh, and Bhutan are only 50-90 percent larger than firms aged five years or less. By contrast, older firms in Vietnam are on average 4.5 times the size of younger firms, in Indonesia older firms are on average 4.8 times as large, and in China 2.4 times 20 This conclusion holds even when restricting the sample to small and micro firms. Among these, the larger did better at business practices within a sample of small firms in Bangladesh, Chile, Ghana, Kenya, Mexico, Nigeria and Sri Lanka (McKenzie and Woodruff []). 21 Note that this calculation is based on the World Bank Enterprise Surveys, which do not include most micro firms, so it most likely underestimates productivity differences that would be observed if firms of all sizes were considered. 22 An important caveat is that it is possible that the value added per worker differences between firms within different size classes are to some extent explained by differences in average worker quality, as workers with higher abilities or skills may self-select into larger firms. xxviii as large (Figure 1.15). One explanation for this is that in India (and also in China), within narrowly defined industries, larger plants have higher marginal products of labor and capital, while in the United States the difference is much smaller (Hsieh and Klenow 2009). A higher marginal product of labor in large firms likely indicates distortions that prevent firm growth, since in a world without distortions, firms would continue to expand until the marginal product of labor or capital equalizes across firms. Figure 1.15: Firm employment and export shares by age Note: The left panel of the figure shows the average number of workers for different age cohorts of firms normalized with respect to the first age cohort (< 5 years old firms), using the enterprise surveys for SAR and comparator countries). The right panel shows the average export share (over total sales) for different age cohorts of firms, also normalized to the first age cohort. Barriers to the growth of firms can likely be found in policies. Across the region, licensing and size restrictions (which have declined in importance but still exist), labor regulations that increase the cost of hiring and firing, financial sector regulations that favor small enterprises, and inadequate bankruptcy laws may have limited the ability of efficient plants to grow and enabled inefficient plants to survive. Problems in enforcing contracts in India, for example, make it costly to hire the right managers, which is crucial for firms’ growth (Bloom et al., 2013). Taxes or labor costs that affect larger firms more than smaller firms may reduce the return on investment in large firms. Impediments to reaching foreign markets, both from trade policy and high costs of logistics, can also impede expansion. xxix 1.3 Annex The following model describes a simple process of reallocation of labor across different sectors, assuming a two sector economy consisting of agriculture and industry. Employment in each sector can change due to net addition of new workers as labor force grows, or due to inter-sectoral migration of labor. This can be represented as follows: (1) where LA is the number of people employed in agriculture, gA is the (constant) net rate of employment growth in agriculture, and M is the migration of laborers from industry to agriculture. Dividing equation (1) by total employment L, and assuming total employment grows at a constant net rate of g yields: (2) Define m as the percentage of population that migrates from one sector to another, such that m = M/Lt. Assuming there are no frictions in the movement of labor from agriculture to industry (or, equivalently, that these frictions remain constant over time), inter-sectoral migration should be driven by the wage gap between two sectors. For example, if the industrial wage rate is higher than the agricultural wage rate, workers should migrate from agriculture to industry. Further, wage rate in a sector is equal to the marginal product of labor in that sector, which in turn is a function of labor productivity. Thus m can be written as: m = f(wA,wI) = f(h(ωA,ωI)) = g(ωA,ωI) (3) where w is the wage rate in a sector, and ω is the labor productivity. Equation (2) can therefore be expressed as follows: (4) This equation can be estimated using the following reduced form model for OECD countries and South Asian economies for which we have sufficient data on employment (India, Pakistan, and Sri Lanka). Proxying labor productivity in each sector with value-added per worker yields the following: where i represents agriculture and j represents industry or services, k represents the country dummy and t represents time. V APW is the value added per worker calculated as the value added at constant prices divided by the number of workers. Li/L is the employment share calculated as the number of employees in a sector divided by total employment. xxx Figure 1.1.16: Scatter plot of quartile 1 of Figure 1.1.17: Scatter plot of quartile 2 employment distribution of firms and income per (median) of employment distribution of firms capita and income per capita 2 2 lnq1 = 1.7469 + .03336 lngdppc R = 2.4% lnmed = 2.1705 + .05022 lngdppc R = 3.2% 4 3 BGD13 3.5 2.5 BGD07 PAK13 IND14 BGD07 BGD13 Log median size IND14 Log q1 size PAK13 3 2 AFG08 BTN09 NPL13 AFG14 PAK07 LKA11 NPL09 2.5 1.5 AFG08 AFG14 BTN09 PAK07 NPL09 LKA11 2 1 NPL13 1.5 .5 4 6 8 10 12 4 6 8 10 12 Log GDP per capita, constant 2005$ Log GDP per capita, constant 2005$ Regression F-test pvalue=0.029 Regression F-test pvalue=0.012 Source: Enterprise survey and WDI Source: Enterprise survey and WDI Figure 1.18: Scatter plot of quartile 3 of employment distribution of firms and income per capita 2 lnq3 = 2.8942 + .05646 lngdppc R = 2.2% 5 BGD07 BGD13 4 PAK13 Log q3 size IND14 AFG08 AFG14 PAK07 BTN09 3 LKA11 NPL09 NPL13 2 4 6 8 10 12 Log GDP per capita, constant 2005$ Regression F-test pvalue=0.038 Source: Enterprise survey and WDI xxxi Source: Authors’ elaboration based on WB Enterprise Surveys Figure 1.19: Export Orientation Index – Figure 1.20: Import Orientation Index – Ranking SAR countries and comparators Ranking SAR countries and comparators VNM MDV VNM 50 60 MDV 40 Import Orientation Index PHL IND IDN PHL IDN 20 IND CHN AFG BGD 0 BTN CHN NPL BGD PAK PAK 0 NPL BTN AFG -20 -50 -40 Source: Authors’ calculations based on WDI FDI, Imports and Exports/GDP ratios, per capita income, and regional patterns 1990-2014 (1) (2) (3) (4) (5) (6) VARIABLES FDI/GDP FDI/GDP Imp/GDP Imp/GDP Exp/GDP Exp/GDP GDP Per Capita (PPP, Constant 2005) 6.91e-05*** 0.000199*** 0.000750*** (1.07e-05) (2.92e-05) (2.51e-05) ECA 0.705 0.0880 -11.63*** -12.98*** -8.802*** -13.82*** (0.521) (0.533) (1.425) (1.432) (1.348) (1.230) LAC 0.622 0.985* -15.10*** -12.91*** -14.36*** -9.476*** (0.552) (0.562) (1.526) (1.541) (1.443) (1.324) MENA -1.688*** -2.729*** -16.64*** -18.68*** -9.538*** -17.09*** (0.652) (0.694) (1.820) (1.878) (1.721) (1.613) North America -2.259 -4.128*** -35.99*** -40.89*** -29.72*** -48.19*** (1.485) (1.520) (4.119) (4.159) (3.895) (3.573) SAR -2.855*** -2.266** -21.09*** -21.21*** -23.57*** -17.62*** (0.858) (0.893) (2.325) (2.396) (2.199) (2.059) SSA 0.178 0.881* -14.41*** -11.90*** -20.93*** -11.50*** (0.516) (0.530) (1.425) (1.464) (1.348) (1.258) Constant 2.238** 1.368 54.82*** 51.91*** 47.44*** 37.05*** (0.939) (0.971) (2.481) (2.531) (2.347) (2.175) Year dummies Yes Yes Yes Yes Yes Yes Observations 4,062 3,979 4,154 4,075 4,154 4,075 R-squared 0.034 0.043 0.055 0.069 0.090 0.260 Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Source: Authors’ calculations based on WDI Changes in FDI, Imports and Exports/GDP ratios, per capita income, and regional patterns 1990-2014 (1) (2) (3) (4) (5) (6) VARIABLES D1.FDI/GDP D1.FDI/GDP D1.Imp/GDP D1.Imp/GDP D1.Exp/GDP D1.Exp/GDP GDP Per Capita (PPP, Constant 2005) 5.29e-05*** 0.000216*** 0.000833*** (1.21e-05) (4.29e-05) (4.04e-05) xxxii ECA 1.888*** 1.428** -3.495* -4.859** -2.665 -7.906*** (0.586) (0.599) (2.093) (2.105) (2.083) (1.984) LAC 0.684 0.860 -12.61*** -10.71*** -11.97*** -6.366*** (0.618) (0.630) (2.238) (2.256) (2.227) (2.126) MENA -0.397 -0.906 -5.583** -5.603** -0.998 -7.491*** (0.731) (0.771) (2.641) (2.714) (2.629) (2.557) North America 0.973 -0.450 -5.025 -10.26* -3.048 -23.23*** (1.644) (1.685) (5.925) (5.999) (5.896) (5.653) SAR -0.669 -0.366 -4.277 -4.890 -16.20*** -10.54*** (0.998) (1.041) (3.475) (3.598) (3.458) (3.391) SSA 1.660*** 2.196*** -0.633 2.479 -10.02*** 0.495 (0.580) (0.599) (2.102) (2.175) (2.092) (2.050) Constant -1.592 -2.283* 4.477 1.022 6.143 -6.049* (1.152) (1.193) (3.796) (3.867) (3.777) (3.644) Year dummies Yes Yes Yes Yes Yes Yes Observations 3,892 3,811 3,896 3,830 3,896 3,830 R-squared 0.007 0.012 0.012 0.018 0.016 0.119 Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Source: Authors’ calculations based on WDI 2. Productivity performance: firms and linkages While productivity can be measured at different levels – macro, sectoral, geographic, etc. – the most robust and intuitive representation is at the level of the firm. The focus on the firm as the unit of analysis and firm dynamics as the driver of productivity growth goes back at least as far Schumpeter (1911), with competition playing a key role in forcing inefficient, unproductive, or unprofitable firms to either improve or exit and transfer their resources to more efficient, productive, or profitable firms, thus boosting economy-wide productivity. Formally, there are two mutually reinforcing mechanisms – spurred on by competition in product and factor markets – that increase productivity. First, greater competition, from either domestic or international sources, pushes firms to become more efficient at doing what they do: for example, by learning from international exposure, investing in innovation, improving business practices, adopting better technology including ICT, and improving the input mix. This is the within-firm component of productivity growth. Second, competition also induces inefficient firms to transfer resources to more efficient ones or exit altogether, boosting economy-wide productivity – this is the between-firm component of productivity growth (Cabral, 2007). Resources can flow from less to more productive uses due to improvements in standard factors like infrastructure, business environment, etc., but also as a result of participation in global value chains (Saia et al, 2015) and agglomeration economies (Desmet and Rossi‐Hansberg 2009; Michaels, Rauch and Redding 2012). The decomposition of changes in productivity into between- and within-firm components (with the former further broken down into contributions from firm entry and exit) has become a standard approach to thinking about productivity dynamics (Olley and Pakes, 1996; Melitz and Polanec, 2012). Unfortunately, none of the countries in the region carry out large, longitudinal firm surveys which would xxxiii directly allow for this type of analysis.23 Therefore, this report approximates the spirit of the decomposition by using cross-sectional firm data to consider elements that are likely to impact productivity across and within firms, in turn. This section begins with the business environment, agglomeration economies and participation in global value chains (GVCs) as determinants of performance across firms, and then considers the role of technology and innovation in determining within-firm productivity. Several broad conclusions emerge. South Asia scores poorly on many indicators of the quality of the business environment, which greatly constrains firms’ productivity in general and particularly limits the growth of firms with high levels of productivity. Productivity gains through the further concentration of economic activity require a reduction in distortions in product and factor markets, particularly barriers that limit the flow of resources between districts and states. Participation in GVCs can raise productivity through exposure to competition and knowledge spill-overs from connections with lead firms; however, South Asian participation in GVCs is largely confined to apparel. Reducing trade barriers, increasing skills, and improving logistics would facilitate greater participation in GVCs. Finally, access to technology varies greatly across South Asian economies, ranging from extensive technology use in India to limited ICT adoption in Bangladesh and Nepal. Even among lead countries, however, the use of e-commerce and other productivity-enhancing online business tools is relatively low. Innovation tends to be concentrated in few, mature firms, and is likely to represent imitation of existing products rather than new products. Greater investment in R&D, improved resource management, and the development of skills that are complementary to technology could play a critical role in increasing innovation and thus boosting productivity. 2.1 Business environment challenges continue to weigh on firm performance Much of the macro and sectoral challenges to productivity in South Asia can be traced to a difficult operating environment for the region’s firms. The business environment (investment climate) has received a great deal of attention in policy and empirical literature as a major constraint hampering firm productivity in the region. Most studies define the investment climate as the environment which determines entrepreneurs’ ability to work efficiently, such as the degree of difficulty in accessing production inputs and dealing with regulatory and legal requirements, and the level of security in running operations and obtaining payments. As argued by Hallward-Driemeier (2007), an inefficient business environment will lead to low and uncertain returns on investment, dragging down overall productivity and possibly more than offsetting technical improvements on the factory floor. 23 There are two exceptions in the case of India, but neither is fully satisfactory for this type of analysis. India’s Annual Survey of Industries (ASI) data is available with panel identifiers starting from 1998-99; however, firms below a certain size threshold (100 employees) are only sampled once every 4-5 years making it difficult to determine whether a firm exited or was not sampled in a given year. The Prowess database by the Centre for Monitoring Indian Economy (CMIE) has data for more than 10,000 manufacturing firms dating back to 1990 but these are mostly large, publicly listed companies. xxxiv The business environment in South Asia – whether measured through expert surveys and de jure requirements (e.g., the Doing Business methodology) or through responses by firms (e.g., the World Bank’s Enterprise Surveys) – is challenging. In a 2006 Investment Climate Assessment of the region, the World Bank argued that South Asian countries under-perform comparators on many investment climate dimensions, including infrastructure and electricity supply, access to finance, employee skills, and corruption (World Bank, 2006). Similar results emerge from the most recent round of Enterprise Surveys, where an average firm in South Asia consistently ranks each investment climate constraint as more binding than does an average firm in China or Vietnam (Table 2.1). While performance varies substantially across countries and indicators – pointing to significant potential for improvement by leveraging best practices from within the region – the overall gap puts South Asia’s firms at a clear disadvantage vis-à-vis select comparators in other parts of the world. Table 2.1 Investment climate constraints in South Asia and comparator countries Sri South South Afghanistan Bangladesh Bhutan India Nepal Pakistan Lanka Asia China Africa Turkey Vietnam (2014) (2013) (2015) (2014) (2013) (2013) (2011) (2012) (2007) (2013) (2015) Access to 49 23 19 15 40 22 33 18 5 16 10 14 Finance Political 76 76 12 16 85 34 13 28 1 3 13 3 Environment Crime 58 8 1 5 14 35 7 10 1 38 8 5 Taxes 56 20 24 31 23 55 41 36 7 8 25 8 Corruption 62 49 4 36 42 64 15 42 1 17 12 5 Informality 33 9 10 17 29 12 28 14 7 11 14 11 Infrastructure 81 55 29 26 79 79 36 42 6 24 25 18 Electricity 66 52 14 21 69 75 26 35 3 21 18 4 Telecom 59 3 15 4 3 14 6 7 4 4 9 8 Transport 43 15 14 10 32 27 12 15 3 4 10 10 Labor 11 3 15 11 3 12 13 12 1 6 6 4 Regulations Workforce 53 16 14 9 9 23 16 13 2 9 10 8 Education Trade & 47 8 9 12 29 30 31 18 4 2 11 24 Customs Note: Cells indicate the percentage of firms who view any given obstacle as a major or severe constraint. Source: Staff calculations using Enterprise Surveys. Lessons from case studies in Volume II of this report echo the findings from the surveys. Difficulties in importing goods, poor trade logistics, and high protection rates have made participation in global markets more costly for firms across the region, while outdated standards and restrictive regulations have limited competition in automotive and agribusiness sectors. Difficulties in accessing well-located and well-serviced industrial land and poor availability of skilled workers have also emerged as important bottlenecks to firm growth. xxxv While firms may have different capabilities to overcome various investment climate constraints, studies show that an average firm in South Asia experiences a sizeable productivity loss from the poor investment climate. For example, Hallward-Driemeier (2007) finds a significant negative impact of customs delays, power outages, poor access to finance, and limited connectivity on total factor productivity and investment rates of garment firms across the region. In particular, the author shows that, if the business environment for firms in India were the same as in China, firm productivity could be one percentage point higher. Analysis that approximates the approach of Hallward-Driemeier (2007) with the most recent round of Enterprise Surveys shows that, by and large, investment climate challenges continue to affect firm performance in the region. Across a wide sample of manufacturing firms in Bangladesh, Bhutan, India, Nepal, Pakistan, and Sri Lanka, both output and value added per worker are systematically lower when firms face greater business environment constraints (Table 2.2). Table 2.2 Impact of investment climate on firm performance Output per worker Value added per worker (log) (log) Losses from power outages (log) -0.024** -0.038*** Losses in transit (log) -0.079*** -0.036*** Improved access to finance 0.342*** 0.403*** Observations 4,566 4,498 R-squared 0.039 0.027 Sector dummies (number) 20 19 *** p<0.01, ** p<0.05, * p<0.1 A restrictive business environment can be particularly damaging to firms that have the most to contribute to productivity growth and job creation. Many investment climate constraints can be particularly burdensome for small firms (Word Bank, 2006), limiting their ability to grow and create employment. Some evidence also shows that higher productivity firms in the region may actually face greater constraints in accessing public services, suggesting that investment climate deficiencies are particularly binding on firms that would grow more rapidly and create more jobs in the absence of distortions (Carlin and Schaffer, 2012). The severity of investment climate obstacles in the region and their adverse impact on productivity have given rise to a series of wide-ranging reforms to address constraints along each aspect of the business environment. For example, the ADB (2006), ADB and World Bank (2004), Afram and Salvi Del Pero (2012), Ferrari and Dhingra (2009), OECD (2009), and World Bank (2006a, 2006b, 2008a, 2008b, 2009, 2010), among others, have proposed a range of policy actions to regional authorities. Improving the investment climate is also high on the regional authorities’ own policy agendas: nearly every country in the region is taking concrete steps to strengthen the business environment (Box 2.1). Box 2.1 Efforts to improve the investment climate in South Asia In Bangladesh, the authorities initiated a number of reforms to address binding constraints impeding xxxvi private sector growth in the recent years. New pieces of business friendly legislations include: (i) the Economic Zones Act of 2010, modernizing the country's economic zones agenda, including the institutional setup, and allowing for more efficient incentives and private participation; (ii) the Competition Law, which is meant to uphold a level playing field for businesses; and (iii) a new Value Added Tax Law that eases the compliance mechanisms for businesses and reduces discretionary exemptions. The authorities also introduced regulatory reforms streamlining business registration, trademark and patent registration, and simplified trade licenses and construction permits: a total of 56 regulatory processes have been reformed in recent years including company, investment, tax, and trademark registrations; trade licenses at local government levels; subordinate rules under three different tax laws for better contract enforcement; and dispute resolution. To foster trade competitiveness, the Government has launched a trade information portal and introduced risk management in clearance process, as well as taking preparatory steps to a new Customs Act, a national single window for trade, and making multi-modal transport effective for trade logistics. Responding to concerns on fragmented policy coordination, these reforms are being carried out in the context of a formal, structured public-private dialogue. In India, the authorities recently launched a new ambitious program of regulatory reform. In 2015, the authorities eliminated the minimum capital requirement and ended the requirement to obtain a certificate to commence business operations. Now Indian entrepreneurs no longer need to deposit 100,000 Indian rupees ($1,629)—equivalent to 111% of income per capita—in order to start a local limited liability company, and can start business five days earlier. Utilities in Delhi and Mumbai undertook significant business process reengineering, combining inspections and procedures to reduce the time required for companies to get connected to the grid and get on with their business. In addition, the Central Government called for all states to adopt automation in registration processes, move towards effective single windows, and implement risk-based inspection regimes that introduce self-certification and third-party audit schemes to lessen the burden of inspections on low and medium risk businesses. In Sri Lanka, the authorities have recently taken steps to eliminate obstacles to FDI, including a) upfront payment of the land lease tax for foreign companies; b) elimination of minimum investment requirements in ICT, R&D and vocational training; and c) implementation of online processing of business visas. Regulatory barriers to trade are being reduced through agreement to ratify the WTO Trade Facilitation Agreement (which is the basis for a medium-term trade reform agenda) and creation of the National Trade Facilitation Committee, which will be the body in charge of leading trade facilitation reform. Finally, a new Secured Transactions Act will enable the use of movables assets as collateral for bank loans, improving access to finance for SMEs. In Pakistan, the authorities have recently embarked on a two-year roadmap to improve the country’s Doing Business ranking to the top 100 by 2018, by preparing a DB Reform Strategy. The strategy provides reform recommendations for all the DB indicators, and also identifies institutions at the provincial and federal level with the mandate to carry out the reforms. The strategy is currently being implemented both at the federal and provincial levels, after having been endorsed by an ‘Ease of Doing Business Committee’ formed by the Government on investment climate reforms. In addition, and complementing the big push on Doing Business, the Government of Pakistan also took a series of legislative actions and implemented regulations to improve access to credit, payment of taxes, and financial intermediation (capital markets and housing finance), as well as reforms in financial transparency and oversight of state owned enterprises. xxxvii While the importance of addressing investment climate constraints in the region is beyond question, the issues are well-known and policy pathways to address them have been mapped out by various institutions, including the region’s own governments. With the exception of identification of specific constraints which are particularly binding on the region’s top performing firms – as revealed by firm interviews and case studies in Volume II of this report – there is little that the current study can add to the vast body of knowledge on the issue. Therefore, the discussion that follows focuses on newer, less-researched determinants and correlates of firm productivity in South Asia: agglomeration economies, value chains, and firm capabilities, including technology and innovation. 2.2 Productivity-boosting agglomeration economies are under-leveraged Agglomeration economies are the benefits that arise when firms and people locate near one another (e.g., in cities and industrial clusters). In South Asia – where concentration of economic activity is rather high – there is plenty of potential for agglomeration economies to increase productivity. However, distortions in goods and factor market prevent resources from flowing to more productive firms, with state and district borders forming boundaries to efficient allocation of resources. Therefore, despite the statistically significant effect of agglomeration on productivity – which in South Asia operates mainly via urbanization (cities) rather than localization (clusters) – the full productivity benefits of agglomeration economies are yet to be realized. The following discussion develops these observations in more detail. 2.2.1 Economic activity in South Asia is highly concentrated Economic activity tends to be geographically concentrated. This stylized fact holds in any country regardless of which industry is considered or concentration measure is used. For example, Rosenthal and Strange (2004) show that in the United States, heavy geographic concentration of industries is not limited to sectors highly dependent on particular raw materials (such as wood in the furniture industry), but extends to sectors where distance or location is much less of an issue, such as the software industry. Michaels, Rauch and Redding (2012) show that the transformation of the American and Brazilian economies over the last 100 years increased the concentration of resources in a few locations. Studies which control for geographical scale and borders, such as Duranton and Overman (2005), also show that economic activity is heavily concentrated. South Asia is no exception. Measures of firm concentration (the locational Gini, which is calculated in the same way as the Gini coefficient used to measure income distribution) in manufacturing in India, Sri Lanka and Bangladesh are quite high (0.48 in Sri Lanka, 0.53 in Bangladesh and 0.67 in India).24 In India, the five largest districts account for 18 percent of total employment, a share that has not moved appreciably over time, although which districts were in the top five has changed (Table 2.3).25 In Bangladesh, the share of the five largest districts – much greater than in India (partially because 24 These are calculated as simple averages of locational Gini coefficients for two-digit ISIC industries. 25 Only two of the top five districts in 1991 remained in the top five in 2009, and employment in the five largest districts in 1991, which comprised 17.4 percent of total employment, amounted to a much lower 14.0 percent in 2009. For the purposes of spatial analysis, the report consistently uses India’s 1989 districts, mapping new districts in later years to their 1989 “parent” districts. xxxviii Bangladesh has only around 60 districts while India has nearly 400) – increased more than 10 percentage points between 1995 and 2012, with most of the increase coming from areas outside Dhaka.26 In Sri Lanka, the five largest (out of 16 total) districts account for three-quarters of total employment, a share that has declined somewhat since the mid-1990s. Table 2.3 Employment in top five districts across South Asia (percent of total employment) Rank District name 1991 District name 2009 Rank in 1991 India 1 Greater Bombay 5.0 Madras 4.7 3 2 Nizamabad 4.1 Bangalore 4.4 4 3 Madras 3.3 Coimbatore 3.7 8 4 Bangalore 2.6 Greater Rewari 2.9 77 5 24 Parganas (North) 2.5 Greater Nagar 2.8 6 Total (top five) 17.4 Total (top five) 18.5 Bangladesh 1995 2012 Rank in 1995 1 Dhaka 36.6 Dhaka 35.4 1 2 Chittagong 15.9 Gazipur 16.6 6 3 Narayanganj 8.6 Chittagong 16.2 2 4 Sirajganj 5.0 Narayanganj 9.3 3 5 Khulna 3.8 Sirajganj 2.8 4 Total (top five) 69.8 Total (top five) 80.4 Sri Lanka 1995 2009 Rank in 1995 1 Colombo 37.4 Gampaha 27.7 2 2 Gampaha 29.7 Colombo 24.1 1 3 Galle 3.9 Kurunegala 9.4 5 4 Kandy 3.8 Kalutara 7.2 9 5 Kurunegala 3.1 Kandy 5.9 4 Total (top five) 77.9 Total (top five) 74.2 In most countries, modern manufacturing has tended to initially develop in very concentrated locations, often on the coast with access to international markets. As development proceeds, large coastal cities become congested, high‐cost locations, while at the same time the scale externalities they offer dissipate as manufacturing processes become more standardized with less need for the learning benefits of large cities. Manufacturing plants move first to suburban or nearby satellite locations and then to secondary cities in the hinterland. Yet, the degree of geographic concentration of manufacturing activities in South Asia has not changed substantially in the last two decades (Table 2.4). Table 2.4 Evolution of spatial concentration of manufacturing in South Asia India: state India: district Bangladesh: district Sri Lanka: district 26 Dhaka continues to dominate in levels, accounting for more than 35 percent of total employment. xxxix Raw index EG index Raw index EG index Raw index EG index Raw index EG index 1994 0.001 -0.000 0.006 0.005 1996 -0.013 0.001 -0.012 0.001 0.001 -0.009 1997 0.029 0.056 0.007 -0.011 1998 0.021 -0.001 1999 0.008 -0.025 0.030 -0.010 ** 2000 0.009 0.018* 0.012 0.021 0.026 -0.011 2001 -0.006 0.005 0.044 0.056 2002 0.044 0.056 2003 0.019 -0.023 2005 0.009 -0.036 ** 2006 -0.009 -0.001 -0.013 -0.005 0.071 -0.504 ** *** 2007 0.073 -1.660 2008 -0.000 -0.330 2009 0.019 0.012 0.016 0.003 -0.004 -0.167 2012 0.047 -0.084 N 351 348 351 348 123 118 263 260 * p<0.1; ** p<0.05; *** p<0.01 The raw concentration index in Table 2.4 measures the degree to which the geographic pattern of employment in the industry departs from the geographic pattern of manufacturing employment in the country as a whole, with larger values indicating greater concentration of activity.27 The Ellison-Glaeser (1997) index corrects for a potential bias in the raw index, as the value is larger in industries with only a small number of very large plants – and allows for comparisons across industries. In any event, neither measure of agglomeration has changed significantly over time (with the exception of one year in India and two years in Sri Lanka), suggesting that more productive locations have generally not been successful in attracting additional resources at the expense of less productive locations, therefore inhibiting overall productivity growth. Ghani et al (2012) provide one explanation for this: the authors suggest that a compositional change may be under way, with larger manufacturing entities moving out of city centers and a greater number of smaller service establishments moving into South Asia’s cities. 2.2.2 Agglomeration economies raise firm productivity Evidence shows that agglomeration is positively associated with firm performance in South Asia (e.g., Glaeser and Kerr, 2009, Ghani et al, 2011, and Mukim, 2011). Proximity to cities had a positive impact on non-farm employment in Nepal (Fafchamps and Shilpi, 2005). Similarly, households in Bangladesh with better access to major urban centers achieved higher returns to nonfarm activities (Deichmann et al., 2008). The concentration of high-return economic activities around Bangladesh’s growth poles (Dhaka and Chittagong) led to higher productivity in the eastern part of the country (World Bank 2008). Therefore, the ongoing urbanization process in South Asia (see World Bank, 2015) is likely to increase productivity due to agglomeration, which may reinforce the rise in productivity as workers move from lower-productivity agriculture to higher-productivity manufacturing and services (see above). , 2 27 This index is calculated as = ∑ =1 ( − ) , where i refers to a two-digit (ISIC) industry, s is location (district or state), and L is employment. xl Evidence from the industry studies in Volume II confirms the importance of agglomeration. The biggest benefits from agglomeration economies were found in the automotive industry, where geographic proximity to the customer has supported efforts to upgrade product, process and function. There is a high, robust correlation between productivity and the propensity of automotive firms of being located next to other automotive firms. While this correlation may in part reflect the selection of higher- productivity firms to participate in clusters, interviews suggest substantial benefits from being located in clusters. The location of leading firms in close proximity to suppliers and clients also has been important in apparel and agribusiness, although it is early to see agglomeration effects in the small electronics sector in the region. Agglomeration is usually discussed in terms of localization (firms in the same industry locating close to one another) or urbanization (firms in diverse industries locating in the same area—see Box 2.2). However, other indicators have also been used to measure the productivity benefits of agglomeration. Two measures of agglomeration at the plant level– market access and proximity to transport hubs – were correlated with productivity in 4 out of the 9 sectors in India investigated by Lall et al (2004). By contrast, they find that measures of localization are correlated with productivity in only two sectors, and measures of urbanization in none. These findings, however, rely on estimating the effects of agglomeration economies jointly with the estimation of the production function, potentially widening uncertainty around the estimates (see, e.g., Beveren, 2010; Combes et al, 2011). Moreover, correlation between measures of urbanization and of market access can be high, as both indicators are derived from the urban population in the same district. Box 2.2 Agglomeration and productivity Economic growth and geographical concentration of economic activities reinforce each other (Baldwin and Martin 2004, and Martin and Ottaviano 2001) through the forces of localization and urbanization. Localization economies are the productivity gains obtained by firms in the same industry locating close to one another and benefitting from sharing inputs, labor market pooling, and knowledge spillovers (Marshal, 1920). Urbanization economies are the productivity gains obtained by firms in other industries locating in the same area (Jacobs, 1969). These can arise when a diverse range of industries in a particular location enables firms to access suppliers from different industries, or when firms can benefit from R&D spillovers and/or access a generally higher-quality labor market. Recently, other sources have been suggested, such as home market effects (concentration of demand encourages agglomeration) and economies of consumption (consumers enjoy variety). According to a survey by Rosenthal and Strange (2004), whether localization or urbanization has the larger impact on a firm’s productivity is one of the oldest debate in this literature. Recent work (e.g., Martin et al, 2011) suggests that localization economies tend to be more effective in raising productivity than are urbanization economies (by 3-8 percent). However, most of these empirical studies are based on developed countries, and there is little evidence for developing countries. In South Asia, much of the economic activity tends to cluster (localize) either naturally or in response to policy distortions. In India’s automotive industry, the physical distribution of activities has been almost entirely subordinated to overcoming logistical difficulties. For example, in the early 2000s, Maruti Suzuki – India’s largest carmaker – relied on some 400 major suppliers located across the country, with some xli almost 2,500km distant from its main plant in Haryana. Its total logistics costs were up to four times as high as its wage bill and it had to carry large buffer stocks. In 2013, buffer stocks were brought down to zero and logistics costs slashed by requiring almost all suppliers to build, warehouse or locate within a few hours radius of the plant; today, approximately 80 percent of Maruti Suzuki’s suppliers are located within a 100km radius of the plant. In Pakistan’s apparel industry, leading firms and their suppliers in leather apparel – already clustered in Sialkot and benefiting from labor pooling, knowledge diffusion, and critical mass of offerings to encourage international buyers to travel to this remote place in Pakistan – privately financed the construction of an international airport and exhibition center to further develop the cluster. On the other hand, empirical evidence shows that urbanization economies appear to have had a larger impact on plant productivity than localization economies from 1995 to 2009 in India and Bangladesh (Table 2.5). The estimation approach follows the two-step strategy of Martin et al (2011): first deriving plant-level estimates of TFP and then assessing the impact of various aspects of agglomeration economies while also controlling for geographical location, industrial diversity, and the degree of competition (with the former measured at the state level and the latter two at district/industry level).28,29 Although statistically significant, the effects are relatively small and, in the case of India, seem to decline over time. In 1991, an increase of 10 percent in the number of employees in sectors other than where the firm operates was associated with a 0.5 percent increase in plant productivity in India; by 2009, the productivity impact halved to 0.2 percent. In contrast, in Bangladesh the effect has ranged from 0.2 to 0.4 percent. Table 2.5 Estimates of agglomeration economies in India and Bangladesh India Bangladesh 28 Note that, due to data limitations, the report cannot adequately address the issue of input endogeneity (especially capital) in the production function following the traditional approaches in the literature, such as Levinshon and Petrin (2003) and Olley and Pakes (1996). However, Van Beveren (2012) showed that differences between most parametric or semi-parametric methods to estimate TFP (including L&P and O&P) and OLS are minimal. Therefore, the approach taken here is to estimate the production function by OLS, allowing the parameters for capital (α) and labor (β) to vary for each two -digit ISIC sector. 29 Localization economies are defined as the log of the sum of all employees from sector s in region z apart from the number of employees in plant i plus one, while urbanization economies are defined as the log of the sum of all employees in region z apart from the employees in sector s plus one (this is done to ensure the inclusion of all plants in the estimation, since the existence of only one plant in a particular region would be discarded as log of zero is not defined). More specifically, the indices are defined as follows for each plant i, sector s, location z, and time period t: ( = ln⁡ − + 1) = ln⁡ ( − + 1) ′ 2 −1 = ln⁡( ∑ ( ) ) ′ − ≠ −1 2 = ln⁡( ∑ ( ) ) ∈ xlii 1991 1994 1996 2000 2006 2009 1995 1997 1999 2001 2005 2012 *** * ** *** Localization 0.01 0.00 -0.01 0.01 0.00 0.01 0.06 0.04 0.01 0.03 0.08 0.02 *** *** ** ** ** ** ** ** *** Urbanization 0.05 0.05 0.00 0.03 0.00 0.02 0.03 0.03 0.03 0.02 0.04 0.03*** Diversity 0.06*** 0.03** 0.11*** 0.05** 0.10*** 0.03 -0.02 -0.12 0.01 0.08** -0.18*** 0.03 ** *** * * *** ** ** Competition 0.02 0.02 0.02 -0.01 0.02 0.01 -0.12 -0.02 -0.02 -0.05 -0.05 -0.02 Observations 41,539 42,565 40,876 25,435 39,462 36,020 3,417 3,178 2,931 3,940 3,155 7,119 R-squared 0.05 0.05 0.02 0.03 0.04 0.03 0.03 0.02 0.01 0.02 0.03 0.01 Robust standard errors in parentheses. Constant and state dummies included but not shown. *** p<0.01, ** p<0.05, * p<0.1 In Sri Lanka, the data allow for even more stringent controls by estimating an unbalanced panel with plant fixed effects (Table 2.6). With this specification, the impact of urbanization economies on plant productivity is positive, particularly in the earlier part of the sample. Looking at the overall picture (from 1995 to 2009), a 10 percent increase in the number of employees in other sectors leads to an increase in productivity of 0.86 percent – higher than the cross-sectional estimates for India and Bangladesh. This suggests that our results using cross-section data might be underestimated and the effects of urbanization economies might be higher than our preliminary estimative. Table 2.6 Estimates of agglomeration economies in Sri Lanka VARIABLES 1995 - 2003 2006 - 2009 1995 - 2009 Localization 0.0315 -0.0300 -0.0002 Urbanization 0.1916*** -0.0410 0.0855*** Diversity 0.1207** 0.0217 0.0560 Competition 0.0159 0.0420 0.0141 Observations 17,125 4,873 21,998 R-squared 0.0044 0.0027 0.0014 Number of id 4,877 3,528 8,405 Robust standard errors in parentheses. Constant term included but not shown. *** p<0.01, ** p<0.05, * p<0.1 These results are for an average firm, while the impact of urbanization on firm productivity is greater for more productive firms and is not significantly different from zero for less productive firms.30 For example, in Bangladesh, the impact of urbanization at the 75th percentile for firm productivity is more than twice as large as at the 50th percentile (Table 2.7). In contrast to earlier results, this approach also identifies significant positive effects of localization economies on firms’ productivity, although results differ qualitatively for the two countries. While less productive firms in Bangladesh benefit from localization effects, in India it is the most productive firms that do so.31 30 To assess this, we estimate quantile regressions which the impact of the regressors conditional to the median or to other quantiles of the response variable rather the mean. The results are estimated at three quantiles (25%, 50% and 75%) for the last available year. 31 Results using other years show similar results. xliii Table 2.7 Agglomeration economies at different quantiles of firm distribution India / 2009 Bangladesh / 2012 VARIABLES 25% 50% 75% 25% 50% 75% Localization -0.0060 -0.0010 0.0109* 0.0179*** 0.0049 -0.0079 ** * *** Urbanization 0.0107 0.0207 0.0119 0.0022 0.0280 0.0674*** Diversity 0.0385*** 0.0505*** 0.0293* 0.0930*** 0.0308 0.0220 *** * Competition 0.0299 0.0124 0.0023 0.0231* 0.0063 -0.0150 32 State Dummies Yes Yes Yes No No No Testing (p-value) 25% = 75% 25% = 50% 50% = 75% 25% = 75% 25% = 50% 50% = 75% Localization 0.7% 25.5% 2.3% 4.7% 17.7% 22.6% Urbanization 85.0% 7.0% 10.3% 0.0% 0.0% 0.0% Observations 36,020 36,020 36,020 7,119 7,119 7,119 Robust standard errors in parentheses. Constant term included but not shown. *** p<0.01, ** p<0.05, * p<0.1 Overall, these findings show that agglomeration economies matter in South Asia, and the magnitude of the estimated impact of agglomeration on firm productivity is similar to that found in previous research focused on developed countries. But unlike the evidence for high income countries, urbanization economies in South Asia seem to matter more than localization economies – although the two are not mutually exclusive. Evidence from the case studies in Volume II underlines the importance of large cities for specialization by firms, as well as the emergence of specialization within smaller cities – e.g. the apparel cluster in Lahore, and the automotive clusters in Pune and Aurangabad. In part this is a natural response to increasing congestion and costs in the primary cities, a transition that also has been observed in China. It is important to emphasize that these correlations do not indicate causality, because data limitations have prevented us from addressing endogeneity when estimating the impact of agglomeration economies on productivity. Furthermore, the analysis did not take into account the potentially negative effects of agglomeration, such as congestion. Addressing such negative externalities might indicate that agglomeration promotes productivity, regardless of whether the effects are through localization or urbanization. 2.2.3 Resources do not flow easily to more productive firms Despite the productivity-enhancing benefits of agglomeration, there appear to be significant barriers to resources moving freely across internal geographical borders in South Asian countries. Duranton et al (2015) propose an empirically-motivated counterpart to the misallocation measure of Hsieh and Klenow (2009), defining misallocation as the (negative of) correlation between firm productivity and some measure of firm performance: be it output, employment, the use of capital, or land and other resources. With this definition, Duranton et al (2015) are able to decompose overall misallocation at the country 32 Model did not achieve convergence using Bangladesh state dummies. xliv level into contributions from different factors of production, as well as distinguish between misallocation of resources within entities (e.g., states or districts) and between them. Calculating misallocation in this way for three South Asian countries for which requisite data are available reveals that, in most cases, district or state borders in the region are “thick” in the sense that impediments to efficient allocation of resources between districts are stronger than distortions within districts. This holds true for all countries in markets for goods, and in India in markets for labor and capital as well (as evidenced by a relatively small contribution of the between-district component to overall efficiency plotted in Figure 2.1). Moreover, the results in Figure 2.1 show that across South Asia, factor (labor and capital) markets are more distorted than goods markets (this is evidenced by more negative values for misallocation of output than for misallocation of labor and capital – a more negative number means less misallocation). Figure 2.1 Decomposition of misallocation into between- and within-district components Within districts Between districts Output Bangladesh Labor Capital Output India Labor Capital Output Sri Lanka Labor Capital -1.2 -0.9 -0.6 -0.3 0 0.3 0.6 Note: A more negative number means more efficient allocation of resources (less misallocation); zero means no correlation between productivity and output or employment; a positive number means less productive firms attract more labor or capital than more productive firms. Numbers in the figure are averages across various years, calculated from the data in Annex Table 2.14. While the empirically-motivated measure of misallocation used here does not readily lend itself to calculating productivity gains from reduced misallocation (as done in Hsieh and Klenow, 2009), regression analysis by Duranton et al (2015) for India suggest that a 1 percent decrease in the index of employment misallocation could raise output per worker in manufacturing by about 0.3 percent, and in services by about 0.9 percent (the difference is because labor is a relatively more important input in services than in manufacturing). Therefore, reducing factor market distortions to improve the ability of more productive firms to access inputs could have important consequences for overall productivity. xlv 2.3 Limited success in linking to global value chains Part 1 of this report argued that global trade and investment integration is important for productivity growth. In the 21st century, this is manifested through global value chains (GVCs) which are characterized by the division of the production process into stages, and the distribution of these stages across different countries. This process, variously known as “production fragmentation” (Arndt and Kierzkowski, 2001), “processing trade” (Görg, 2000), “vertical specialization” (Hummels, Rapoport and Yi, 1998), “slicing up the value chain” (Krugman, Cooper and Srinivasan, 1995), or “the second unbundling” (Baldwin, 2006) has been made possible by major changes in logistics and managerial organization in the last third of the 20th century. At the firm level, various aspects of GVC participation, such as imports of parts and components, entry into export markets, and knowledge spillovers from tie-ins with lead firms, have been associated with higher productivity. However, South Asia’s participation in GVCs largely remains confined to apparel. And even in this sector, the sophistication of the region’s products declined between 2000 and 2010. To take better advantage of the productivity-enhancing opportunities offered by participation in GVCs, South Asia needs to further develop capabilities that matter for GVCs, including human capital, institutions, logistics, and removal of trade barriers. The following discussion develops these observations in more detail. 2.3.1 GVC participation supports productivity GVCs make it possible for firms in developing countries to participate in producing the world’s most complex and sophisticated products, by specializing in a piece of the production process where firms have a comparative advantage and produce at the necessary large scale to be competitive globally (Figure 2.3). Lead firms, typically located in advanced economies, perform the higher value-added activities (such as design, branding, and retail), but outsource most or all of the manufacturing to a global network of producers. Beyond their direct contributions, leading firms also have major positive effects through the knowledge and support they provide to suppliers and the competitive pressure they put on all firms in the industry. Their example and competition compel other firms to improve, and signal to the international investor community what can be achieved in the country. Foreign firms’ subsidiaries in South Asia play a particularly important role in complex, capital- and knowledge-intensive activities such as car assembly (e.g. Maruti Suzuki in India and Hyundai in Pakistan) and electronics (Samsung in India and Tos Lanka, a subsidiary of Toslec from Japan, in Sri Lanka) The relationships between foreign and domestic firms provide a vital first step towards increasing productivity and producing goods that meet world-market specifications with regard to technological content, quality and design (Helpman, 1984). Nowhere is this more evident than in East Asia, where the transfer of technology and knowledge facilitated through GVCs made it possible for countries at initially low levels of income – such as China, Hong Kong, South Korea, Singapore, and Chinese Taipei – to move up the ladder in terms of productivity, capital intensity and quality (Kimura, 2005; Konings, 2007; Kee and Tang, 2015). xlvi Figure 2.3 Structure of the global value chain for apparel Source: World Bank (2016). A large literature documents firm-level productivity benefits from various aspects of GVC participation such as access to larger markets, learning-by-exporting, and knowledge spillovers from FDI – although recent evidence suggests that the contribution of these forces to productivity growth declined in the period 2004-2011 compared to 1995-2003, primarily on account of the global trade slowdown (Constantinescu, Mattoo and Ruta, 2016). Causality is difficult to establish, as firms that are already more productive are more likely to enter export markets than less productive firms (Melitz, 2003; Clerides et al, 1998; Wagner, 2007; Abraham et al, 2010; Amiti and Konings 2007; Goldberg et al. 2010; Atkin et al. 2014). Empirical studies have used a variety of statistical techniques to control for this selection bias. In a sample of Indonesian firms, Arnold and Javorcik (2009) find evidence of increased labor productivity due to capital investment as well as organizational and management restructuring following acquisition by a foreign affiliate. In South Asia, evidence suggests that FDI-receiving firms in Bangladesh are more productive (Kee, 2005). In a review of such studies, Havranek and Irsova (2011) conclude that a 10 percent increase in foreign presence is associated with a 9 percent increase in the productivity of local suppliers through their exposure to foreign firms. This evidence suggests that firms often participate in GVCs first and then become more productive, rather than the other way around. Classifying firms with a share of imported raw material in total raw material greater than 10 percent as GVC participants, we find that the TFP of GVC participants in India, Bangladesh and Sri Lanka tends to exceed that of non-participants (Figure 2.4). The picture remains more or less the same when GVC participants are classified as firms with a share of imported raw material in total raw material greater than 20 percent. The same pattern is identified for participation in international markets more generally: firms with a trade share (exports plus imports) greater than 10 percent of value added are xlvii associated with, on average, higher levels of TFP across India, Bangladesh and Sri Lanka (Figure 2.11). And this picture remains robust to different trade shares being used to define GVC participant firms. It should therefore come as no surprise that GVC participants, as reflected in evidence from India and Bangladesh, are associated with a higher share of exports in value added.33 Figure 2.4 GVC participation and firm TFP Bangladesh India .6 .6 .4 .4 Density Density .2 .2 0 0 -10 -5 0 5 Residuals -10 -5 0 5 10 ltfpl_15 Non-GVC participants GVC participants Non-GVC participants GVC participants Various efforts have been made to pinpoint the source of productivity benefits from exporting. Increases in productivity following entry into foreign markets are attributed to greater scale economies in Sub-Saharan Africa (Van Biesebroeck, 2005) and Slovenia (De Loecker, 2007). Evidence from a randomized trial in Egypt indicates that substantial productivity gains from exporting were due to knowledge transfers (Atkins et al, 2014). In South Asia, studies have found that “learning-through- exporting” has boosted productivity of firms in India that enter export markets (Mukim, 2011), in part through scale effects.34 Similarly, firms in Bangladesh with more export experience exhibit higher productivity (Fernandes, 2008). Among domestic producers of apparel inputs in Bangladesh, even those firms that do not export and do not supply to exporters can experience productivity spillovers by learning from the experience of other firms that are part of a shared supplier network that supports exporters (Kee, 2015). On the import side, access to imported intermediate inputs can also boost firm productivity: for example, when India liberalized its tariff regime, access to a greater range of intermediate goods at lower overall prices made manufacturing firms more productive (Goldberg et al., 2010). Case studies document the important productivity benefits from interactions with foreign firms in the context of global value chains. For example, the Desh-Daewoo joint venture, which included the intense technical and managerial training of 130 Bangladeshi in Daewoo’s Pusan plant in 1979, established the foundation for the next generation of Bangladeshi entrepreneurs. 33 In India, the share of exports in value added is, on average, 12.9 percent for GVC participant firms and 2.4 percent for non- participant firms. The corresponding figures for Bangladesh are 47.6 percent and 17.4 percent respectively. 34 On the other hand, a recent a study based on a panel of 10,685 Indian manufacturing firms between 1990 and 2011 found that businesses tend to experience productivity growth a year prior to entering export markets rather than after entering – pointing to potential reverse causality (Gupta et al. 2013). xlviii 2.3.2 South Asia’s success in global and regional GVCs is limited to apparel South Asia has the second-highest level of GVC exports in total exports among developing regions, but almost entirely because of the large share of final and intermediate apparel products (reinforcing the findings on global market shares in Part 1 of the report). Around half of GVC exports from South Asia are in final apparel whereas East Asia specializes in electronics, and Europe, North America and Latin America in autos (Figure 2.5). In imports, South Asia is relatively less integrated in GVCs: firms making final products tend to obtain inputs from domestic sources (or themselves), indicating a lower level of GVC integration than implied in the export data.35 In the case of apparel, however, local sourcing of intermediates combined with the predominance of final apparel exports is consistent with FDI-led GVC activity in South Asia, i.e. global lead firms set up factories and use local materials to manufacture and export final apparel. Figure 2.5 GVC participation across regions Within the region, countries vary greatly in terms of the extent to which they are integrated into GVCs (Box 2.4). In 2013, the share of total merchandise exports in major GVC products (apparel, autos, electronics and footwear) was around 80 percent for Bangladesh, 45 percent for Sri Lanka, 40 percent for Pakistan, and 15 percent for India, while participation of the remaining regional countries is negligible (Table 2.8). By this metric, Bangladesh has one of the highest GVC participation rates in the world, although it reflects the fact that Bangladesh exports little else (Figure 2.6). India’s participation in GVCs, by the same token, is low precisely for the reason that it has a more diversified export basket that consists of a wide variety of products – some of which may also have some of the characteristics of GVC production, but are not included in this analysis.36 Box 2.4 Measuring GVC participation There are a number of ways to measure participation in GVCs. Some analyses, especially those which 35 Bangladesh, India, Pakistan, and Sri Lanka (the “South Asian 4”) run a substantial trade surplus in final GVC goods ($68.0 billion of exports vs. $23.8 billion of imports) and have approximately balanced trade in intermediate goods ($24.3 billion of exports vs. $25.1 billion of imports): they are significant net importers of electronics intermediates, modest net importers of automotive intermediates, and net exporters of apparel and footwear intermediates. 36 India’s largest exports at the detailed (HS6) level are refined petroleum, diamonds, jewelry e.g. of gold, pharmaceuticals for retail sale, and processed rice. All of these lie outside the scope of this analysis. xlix focus on tracking global flows of value-added through input-output methods, essentially view all trade as GVC-oriented (Mattoo, Wang, and Wei, 2013). A country that only exports crude oil or metallic ores may have a high degree of GVC participation of a sort, since these crude materials are eventually transformed into sophisticated goods or parts of other goods in some other country.37 However, linkages with lead firms that lead to technology transfer or deeper interactions with final markets are more likely to take place when countries are engaged in the middle or later stages of the production process. GVCs in vehicles, electronics and apparel/footwear are characterized by a lead-firm network structure, and have been much studied. The share of total global merchandise exports accounted for by these three GVCs has fluctuated between around 14 percent and 28 percent since 1990. Studying the similarities and differences in the organization of these three GVCs can improve our understanding of GVCs, or as they are sometimes called “global supply chains” (USITC, 2011). These three sectors differ in the methods used to coordinate activity over long distances, and in the extent to which they tend to be coordinated by traditional manufacturers (autos), owners of brand names with strong research capabilities (electronics), or buyers of final products working with global middlemen (apparel). This report uses a modified version of the definition of the three classic GVCs in Sturgeon and Memedovic (2011). Products are classified as belonging to one of the three GVCs based on a combination of expert opinion and their position in the U.N. Statistical Division’s Broad Economic Categories (BEC), which help to distinguish between intermediate and final goods. This leads to a list of over 400 traded goods, identified in the SITC Rev. 3 classification at the four-digit or five-digit level. Each of the GVCs is then divided into two sub-sectors to reflect intermediate and final goods (e.g. intermediate electronics and final electronics), making six GVC sectors in all. For the purposes of this analysis of South Asia, the Sturgeon and Memedovic (2011) categories are modified in two ways. First, the footwear sector, both intermediate and final, is separated from apparel, making eight categories instead of six. Second, the definition of the “autos” sector, which originally included only passenger motor vehicles and motorcycles, is broadened so as to encompass other road vehicles (e.g. trucks, buses and trailers). 37 In particular, exporters of primary products experience the sort of GVC participation described as “forward linkages” in international input-output databases. Countries which export final goods requiring large amounts of imported intermediate goods are said to experience “backward linkages.” l Figure 2.6 South Asia GVC exports by country (share of total exports) 1.00 0.90 0.80 Intermediate footwear 0.70 Intermediate electronics 0.60 Intermediate autos 0.50 Intermediate apparel 0.40 Final footwear Final electronics 0.30 Final autos 0.20 Final apparel 0.10 0.00 Bangladesh 2013 India 2013 Pakistan 2013 Sri Lanka 2013 In terms of products, each of the four South Asian countries have significant export positions in final apparel, covering nearly the full range of garment products, while the intermediate apparel – dominated by cotton textiles – is particularly important in India and Pakistan. Bangladesh’s exports of final apparel in 2013 – which have nearly tripled since 2007 – amounted to over US$26 billion, making it the second largest exporter of final apparel in the world next to China. Sri Lanka and Bangladesh, due in particular to effective import facilities for exporters, perform at East Asia’s level in terms of exports per capita ($216 and $147, respectively), while India and Pakistan are at an order of magnitude lower ($10 and $23, respectively). Annual growth rates of exports over 2003-2013 in India (9.6 percent) and Sri Lanka (5.3 percent) were modest, and in Pakistan were sluggish (2.7 percent). These rates of growth, however, may not necessarily reflect productivity improvements because the region benefited from the 2005 elimination of the Multi-Fiber Agreement (MFA), which had restricted textile imports from developing countries to developed countries (World Bank, 2015). Table 2.8 Per capita GVC exports from South Asia (USD) Sri Afghanistan Bangladesh Bhutan India Nepal Pakistan Lanka China Vietnam 2013 2013 2012 2013 2013 2013 2013 2013 2013 Final apparel 0.0 170.1 0.0 13.1 2.7 24.7 210.3 125.4 189.1 Final autos 0.0 0.0 0.0 6.5 0.0 0.1 1.9 23.6 4.7 Final electronics 2.0 0.0 0.5 3.7 0.2 0.6 6.9 252.8 274.1 Final footwear 0.0 3.2 0.0 1.8 0.0 0.6 1.1 35.5 93.6 Intermediate apparel 0.0 0.5 0.9 6.1 2.3 30.4 6.2 42.3 29.9 Intermediate autos 0.0 0.1 0.1 5.0 0.0 0.3 2.6 39.5 40.3 Intermediate 0.0 0.1 0.0 1.0 0.0 0.0 1.1 67.5 16.2 li electronics Intermediate footwear 0.0 0.1 0.0 0.3 0.0 0.0 0.4 1.9 3.6 Total 2.0 174.1 1.5 37.5 5.2 56.1 230.5 588.5 651.5 Note: Data for Bangladesh 2013 are mirror data. Most South Asian countries have negligible exports of GVC products other than apparel, with India accounting for almost all of the region’s exports of autos and electronics. India has a large auto parts industry, with growth assisted by increased exposure to international competition after the lowering of trade barriers. India’s auto parts exports doubled in the last ten years to reach US$6.4 billion in 2013, reaching sophisticated markets such as the United States, the United Kingdom, Italy, and Germany.38 India already exports more auto parts than Indonesia, Vietnam or Morocco, but only about one-tenth of China (Table 2.8). India is also one of the largest and most rapidly growing developing country exporters of final autos: it exports to middle-income countries (e.g. South Africa, Mexico and Algeria) as well as high-income countries (e.g. the United Kingdom and Australia), and, at current growth rates, its exports of final autos may exceed those of China well before the end of the decade. In final electronics, India’s exports have quadrupled in the six years ending in 2013 – but growth is coming from a small base with China’s exports dwarfing India’s by a factor of 20. India’s rapid exports growth has come largely from penetrating developing country markets with lower- priced products (Figure 2.7).39 Seventy-one percent of India’s exports of passenger motor vehicles go to the Middle East/North Africa, Latin America-Caribbean, rest of South Asia, and Sub-Saharan Africa. Similarly, the largest destinations for Indian cell phone exports are Argentina, Russia, South Africa, and the United Arab Emirates. So on average, India’s exports of GVC products are skewed more towards more middle income countries than exports of, say, Germany or Japan, which are geared more towards OECD markets. Figure 2.7 Unit values of India’s auto and electronics exports 38 India exports chassis and engines, including spark-ignition auto engines, diesel engines and aircraft engines, as well as a variety of smaller parts. 39 It is a stylized fact of international trade that rich countries import goods with higher unit values than poorer countries, presumably because their consumers can afford higher-quality varieties of products (Ferrantino, Feinberg and Deason, 2012; Manova and Zhang, 2009; Bastos and Silva, 2010). lii Trade data don’t provide a full picture of GVC integration, because many GVC-related sales take place within national borders. Measures of foreign value added in exports that draw on multi-regional input- output tables are useful indicators of the extent to which countries are engaging fully in the international division of labor that GVCs make possible.40 The OECD-WTO TiVA (Trade in Value Added) data, for example, show that the share of foreign value added in India’s exports rose from 9.4 percent in 1995 to 24.1 percent in 2011, exceeding Indonesia’s share but continuing to lag behind Korea, China, Malaysia and Vietnam (Figure 2.8). This suggests that India is less integrated in GVCs than comparator countries in East Asia when domestic value added is taken into account. Focusing on textiles and apparel as a unified sector, the pattern is similar: India relies less on foreign inputs than its comparators, but that reliance has increased in recent years. By contrast, China’s foreign value added in textiles and apparel has declined, perhaps indicating the country’s growing position in upstream activities. Figure 2.8 Foreign value added in exports, India and comparators Countries in South Asia vary in the extent to which they are “upstream” (specialize in intermediate goods) or “downstream” (specialize in final goods) GVC participants. Pakistan is the furthest upstream (54 percent of its GVC exports are intermediates), followed by India (37 percent). By contrast, Sri Lanka is much further downstream (4.6 percent of its GVC imports are intermediates), and Bangladesh is even further downstream (0.4 percent). These characterizations have been rather stable over time and may be related to the market structure of GVC sectors. In apparel, Bangladesh and Sri Lanka are dominated by large, formal firms that are geared towards the global market. On the other hand, India and Pakistan have a sizeable informal apparel sector where firms employ fewer than 10 workers. Yet, none of the four countries are fully specialized. For example, Bangladesh imports significant amounts of cotton yarn from India but does not export very much fabric because the transformation of yarn into fabric is to a significant extent absorbed by the domestic apparel industry. Furthermore, both Pakistan and India have significant exports of both intermediate and final goods, so they are upstream in some product lines and downstream in others. 40 In the archetypal case of China, foreign value added is highest in those sectors with the highest degree of foreign investment and in more technologically progressive sectors such as computers and telecommunication equipment (Koopman, Wang, and Wei 2012). By contrast, products like steel and ceramics tend to have higher domestic value added in exports. Similarly, in international comparisons, the share of foreign value added is higher for East Asian countries and Mexico, which are deeply imbedded in GVCs, and lowest for primary-product exporters like Russia, Brazil, and Saudi Arabia (Koopman, Powers, Wang and Wei 2010). liii The sophistication of South Asia’s exports has improved in some respects. One way to assess sophistication is by measuring the typical income level associated with countries that export a similar basket of goods as the country in question (analogous to PRODY).41 In the decade between 2000 and 2010, product sophistication of final apparel increased in India, Sri Lanka and Bangladesh but declined in Pakistan (Figure 2.9). However, product sophistication for cloth, yarn and other apparel inputs converged over the decade of the 2000s, with sophistication increasing in India and Pakistan and declining in Sri Lanka and Bangladesh. Figure 2.9 Product and market sophistication of South Asia’s textile and apparel exports Indicators of market sophistication, measured by the average income level of the destination markets, are more varied across countries in South Asia than indicators of product sophistication.42 Although final apparel in Sri Lanka and Pakistan is directed at higher-income markets than apparel from Bangladesh and India, market sophistication declined between 2000 and 2010 in all four countries (Figure 2.7). This could reflect either increased sales to middle-income markets or more intense competition in high- income markets, or both. In the case of intermediate apparel market sophistication, Pakistan and Bangladesh have shown gains between 2000 and 2010, indicating an increasing ability to penetrate high- 41 PRODY (as well as EXPY, described in an earlier section) is an index that measures the quality of export baskets. The index, proposed by Hausmann et al (2007), is a weighted average of per capita incomes of countries producing a given product (with global export shares as weights). A higher PRODY for a given product means that this product is associated with a higher level of per capita income (it is more likely to be exported by a richer country). For South Asia, this means that the product sophistication of each South Asian country’s bundle corresponds to a level of income significantly higher than the South Asian countries themselves. 42 Both sophistication measures are normalized with the US = 1. The relative income levels for South Asian countries are as follows: Bangladesh 0.04, Bhutan 0.11, India 0.07, Maldives 0.16, Nepal 0.03, Pakistan 0.06, Sri Lanka 0.11. liv income markets. India and Sri Lanka, on the other hand, have shifted exports relatively toward lower- income markets. Even though it is more feasible than ever to produce complex goods on a “made-in-the-world” basis, considerations of transport and other transactions costs, as well as timely delivery, often cause value chains to cluster on a regional basis. The best-known of these clusters are the East Asian RVCs in electronics and the US-Germany-Japan automotive RVCs. In South Asia, a regional value chain is emerging in intermediate apparel: intra-South Asian apparel trade amounted to $2.5 billion in 2013, up sharply from $400 million in 2003, and 24 percent of imported intermediate apparel inputs come from within the region, up from 18 percent ten years ago. In the region, Bangladesh and Sri Lanka have the highest share of final apparel goods (86 and 44 percent of total exports), and source many apparel inputs from Pakistan and India who focus relatively less on final products (18 and 6 percent of total exports). In 2013, two-thirds of India’s exports of knit and crochet fabric were destined for Sri Lanka and Bangladesh, while nearly half of Pakistan’s exports of woven cotton denim were destined for Bangladesh and Sri Lanka (Figure 2.10). Figure 2.10 Regional GVCs in apparel in South Asia India's exports of knit and crochet Pakistan's exports of woven cotton denim fabric, nes (SITC 65529), 2013 > 200g (SITC 65243), 2013 Mauritius 3% Other Egypt 13% Other 3% Bangladesh 27% United Sri Lanka Total 42% Total States 46% 2013 exports: Cambodia 2013 17% 3% $210 Sri Lanka exports: Banglades h million 5% Egypt Turkey $496 18% 9% 14% Sri Lanka's exports of woven cotton Sri Lanka's imports of woven cotton dyed dyed > 200g (SITC 65242), 2013 > 200g (SITC 65242), 2013 Taiwan Other India Indonesia Other 2% Hong Kong 1% 1% 1% 2% Egypt 8% 2% Turkey Pakistan 2% 18% China Total 45% 2013 Total Banglades h exports: 2013 $29.8 India 92% imports: million 26% $163.6 There also is evidence of an “East Asia/South Asia” regional value chain. 70 percent of South Asia’s imported apparel inputs come from East Asia, having grown about as rapidly as the intra-South Asian trade. Overall, South Asia sends 26 percent of its exports of GVC intermediates to East Asia, and lv purchases 68 percent of its extra-regional inputs from East Asia -- this extent of orientation of South Asia’s exports of intermediates to East Asia is well above global averages in apparel, autos and electronics. South Asia’s reliance on intermediates from East Asia cuts across sectors, although it is not quite as large as for exports (Figure 2.9). Figure 2.9 Intermediate imports and exports to and from East Asia Sub-Saharan Africa South Asia North America Middle East and North Africa Intermediate imports sourced from East Asia Latin America and Caribbean Intermediate exports destined for Europe and Central Asia East Asia East Asia and Pacific 0% 10% 20% 30% 40% 50% 60% 70% 80% 2.3.3 Most policy determinants of GVC participation are lacking Although a number of studies have looked at the determinants of production fragmentation (e.g., Hillberry, 2011) and supply chain trade (e.g., Rahman and Zhao, 2013), the literature is yet to give a clear picture of the drivers of GVC participation and competitiveness. At the level of the firm, local businesses need to have reasonably high productivity, a capacity to absorb new technologies (skill and capital intensity), and ideally experience with trading across borders in order to be qualified suppliers.43 The lack of such firm-level capabilities can inhibit the extension of GVCs to certain countries (see Farole and Winkler, 2014 for the case of Africa). Other studies have tested the importance of specific drivers of GVCs, including trade policy (Orefice and Rocha, 2014), transport (Hummels and Schaur, 2014), trade logistics (Saslavsky and Shepherd, 2012), and time zones (Dettmer, 2014). While these and other studies give us a sense of what factors are likely to be important in determining GVC dynamics, the question of which specific drivers matter most for country-level participation in GVCs remains open. In a recent study, Farole and Pathikonda (2015) find a greater intensity of GVC products compared to non-GVC products44 across a sample of over 100 countries for a range of capabilities that are most common in the theoretical, policy, and empirical literature on GVC trade.45 They divide capabilities into three categories: (a) fixed (b) long-term policy variables and (c) short-term policy variables. Fixed capabilities include proximity to markets and natural (resource) capital. Policy variables that can be changed gradually over a relatively long period of time include human capital, physical capital and institutional capital. Policy variables that can be changed directly through a policy shift or negotiations in the short to medium term include logistics/connectivity, wage competitiveness, market access and access to inputs. 43 See, for example, Corcos et al (2013), Defever and Toubal (2013), and Jabbour (2012) 44 As defined by combining lists generated by Athukorala (2010) and Sturgeon and Memedov (2011) 45 102 countries, which together represent 81% of world trade in 2012. lvi South Asian countries vary on these capabilities (Table 2.9). Sri Lanka scores highest on the level of human capital as measured by the average years of schooling. India’s total natural capital far outpaces other countries, with Pakistan a distant second. Bhutan’s institutions are rated ahead of the other countries, followed by India, with Pakistan and Bangladesh lagging behind considerably. Geographically, Afghanistan and Pakistan in the region’s north-west appear the most disadvantaged, whereas Sri Lanka seems closest to markets. India is a much more sophisticated logistical hub, whereas some countries – Nepal, Bhutan, and Afghanistan – are in some of the most challenging logistics environments in the world. The minimum wage is not very different across the “South Asian 4”. In terms of access to foreign markets, India faces fewer trade barriers than Bangladesh, Sri Lanka and Pakistan. On the flipside, in terms of trade barriers on imported inputs, India appears to be more restrictive. To put these figures in context, Figure 2.12 shows the standardized capability levels compared to two regional blocs: ASEAN which is already of major importance in GVCs and SACU (Southern African Customs Union) which is a potential competitor to South Asia looking ahead. Data on India is shown separately from the rest of South Asia, given its size and level of industrialization. Table 2.9 GVC capabilities and endowments in South Asia (2010) Catego Capability Indicator Afghani Bangla Bhutan India Sri Maldiv Nepal Pakista ry stan desh Lanka es n Fixed Proximity to Proximity to 0.59 0.43 0.47 0.51 0.33 0.32 0.48 0.57 markets Markets (GDP- weighted distance index; 0 to 1) Natural capital Total value of 197.74 8.92 2,959.7 40.71 0.33 66.83 522.55 natural capital Long- Human capital Average years 3.85 5.91 6.24 10.06 6.02 4.23 5.02 term of schooling (>15 years old population) Physical Capital stock 2,764.3 capital ($ per capita) Institutional Rule of Law -1.90 -0.79 0.12 -0.04 -0.08 -0.33 -1.01 -0.74 capital (Rating from - 2.5 to 2.5) Short- Logistics Logistics 2.24 2.74 2.38 3.12 2.29 2.40 2.20 2.53 term /connectivity Performance Index (Rating; 1 to 5) Wage Minimum 38.57 28.37 38.55 75.90 41.59 competitivenes wage for a s 19-year old worker or an apprentice (US$) Market access Overall Trade 23.49 16.87 1.68 8.37 18.47 34.23 1.68 14.73 Restrictivenes s Index (of Trading Partners) Access to Overall Trade 14.90 7.42 20.20 12.63 7.37 inputs Restrictivenes s Index lvii Note: The values for natural capital, wage competitiveness and the two access indices correspond to, respectively, 2005, 2014 and 2009. In the indices measuring market/input access, a higher number indicates lesser access. India has several advantages relative to ASEAN and SACU: natural capital, wage competitiveness, and proximity to markets. Yet, it lags behind the two other blocs in terms of physical capital, human capital, institutions and logistics. In terms of access to imported inputs, India erects higher barriers compared to both ASEAN and SACU. At the same time, on the export side, India faces lower barriers relative to ASEAN but higher barriers relative to SACU. Similarly, Pakistan, Bangladesh, Sri Lanka, Nepal, Bhutan and Maldives are on average more wage competitive than ASEAN and SACU and are closer to markets. Their level of natural capital is higher than SACU, but lower than ASEAN. This group of South Asian countries (excluding India), on average, fares worse than SACU and ASEAN countries on other capabilities, including logistics, institutions, human capital and access to markets overseas. Last, but not least, they have better access to imports of intermediate inputs than ASEAN on average, but lower than SACU on average.46 Figure 2.12 GVC capability gaps vs. ASEAN and SACU India vs. ASEAN India vs. SACU Inputs Access MAOTRI Market Acces… Market Access MAOTRI Inputs Acces… Wage Competitiveness Minimum Wage Logistics Index Rating Logistics Index Rating Institutions Index Rating Institutions Index Rating Physical Capital Physical Capital Human Capital Human Capital Natural Capital Natural Capital Proximity to Markets Proximity to Markets -2 -1 0 1 2 3 4 -2 -1 0 1 2 3 4 South Asia (excluding India) vs. ASEAN South Asia (excluding India) vs. SACU Inputs Access MAOTRI Market Acces… Market Access MAOTRI Inputs Acces… Wage Competitiveness Minimum Wage Logistics Index Rating Logistics Index Rating Institutions Index Rating Institutions Index Rating Physical Capital Physical Capital Human Capital Human Capital Natural Capital Natural Capital Proximity to Markets Proximity to Markets -1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5 Note: A positive number indicates an advantage for South Asia vis-à-vis the other regions. 46 These comparisons remain unchanged when South Asia (excluding India) is defined by the country-group formed by Bangladesh, Pakistan and Sri Lanka. lviii 2.4 Firm capabilities are constrained An average firm in South Asia does not operate at optimum efficiency. Counter-intuitively, the region’s firms over-employ relatively scarce capital and under-employ labor, South Asia’s abundant resource. Performance varies more widely with regard to knowledge inputs – ranging from extensive technology use in India to limited ICT adoption in Bangladesh and Nepal – but even among leaders, the use of e- commerce and other productivity-enhancing online business tools is relatively low. When it comes to innovation, the region again has both leaders and laggards, but generally innovation tends to be concentrated in few, mature firms and even then, it is more likely to be of the imitation, catching-up to the frontier variety. Having said this, innovation – to a large extent enabled through ICT adoption – is an important driver of productivity at the firm level. This suggests that further investments in firm capabilities, including better resource management, improving skills, deepening technology adoption, and nurturing innovation, could raise productivity in South Asia. The following discussion develops these observations in more detail. 2.4.1 Firms lack in managerial quality and do not use resources efficiently Studies find that organizational factors, including skills and management practices, hold back firm productivity in South Asia (Bloom et al. 2010). For example, management practices in India are weaker than those in the United States, Brazil, and China (Bloom and Van Reenen, 2010), as well as many other developing countries (Figure 2.14) due to Indian firms’ tendency not to collect and analyze data, set and monitor clear performance targets or explicitly link pay with. Conversely, improved management practices in South Asia have led to increased profitability and productivity. Bloom et al. (2011) perform an experiment providing management consulting to large textile firms in Maharashtra, India, and find that the intervention led to significant improvements in the firms’ productivity levels. McKenzie and Woodruff (2015) show that business practices matter as much for microenterprises as for larger firms – boosting productivity, profits, survival rates, and sales growth – and the positive effects of improved practices are robust to numerous measures of owners’ human capital. Formal managerial training is limited in South Asian firms. For example, only 43 percent of nonproduction workers in the automotive sector in India are formally trained, compared to nearly 70% in China. lix Figure 2.14 Weak performance on management capabilities Source: World Management Surveys An indirect way to assess managerial capabilities is to look at how firms utilize resources available to them. For example, profit maximization requires that firms employ resources such as labor and capital until the additional contribution of these factors to firm revenue (marginal revenue product) equals the going wage/rental rate. Of course, firms often do not observe the marginal products directly; however in the absence of major distortions (physical, regulatory, or information asymmetries), an average firm should employ something close to an optimal factor mix – if this were not so, firms could gain by hiring (relatively) cheaper resources and factor owners could gain by commanding returns above their productivity levels until marginal products and costs were equalized. In South Asia, many firms do not appear to employ efficient combinations of factors. Case studies reveal low capacity utilization in India and Pakistan among apparel makers (operating at 6.5 months annually vs. the global average of 9 months) and auto makers (66 and 44 percent, vs. more than 75 percent in China). Only four of the 18 OEMs in the auto sector in India and Pakistan operate at the industry standard for efficiency of 100,000 units per model. Previous research on India and Sri Lanka has attributed weak performance of the manufacturing sector to firms consistently under-employing relatively abundant labor and over-employing relatively scarce capital (Fernandes and Pakes, 2008; Dougherty et al, 2009; Hasan et al, 2013).47 In India, these results suggest that the optimal level of employment was 3-6 times the actual level, whereas in Sri Lanka, the optimal level of labor was 1.1 times the manufacturing employment in 2003.48 Repeating this analysis with the most recent data 47 The analysis is based on the 2002 and 2005 rounds of the Investment Climate (Enterprise) Survey in India, and the 2005 round of the Investment Climate (Enterprise) Survey in Sri Lanka. The authors measure factor under/over- utilization by estimating a Cobb-Douglas production function to calculate the marginal products of labor and capital, and comparing these marginal products with factor costs (wages for labor and interest and depreciation rates for capital). The estimation approach followed Olley and Pakes (1996) to correct for entry/exit (selection) and the endogeneity of input use (inputs may be selected at the same time as output). 48 The reported under-utilization rates do not imply that employment would rise by the indicated factor if firms were able to use factors optimally: as firms would hire more labor, wages would rise, putting downward pressure on optimal employment levels. lx available for Bangladesh, India, Nepal, Pakistan, and Sri Lanka reveals that under-utilization of labor remains a persistent feature of the operating environment for firms in most of South Asia.49 The optimal level of employment of firms in India and Sri Lanka is 1.7 and 2.2 times current employment levels, respectively, while estimates for Nepal and Pakistan suggest under-utilization on the order of 14-16 times the existing workforce. Bangladeshi firms, on the other hand, appear to over-utilize labor: firms hire approximately 18 percent more workers than would be optimal at the prevailing wage rate.50 Although the data include businesses as small as 5 employees, larger firms make up the majority of the sample. However, this makes the findings of even greater concern, since larger firms could be expected to have greater capacity to manage resource use, leverage economies of scale, and be closer to the knowledge frontier. Instead, most large firms in South Asia do not operate close to what would be considered optimum efficiency levels given the prevailing factor prices, costing themselves lost profits and bringing down aggregate productivity. 2.4.2 Knowledge and technology adoption are low Firms rely on technology to enhance efficiency of production processes and to connect more effectively with customers and suppliers. In particular, information and communication technology (ICT) is a major potential enabler of productivity growth, especially in countries/locations further away from the technological frontier. In this report, ICT is defined as the use of computers and other electronic equipment and systems to collect, store, use, and send data electronically. Therefore, ICT is an umbrella term that includes any communication device or application, such as cellular phones, computer and network hardware and software, as well as various services and applications associated with them, such as internet and videoconferencing. 49 These estimates are based on the 2010 Enterprise Survey data for Pakistan, 2011 for Sri Lanka, 2013 for Bangladesh and Nepal, and 2014 for India. 50 To obtain these estimates, similar to the approach of Fernandes and Pakes (2008), marginal products of labor and capital are estimated using country-specific Cobb-Douglas production functions with sector dummies, and then compared with the prevailing wage and rental rates for each country. However, due to the absence of panel data, the estimation approach does not explicitly account for firm selection and endogeneity of input choices. While the extent of the implied basis is difficult to judge, results reported by Fernandes and Pakes (2008) for India show that coefficient estimates obtained by OLS lie within two standard errors of results obtained with the Olley-Pakes estimator when firms are allowed to optimize labor volumes in every period, even when controlling for firm exit. lxi Figure 2.15 ICT adoption across countries in South Asia India Pakistan Bangladesh Nepal South Asia Africa share of firms w hich Have internet connection Regularly use computer Purchase or develop software 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% In high-income countries, penetration of ICT is by now nearly universal: 97.9 percent of businesses with ten or more employees in OECD countries have an internet connection (OECD 2012). Many developing countries also have a high percentage of firms using ICT: for instance, the percentage of Turkish and Mexican firms using the internet is over 90 percent (OECD 2012). However, adoption of ICT in South Asia is uneven. Only half of Nepalese and Bangladeshi firms use computers in their business, which is lower than the average in Africa (Figure 2.15).51 On the other hand, nearly all Indian firms use computers, similar to what is observed in the European Union. Twice as many Indian firms use computers and software than firms in Nepal and Bangladesh, and 30 percent more firms use computers and software in India than in Pakistan. Pakistani firms lie in between the two sets of countries, with around 71 percent of firms having at least one worker using a computer. ICT adoption rates also vary considerably by sector within and across countries. For example, the use of ICT in electronics is much higher than in apparel in most South Asian countries, ICT use is higher in sectors more exposed to international competition, for example apparel and some automotive sectors, than sectors that are less exposed to competition, for example agribusiness.52 Box 2.6 Data on ICT adoption in South Asia The analysis of ICT and innovation practices is based on the data collected by the Enterprise Surveys, which were implemented mainly in 2013 and 2014.53 Overall, the dataset has around 5,500 observations unevenly distributed between four countries in South Asia (e.g., the number of firms surveyed in India exceeds the number surveyed in the other three countries). More than 4,000 51 In the discussion that follows, South Asia’s performance is benchmarked against the average in the Africa region, which includes ten other countries where the same innovation questionnaire was implemented: DRC, Ghana, Kenya, Namibia, Nigeria, South Sudan, Sudan, Tanzania, Uganda and Zambia. Although South Asia and Africa are very different in many respects, levels of GDP per capita are similar among the sampled countries. 52 See table in Volume II. The source is the World Bank Enterprise Surveys, 53 India is the only country in which an enterprise survey took place in 2014 instead. lxii manufacturing firms were surveyed, compared to 1,266 service firms. The survey is mainly representative of medium, and to a lesser extent larger, firms although a few firms in the survey were de facto micro when surveyed as they had less than 5 employees. The innovation module from the World Bank Enterprise Survey (ES) includes questions on ICT use, although the survey does not report how much firms invest in ICT. Specifically, the ICT section of the innovation module provides information on two aggregate dimensions of ICT use: (i) computer and software use and (ii) different types of internet use. Computer and software use is a critical channel to improve the production process, and internet use can be a critical tool to improve performance via reducing information costs, enabling e-commerce and facilitating communication. The table below provides an overview of the main questions of interest formulated in the survey: Box Table: ICT questions from the Innovation Module Computer and Percentage of workers’ using a computer regularly Software Use Whether a firm has purchased or developed any software in house Whether a firm has IT staff Total cost spent on hiring an external computer or software consultant Internet Use Communication: whether a firms uses internet for internal communication among its employees or with clients and suppliers E-commerce: whether a firm uses internet for online purchases or sales Information: whether a firm uses the internet for managing inventory, marketing products or researching of developing ideas on new products and services In order to obtain a meaningful measure of ICT use, and given the lack of information on what specific sub-dimensions of ICT use are more important for performance, this report calculates a synthetic index for each ICT use using the average of the normalized sub-components in each country. For computer use, we use four variables: two indicator variables and two continuous variables (percentage of workers’ using a computer regularly and total cost spent on ICT consultants). Given the potential for large sector differences in these continuous variables, we re-scale the values as the ratio to the sector mean and then we normalize them by subtracting the sample mean and divide them by the standard deviation. Seven indicators are available for internet use: two for communication (whether a firm uses internet for internal or external communication), two for e-commerce (buying or selling online) and three for information (management of inventory online, marketing online and research online) – which are then normalized and averaged an internet use index. Finally, an aggregate ICT total index is calculated as the average of the computer and internet indexes. lxiii Computers Computers and software Software Overall ICT index E-commerce Internet use Online marketing Online inventory management As expected, almost every firm using a computer also uses the internet (Figure 2.15). Only one out of five firms with computers in Nepal does not use the internet, the lowest ratio in the four countries. Similar to computer use, internet use is higher in larger and more foreign-exposed firms. However, while the difference in the share of internet use in large and small firms in India is around 6.5 percentage points (99.6 percent to 93.1 percent), it increases to 23.4 percentage points in Nepal and 18.1 percentage points in Bangladesh. In Nepal, internet use appears to increase with age: only half of firms less than 5 years old use the internet, while 88 percent of those older than 20 years do. The picture in Pakistan is, however, the opposite; younger firms are more likely to use the internet. In all countries, service sector firms are more connected to the internet than are firms in manufacturing, although the difference is not large. Figure 2.16 Types of internet use by firms across countries India Bangladesh Pakistan Nepal South Asia Africa share of internet -co n ne ct e d firms w hich Market online Manage inventory online Purchase online Sell online 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% lxiv Firms appear to use the internet more to reach customers rather than connect with suppliers (Figure 2.16). Given the high rates of connectivity in South Asia (and particularly in India), it is striking that South Asian firms are doing less e-commerce than African countries in most cases. In all four countries, the share of firms selling goods on the internet is higher than the share of firms purchasing online, which is a different pattern to what is observed in Africa. South Asian firms tend to do less marketing on the internet than African countries; 58 percent of firms in African countries do marketing online, but only Indian firms (the highest share in the region) advertise on the web at the same rate. Although internet use for marketing increases with size, large firms do less marketing online than large firms in African countries, with the exception of Nepal. Overall, and despite greater internet access in South Asia, the use of important internet management strategies such as e-commerce and marketing are still much lower than its potential. Overall, data reveal three important patterns with regard to ICT use in South Asia. First, ICT adoption varies significantly across countries: India scores very high on multiple dimensions of technology use and Pakistan is in line with global peers, but ICT adoption in Bangladesh and Nepal is very low and lower than in African countries. Second, India is a regional leader in computer and software use among firms, suggesting that there are potential spillovers from the country’s strong software industry. Third, despite prevalent internet use, the adoption of internet commercialization practices for marketing products (e- commerce) is relatively low, with the difference particularly stark in the case of India. What determines such different rates of ICT adoption? The literature identifies four sets of factors: firm characteristics, market structure, demand-side variables, and complementary factors such as skills, other technologies, and agglomeration economies that may facilitate diffusion (Box 2.7). All seem to be important in South Asia. Larger firms and exporters are more likely to adopt ICT practices, with the size of each coefficient inversely related to the country’s economic development, suggesting larger effects in poorer countries in the region (see Table 2.16 in the Annex for detailed regression results). 54 Younger firms are more likely to adopt ICT practices only in the case of India, which contrasts with the results of Commander et al. (2006), and foreign owned firms are not more ICT intensive. The link between importing and ICT adoption, suggested in Hollenstein (2004) or Haller and Siedschlag (2011), is only statistically significant for India and Bangladesh. Box 2.7 Determinants of ICT adoption The literature has identified a number of firm characteristics that are important for adoption of ICT at the firm level (some of these parallel the observations from the World Bank Enterprise Surveys discussed in the preceding section). For example, a number of studies find a positive correlation between firm size and the adoption of ICT (Teo and Tan, 1998; Thong 1999; Fabiani et al., 2005; 54 All equations control for sector effects using 2-digit ISIC dummies, and we estimate the different models individually for each country using ISIC-2 digits controls and for the whole region using also country dummies. Given the large number of observations in India, the pooled sample for the South Asia region is largely dominated by India. All country individual estimates use sampling weights. Based on the earlier discussion we use two specifications, one parsimonious specification with only a few key variables and an extended version that includes all the elements described earlier. lxv Giunta and Triveri, 2007; Haller and Siedschlag 2011). Walczuck et al. (2000) have pointed out that small firms in the Netherlands are not adopting internet at the same speed as larger firms.55 Beyond size, some studies suggest that adoption and use of ICT is higher in younger firms (Commander et al. 2006; Haller and Siedschlag 2011). Some studies have examined the role of education or skills (human capital) in the process of adopting new technologies (Bartel and Lichtenberg, 1987; Chun, 2003), while others have shown that the demand for educated workers rises with the use of the new technology (Berman et al, 1994; Doms et al, 1997 and Haskel and Heden, 1999, Bugamelli and Pagano, 2004). The environment in which the firm operates also matters for ICT adoption. Firms facing stronger competition are more inclined to innovate and adopt new technologies, such as ICT, in order to improve their performance and chances of survival. Some studies show that competitive pressure is positively associated with ICT adoption (Dasgupta et al. 1999; Kowtha and Choon, 2001; Hollenstein, 2004; Kretschmer, Miravete, & Pernías 2012). Firms exposed to international competition in export markets may be more inclined to adopt new technologies (Hollenstein, 2004; Lucchetti and Sterlacchini, 2004; Bayo-Moriones and Lera Lopes; 2007; Giunta and Trivieri 2007; Haller and Siedschlag 2011). Similarly, foreign-owned firms are more likely to be early adopters of new technology as well as potentially important channels of new technology diffusion (Keller, 2004; Narula and Zanfei, 2005). Consistent with the complementarity between technology and skills observed in OECD countries (Berman et al, 1994; Doms et al, 1997 and Haskel and Heden, 1999, Bugamelli and Pagano, 2004), skills matter critically for technology adoption in South Asia. The share of high school graduates in firm employees is positively and significantly associated with ICT adoption in the region pooled sample, and in all the country estimations except for Pakistan. In Bangladesh and Nepal, access to finance also matters, while in India, access to foreign technology via licensing is an important channel. Agglomeration matters for ICT diffusion only partially: in India (but not in other countries) firms in the main business cities are more likely to adopt ICT, while in Nepal city size matters a great deal. Results for individual components of ICT adoption largely parallel those for the aggregate index. These findings suggest different policy priorities for countries within the region. Nepal and Bangladesh could concentrate on supporting the adoption of internet and computers in the private sector. The public sector can play a role by investing in infrastructure, especially in Nepal where diffusion is higher in larger cities, by helping to train skilled workers, and by supporting the diffusion of technology. Once this more basic ICT adoption is mainstreamed across all firms in these countries, the focus should shift towards greater integration of ICT practices that improve management and performance, in our indices represented by the use of software and the use of the internet for the commercialization of products. In the case of India, where the use of ICT is pervasive in the private sector, the focus could be on the use of the internet for the commercialization of products, facilitated by improved access to finance. Given the large extent of software development and the availability of IT engineers, it is likely that improving 55 Other studies find a weak or insignificant relationship (Teo et al., 1997, Lefebvre et al., 2005, Love et al., 2005). Hollenstein (2004) argues that the relationship between ICT adoption and firm’s size might be non -linear, which explains partially the cause of weak or insignificant relationship. lxvi access to finance and establishment of broad-based online financial transactions platforms could help broaden e-commerce use. 2.4.3 Innovation is widespread but novelty is limited As with ICT use, adoption of innovation practices differs significantly across South Asian countries. Innovation leaders (Bangladesh and India) have a larger percentage of firms conducting R&D than the average of the Eastern Europe & Central Asia (ECA) and Africa, while laggards (Nepal and Pakistan) display a lower percentage (Table 2.10). The same holds true for R&D expenditures per employee, although on the whole South Asian performance is well below ECA. The incidence of R&D in India is the highest in the region, while the concentration rate is low. Thus, even though the average intensity of R&D (i.e., R&D spending per worker) is low compared to the average of the ECA region, many firms are investing in R&D. India has the potential to become one of the main global centers for automotive research and development; a large share of Bosch’s global R&D is done in Bangalore, where it employs 15,000 personnel. Investment in innovation differed significantly across firms. Small, non-exporting, national and very young firms are more R&D intensive in India, while in Bangladesh, large, exporting, foreign and old firms are significantly more R&D intensive. In Pakistan, there is a very large concentration of R&D activity in a very small number of firms. Investment in innovation in the GVC sectors discussed in Volume II is greater in large firms. In the automotive sector, for example, field interviews show that large firms spend more on innovation and R&D than do smaller firms. And in agribusiness, 65 percent of large firms reported expenditures on innovation, compared to only 49 percent of SMEs. Some large multinationals have set up R&D centers with world class capabilities in South Asian countries to support global operations. Overall, however, investment in innovation is low in the industries discussed in Volume II. For example, innovation expenditures in electronics were low in a global context: only 1.1% of sales in large firms employing more than 100 workers, and 4.7% in small firms. Nevertheless, the return to innovation in South Asia is high. In agribusiness, for example, every 1% spent in innovation generated around 2% of participation of new products in firms’ sales. Table 2.10 Knowledge capital intensity Type Indicator Bangladesh Pakistan India Nepal South Asia ECA Africa R&D percent firms 19 6 56 4 21 9 19 US$ per worker 8 - 14 6.5 14 498 18 (median) Equipment percent firms 75 17 68 23 46 - 29 US$ per worker 92 197 227 130 179 - 180 (median) Licensing percent firms 5 3 4 1 3 - 8 US$ per worker 6 82 25 234 27 - 21 (median) Training percent firms 19 6 56 4 21 9 19 US$ per worker 12 107 21 73 21 - 47 (median) Note: Intensity is calculated only for firms engaged in R&D (only 23 firms reported this information in Pakistan). Source: Authors’ elaboration from Enterprise Survey (2014) Box 2.8 Innovation activities and outputs in the Enterprise Survey Innovation can be measured by looking at innovation inputs, innovation outputs, or both. However, lxvii the subjective nature of many of the questions used in innovation surveys presents a challenge. The Oslo manual, which is the main reference for these type of surveys, defines innovation as “…..the implementation of a new or significantly improved product (good or service), or process, a new marketing method, or a new organizational method in business practices, workplace organization or external relations.” Most surveys use this definition to identify innovations by directly asking firm managers and owners whether they have implemented “new” or “significant” changes or improvements in the last three years. This is problematic since “significant” improvement is a highly subjective term, and is also self-reported. In general, any sound analysis of innovation activity should combine a focus on both knowledge capital inputs and innovation outputs. In the World Bank Enterprise Surveys, the following information on various sources of knowledge capital and innovation outputs is available:56 Innovation or knowledge inputs: i. Research and development: source of R&D (internal versus external) as well as expenditure. ii. Capacity building: training provided as a result of new innovations as well as expenditure. iii. Purchase/licensing of inventions or other knowledge forms. Firms reported on the expenditure on the purchase of inventions or intellectual property that helped them come up with innovations. iv. Acquisition of business intelligence. Interviewed firms reported on what were the key sources of information and ideas for their innovative activities. v. Intellectual property. Firms reported on whether they applied for patents, utility models, trademarks, or copyright design, or registered an industrial design. The Enterprise Surveys innovation module differentiates between two types of technological innovations (product and process) and two types of non-technological innovations (organization and marketing): i. Product innovations. These are essentially new, redesigned, or substantially improved goods or services. In the context of the survey, there are 3 metrics used:  Products new to the firm;  Significantly improved products;  Products new to the market. ii. Process innovations. These include the implementation of new or significantly improved production or delivery methods (including significant changes in techniques, equipment, and/or software). Minor changes or improvements; an increase in production or service capabilities through the addition of manufacturing or logistical systems that are very similar to those already in use; ceasing to use a process; simple capital replacement or extension; changes resulting purely from changes in factor prices, customization, regular seasonal and other cyclical changes and trading of new or significantly improved products are not considered to be innovations. The aspects considered include:  Innovation methods for manufacturing products or offering services, 56 The distinction between types of innovations, while clear in theory, can be a matter of some confusion for survey respondents. For example, new marketing processes, like discounts, new packaging or new client segments, are sometimes confused with process or product innovations. The fact that interviewees provide a recorded description of product and process innovations allows the user to verify the identified innovations and reclassify wrongly attributed cases to their respective category, and to invalidate cases that do not constitute an innovation at all. This exercise has been conducted by Cirera et al. (2015), who kindly provide us with “clean innovation data.” For the overall sample of South Asia firms, the cleaning exercise conducted by the authors has decreased the rate of product innovation - from 53 percent to 51 percent - and the rate of process innovation - from 64 percent to 58 percent. Although the cleaning exercise reduced innovation rates for most countries, in Nepal the rate of product innovation increased from 10 to 12 percent as the result of reclassification from process innovations. lxviii  Innovative logistics, delivery, or distribution methods for inputs, products, or services,  Innovative supporting activity for processes, such as maintenance systems or operations for purchasing, accounting, or computing. iii. Organizational innovations. These include the implementation of a new organizational method in business practices, workplace organization, or external relations. The form of innovation is grouped into two categories. First, structural innovations affect responsibilities, accountability, command lines, and information flows, as well as the number of hierarchical levels, the divisional structure of functions (research and development, production, human resources, financing, etc.) or the separation between line and support functions. And second, procedural innovations consist of changes to routines, processes, and operations of a company. Thus, these innovations change or implement new procedures and processes within the company, such as simultaneous engineering or zero buffer rules. iv. Marketing innovations. These include changes made to incorporate advances in marketing science, technology or engineering to increase the effectiveness and efficiency of marketing, to gain competitive advantage. Studies show that training and R&D are complementary in supporting productivity outcomes; however, firms across South Asia spend relatively little on training.57 Still, leaders do substantially more training than laggards, although training amounts are challenging to compare because so few firms in the laggard group extend training to their employees. Turning to innovation outputs, the region’s leaders exhibit innovation rates around 80 percent, well above the average of the ECA and Africa regions, while Pakistan and Nepal have innovation rates of 15 percent and 21 percent (Figure 2.17). Process innovation is more important in Bangladesh and India, while product innovation is more important in Nepal and Pakistan. Consistent with the earlier discussion of firm capabilities, firms in all countries are more likely to innovate in marketing than in organizational issues. 57 In a study for Mexico, Lopez-Acevedo and Tan (2003) found that training had large and statistically significant wage and productivity outcomes, that joint training and R&D yielded larger returns than investments in just one or the other, and that both training and technology investments enabled firms to improve their relative position in the wage and productivity distribution. lxix Figure 2.17 Innovation practices in South Asia and comparator countries share of firms engaged in innovation activities Technological: Product & Process Process Product Marketing Organizational Bangladesh India Nepal Pakistan ECA Russia Turkey Africa 0% 20% 40% 60% 80% 100% As suggested by the literature (e.g., Hall and Lerner, 2009; Kerr and Nanda, 2014), technological and non-technological innovation rates are significantly higher for larger firms, consistent with evidence from ECA and Africa.58 On the other hand, the evidence on the relationship between innovation and firm age is mixed: in India, younger firms display significantly higher rates of organizational innovation and marketing while the opposite is true in the rest of the region. Trader firms (i.e., exporter, importer or both) have higher rates of technological and non-technological innovation, again consistent with the literature (Lileeva and Trefler, 2010), although the differences are statistically significant in few cases. In all countries except Pakistan, exporters are more innovative than importers in terms of creating new products; but importers and two-way traders are more innovative than exporters regarding process and organizational innovation (with the exception of Nepal). With the exception of Pakistan, foreign-owned firms are also more innovative (Brambilla, 2009; Aghion et al., 2013), although the differences are relatively minor.59 Few firms engage in disruptive innovative activities such as introducing new products to the country or to the world, while the majority of firms conduct incremental innovations or pure imitation - either by upgrading the quality of existing goods or introducing new products to the firm (Cirera et al., 2015). This tendency, observed in ECA and Africa, also holds in South Asia: despite high average innovation rates in the region, there is only a low degree of novelty in innovation (Table 2.11). Most of the inventions of 58 These differences are statistically different for almost all types of innovation in India and Nepal, for product and marketing innovation in Bangladesh and for organizational innovation in Pakistan. 59 Foreign-owned firms are those that have over 25% ownership by private foreign individuals, companies or organizations. The differences are statistically significant for product and process innovation in Bangladesh, product and process innovation simultaneously in India, marketing in Nepal, and process innovation and marketing in Pakistan. lxx the region’s innovation leaders (Bangladesh and India) involve the imitation of existing products and/or processes. In these countries, firms that introduce radical innovations are young, middle or larger sized, exporters, and domestically-owned. At the other end, Nepal and Pakistan show very low innovation rates in general, including imitation activities. In these countries, innovating firms tend to be older, larger, non-exporters, and more likely foreign. In general, firms that have a higher level of R&D intensity tend to introduce more radical innovations. Table 2.11 Types of innovation undertaken (percent of all firms) South Bangladesh Pakistan India Nepal ECA Africa Asia New to firm 44 8 54 12.3 30 18 25 Of which new product 20 38 56 0 29 74 68 Of which product upgrade 80 62 44 100 72 26 32 Imitator(new to firm/local 37 6 47 12 10 20 market) 26 New to national market 4 2 4 0.3 3 6 3 New to international market 3 0.5 2 0.03 1 2 2 Source: Authors’ elaboration from Enterprise Survey (2014) Most of the product and process innovations in the region are in-house (Figure 2.18). External cooperation is in most of the cases linked to other firms, although cooperation with the private sector plays a more important role for process innovation than for product innovation. Successful firms do report that in-house R&D capability has been an important driver of competitiveness, enabling them to compete on quality as well as cost. Nevertheless, this high reliance on in-house innovation development - larger than in Africa and much larger than in the ECA region –implies limited scope to introduce more novel products, and likely underpins the large imitation rates in India and Bangladesh. Figure 2.18 In-house vs collaborative innovation How was product innovation developed? How was process innovation developed? (% of product innovation firms) (% of process innovation firms) Nepal Nepal India India Pakistan Pakistan Banglades Bangladesh h 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100% All In House Cooperation other Firms All In House Cooperation other Firms University or Government Don't Know University or Government Don't Know Innovation activity is also spatially concentrated, but only when considering the novelty of innovation. While in general, R&D investments and innovation are more concentrated than employment (Carlino lxxi and Kerr, 2014) with patents mostly originating in few and large cities (Forhnahl and Brenner, 2009; Bairoch, 1988), innovation activities in South Asia are less concentrated than employment, except in Pakistan (Table 2.12). However, higher degrees of novelty, such as products that are new to the national or international market, are more concentrated than employment. Thus, agglomeration in South Asia may matter mostly for more radical innovation than for imitation. Table 2.12 Concentration of economic and innovative activities (Herfindahl index) Employment Firms Innovators Innovators Radical R&D share share (product or (product) (national/ process) international) Bangladesh 0.390 0.280 0.280 0.280 0.400 0.240 India 0.017 0.006 0.006 0.009 0.040 0.007 Nepal 0.510 0.310 0.270 0.220 0.630 0.700 Pakistan 0.260 0.170 0.440 0.550 0.420 0.290 Source: Author’s elaboration from Enterprise Survey (2014) 2.4.4 Returns to innovation are high Innovation is a key determinant of firm-level productivity. Process innovation increases firm productivity through more efficient use of intermediate inputs and factors of production, while organizational innovation encourages the reallocation of inputs and factors of production across activities within firms. Product innovation increases learning-by-doing and helps firms to offer new and upgraded products, while marketing innovation and innovative branding strategies allow firms to differentiate their products from those of their competitors and gain market share. Precise estimation of the impact is clouded by complementarities across different innovation concepts and issues of causality and endogeneity. However, once these are addressed in a convincing fashion, the estimated impact of innovation on productivity can be substantial. In a survey of evidence, Mohnen and Hall (2013) show that the most common value of the elasticity of firm-level productivity with respect to the intensity of product innovation – measured as the contribution of new products developed in the last three years to total sales –is 0.25, suggesting that an increase of 10 percent in the latter raises productivity by 2.5 percent. This relationship is stronger in manufacturing than in services (Criscuolo, 2009). The approach pioneered by Crepon et al (1988), which links knowledge inputs (e.g R&D and ICT adoption) to innovation outputs (e.g. better machines or more efficient managerial practices), and innovation outputs to productivity, generates several insights concerning the sources and effects of innovation in South Asia (Box 2.8). Beginning with knowledge inputs, the most important determinant of R&D adoption for all countries in South Asia is firm size, with larger firms more likely to engage in R&D activities. Having a license to use foreign technology increases R&D in all countries but Bangladesh. Exporters in India and older firms in Pakistan are also more likely to engage in R&D activities than non- exporters and young firms, respectively. Further, financial constraints are associated with lower investments in R&D activities for all countries except Bangladesh. Market structure appears to affect R&D only through informal sector competition in India, while other variables related to market structure lxxii are not significant– perhaps because less than 9 percent of the sample firms compete in an oligopolistic or monopolistic market. Box 2.8 The CDM model The CDM model explores the relationship between the basic determinants of firms’ investment in knowledge and productivity. Firms invest in knowledge inputs that can be transformed into innovation outputs. At a later stage, these outputs affect firm-level productivity, contingent on the capacity of firms to transform innovation outputs into improvements in product quality and efficiency (Crepon, Duguet, and Mairesse, 1988). The CDM model requires the estimation of three main components: (i) the knowledge function, which involves estimating the determinants of R&D and ICT adoption, (ii) the innovation equation, and (iii) the productivity equation. The model is a recursive system of four blocks of equations, where each endogenous variable is determined sequentially. Firms first decide the intensity of two input choices – R&D and ICT. These input choices along with other factors feed into different types of innovation outcomes (product and/or process, or innovation sales). Finally, innovation drives productivity (measured as output per worker) at the firm level through an augmented Cobb-Douglas production function which includes innovation outcomes as inputs. The determinants of the adoption of knowledge inputs include firm characteristics, market conditions and structure, and technology push factors. The first set includes variables capturing firm size, age, and financial constraints (measured by the share of internal sources used to finance working capital). With regard to market structure, early empirical evidence provided by Porter (1990), Geroski (1990), Baily and Gersbach (1995), Nickell (1996), and Blundell, Griffith and Van Reenen (1995) supports the view that competitive pressures encourage innovation, while more recent evidence by Aghion, Bloom, Blundell, Griffith, and Howitt (2005) shows that the relation is inverted U-shaped. In terms of composition, Cusolito (2009) shows that competition induces firms to specialize vertically by upgrading the quality of existing goods. To account for these effects, the model includes variables measuring whether competition from informal firms is an obstacle for the firm, if the sector in which the firm operates has a duopoly structure, and the extent of integration into international markets through trade. With regard to technology, the model considers whether the firm recently upgraded some of its working capital, and whether the firm has a license to use foreign technology, as these variables can make investments in knowledge capital more attractive. Moving on to the determinants of innovation outputs, R&D drives the intensity of innovation (i.e., the share of company’s sales that can be attributed to the introduction of product or process innovation) but does not impact the probability of adopting a technological innovation. ICT is significantly related to innovation intensity only in India and in the adoption of technological innovations in Nepal, but not in the case of Pakistan or Bangladesh. Lack of complementary factors such as skilled labor reduces innovation intensity, although marginally, in all countries except Bangladesh. However, other constraints, for example access to external sources of funding, do not appear to play a significant role. Knowledge spillovers have a positive effect on innovation-induced turnover for leading firms, but they are insignificant for laggards. Further, agglomeration or urbanization effects do not appear to be an important determinant of innovation, with most of the innovation activity occurring outside business cities in Nepal. Demand pull factors, which reflect consumers’ willingness to pay a higher price for a lxxiii given quantity, are important in explaining innovation-induced sales gains in India, Pakistan and Bangladesh. Results from the final stage, which links innovation outputs to labor productivity, show that the impact of innovation on productivity in Nepal and Bangladesh is positive, statistically significant, and larger than in OECD countries (Table 2.13). In India, the large number of observations allows for separate estimation of product and process innovation, with both coefficients positive and statistically significant. The degree of novelty does not introduce any additional effect on productivity, and the returns are the same as imitation. Thus, the evidence suggests that there are positive returns to imitation in South Asia, mostly coming from very incremental innovations in Bangladesh and India, but radical innovations do not increase firm performance above and beyond the gains from imitation.60 Table 2.13 Impact of innovation practices on firm productivity Bangladesh India Nepal Log(L) 0.1429*** 0.1416*** 0.0886*** 0.0901*** 0.3769*** 0.3586*** Log(K/L) 0.2827*** 0.3006*** 0.1567*** 0.1567*** 0.2369*** 0.2421*** Product &/or process 0.5544* 0.6902** 1.3959*** 1.5707*** Product innovation 1.2050*** 1.2146*** Process innovation 0.9759*** 0.9739*** Prod &/or proc*national 0.0094 0.2742 Prod &/or proc*intl -0.0103 0.4718 Product*national 0.0233 Product*intl -0.0972 Process*national -0.0273 Process*intl -0.1172 (0.128) Observations 990 990 3,481 3,481 470 470 *** p<0.01, ** p<0.05, * p<0.1. Constant and sector dummies included but not shown. 60 The estimates are robust to alternative methodologies. Given the fact that the GSEM methodology is more robust with large samples and well specified models, we also estimate the same models using three-stage least squares (3sls). One disadvantage of this methodology is that it uses the sample of the stage with lower number of estimates and also does not allow for a mixed process, since all the stages have to be estimated linearly. On the other hand, one advantage is that it is computational less demanding than FIML and still addresses the issue of endogeneity instrumenting at each stage. The results for the returns to innovation, although larger in size to GSEM, are identical in terms of statistical significance. lxxiv 2.5 Annex Table 2.14 Output and factor misallocation in South Asia Bangladesh: districts India: districts Sri Lanka: districts Misallocation Within Between Misallocation Within Between Misallocation Within Between 1991 Output -0.907 -0.842 -0.065 Labor -0.042 -0.059 0.017 Capital -0.341 -0.318 -0.024 1994 Output -1.322 -1.245 -0.077 Labor -0.003 -0.039 0.036 Capital -0.391 -0.375 -0.015 1995 Output -1.791 -1.692 -0.099 -0.820 -0.689 -0.131 Labor -0.015 0.066 -0.081 0.038 0.123 -0.085 Capital -0.030 0.032 -0.061 -0.046 0.073 -0.119 1996 Output -1.034 -0.990 -0.044 -1.010 -0.898 -0.113 Labor 0.019 0.004 0.015 0.043 0.125 -0.082 Capital -0.134 -0.057 -0.077 1997 Output -3.021 -2.902 -0.119 -0.869 -0.714 -0.154 Labor 1.212 1.233 -0.022 0.093 0.169 -0.076 Capital -0.258 -0.177 -0.081 0.074 0.180 -0.106 1998 Output -0.913 -0.777 -0.136 Labor 0.131 0.221 -0.090 Capital 0.876 0.980 -0.104 1999 Output -1.275 -1.175 -0.100 -1.447 -1.307 -0.140 Labor 0.542 0.614 -0.072 0.017 0.115 -0.098 Capital -0.209 -0.136 -0.073 -0.126 -0.030 -0.097 2000 Output -0.883 -0.763 -0.119 -1.318 -1.192 -0.126 -0.725 -0.516 -0.209 Labor 0.064 0.14 -0.076 0.020 -0.021 0.041 -0.013 0.138 -0.151 Capital 0.016 0.072 -0.056 -0.369 -0.288 -0.081 -0.026 0.136 -0.162 2001 Output -1.130 -0.948 -0.182 Labor 0.057 0.160 -0.103 Capital -0.087 0.065 -0.152 2002 Output -1.644 -1.477 -0.167 Labor 0.132 0.249 -0.117 Capital -0.063 0.046 -0.108 2003 Output -1.677 -1.632 -0.045 Labor 0.047 0.067 -0.020 Capital -0.265 -0.257 -0.008 2005 Output -1.594 -1.488 -0.105 -1.075 -1.030 -0.045 Labor -0.083 0.01 -0.093 0.017 -0.046 0.063 Capital 1.473 1.681 -0.208 -0.233 -0.230 -0.003 2006 Output -1.964 -1.780 -0.184 Labor -0.186 -0.161 -0.025 lxxv Capital -0.417 -0.389 -0.027 2007 Output -1.523 -1.448 -0.075 Labor -0.181 -0.168 -0.013 Capital -0.012 0.027 -0.039 2008 Output -1.368 -1.237 -0.132 Labor -0.068 0.002 -0.070 Capital -0.390 -0.281 -0.109 2009 Output -1.226 -1.103 -0.123 Labor 0.066 0.135 -0.069 Capital 1.585 1.542 0.043 2010 Output -1.362 -1.266 -0.097 Labor 0.019 -0.003 0.022 Capital -0.071 -0.007 -0.065 2012 Output -0.557 -0.505 -0.052 Labor 0.111 0.135 -0.025 Capital 0.069 0.096 -0.027 Note: A more negative number means more efficient allocation of resources (less misallocation); zero means no correlation between productivity and output or employment; a positive number means less productive firms attract more labor or capital than more productive firms. Table 2.15 ICT index by country, sector, and size Categories Bangladesh India Nepal Pakistan South Asia Africa (Average) Aggregate 58 131 51 84 81 65 Main Sectors Services 31 137 58 69 74 64 Manufacturing 68 130 37 95 83 69 Specific Sectors Food 47 136 19 91 73 90 Apparel 101 106 40 113 90 48 Electronics 186 142 65 103 124 109 Automobile 50 127 100 94 93 108 Others 46 132 53 81 78 64 Size Small 21 108 39 58 57 56 Medium 44 129 101 91 91 87 Large 124 158 150 131 141 126 Business Center Main Business City 61 135 54 85 84 71 Other City 52 113 26 58 62 49 City Size Over 1 million 59 137 66 85 81 68 Under 1 million 55 126 41 77 74 56 Note: ICT index is standardized to average 100 and deviation 100. lxxvi Table 2.16 Determinants of overall ICT adoption South Asia Nepal Bangladesh India Pakistan Firms Size 0.1807*** 0.3598*** 0.1654*** 0.1198*** 0.1908*** (0.0072) (0.0529) (0.0244) (0.0170) (0.0412) Age -0.0033 -0.0055 0.0021 -0.0759*** -0.0106 (0.0108) (0.0602) (0.0308) (0.0276) (0.0812) Exporter 0.1603*** 0.2770** 0.4836*** 0.1230*** 0.2857** (0.0223) (0.1368) (0.1275) (0.0426) (0.1132) importer 0.2058*** -0.1158 0.1757* 0.3120*** 0.0431 (0.0273) (0.1121) (0.0937) (0.0655) (0.1278) Foreign 0.1074 -0.0688 -0.0127 -0.1115 -0.5897* (0.0799) (0.3687) (0.1162) (0.1373) (0.3173) New Capital t-1 0.0174 0.2311** 0.1194** -0.0138 -0.1777 (0.0166) (0.0910) (0.0607) (0.0342) (0.1101) Informal Sector Obstacle -0.0146 0.0013 -0.1220** 0.0417 0.2723** (0.0240) (0.0885) (0.0537) (0.0575) (0.1088) license tech foreign 0.0528* -0.0136 0.0291 0.1753** 0.1072 (0.0288) (0.1703) (0.0775) (0.0880) (0.1210) share working cap -0.0004* -0.0012 -0.0004 -0.0001 -0.0041** (0.0002) (0.0011) (0.0006) (0.0006) (0.0021) duopoly/monopoly 0.0499* 0.4256** 0.2233** 0.0540 -0.1430 (0.0274) (0.1867) (0.1117) (0.0508) (0.1164) Business City 0.0163 0.1664 0.0562 0.0990** 0.0633 (0.0209) (0.1018) (0.0689) (0.0498) (0.1344) City > 1 million 0.1332*** 0.4064*** -0.1652* 0.1399 -0.4271 (0.0510) (0.1483) (0.0943) (0.1051) (0.3941) City 250.000 to 1 million 0.1050** 0.4660*** -0.0674 0.0888 -0.5418 (0.0503) (0.1548) (0.0939) (0.1083) (0.4103) City 50.000 to 250.000 0.0583 0.3629*** -0.1528 0.1335 -0.1968 (0.0506) (0.1234) (0.1034) (0.1092) (0.4049) High School workers 0.0017*** 0.0045*** 0.0058*** 0.0009* 0.0013 (0.0003) (0.0013) (0.0013) (0.0006) (0.0014) Observations 5116 470 967 3318 361 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1. Estimates use sampling weights, and country dummies are included in regional pooled estimates. Constant term not shown. lxxvii Table 2.17 Determinants of R&D adoption and ICT intensity Nepal Bangladesh India Pakistan R&D ICT index R&D ICT index R&D ICT index R&D ICT index Firms Size - in log 0.3303*** 0.3447*** 0.2848*** 0.2537*** 0.2251*** 0.1574*** 0.1784** 0.2181*** (0.106) (0.023) (0.039) (0.015) (0.021) (0.008) (0.071) (0.022) Log Firm Age 0.2336 0.0526 0.0577 0.0212 0.0071 -0.0136 0.4247** 0.0946** (0.199) (0.039) (0.069) (0.025) (0.030) (0.012) (0.171) (0.044) Firm Exports 0.1673 0.1565** -0.0011 0.3138*** 0.1286** 0.1035*** 0.0109 0.3080*** (0.304) (0.071) (0.148) (0.056) (0.063) (0.026) (0.247) (0.076) - Working capital -0.0061* 0.0003 -0.0004 -0.0007 0.0019*** -0.0002 -0.0086* -0.0038** (0.004) (0.001) (0.001) (0.001) (0.001) (0.000) (0.005) (0.002) Duopoly/ Monopoly 0.3389 0.2058** -0.0602 0.2111** 0.0093 0.0299 0.0430 0.0114 (0.493) (0.099) (0.245) (0.086) (0.080) (0.033) (0.253) (0.077) New Capital previous Year 0.3052 0.1820*** -0.0271 0.0174 0.0155 -0.0249 -0.0733 -0.0764* (0.230) (0.050) (0.098) (0.046) (0.050) (0.018) (0.226) (0.045) Informal Sector as - Obstacle 0.1104 0.0448 -0.2199 -0.0557 0.2187*** -0.0074 -0.0860 0.0814 (0.269) (0.041) (0.181) (0.065) (0.070) (0.013) (0.299) (0.058) License foreign 1.9250*** -0.1022 -0.1771 0.1327*** 0.1884** 0.0277 0.8201*** -0.0831* (0.462) (0.135) (0.147) (0.051) (0.087) (0.029) (0.249) (0.049) - - - Constant -2.9221*** -1.4913*** -1.6676*** 1.3925*** 1.3279*** 0.5659*** -1.3310 -0.6241* (0.620) (0.499) (0.297) (0.108) (0.149) (0.060) (0.941) (0.353) Observations 470 470 990 990 3,481 3,481 499 499 ISIC- ISIC- ISIC- ISIC- ISIC- Sector dummies ISIC-1digit ISIC-1digit ISIC-2digit 2digit 2digit 2digit 1digit 1digit Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 lxxviii Table 2.18 Determinants of innovation Nepal Bangladesh India Pakistan Technol. innovation Technol. innovation Technol. innovation Technol. innovation innovation sales innovation sales innovation sales innovation sales Firms Size - in log -0.5037*** -0.0190 0.2318 0.0055 0.2800*** -0.0504*** 0.3496*** 0.0124 (0.118) (0.035) (0.228) (0.054) (0.028) (0.016) (0.048) (0.010) Invests in R&D 0.2946 0.1817*** -1.1062*** 0.4595*** -1.8674*** 0.1978** 0.1914 0.0924*** (0.375) (0.025) (0.339) (0.033) (0.066) (0.096) (0.451) (0.035) ICT Index 1.9345*** 0.0497 -0.5478 -0.1465 0.2354 0.1908*** -1.5988*** -0.0468 (0.197) (0.096) (0.911) (0.203) (0.239) (0.062) (0.122) (0.043) Log Firm Age -0.1699* -0.0126 0.0246 -0.0085 -0.0697 -0.0088 0.2090** 0.0152 (0.089) (0.010) (0.063) (0.015) (0.043) (0.006) (0.085) (0.011) Education as Obstacle -0.0404 -0.0280* 0.2207* -0.0045 -0.0412 -0.0218* -0.1052 -0.0310** (0.117) (0.016) (0.124) (0.022) (0.041) (0.013) (0.109) (0.015) Firm Exports -0.4398*** -0.0475* 0.3432 -0.0309 0.1827*** 0.0290* 0.4155*** -0.0076 (0.151) (0.027) (0.288) (0.076) (0.053) (0.016) (0.159) (0.021) Demand Pull Effect -0.0611 -0.0173 0.1592* -0.0526*** 0.0376 0.0220** 0.0413 0.0264** (0.098) (0.015) (0.095) (0.019) (0.037) (0.010) (0.061) (0.013) Working capital -0.0005 -0.0001 -0.0017 0.0008** -0.0005 -0.0001 -0.0059* -0.0001 (0.002) (0.000) (0.001) (0.000) (0.001) (0.000) (0.003) (0.000) Duopoly / Monopoly -0.1417 0.0355 -0.2531 0.0467 0.0796 0.0133 -0.0351 0.0061 (0.226) (0.030) (0.350) (0.069) (0.065) (0.015) (0.140) (0.017) business -0.2419** -0.0294* -0.0306 0.0253 0.0177 0.0020 0.1533 0.0204 (0.118) (0.017) (0.097) (0.020) (0.023) (0.010) (0.154) (0.022) spillover -2.8679 -0.1570 2.4450* -0.7407*** 8.4422 3.1811** 1.5389 0.3170 (1.877) (0.274) (1.482) (0.284) (7.763) (1.536) (1.258) (0.205) Constant 2.2236*** 0.1590 -0.3917 -0.0933 -0.8571*** 0.2784*** -1.7833*** -0.0998 (0.433) (0.189) (1.426) (0.296) (0.241) (0.063) (0.569) (0.087) Observations 470 470 990 990 3,481 3,480 499 502 Sector dummies ISIC-1digit ISIC-1digit ISIC-2digit ISIC-2digit ISIC-2digit ISIC-2digit ISIC-1digit ISIC-1digit Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Note: “Technological innovation” is a dummy variable with value 1 if any new or significantly improved product, service or process introduced by this establishment in last three years. Innovation sales is the share of sales that can be attributed to the introduction of a new or upgraded innovation process. lxxix 3. The way forward 3.1 Potential for increased growth through policy reforms The potential benefits of becoming more competitive (productive) are discussed at the macro, sectoral and firm level in turn below. The policy actions underlying these scenarios are discussed in the next section. 3.1.1 Macro benefits: faster exports growth through higher productivity South Asia has tremendous potential to increase incomes through policies that enhance productivity and gain market share in exports. One approach to assessing the economy-wide implications of such policies is with a global computable general equilibrium (CGE) model, which is used here to consider how a reduction in both international and domestic trade costs could raise income growth in South Asia through 2030. The key assumptions are calibrated to the latest Global Trade Analysis Project (GTAP) dataset with a 2011 base year (the chapter appendix provides a brief introduction to the model). Most importantly, productivity in South Asia is calibrated to contribute an average of 2 percentage points to total GDP growth through 2030 – consistent with what the region was able to achieve during its best recent decade of growth in the 2000s but well above the experience thus far in the current decade (about 0.9 percentage points per year). In this baseline scenario, South Asia’s relatively young population and strong productivity growth achieve a substantial rise in per capita income through 2030. Unlike China, where total population is projected to decline after 2025, South Asia’s population is expected to continue growing, with India becoming the world’s most populous country shortly after 2020. The number of skilled workers – those with a secondary degree or higher – is expected to rise by 84 percent by 2030, ranging from 42 to 125 percent across the countries. The shift from agriculture to industry and services is expected to accelerate, driven by rising incomes and much lower wages in agricultural versus non-agricultural activities. While the agricultural workforce only increases by 10 percent, the labor force in non- agricultural activities rises by some 60 percent (the average increase in the labor force is 30 percent). Consequently, the share of the region’s workers employed in non-agricultural activities increases from 40 to 50 percent. Real GDP in the region would rise by 6 percent per year (tripling by 2030), and South Asia would have one of the highest GDP growth rates in the world, more or less mirroring the growth in East Asia. Per capita income would rise from around $1,400 in 2011 to around $3,400 in 2030 (in 2011 prices and market exchange rates), but still well below the average of $55,000 in high-income countries. Despite rapid gains in productivity, increases in the number of workers and in the volume of capital (particularly the latter) represent the most important source of growth, as in the past. Under these assumptions, South Asia becomes the world’s fastest-growing region in terms of exports. Merchandise exports (in constant US dollars) rise by 264 percent between 2011 and 2030, compared with a 138 percent increase for all developing countries and 83 percent for the world. The majority of this growth comes from manufacturing and service exports, whereas exports of agricultural goods lxxx (excluding processed food) rise by a much more moderate 27 percent. This reflects a limited rise in the agricultural labor force and availability of land, some deceleration in yield growth, and relatively high income growth which shifts demand towards higher valued agricultural goods (fruits, vegetables, dairy etc.) and non-agricultural goods. On the other hand, services exports more than triple and manufacturing exports rise by nearly 300 percent, as South Asia’s rapid labor force growth, rising skill endowments, and rapid productivity growth (implicit in the baseline growth scenario) capture a growing share of the global market in higher value added products (Figure 3.2). Overall, the region more than doubles its global export share in manufacturing, and increases its global share of services exports by 75 percent. Figure 3.2 South Asia’s global market share, 2011 and 2030 (under various scenarios) 2011 BaU Trade margins Domestic margins Iceberg costs Services Manufacturing Agriculture Electronics Motor vehicles Wearing apparel 0.0 5.0 10.0 15.0 20.0 25.0 Within manufacturing, more skill-intensive sectors account for a larger portion of the overall growth: export growth rates range from 193 percent for textiles to 220 percent for wearing apparel to 400 percent for motor vehicles to 435 percent for electronics. Starting from a relatively low base, by 2030 South Asia more than triples its share in global exports of electronics and motor vehicles, and comes close to doubling its already significant market share in wearing apparel (excluding textiles and leather). Within the two former sectors, nearly all the growth comes from India while other countries -- despite rapidly increasing their exports -- remain small players in global markets. Even though India’s exports of electronics and motor vehicles increase more rapidly than China’s, by 2030 India’s exports of motor vehicles only just approach China's current levels, and its 2030 electronics exports remain an order of magnitude below what China exports today. In wearing apparel, performance is more equal across the region. By 2030, Bangladesh and India account for 7 and 8 percent, respectively, of global exports in this category, while Pakistan and Sri Lanka each add another 2 percent of global exports. If productivity growth were instead closer to what the region has been able to achieve in the current decade, the growth of exports and incomes would be significantly slower. Exports would rise by 5.7 percent year, compared to 6.9 percent in the baseline, and gains in global market share of manufacturing and services would be significantly lower. However, even with lower productivity growth, lxxxi real GDP in the region would still expand by 5.0 percent per year – on par with the developing country average and well above the 2.0 percent GDP growth in high income countries. The region still increases its share of global export markets, although its performance suffers vis-à-vis China and the rest of East Asia. Slower productivity growth reduces South Asia’s share of the global wearing apparel market by nearly 4 percentage points, and its shares of automobiles and electronics by 0.5 percentage points each, compared to the baseline scenario. On the other hand, productivity enhancing improvements in trade facilitation and the functioning of domestic markets could generate additional gains in exports. A scenario which lowers the region’s high logistics costs due to weak port infrastructure, burdensome customs regulations, and inefficient warehousing (e.g., comparable to improved performance on the domestic component of the Logistics Performance Index (LPI)) could raise total export growth over the forecast period from 256 percent in the baseline to nearly 340 percent, and increase the trade to GDP elasticity from 1.1 to 1.5. A different scenario which targets the international component of the LPI through more rapid implementation of ongoing improvements in the port-to-port trade and transportation costs would generate lower impacts, as according to the GTAP database the FOB/CIF margins are already low (4 to 8 percent on average). Exports increase by 11 percent in 2030 relative to the baseline, or roughly half the increase in exports generated by the first scenario. The scenarios give rise to increased labor demand in manufacturing. Employment rises relative to the baseline in agribusiness and falls in sectors such as agricultural crops, fossil fuels, and trade, transport and business services–the latter due to increased efficiency in providing trade and transport services. The results differ across scenarios, in part reflecting the initial level of openness to trade—both on the import and export side. As the aggregate supply of labor is the same across simulations, rising demand for labor due to lower trade costs increases wages.61 Reductions in logistics costs and in domestic trade costs raise the average wage by around 12 percent, while the fall in international trade costs increases the average wage by only 1 percent. The average wage of unskilled workers rises by 17 percent, compared to 8 percent for skilled workers, compared to the baseline. The efforts required to achieve the cost reductions differ across the three scenarios. Thus, while the scenarios indicate the magnitude of the effect of policy improvements, decisions on what policies to undertake would also require identifying the needed measures and quantifying their costs. 3.1.2 Sectoral benefits: more jobs, higher earnings, greater inclusion An additional perspective on how productivity improvements could affect the domestic economy can be gained by estimating an “elasticity of substitution” for South Asia’s exports – that is, an expected increase in the region’s exports given a change in relative prices vis-à-vis South Asia’s competitors (see 61 By design, there is no change in the aggregate level of employment across scenarios. More plausibly, the changes induced by the scenario would most likely lead to changes in labor force participation rates: in aggregate as real wages increase; and perhaps across gender to the extent that there is potential expansion of sectors such as wearing apparel with a higher concentration of female employment. lxxxii the appendix for the methodology used). The advantage of this partial equilibrium approach, compared to the general equilibrium approach used immediately above, is that estimates can be obtained at a detailed product level and with much higher precision. To keep the exercise manageable, the analysis focuses on apparel, which accounts for 12 percent of the region’s merchandise exports and employs 3 percent of the region’s workers, and only on exports to the United States and EU-15 markets. South Asian exports could increase sharply if prices were to rise more rapidly in China than in South Asia (Table 3.1 combines the coefficients estimated by the procedure outlined in the appendix with import shares of each country to generate the elasticity of substitution). A 10 percent increase in Chinese prices would reduce US imports from China by 7.9 percent (almost $700 million), while exports from Bangladesh, India, and Pakistan would increase by 13.6 percent ($519 million), 14.6 percent ($414 million), and 25.3 percent ($336 million).62 South Asia’s emerging competitors could benefit even more: Vietnam’s exports could increase by 37.7 percent ($2.2 billion) and Cambodia’s by 51.3 percent ($1.1 billion). The same relative price increase in the EU markets would have little effect on exports from Bangladesh and Pakistan, but would raise exports from India and Sri Lanka by 19.0 percent and 22.5 percent, respectively –consistent with the current production relationships between these countries and the EU. These results suggest that competition in the apparel markets is intense, with buyers highly sensitive to price changes. Thus, policies that increase productivity in apparel may be effective in generating large export gains. Table 3.1 Elasticity of substitution for US and EU apparel imports Bangladesh Cambodia India Pakistan Sri Lanka Vietnam US 1.36*** 5.13*** 1.46*** 2.53*** 0.02 3.77*** EU -0.24 2.53 1.90*** -0.06 2.25*** 1.64*** Source: Author’s calculations using data from OTEXA. Notes: SUR with homogeneity and symmetry and fixed effects and weights. Estimates based on Table 3.3. *** = statistically significant at 1 percent. A negative value means a decline. Increases in exports in the apparel and textile sector would boost South Asian employment. As the textile and apparel sectors are relatively labor intensive, the rise in output would generate a larger increase in employment than on average across sectors (the procedure used to estimate labor demand is explained in the appendix). These jobs are also more likely to be formal. Evidence from Bangladesh suggests that permanent employment is more sensitive to increases in output than is informal employment: although formal workers are more expensive, they are also more productive (Diaz-Mayans and Sanchez, 2004).63 Increased employment and wages also would particularly benefit women (Box 3.1). The demand for female labor in Bangladesh’s garment sector is more elastic than the demand for male labor—a 1 percent increase in foreign sales is associated with 0.04 percent increase in female and 0.02 percent 62 Sri Lanka would experience an increase in exports of less than 1 percent in this scenario. 63 Results for India – the only other country where the data distinguish between permanent and temporary employees – suggest that the elasticities for the two types of labor are about the same. lxxxiii increase in male labor demand. In 2012, a 1 percent increase in expected wages was associated with an 89 percent increase in the probability of female labor force participation in Sri Lanka, 31 percent in Bangladesh, 19 in India, and 16 in Pakistan – although the importance of this channel seems to have declined over time (see appendix for the estimation procedure and Table 3.2 for results).64 This is particularly important for South Asia, which has a large pool of potential female workers as its female labor participation rate is only 32 percent, compared to 58 percent in Latin America and the Caribbean, 62 percent in Europe and Central Asia, and 67 in East Asia (World Bank 2014c). Since apparel is a relatively low-skilled industry, employment opportunities in apparel that pay more than agriculture could potentially draw South Asia’s non-participating women into the labor force. Table 3.2 Marginal effects of female labor participation with respect to the expected wage 1995 2000 2005 2012 *** *** Bangladesh n.a. 1.646 0.141 0.306*** India 0.551*** 0.426*** 0.410*** 0.189*** Pakistan 0.085*** 0.194*** 0.188*** 0.163*** Sri Lanka 1.011*** 0.939*** 0.696*** 0.892*** Note: Expected wages are measured in logarithms. Bangladesh—the last column is 2010, India—the first column is 1994, second—2001, last—2010, Pakistan—the first column is 1996, second—2001, Sri Lanka- first column is 1996, third—2006. Box 3.1 Why focus on women? There is growing evidence that gender gaps in the labor market and low female labor force participation rates have a major impact on incomes. Lower employment and wage rates for women are estimated to reduce GDP per capita by as much as 27 percent in some regions (Cuberes and Teignier, 2012). Another study estimates that raising the female labor force participation rate to the same level as males would raise the GDP in the United States by 5 percent, in Japan by 9 percent, in the United Arab Emirates by 12 percent, and in Egypt by 34 percent (Aguirre et al., 2012). A third study finds that countries with a comparative advantage in female-labor intensive goods are characterized by lower fertility, likely indicating women delay marriage and child-bearing, which can result in better pregnancy outcomes and better health (Do, Levchenko, and Raddatz (2014). At the microeconomic level, some studies show that female labor force participation and employment is beneficial for a number of household indicators, including children’s health and education and decision-making about fertility and marriage.  In India, a randomized experiment finds that an increase in labor market opportunities for women raised their labor force participation and their probability of going to school instead of getting married or having children, along with better nutrition and health investments for 64 There are a number of potential explanations for this, most involving some sort of a U-shaped relationship between female labor force participation and economic development (Goldin, 1995; Verik, 2014). For example, female participation rates may be highest in the poorest countries, where many women are engaged in subsistence activities. Countries with somewhat higher incomes could have lower female participation rates, because of the transition of (mainly) men to industrial jobs. At some point in the developmental process, however, higher female education levels, lower fertility, and a larger share of services in output that opens up opportunities for women results in higher female participation rates. There is also evidence for India that labor market outcomes depend in part on differences in the level of urbanization, with relatively few employment opportunities for women in the growing areas that are more urbanized than villages but less so than large cities (Chatterjee et al, 2015). Other factors might be limited availability of transportation to work, bad working conditions, and a lack of institutions for early childhood education. lxxxiv school-aged girls (Jensen (2012).  Also in India, a recent study on women employed in the textile industry finds that those with a longer history of employment tended to delay marriage and have a lower desired fertility rate. Moreover, these effects had spillovers within the family—the younger sisters of women who worked in textiles also married later, and their younger brothers were less likely to drop out of school (Sivasankaran (2014).  In Bangladesh, a study shows that the growth of the garments sector was associated with a 0.27 percentage points increase in girls’ school enrollment over 1983-2000—a more sizeable effect than a simultaneous supply side intervention of providing a subsidy for girls to remain in school (Heath and Mobaraq (2012). Girls who live near a garment factory are 28 percent less likely to be married and 29 percent less likely to have given birth than those living in villages farther away from a factory.  Also in Bangladesh, a recent study finds that formally employed women had fewer children and possessed greater decision-making power over their own health expenses and formal savings (either through insurance or a bank account). (Kabeer et al. (2013). Our own estimates confirm that South Asian households with women working, especially in the textile and apparel sector in India and Pakistan, tend to have fewer young children on average than women working in agriculture and women who are not in the labor force or unemployed. Also, in Sri Lanka they spend almost twice as much (SLR 1,112) a month on education per student than households with women working in agriculture (SLR 657) (Sri Lanka household survey, 2008). Box Figure: Working in garments with fewer children (Number of children 0–5-years old in a household by female sector of employment) 1.40 Agriculture 1.20 1.00 Textile and apparel 0.80 0.60 0.40 Other Industries 0.20 0.00 Unemployed or not in the labor Bangladesh India Pakistan Sri Lanka force Source: Authors’ estimates from household data. Combining the empirical evidence presented so far in this section – the price sensitivity of South Asian apparel exports to high income markets and the responsiveness of employment to apparel output – provides an estimate of the potential number of jobs that South Asia could generate through greater apparel exports. For the U.S. market, a 10 percent increase in Chinese apparel prices would increase apparel employment in Pakistan for males by 8.93 percent—by far the largest increase among South Asian economies—followed by Bangladesh and India (Figure 3.4). The gains for Sri Lanka are less than 1 percent, but it is important to keep in mind that these estimates are for exports to the United States only. For the EU market, a 10 percent increase in Chinese apparel prices would increase male apparel employment by 8.55 percent in Sri Lanka and 4.30 percent in India, but Bangladesh and Pakistan would experience small decreases because their products are not close substitutes for Chinese apparel in the EU. All of the results are qualitatively similar for females. lxxxv Figure 3.4 Employment effects of 10 percent rise in China apparel prices, by destination and gender 3.1.3 Firm benefits: greater density of successful firms At the firm level, the potential is exhibited by the achievements of leading firms in the region which have managed to rise to standards of global excellence. The experience of these leading firms, as shown in the case studies presented in Volume II, demonstrates that world class levels of operational performance, efficiency, and innovation can be achieved with the right management, scale/technology and worker training. These firms managed to flourish because they were operating in countries and sectors (e.g. apparel in Bangladesh and Sri Lanka) or sub-sectors (e.g. auto parts in India) where the policy environment was conducive or because they were able to internalize some of the constraints in their external environment (e.g. through vertical integration in the case of agribusiness as well as apparel in Pakistan and India). These examples show, at the firm level, the impact that more conducive and supportive policies could have by increasing the number of successful firms and broadening their impact. Leading firms play a critical development role by being major sources of productive employment, exports and innovation; helping improve the performance of suppliers by providing them with access to high value markets, technology, skills and financing; increasing competitive pressures on other firms; and through example providing a source of inspiration for local players and sending a strong message to potential international investors. This section summarizes the experience from more than 50 leading firms in South Asia interviewed in the context of the industry case studies – apparel, automotive, electronics and agribusiness – which comprise Volume II of this report. The case studies reveal that many of South Asia’s top firms are indigenous. These include most of the leading firms in apparel (e.g., US Apparel from Pakistan, Orient Craft from India, Pacific Jeans from Bangladesh and MAS from Sri Lanka) and a growing number of auto parts suppliers (e.g. Bharat Forge, Hi-Tech Gear and HTGL from India). Notable leading South Asian firms in agribusiness include Fauji lxxxvi Foundation (food conglomerate from Pakistan), Dilmah (high value tea from Sri Lanka) and KRBL (Basmati rice from India). Even in the relatively new (to the region) electronics sector, there are examples of emerging world class South Asian firms (e.g., Dixon Technologies and Micromax from India). Most of these firms started from modest beginnings but expanded substantially over time; for example, Dilmah started with 18 staff in 1974 and has grown to 35,000 employees. The experience of these leading firms demonstrates that world class levels of operational excellence, efficiency, and innovation can be achieved with the right management, scale/technology and worker training. For example, the Samsung plant in NOIDA (outside New Delhi, India) ranks second in terms of efficiency out of thirty comparable Samsung plants from around the world, Dilmah and KRBL are recognized as premium tea and rice brands globally, while MAS has developed a range of high performance sportswear based on their innovative synthetic fabric. Some of these firms are turning global through the acquisition of other leading firms abroad: for example, Bharat Forge – a company that has managed to break into design, engineering, R&D, testing, calibration and other higher value- added services, and integrated these with their existing manufacturing product lines – has acquired automotive companies in Germany. In order to acquire these capabilities, South Asia’s leading firms pursued international integration, reaped the productivity gains generated by locating close to suppliers and clients, invested in skills and improved management practices, and benefitted from public investments in trade logistics and innovation. For example:  Global linkages: Many of South Asia’s leading firms actively sought to connect with global leaders through supplier linkages. Examples include Bharat Forge and MSSL with Maruti- Suzuki, Hi-Tech Gear with Hero-Honda, and MAS with Victoria Secret). Over time, these companies challenged themselves further through exposure to export markets and/or very competitive domestic markets (e.g., auto parts following the reduction in import tariffs and electronics).  Agglomeration economies: Geographic proximity to the customer appears to have aided efforts to upgrade products, process and functions for these firms. Their close location enabled MSSL to hold frequent meetings with the OEM on existing products; at times during the course of these discussions, a new need would reveal itself. This would lead to subsequent meetings to identify the OEM’s requirements.  Skills: To compete in global markets, these firms made significant investments in acquiring skilled manpower at all levels and meeting international standards. For example, workers at Tos Lanka undergo training in Japan for a period ranging from three months to one year. The Chairman of Bharat Forge said: “We have leveraged our tie-ups with leading academic institutions to create a strong talent pipeline. Our efforts have resulted in creation of an over 7,000-strong global pool of skilled engineers and technicians”. According to a senior executive from Hi-Tech Gear: “We train workers and lose them to OEMs. But we still train because the ones who stay are crucial for our productivity. Unskilled workers are cheaper but costs match up when their mistakes are financially accounted for.”  Innovation: For MSSL, once they had successfully acquired a new technology and delivered to the customer, they would ask their engineers, “How can we leverage this technology for lxxxvii adjacent products? What more could we do with it? What would that take?” There are several instances where MSSL upgraded products or began to produce inputs to their existing products. It expanded from producing basic plastic components to building tooling and injection molding machines to deliver on a range of complex plastic products. This enabled MSSL to deepen relationships with and increase sales to existing customers, enter new product categories, and expand their participation in global value chains.  Public investments in trade logistics and innovation capacity have also been important to these firms’ success. For example, Pacific Jeans from Bangladesh said that the system of bonded warehouses and back to back letters of credit provided by the government in the 1970s got the industry going by providing it access to critical imported inputs. In the case of KRBL, the Indian government played a critical role in the development of the PUSA-1121 Basmati rice variety. 3.1.4 3.2 Need for greater emphasis on trade policies, spatial policies and firm capabilities We consolidate in this sub section the policy recommendations emanating from the analysis in Volume I and Volume II of this report. Our recommendations fall into three broad categories aimed at the three critical, and underutilized, drivers of competitiveness: (i) Policies to better connect to Global Value Chains, (ii) Policies to maximize agglomeration benefits and (iii) Policies to support innovation and productivity. We discuss each of them in turn below. 3.2.1 Policies to better connect to Global Value Chains Consistent with the analysis of Section 2.4 in this Volume (“Limited success in linking to global value chains”), trade related issues have been found to be the most important constraints on competitiveness/productivity in the four industry case studies presented in Volume II – e.g. difficulties exporters face in importing inputs in a timely manner at world market prices (apparel and electronics), poor trade logistics and inverted tariffs (electronics) and high effective protection rates (automotive assembly and agribusiness). We discuss below each of the four main trade-related policy recommendations emanating from both Volumes: 1. Gradually reduce import tariffs and non-tariffs barriers towards a common low baseline to increase exposure and access to global good practices and remove inverted tariff structures. Reducing barriers to external trade in South Asia would reap substantial productivity gains. Rates of protection can be quite high in South Asia. Tariffs should be gradually reduced in the cases where high tariffs shield industries from international good practices. Such cases have been found and discussed in the automotive assembly industry (23% to 100% on final autos in all regional countries except Sri Lanka) as well as in the agribusiness industry (7 to 30% applied, 16 to 190% bound tariffs). Non-tariff barriers in the region are also pervasive, and para-tariffs can be high (the average import tax rates in Bangladesh and Sri Lanka are more than double the customs duty average if para-tariffs are included). Restrictions on services trade are higher in Bangladesh, India, Nepal and Sri Lanka than in high income economies, and even China, although services restrictions are low in Pakistan. Reducing tariffs on final goods would improve incentives for innovation, and shift labor and capital from low-productivity firms that cannot lxxxviii survive in a more competitive environment to high-productivity firms. The auto-part industry shows how a gradual reduction can lead to increased growth and competitiveness (figure below): 60% 50% 12000 $Million, current prices) Domestic product (US 50% 40% 40% 10000 40% 35% 35% 35% 8000 30% 30% 25% 6000 20% 20% 15% 4000 10% 2000 0% 0 Nominal tariff (in %) Domestic product (US $Million, current prices) Tariffs should also be reduced on intermediate goods in the cases where they are higher than on final goods leading to an “inverted tariff structure” which discourages domestic manufacturing. Such cases were found on yarn in Nepal; 15-21% tariffs on intermediate apparel goods in Bangladesh, Pakistan and Maldives; more than 30% tariffs on auto parts in Pakistan; in India, 7.5% tariff on materials for medical equipment while final goods face tariff of 5%; in Pakistan, finished poultry products are imported at zero duty from Malaysia and at 16 percent duty from China, yet duties on the inputs for local poultry processors are 15-30 percent, in addition to the sales tax (GST) of 17 percent; in India electronics, high tariffs on ‘dual use’ materials under ITA, as process for obtaining exemption from duty is cumbersome; and in India automotive, zero tariffs on final goods under bilateral trade agreements (e.g. with Thailand) while intermediate inputs still face tariffs. 2. Reform the Duty and Tax Remission for Export (DTRE) schemes to facilitate access to imported inputs for exporters. A key policy reform which should not wait for an overall reduction and convergence in tariffs is the reform of the Duty and Tax Remission for Export (DTRE) schemes which are plagued with red tape in the region. These schemes are supposed to enable exporters to import key inputs free of duties and taxes, but in practice they seldom work with the result that exporters are limited to exporting products with locally sourced inputs which greatly constrain their capacity to expand and/or improve quality. The apparel case study is the best illustration of the importance of such schemes as shown by the superior performance of the Bangladesh apparel industry (the apparel industry is the only industry which enjoys extensive access to bonded warehouses and accounts for 90% of its exports) and the Sri Lanka apparel industry (which does not require such scheme as Sri Lanka has zero tariffs on textiles) as compared to the India and Pakistan apparel industries where such schemes are plagued with red tape (figure below). It is thus no surprise that India’s and Pakistan’s apparel export associations have put liberalizing the import regimes for inputs at the top of their list of policy recommendations. lxxxix Difficulties with duty drawback schemes were also found to affect the electronics industry in India (the extremely cumbersome procedures around notification 25/99 discourage firms from using it) and the auto parts industry in Pakistan. 3. Strengthen the soft and hard infrastructure to further facilitate external trade. More generally, the poor efficiency of customs and other clearance procedures for traded goods, as well as inadequate logistic services, are major impediments to firms’ ability to sell to external markets and source inputs efficiently. In India, the average time reported to clear customs varied from 2 to 10 days for large firms, and 14 to 21 days for SMEs. Firms are often forced to hold higher inventories to compensate for lengthy and unpredictable delays in customs, and nevertheless may be forced to delay shipments, both of which can severely erode competitiveness. Important steps to speed clearance at the border include providing for fully electronic submission of documents (single electronic window), putting in place a risk based inspection system for imported containers (reducing the need to physically inspect all of them), improving coordination of border management agencies, and establishing an effective and quick grievance redress mechanism - the current administrative mechanisms take very long, and firms are scared of reprisals. Improvements in port efficiency and service sector infrastructure will also be critical, in particular the capacity for ports to handle larger, more sophisticated vessels. When asked as to why he was not investing in Bihar where most of its labor comes from, a leading apparel player from Rajasthan answered: “fix the Calcutta port”. xc 4. Improve internal connectivity and domestic markets Limits on internal connectivity also constrain firm productivity in South Asia. For example, Indian firms reported that while it takes 11 days for a container to travel from Shanghai to Mumbai, it takes 20 days to travel from Mumbai to Delhi. Firms in auto components, textiles, electronics, and heavy engineering report maintaining 27 percent higher inventories to cope with these internal obstacles. Poor infrastructure is one reason for these delays, but a careful survey shows that a quarter of the journey time is spent at check posts, state borders, city entrances, and other regulatory stoppages. In India, differences in tax regimes between states are important reasons for the need for, and time consumed by, internal clearances. Implementation of a unified goods and services tax would reduce such requirements, eliminate the cascading effect of the Central Sales Tax (CST), and ensure that inter-state and intra-state transactions incur the same tax liability by allowing firms to claim full credit on input purchases. Firms also cite problems in domestic product markets, including controls on prices, inappropriate product standards, other constraints on markets and administrative requirements related to the transport of goods as important constraints on production. Agribusiness firms state that restrictions on markets limit their operations. Outdated regulatory barriers hinder the development of storage and processing infrastructure. In particular, stock limits and price caps can be imposed with penalties that include potential jail sentences of up to seven years, which severely limits private sector interest in participating in these markets. Market committees impose strict controls on the marketing of agricultural produce. Produce can only be traded through the market, or for some crops (e.g. sugar cane) direct purchases are allowed but are subject to a fee. As a consequence, there is no competition from private markets, services are poor, and the setting of fees is opaque. Price caps combined with minimum support prices on commodity products discourage investments in higher quality products, and subsidies on fertilizers and water tend to benefit larger farmers and sustain low productivity and environmentally damaging practices. Product market standards that are unnecessarily restrictive, do not reflect the latest technology, or are seriously out of line with international standards particularly limit productivity in sophisticated industries. For example, the failure to update technical standards to those accepted by the EU and the United States means that automotive firms in India are not prepared for the latest international standards, and thus face difficulties in competing in global markets. Moreover, frequent changes in regulations, for example pertaining to emission norms in India, have imposed heavy losses as firms have to change their technology. Automotive firms cite the need for a clearly articulated, long-term plan on the stance on Euro 4 and Euro 5 introduction, consumption test conditions to orient choices for transmission solutions, and crash-test alignment on foreign requirements milestones to plan changes. Firms also call for greater representation of the private sector in the formulation and implementation of standards. In the agribusiness industry, processors and traders consulted for this report felt that food safety regulations often are rigid, not in pace with scientific advancements, and not in line with WTO’s Agreement on Sanitary and Phyto-Sanitary measures. Overlapping responsibilities among government xci bodies responsible for food safety, coupled with a lack of coordination, impairs transparency and the ability of firms to comply with regulations in Bangladesh, India, and Pakistan. Enforcement of regulations is reportedly inefficient or lacking, and food safety laboratories are not recognized by international bodies and lack the capacity for certain tests, such as for pesticide, mycotoxin and antibiotic residues. As a result, the system fails to effectively protect consumers and impedes firms’ access to foreign markets. 3.2.2 Policies to maximize agglomeration benefits Agglomeration benefits, or the benefits that accrue to firms and workers from locating close together in cities or clusters, are important for productivity. While measures of concentration are high in South Asia, concentration has not increased substantially over the past two decades, suggesting that more productive locations have generally not been successful in attracting additional resources at the expense of less productive locations. This reflects significant barriers to movements of goods, labor and capital across internal borders in South Asian countries. Indeed, impediments to efficient allocation of resources between districts are stronger than distortions within districts, indicating significant barriers to firms reaping the benefits of agglomeration. The case studies in volume II confirm that, when it is allowed to happen, agglomeration is positively associated with firm performance in South Asia. For example, interviews suggest that automotive firms gain substantial benefits from being located in clusters given the importance of frequent technical interactions, and location next to other automotive firms is highly correlated with productivity levels. Another interesting case is the light manufacturing cluster in Sialkot (Pakistan) where agglomeration benefits more than compensated for a challenging investment climate in this distant location (1000 km from the Karachi port. The Sialkot cluster derives its competitive advantage from the ability of firms to hire workers from a large pool of skilled labor as well as from the ability to offer a one stop solution to global buyers – the private enterprises in the cluster financed the development of an international airport which provides direct connections to Dubai as well as the development of new industrial zones to accommodate their growth and help comply with ever more string social and environmental norms. Nevertheless, urbanization economies appear to be more important for productivity in South Asia than clusters – although the two are not mutually exclusive. Evidence from the case studies in Volume II underlines the importance of large cities for specialization by firms, as well as the emergence of specialization within smaller cities. Restrictions on land markets in South Asia discourage domestic and foreign investment and limit the benefits firms can gain through agglomeration and clustering. Firms mention the difficulty in accessing industrial land in Bangladesh as an important constraint on development. Inadequate space in well- serviced clusters also impairs productivity in apparel SMEs that are stranded in congested city centers across South Asia. Difficulties in the land market that will take a long time to resolve (e.g. lack of secured land titles), the need to provide infrastructure and overcome coordination issues, as well as the need to overcome negative externalities (e.g. pollution) and foster positive externalities (e.g. help attract leading investors and generate agglomeration economies and cluster effects) underlines the importance of public interventions in improving access to land. Historically, public support has been xcii provided in most countries through industrial zone developments. These have had a mixed record of success, as many of the public zones were not in appropriate locations or have been poorly managed (e.g. PSIC zones in Pakistan Punjab), while not enough quality industrial land was provided in the most suitable areas. The lack of well-located and well-serviced industrial land limits export oriented FDI in electronics and apparel – this is problematic as export oriented FDI have a choice of countries in which to invest, for example Vietnam which has readily available industrial land in prime locations. A notorious example is Samsung’s decision to withdraw a planned $1.25 billion investment in Bangladesh which would have employed 50,000 workers because the company could not obtain 250 acres in an export processing zone – Samsung is now a major investor in Vietnam where it contributed to launch the electronics industry. Conversely and unfortunately, Indian States have competed fiercely to attract major OEMs with tax incentives and land deals – with the risk of leading to a “fiscal race to the bottom”, sub-optimal investment locations and industry fragmentation. So it is ironic that governments are not providing sufficiently good land to export oriented FDI (which have a choice of country to invest in) while they provide too much land and incentives to market oriented FDI which would have invested in any case. Cooperation with the private sector can play an important role in improving the efficiency and availability of clusters for industrial development. Industrial zones can help SMEs cluster around their main customers (e.g. automotive and electronics) as well as have access to common facilities for R&D and testing facilities, waste disposal, and recycling – for example, the Combined Effluent Treatment Plants in the upcoming leather and apparel parks in Punjab, Pakistan. An important lesson for South Asia can be gained by China’s cooperation with private firms to develop “Plug and Play” industrial zones, which provide SMEs with ready to use standardized industrial buildings, and provide decent worker housing close to the factories. India has developed an interesting public private partnership solution to address the coordination and financing issues associated with moving an urban SME cluster to an industrial estate outside the city. In the Scheme for Integrated Textile Parks, ILFS (a company of mixed public and private ownership) helps SME clusters set up Special Purpose Vehicles, find appropriate land and secure the required financing. ILFS also provides managerial and technical training to the members of the cluster. It is interesting to note that many of the modern industrial clusters in South Asia remain located within or near large urban centers (e.g. Dhaka and Chittagong for apparel in Bangladesh, Delhi/Noida and Chennai for electronics in India, Karachi and Lahore for automotive in Pakistan). This helps explain the result at the aggregate level that agglomeration benefits in South Asia come primarily from urbanization rather than specialization effects (see section 2.3 of Volume 1) as the cases show that most specialization happen within the context of large cities. The cases also show the emergence of specialization within smaller/specialized cities – e.g. the light manufacturing cluster in Sialkot, the automotive clusters in Pune and Aurangabad. This is the prelude to the next wave of economies of agglomeration which should be driven by smaller/specialized cities like it happened in China and more developed regions as primary cities become too congested and expensive. xciii To enable this natural/desirable evolution and as discussed in the World Bank’s South Asia Urbanization Flagship Report (2016), South Asian governments should continue to invest in infrastructure to better connect and equip secondary cities as well as pursue the decentralization agenda. One critical aspect of this decentralization will be to delegate authority over land markets (including over property tax) to (elected) local governments to provide them with the authority, the resources and incentives to promote industrial development by facilitating private sector led industrial zones. This is indeed the path followed by China, which started with five Special Economic Zones launched by the Central Government followed by thousands of industrial zones launched by the private sector with the support of local governments which financed the infrastructure and facilitated access to land. In addition to land market reforms and as suggested by the evidence given in section 2.2.3 on the degree of misallocation of labor and capital, and the importance of agglomeration economies for firm productivity, should focus policy makers’ attention on removing constraints on the growth of firms. The removal of policy-induced distortions that limit the flexibility of labor and capital markets could enable more productive firms to grow. In particular, policies to increase the flexibility of labor markets, especially for women – who face particularly high discrimination in South Asia’s labor markets (World Bank, 2012a) – are likely to substantially reduce misallocation of labor and improve productivity. Additional flexibility could also improve labor mobility, which is relatively low in the region: Glaser, Chauvin and Tobio (2011) found that only 0.4 percent of the population in South Asia lived in a different state five years earlier, compared with 9 percent in the United States. Labor market policies in the region remain an important constraint, especially as per capita incomes rise (Figure 2.2). While hiring rules in the region are rather flexible, dismissal procedures in South Asia are among the most onerous in the world (World Bank, 2012b). Minimizing the misallocation of labor and capital, and maximizing the benefits of agglomeration economies therefore go hand-in-hand. Policies directed at improving urban governance and bridging the region’s infrastructure gap will ensure that firms and workers will be matched more easily. Achieving this will require tackling congestion issues head-on (Box 3.2). In particular, investments in improved urban connectivity (going beyond roads to make investments in public transit),65 provision of quality affordable housing and other basic infrastructure services, and reducing the negative social impact of agglomeration (e.g., crime),66 should be high on the policymakers’ agenda. Box 3.2 Leveraging urbanization in South Asia The World Bank’s 2016 South Asia flagship “Leveraging Urbanization” urges policy action to reduce the high costs of urban congestion in the region. To address key congestion constraints, the report calls the attention of policymakers to three fundamental urban governance deficits: an empowerment deficit, a resource deficit, and an accountability deficit.  Empowerment. Most urban local governments in South Asia suffer from unclear institutional 65 Duranton and Turner (2011) find that in US cities, improved road provision eases traffic congestion only in the short run, which means that expansion of roads is unlikely to relieve congestion in the long run. Public transport improvements appear to be the most powerful tool to alleviate the inconvenience of commuting in urban areas. 66 Using Brazilian city-level data, Lage de Sousa (2014) showed that migration is negatively affected not only by local crime rates but also by those occurring in neighboring areas. xciv roles, limited functional and revenue assignments, and limited control over human resources. Empowering urban local governments in South Asia will require a dedicated commitment to clarifying intergovernmental fiscal legal frameworks by amending existing laws, enforcing them, and in some cases, establishing new and simple laws. Significant effort will also be required to establish and align incentives for urban management, governance, and finance.  Resources. Revenue mobilization and management are difficult for most urban local governments in South Asia. Revenue mobilization is constrained by established fees and tax rates, narrow tax bases, and weak administrative capacity to fully utilize the existing revenue opportunities. Budgetary transfers, while officially unconditional, often come with higher-level rules and “guidance” on use. Improved design, implementation, and effectiveness of intergovernmental fiscal transfers are required to close the resource gap.  Accountability. While formal administrative accountability systems generally exist in the region, many are fairly weak or little used. Even though audits are legally mandated, poorly performing local governments continue to receive transfers without penalty. Bridging the accountability deficit requires the development of better systems and practices and building the capacity of both government (at all levels) and citizens, including nurturing the social contract between local governments and citizens and clarifying fiscal relations between local governments and higher tiers of government. The report also raises three additional, and interrelated, areas for policy action:  Connectivity and planning. Decision makers should focus on strengthening transport links that improve connectivity between urban areas (e.g., between large and secondary cities, and secondary cities and towns), adopt forward-looking planning approaches to guide expansion on city peripheries (where it is most rapid), revitalize city cores by investing in better-quality public urban spaces to enhance pedestrian walkability and livability, and adopt granular spatial planning approaches that permit greater variation in land uses and development intensities.  Land and housing. City and suburban governments need to go beyond slum upgrading and embrace measures to stimulate the supply of affordable housing and offer more options to both low- and middle-income households. The supply of affordable housing can be increased over time through more permissive land-use and development regulations. Also needed is Infrastructure investment to open up land for residential development, easy-to-use land titling and registration systems, and greater access to construction and mortgage finance. In addition, government regulations need to be revised to foster the provision of more affordable rental housing.  Resilience to disasters and climate change effects. Cities in South Asia are particularly exposed to disaster shocks. The first step in developing a resilience strategy is to accurately identify and quantify the national, subnational, and city risks, and build national geo-referenced hazard exposure databases. With the help of urban planners, engineers, and academics, cities should revisit the design and enforcement of building codes and land-use plans to avoid further building in risk-prone areas and to reinforce structures so they are resilient to various hazards. In addition, national disaster risk-financing frameworks need to be developed based on risk layering to match risks with appropriate financing instruments. Source: Ellis and Roberts (2016). 3.2.3 Policies to support innovation and productivity The findings from both Volume I and Volume II suggest different approaches to innovation policy for countries that are innovation leaders versus laggards. For leaders, the critical challenge is how to xcv generate novel, and if possible radical, innovations. Here, a focus on enhancing complementary factors - skills and finance – but more importantly breaking the nature of inward innovation development by supporting cooperation with other firms and institutions is warranted. On the other hand, for laggards the policy focus needs to concentrate on increasing the number of firms engaged in incremental innovation. Firms’ own capabilities, including the efficiency of operations, access to knowledge, the skill level of the workforce, and the ability to innovate, are, together with competition, primary drivers of productivity. Efficiency levels are low in South Asia, and the average firm over-employs relatively scarce capital and under-employs labor, South Asia’s most abundant resource. The uptake of important knowledge inputs varies greatly among South Asian firms. Information and communication technology (ICT) is a major potential enabler of productivity growth, especially in countries/locations further away from the technological frontier. Only half of Nepalese and Bangladeshi firms use computers in their business, which is lower than the average in Africa, while nearly all Indian firms use computers, similar to what is observed in the European Union. ICT adoption rates also vary considerably by sector within and across countries. Even among leaders, however, the use of e-commerce and other productivity-enhancing online business tools is relatively low. The ability to innovate varies greatly across countries and across firms within countries. The share of firms involved in R&D investment in India exceeds the average in the Eastern Europe & Central Asia and Africa, while the share of firms in Nepal and Pakistan involved in R&D is lower than in these two regions. Innovation in South Asia is largely incremental (e.g. upgrading the quality of existing goods or introducing new products to the firm) or involves efforts to imitate technology in more advanced countries, rather than introducing entirely new products or processes. Investment in innovation differs significantly across firms, and tends to be concentrated in a few mature firms. While some large multinationals have set up R&D centers with world class capabilities in South Asia, investment in innovation is low in the industries discussed in Volume II. Investments in R&D are an important determinant of innovation and productivity in the region, and there are multiple examples of public R&D interventions catalyzing firm growth, especially in agribusiness where research needs to be localized and spread widely among a large number of farmers (Box 3.3). Box 3.3 Public support to the development of Pusa-1121 basmati At the turn of the century, managers at KRBL attended a demonstration by the Indian Agricultural Research Institute (IARI) where a new “evolved” variety of basmati rice, numbered 1121, was presented. KRBL staff were shown the extraordinary cooking characteristics which resulted in the longest cooked grain of any basmati type. Subsequently, KRBL acquired a small sample of 3.5kg from IARI, and in 2001 began growing it for multiplication even before the line had entered national trials. Three seasons later, when the variety was officially released as Pusa-1121, KRBL had 20,000 tons ready. Over the next three seasons a portion of the crop was saved for multiplication and a portion milled for test marketing. KRBL had already established a network of farmers through their attempts at contract production. The knowledge that KRBL would buy 1121 in the local wholesale markets took the marketing risk away from the farmers growing the new variety. The results of testing were overwhelmingly positive, both from the growers, who recognized the higher returns with higher yields on a shorter growing cycle with a lower water requirement, and from the consumers in the Gulf markets who found that a cup of milled rice gave 4.5 cups of boiled rice as against the more typical 4 xcvi cups. From there, adoption of the new variety spread rapidly to cover 84 percent of basmati plantings in Punjab and 68 percent in Haryana by 2013. However, overall (public and private) investment in R&D in South Asia is low and has remained relatively unchanged over the past decade, while Latin America and particularly East Asia have increased their investments in R&D. The growing gap is particularly worrisome in light of empirical evidence which shows that social returns from R&D are at least twice as high as private returns (Bloom et al, 2013) and may be even higher in developing countries further away from the technological frontier (Griffith et al., 2004). Therefore, efforts to increase investments in science, technology, and innovation – combined with rigorous evaluations of existing and new programs to ensure efficiency of service delivery and cost effectiveness – are important to boost innovation, productivity, and growth. Skills matter critically for technology adoption in South Asia, and worker skills are an important complement to firm investments in technology, research, and management capabilities. The share of high school graduates in firm employees is positively and significantly associated with ICT adoption in the region as a group, and in all the country except for Pakistan. Many firms in the industries surveyed for volume II cited the low level of skills as a major constraint on productivity. For example, skills were viewed as a key factor in the success of firms in the automotive industry. The lack of adequate managerial skills was seen as a serious problem. For example, only 43 percent of nonproduction workers in the automotive sector in India are formally trained, compared to nearly 70 percent in China. Many firms in South Asia are at a disadvantage vis-à-vis competitors in other countries when it comes to innovation, managerial capabilities, technology adoption, and worker skills – while evidence shows that relaxing these constraints can lead to substantial improvements in productivity. To increase innovation, the policy agenda differs across South Asian countries. Leading countries (Sri Lanka and India) should emphasize supporting cooperation among firms and institutions to achieve novel innovations. By contrast, countries where the observed adoption of innovation is lower (Nepal and Pakistan) should concentrate on increasing the number of firms engaged in incremental innovation. Across South Asia, R&D expenditure is increasingly falling behind other regions, and efforts to increase public and private investments in science, technology, and innovation – combined with rigorous evaluations of existing and new programs to ensure efficiency of service delivery and cost effectiveness – could help boost innovation, productivity, and growth. In addition, authorities should focus on enhancing inputs— technology, skills, and finance— that are complementary to R&D investments. Given the varying rates of technology adoption in the region, increasing the limited adoption of internet and computers among private firms, and then turn to increasing the use of ICT to improve management and performance is particularly important in Nepal and Bangladesh. By contrast, the use of ICT is common in Indian firms, so efforts should be devoted to increasing the use of the internet for the commercialization of products (e-commerce). One reason for low investments in R&D is scarcity of complementary factors such as education and quality of the private sector, including managerial capabilities (Goni and Maloney, 2014). Poor xcvii managerial and worker skills are two major culprits limiting innovation. Governments across the world are experimenting with programs that boost managerial capabilities at the firm level by providing SMEs with access to consulting services (similar to agricultural extension services), and there is much scope for authorities in South Asia to learn from ongoing efforts and implement their own pilot initiatives. Strengthening education systems would also improve the basic skills necessary for adopting technology that underpins innovation and productivity. Public efforts to enhance skills should involve establishing educational partnerships, as well as upgrading university and vocational curricula in relation to procurement, supply chain and marketing competencies, including e-marketing and e-commerce. Increasingly aware of the importance of investments in managerial capabilities, governments across the world are experimenting with programs that boost these capacities at the firm level. In Latin America and Africa, authorities are piloting interventions which provide SMEs with access to individualized consulting services (similar to the approach by Bloom et al, 2013), as well as more novel approaches of providing group consulting services, which can be delivered at lower cost and leverage group-learning dynamics (similar to agricultural extension services). Since much of the original research into the importance of managerial capabilities for firm performance originates in South Asia, there is much scope for authorities to learn from ongoing efforts and implement their own pilot initiatives. Worker skills are an important complement to firm investments in technology, research, and management capabilities. Governments can and should play a leading role in the development of technical, managerial, and vocational skills. Establishing educational partnerships, as well as upgrading university and vocational curricula in relation to procurement, supply chain and marketing competencies including e-marketing and e-commerce, will be important to create a generation of business managers who can successfully communicate with global firms. Firms should forge more robust linkages with local universities and technical universities. Training, however, cannot be limited to pre-service training from public TVET institutes, which has shown mixed success in India and Pakistan as compared to China where vocational training benefit from extensive industry participation (Box 3.4). On-the-job training (including apprenticeships) is also a very effective way to acquire skills (often superior to government- led programs and own investment of the worker) although it has some skill bias towards existing needs. Company-led training programs by Samsung, LG and Intel in Vietnam and by Daewoo in Bangladesh have addressed important skill gaps. Large Pakistani apparel firms report that they carry out in-house training for most of their workers (Nabi & Hamid, 2013). Box 3.4 China's approach to skilling its workforce China has taken effective steps to deal with the demand-side challenges associated with training and skilling of its industrial workforce. Over the years, the Chinese government has invested extensively in vocational education. As a result, nearly 50 percent of the secondary level students in China have access to vocational education. The quality of training in Chinese vocational institutions is good, mainly due to extensive industry participation, favorable government policies and a flexible curriculum. The key stakeholders in the ecosystem work hand-in-hand. Chinese courses mandate students to undergo one year of training to get the diploma- ensuring faster absorption into the job market. Similarly, to make sure that the faculty keeps abreast of the latest industry practices, the Chinese xcviii government has made it compulsory for vocational trainers to spend at least a month every year in manufacturing companies. Additionally, China has made it very easy for vocational students to move back into general academic programs by sufficiently covering general academic skills in vocational curricula. Chinese firms take employee training, This is reflected in the fact that Chinese manufacturers spend twice the amount on training and development than their Indian counterparts. Source: BCG (2013). xcix 3.3 Annex Models and estimation procedures General equilibrium model The forward-looking analysis in this report is based on a global recursive dynamic computable general equilibrium (CGE) model. The CGE approach has several key advantages that can be used to complement other forms of analysis. First, it is deeply structural with multiple sectors and globally integrated markets that determine bilateral flows of goods and services. Second, it is both country and globally consistent. Third, its richness allows for multiple ways to introduce productivity growth. The model is calibrated in the base year (2011) to Version 9 of the GTAP database —the latest available version. The database has been configured for 15 regions—including all of those available for South Asia: Bangladesh, India, Nepal, Pakistan, Sri Lanka and rest of South Asia aggregate region— and 32 economic activities: 12 in agriculture and food, fossil fuels and other mining, key manufacturing sectors (textiles, wearing apparel, electronic equipment and motor vehicles) and 6 service sectors. The model is constructed as a time sequence of comparative static equilibria with dynamic equations linking the periods. In each static equilibria, the model is a relatively standard CGE model. Production is modeled as a nested series of CES functions, with a different nesting for crops, livestock and other sectors. Energy is treated as a special input—it is a complement to capital in the short run, but a substitute in the long run. One additional feature of the production structure is that it has a vintage structure. Substitutability is assumed to be lower with installed capital, and there is a degree of capital mobility friction with installed capital. Thus a negative shock to a sector only leads to partial adjustment to the capital stock in that sector. Household demand is based on the constant-differences-in-elasticity (CDE) utility function that is non- homothetic.67 The government sector purchases goods and services, and collects taxes (on sales, imports, exports, factors and household income). The government deficit is assumed to be held constant—and the direct tax on households adjusts to meet the fiscal target. Investment is savings driven and is equal to the sum of household, public and foreign savings. The baseline assumes a targeted path for investment as a share of GDP; given that government and foreign savings are assumed to be fixed at base year levels, household savings are allowed to adjust to meet the investment target. Trade is modeled using the ubiquitous Armington assumption (Armington, 1969), whereby goods are differentiated by region of origin. In the current version of the baseline, preferences between domestic and foreign consumption are held constant and thus trade shares, in the absence of changes in relative prices, move in rough proportion to GDP (with adjustments largely due to compositional effects). With a fixed capital account balance, the real exchange rate adjusts to ex ante movements in the trade balance. Bilateral trade is identified with four different prices and three wedges. The first wedge is between the producer price and the exporter’s border (or FOB) price—this is an export tax (or subsidy). The second wedge reflects the costs of international trade and transport and converts the FOB price to the CIF price at the border of the importer. The importer adds to this a tariff, the third and final wedge. Factor markets are assumed to clear. The model also allows for a segmented labor market (between agriculture and non-agriculture) using a Harris-Todaro specification (Harris-Todaro, 1970). In the case of a segmented market, migration reflects changes in the relative return to labor across the two labor markets. Capital markets clear at the national level. However, installed capital is only partially mobile in sectors with deficient growth. Finally, land is only partially mobile across (agricultural) activities. Dynamics is composed of three broad elements. The aggregate supply of labor is assumed to grow at the rate of growth of the working age population—defined as the population aged between 15 and 64. The growth of skilled labor is aligned with the growth rate of higher education (those with secondary degree or higher) as projected by IIASA for the shared socio-economic pathways (SSPs), with 67 It allows for much greater richness in cross-price substitutability than the ubiquitous linear expenditure system (LES). But nonetheless, both systems suffer from relatively poor dynamic behavior. In the case of the CDE, income elasticities are relatively stable relative to income growth. c unskilled labor measured as the difference between aggregate and skilled labor. Capital is equated to the previous period’s capital stock, less depreciation, plus the previous period’s investment—the standard motion equation for capital. Productivity is labor-augmenting: it is uniform between skilled and unskilled labor, but the model allows for it to be differentiated across sectors. In the current baseline, it is assumed to be uniform across all sectors and is calibrated to target a given growth path for real income per capita. Estimating the elasticity of substitution for exports The empirical approach employed here is similar to Feenstra (1994), adapted to be used with panel data rather than his cross-sectional approach. The methodology therefore combines elements from a standard gravity model, direct estimation of elasticities, and Feenstra’s model. As in any typical demand equation, the quantity of apparel that buyers want to purchase from each country depends on the price Pi that country offers, the prices Pj that other countries offer, and the total income Y of the buyer. For tractability, the analysis focuses on two major destinations (US and EU-15, which are the two top import markets for apparel with 63 percent of global imports in 2012 according to UNSD, 2014a), two market leaders in apparel exports (China and Latin America), and two emerging competitors of South Asia (Cambodia and Vietnam). The system is estimated with three equations: US or EU imports (denoted by A) from focus country i (Bangladesh, Cambodia, India, Pakistan, Sri Lanka, and Vietnam), U.S. imports from China, and U.S. imports from Latin America. The estimates (shown in Annex Table 3.3) are in line with expectations. The relationship between own price and quantity demanded in foreign markets is negative, with larger absolute values suggesting more elastic demand (e.g., Cambodia, which produces lower-value goods, such as t-shirts, has a more elastic demand than Sri Lanka, which produces higher-valued goods, such as women’s undergarments). Among all countries, China is the most vulnerable in terms of quantity drops if its prices rise, suggesting that if/when prices in China were to rise, demand could shift to other destinations in large amounts. Estimating labor demand This section modifies the classic labor demand model (Hamermesh, 1993) by controlling for structural differences in labor productivity related to the size of firms, macro/global changes over time, and imposing symmetrical cross-wage elasticities. 68 Labor is heterogeneous across males and females, all factor prices including wages and rental rates are exogenous to the firm, and output proxies for exports since most apparel firms in the region are export oriented. This model is similar in spirit to Grossman (1986), who proposes that inter-sectoral labor mobility is responsible for how import competition affects jobs.69 Along the same lines, Revenga (1997) studies the impact of trade liberalization on wages and employment in Mexico’s manufacturing sector, and Currie and Harrison (1997) conduct a similar study for Morocco. The results—which focus on Bangladesh, India, Pakistan, and Sri Lanka—show that labor demand in apparel is not particularly sensitive to the wage rate (see Annex Table 3.4 for detailed estimation results). In apparel, a 1 percent increase in the wage is associated with 0.06 percent (Sri Lanka) to 0.68 percent (Pakistan) decrease in male employment and 0.01 percent (Sri Lanka) to 0.62 percent (Pakistan) decrease in female employment. In textiles, the results are similar, except for India where the change is bigger than in apparel for men and women. The cross-wage elasticities are positive, which means that men and women are substitutes across all industries including textiles and apparel: a 1 percent increase in male 68 Cross-wage elasticity is modeled as elasticity of male (female) employment with respect to change in female (male) wage. 69 Regarding the intersectoral reallocation of labor, Seddon and Wacziarg (2001) and Levinsohn (1999) provide further readings. ci wage is associated with an increase in female employment in the range of 0.07 (0.09) percent in apparel (textile) in Sri Lanka to 0.33 (0.28) percent in apparel (textile) in Pakistan.70 Estimates of labor demand are based on a two-step procedure. First, standard Mincer-type equations are used to establish whether working in apparel carries a wage premium over agriculture (a labor-intensive, low-skilled alternative), especially for women. Individual wages are estimated as a function of age, education, and a set of industry and occupation dummies, controlling for self-selection as in Heckman (1978). Second, building on Becker (1965, 1973, 1974), whose seminal work developed a framework for the analysis and the classic labor supply model (Hausman, 1980; Blundell and MaCurdy, 1999), we explore whether expected higher wages—which could be induced by a greater availability of jobs in apparel in response to an increase in apparel exports—would attract more women into the labor force. Here, female labor supply is a function of the expected wage as well as marital status, education, household size, education of household head, number of children between the ages 0 and 5 and between the ages of 6 and 18, and the rural/urban location dummy. Although Klasen and Pieters (2012) use India data to estimate the female labor supply, this is the first study to do this exercise for the region. Simulation assumptions The key CGE assumptions are calibrated to the latest Global Trade Analysis Project (GTAP) dataset with a 2011 base year. The CGE model is first used to construct a business-as-usual scenario through 2030 that assumes no additional policy reforms or unforeseen shocks. The scenario provides a view of the global economy in general – and South Asia in particular – based on the so-called Middle of the Road scenario of the Integrated Assessment Modeling Consortium (IAMC).71 In this scenario, real GDP in the region would rise by 6 percent per year (tripling by 2030), supported by rapid population growth, an increasing share of skilled workers, and the relatively optimistic productivity growth assumptions embedded in the OECD GDP projections for South Asia. The sources of growth are projected to change somewhat from recent experience (1). The share of GDP growth accounted for by increases in labor and capital falls from 70 percent in 2011-15 to a little over 60 percent in 2026-30. Increases in labor productivity make a greater contribution to GDP growth in the later period, averaging a contribution of 2 percentage points to total GDP growth through 2030. Still, as in the past, increases in the number of workers and in the volume of capital (particularly the latter) represent the most important source of growth. Since the baseline scenario holds the policy environment constant, the following discussion introduces three new scenarios that model improvements in trade facilitation and the functioning of domestic markets in order to quantify the possible effects of policy interventions aimed at enhancing productivity: 70 In India the results were not statistically different from zero. 71 The IAM community has developed five distinct scenarios (referred to as social economic pathways) with different storylines, for example equitable and environmentally sustainable growth or a fragmented world with poor global governance, low growth and persistent high poverty levels. All projections are available at the IIASA website: http://www.iiasa.ac.at/. The SSP2 (Middle of the Road) scenario used here is based on the OECD GDP projections (see Chateau et al 2012) and the medium variant of the 2010 revision of the UN population projections. The growth projections have been modified to gap fill for the missing countries, re-base to a different base year (holding the growth rates constant) and annualize the projections that were made available at 5-year intervals. cii  The first scenario lowers the ‘iceberg’ cost of trade. Iceberg costs ref er to logistic difficulties, for example due to weak port infrastructure, burdensome customs regulations or inefficient behind the border services (e.g., warehousing, transportation), that raise the costs of trade. In the model, iceberg costs are measured by a lower quantity of imports arriving from a source country than the quantity exported. In this simulation, the iceberg parameter is modified so as to lead to a 10 percent rise in the quantity of exports arriving at destination (compared to the baseline), but only for bilateral trade between South Asian countries and the rest of the world. The magnitude of this change is consistent with the potential for improving logistics in the region, given its poor performance on the domestic component of the Logistics Performance Index (LPI).  The second scenario assumes a more rapid implementation of ongoing improvements in the port- to-port trade and transportation costs in the region, as measured by a decline in the trading margin between the FOB and CIF prices. While in the baseline scenario, the margin is assumed to fall by 1 percent per year across all trade relationships, in this simulation the margin declines by 2 percent per year for all trade relationships involving the South Asian countries—for both their exports and their imports. This would be consistent, for example, with South Asia taking advantage of improved international transportation networks (e.g., a denser network), enhanced competition, and some technological upgrading of ports to handle larger, more sophisticated vessels – consistent with improvements in the international component of the region’s LPI.  In the third scenario, a similar assumption is used to lower the domestic cost of trade. We assume that the cost of moving goods from the farm or factory-gate to local markets and the border for export is reduced from an average of 10 percent of the total value of goods to as little as 5 percent.72 This scenario would be consistent with improved functioning of domestic markets due to better product market regulation and reduced distortions involved in the movement of goods within countries. 72 The current version of the GTAP database incorporates the domestic trade margins in the input-output table. The decline in the domestic trade margin is achieved by reducing the input-output coefficient of selected service sectors (if there are not enough services to justify a domestic margin of 10 percent, service requirements are halved). Note that this scenario has no direct impact on the cost of imports as it does not directly affect the end-user price of imports. This is somewhat contrary to reality, as one would anticipate that improvements in domestic margins would also lead to reducing the end-user price of imports. ciii Figure A1 South Asia’s sources of growth, baseline, 2011-2030 civ Table 3.3 Apparel demand function estimates (SUR Weighted Fixed Effects using Shares) (1) (2) (3) (4) (5) (6) VARIABLES Bangladesh Cambodia India Pakistan Sri Lanka Vietnam 1: X Own Price -0.046*** -0.057*** -0.046*** -0.053*** -0.046*** -0.049*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 1: Q World -0.007*** -0.005*** -0.010*** -0.012*** -0.011*** 0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 2: China Own Price -0.058*** -0.068*** -0.053*** -0.059*** -0.052*** -0.063*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 2: Q World 0.041*** 0.036*** 0.042*** 0.041*** 0.045*** 0.032*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 3: LAM Own Price -0.037*** -0.034*** -0.035*** -0.040*** -0.046*** -0.017*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) A: China-X 0.003*** 0.015*** 0.005*** 0.006*** -0.005*** 0.025*** 0.000 0.000 0.000 0.000 0.000 0.000 B: LAM-X -0.018*** -0.018*** -0.013*** -0.012*** -0.012*** -0.020*** 0.000 0.000 0.000 0.000 0.000 0.000 C: China - LAM -0.006*** -0.007*** -0.005*** -0.006*** -0.006*** -0.007*** 0.000 0.000 0.000 0.000 0.000 0.000 Rest of World P 0.061*** 0.060*** 0.054*** 0.059*** 0.063*** 0.044*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 3: Q World -0.025*** -0.025*** -0.029*** -0.030*** -0.029*** -0.022*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Constant 0.393*** 0 0 0.555*** 0 0 (0.018) 0.000 0.000 (0.017) 0.000 0.000 Observations 264,293 244,909 284,071 257,613 260,567 264,175 Notes: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Each column represents the results from a 3- equation system with homogeneity and symmetry constraints imposed. The first equation (with the “1” prefix) is for the country “X” listed at the top of each column. The second equation (with the “2” prefix) represents China. The third equation (with the “3” prefix) represents Latin America. P represents prices. Q represents quantities. LAM represents Latin America. The dependent variable is the share of imports in each 10-digit HTS good from the country specified in each of the three equations in each system. The “A” prefix represents variables that appear in, and are constrained across, equations 1 and 2. The “B” prefix represents variables that appear in, and are constrained across, equations 1 and 3. The “C” prefix represents variables that appear in, and are constrained across, equations 2 and 3 (China and Latin America). The “Rest of World P” variable appears in all three equations and is constrained to have the same coefficient in all three equations. This variable is a proxy for all other possible input factors available to the Buyers when making purchasing decisions. Note: *** p<0.01, ** p<0.05, * p<0.1 cv Table 3.4 Labor demand in textiles and apparel All industries Textile industry Apparel industry Panel A: Bangladesh Male Female Male Female Male Female Log male wage -0.137*** 0.0268** -0.348*** 0.099*** -0.203*** 0.144*** (8.803) (2.506) (14.69) (7.428) (8.301) (7.305) Log female wage 0.027** -0.179*** 0.099*** -0.155*** 0.144*** -0.324*** (2.506) (11.84) (7.428) (8.590) (7.305) (13.91) Log output 0.230*** 0.292*** 0.407*** 0.285*** 0.311*** 0.323*** (45.48) (45.04) (41.44) (23.83) (41.73) (45.63) Small firm -0.830*** -0.618*** -0.548*** -0.618*** -0.479*** -0.699*** (40.47) (23.39) (15.18) (13.76) (10.76) (16.53) Large firm 0.636*** 1.828*** 0.374*** 1.345*** 0.669*** 1.587*** (28.81) (64.25) (8.497) (24.46) (17.96) (44.81) Constant 1.287*** 0.099 1.373*** 0.636*** -1.246*** -0.259 (8.050) (0.554) (5.769) (2.613) (4.521) (0.991) Year dummy Yes Yes Yes Yes Yes Yes Observations 10,656 10,656 3,278 3,278 5,228 5,228 R-squared 0.756 0.812 0.776 0.722 0.586 0.786 Panel B: India Log male wage -0.119*** 0.001 -0.010*** -0.002 -0.120*** -0.030 (9.595) (0.228) (2.814) (0.0921) (2.797) (1.041) Log female wage 0.001 -0.0754*** -0.002 -0.106*** -0.030 -0.139*** (0.228) (7.964) (0.092) (3.713) (1.041) (3.700) Log output 0.158*** 0.126*** 0.137*** 0.0772*** 0.176*** 0.172*** (18.96) (13.46) (11.07) (5.462) (10.33) (11.27) Small firm -0.390*** -0.395*** -0.555*** -0.541*** -0.568*** -0.506*** (28.28) (26.97) (14.77) (13.57) (8.615) (7.666) Large firm 0.396*** 0.440*** 0.532*** 0.557*** 0.570*** 0.651*** (29.62) (29.39) (14.24) (15.26) (13.67) (13.94) Constant 1.604*** 2.824*** 2.231*** (13.13) (10.10) (6.379) Fixed effects Yes Yes Yes Observations 156,102 25,388 12,630 R-squared 0.146 0.170 0.228 cvi All industries Textile industry Apparel industry Panel C: Pakistan Male Female Male Female Male Female Log male wage -0.540*** 0.277*** -0.677*** 0.277*** -0.676*** 0.328*** (33.47) (21.14) (11.89) (5.231) (17.23) (9.855) Log female wage 0.277*** -0.539*** 0.277*** -0.653*** 0.328*** -0.624*** (21.14) (39.23) (5.231) (12.91) (9.855) (18.66) Log output 0.128*** 0.108*** 0.121*** 0.115*** 0.353*** 0.336*** (9.133) (7.607) (3.818) (3.825) (6.986) (6.880) Small firm -0.690*** -0.681*** -0.957*** -0.838*** -0.439** -0.355* (9.257) (9.003) (3.268) (3.004) (2.138) (1.791) Large firm 1.217*** 1.224*** 1.546*** 1.465*** 1.123*** 1.097*** (15.32) (15.17) (5.794) (5.762) (6.442) (6.507) Constant 3.053*** 2.811*** 3.615*** 3.501*** 0.824 0.729 (17.32) (15.89) (5.875) (5.980) (1.390) (1.277) Year dummy No No No No No No Observations 720 720 129 129 88 88 R-squared 0.780 0.711 0.692 0.644 0.813 0.837 Panel D: Sri Lanka Log male wage -0.022*** 0.081*** -0.050*** 0.072*** -0.056*** 0.094*** (31.59) (110.2) (14.36) (45.18) (20.10) (51.28) Log female wage 0.081*** -0.034*** 0.072*** -0.025*** 0.094*** -0.014*** (110.2) (21.64) (45.18) (13.71) (51.28) (6.038) Log output 0.264*** 0.045*** 0.249*** 0.122*** 0.380*** 0.350*** (81.53) (9.567) (29.15) (13.63) (44.20) (45.53) Small firm -0.428*** -0.827*** -0.322*** -0.950*** -0.152*** -0.786*** (28.08) (39.08) (6.967) (19.62) (3.336) (19.38) Large firm 0.629*** 1.684*** 0.953*** 1.130*** 0.666*** 1.067*** (42.07) (80.59) (18.97) (21.44) (16.52) (29.74) Constant -2.689*** 1.322*** -1.571*** 0.478*** -4.757*** -3.627*** (46.18) (16.28) (10.21) (3.233) (29.59) (25.67) Year dummy Yes Yes Yes Yes Yes Yes Observations 20,027 20,027 3,410 3,410 4,401 4,401 R-squared 0.705 0.669 0.749 0.669 0.696 0.819 z-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Note: Bangladesh - Year 1998 and medium size dummies are omitted. Years 1995/96, 1997/98, 1999/2000, 2001/02, 2005/06, and 2012/13. India - Each pair of male/female regression results were estimated as a single equation that included year dummy variables. Estimation (1) included industry dummy variables for textiles and apparel (separately) and their interactions with output. Year 1998 and medium size dummy variables are omitted. Years of panel 1997/98–2007/08. All regressions also included firms-specific fixed effects. Pakistan - Medium size dummy is omitted. Year 2005/06. Sri Lanka - Year 1995 and medium size dummies are omitted. Years 1995-2009 cvii 4. References Abernathy, F., A. Volpe, D. Weil, David, (2006). “The future of the apparel and textile industries: prospects and choices for public and private actors,” Environment and Planning 38(12): 2207-2232. Abraham et al, 2010 Acemoglu, D. and V. Guerrieri. 2008. “Capital deepening and non-balanced economic growth.” Journal of Political Economy 116(3): 467–498. Afram, G. and A. Salvi Del Pero. 2012. Nepal’s Investment Climate: Leveraging the Private Sector for Job Creation and Growth. Washington, DC: World Bank. Aghion, P., N. Bloom, R. Blundell, R. Griffith and P. Howitt (2005) “Competition and Innova tion: an Inverted-U Relationship”, Quarterly Journal of Economics, 120(2), pages 701-728. Amity, M., and J. Konings (2007). “Trade Liberalization, Intermediate Inputs, and Productivity: Evidence from Indonesia.” American Economic Review 97 (5): 1611–38. Armington, P. 1969. Arndt and Kierzkowski, 2001 Arnold, J., B. Javorcik, and A. Mattoo (2011). “Does Services Liberalization Benefit Manufacturing Firms? Evidence from the Czech Republic.” Journal of International Economics 85 (1): 136–46. Arnold, J., B. Javorcik, M.Lipscoomb and A. Mattoo (2015) “Services Reform and Manufacturing Performance: Evidence from India”, Economic Journal, forthcoming, Asian Development Bank and World Bank. 2004. Sri Lanka: Improving the Rural and Urban Investment Climate. Washington, DC: World Bank. Asian Development Bank. 2006. South Asia Economic Report. Manila: ADB. Atkins, D., A. Khandelwal, and A. Osman. 2014. “Exporting and Firm Performance: Evidence from a Randomized Trial,” NBER Working Paper 20690. Aw, B., X. Chen, and M. J. Roberts. 2001. “Firm-Level Evidence on Productivity Differentials, Turnover, and Exports in Taiwanese Manufacturing.” Journal of Development Economics 66: 51–86. Ayygari, M. A. Demirguc-Kunt and V. Maksimovic (2007) “Firm Innovation in Emerging Markets”, World Bank Policy Research Working Paper No. 4157. Baldwin 2006 Baldwin, R. E., & Martin, P. 2004. “Agglomeration and regional growth.” Handbook of regional and urban economics, 4, 2671-2711. Baldwin, R., R. Forslid, P. Martin, G. Ottaviano, and F. R. Nicoud. 2003. Economic Geography and Public Policy. Princeton: Princeton University Press. Banerjee, A. and E. Duflo. 2006. “Growth theory through the lens of development economics” in P. Aghion and S. Durlauf, eds., Handbook of Economic Growth, Vol 1A, pp. 473–552. North Holland: Elsevier Academic Press. cviii Banerjee, A. and E. Duflo. 2012. “Do Firms Want to Borrow More? Testing Credit Constraints Using a Directed Lending Program.” Banerjee, A. and K. Munchi, 2004. "How Efficiently is Capital Allocated? Evidence from the Knitted Garment Industry in Tirupur.” 2004. Review of Economic Studies 71(1):19-42 Banerjee, A., E. Duflo and N. Qian. 2012. ”On the Road: Access to Transportation Infrastructure and Economic Growth in China”, MIT Baumol, W. J. 1967. “Macroeconomics of unbalanced growth: the anatomy of urban crisis.” The American economic review 57(3): 415–426. Baum-Snow, N., L. Brandt, J. V. Henderson, M. A. Turner and Q. Zhang. 2014. ”Roads, Railroads and Decentralization of Chinese Cities”. Bloom, N., A. Mahajan, D. McKenzie, and J. Roberts. (2010). “Why Do Firms in Developing Countries Have Low Productivity?” American Economic Review: Papers & Proceedings 2010, 100:2, 619–623. Bloom, N., M. Schankerman, and J. Van Reenen. (2013). “Identifying Technology Spillovers and Product Market Rivalry” Econometrica 81: 1347–1393 Bloom, N., and J. Van Reenen. (2007) “Measuring and Explaining Management Practices Across Firms and Countries,” Quarterly Journal of Economics, 122(4), 1341-1408. Bloom, N., B. Eifert, A. Mahajan, D. McKenzie and J. Roberts (2013) “Does Management Matter? Evidence from India”, Quarterly Journal of Economics, 128(1), pages 1-51. Bollard, A., P. Klenow and G. Sharma. (2013). India’s Mysterious Manufacturing Miracle.Review of Economic Dynamics 16, 59-85. Bond, S. R. 2002. Dynamic panel data models: a guide to micro data methods and practice. Portuguese economic journal, vol. 1(2), pp. 141-162. Bryan, G., S. Chowdhury and A. M. Mobarak. "Under-Investment in a Profitable Technology: The Case of Seasonal Migration in Bangladesh," Econometrica, 82(5): 1671-1748. Cabral, L. (2007), “Small firms in Portugal: a selective survey of stylized facts, economic analysis, and policy implications”, Portuguese Economic Journal 6(1): 65–88. Carlin, W. and M. Schaffer. 2012. “Understanding the Business Environment in South Asia.” World Bank Policy Research Working Paper 6160. Chatterjee, U., R. Murgai, and M. Rama. 2015. “Employment Outcomes along the Rural -Urban Gradation,” Economic and Political Weekly 50(26-27). Chauvin, J. P., Glaeser, E., & Tobio, K. (2011). Urban economics in the US and India. Presented at the Economic Geography Conference, Seoul, Republic of Korea, June 29. Clerides,S.K., S. Lack and J,R. Tybout. 1998. “Is Learning by Exporting Important? Micro-dynamic Evidence from Colombia, Mexico, and Morroco,” The Quarterly Journal of Economics 113: 903-947. Cohen, W. and S. Klepper (1996a) “Firm Size and the Nature of Innovation Within Industries: the case of Product and Process R&D”, Review of Economics and Statistics 78(2): 232-43 cix Combes, P. P., Duranton, G., and Gobillon, L. 2011. The identification of agglomeration economies. Journal of Economic Geography, vol. 11(2), pp. 253-266. Datta, S. 2012. “The Effect of Improved Highways on Indian Firms”, Journal of Development Economics Deichmann, U., F. Shilpi and R. Vakis. 2008. “Spatial Specialization and Farm-Nonfarm Linkages,” World Bank, Washington, DC. Dekle, R. and G. Vandenbroucke, 2010. "Whither Chinese Growth? A Sectoral Growth Accounting Approach," Review of Development Economics, Wiley Blackwell, vol. 14(s1), pages 487-498, 08. Dekle, R. and G. Vandenbroucke. 2012. “A quantitative analysis of china’s structural transformation.” Journal of Economic Dynamics and Control 36(1), 119–135. Del Mel, S., D. McKenzie and C. Woodruff (2008), ‘Returns to Capital in Microenterprises: Evidence from a Field Experiment,’ Quarterly Journal of Economics, 113(4), 1329-1372. Desmet, K. and E. Rossi‐Hansberg. 2009. "Spatial growth and industry age," Journal of Economic Theory 144(6): 2477‐2502. Donaldson, D. 2014. “Railroads of the Raj: Estimating the Impact of Transportation Infrastructure”, American Economic Review Dougherty, S., R. Herd and T. Chalaux. 2009. “What is Holding Back Productivity Growth in Indi a? Recent Microevidence,” OECD Journal: Economic Studies. Duggan, V., S. Rahardja, and G. Varela. (2013). “Service Sector Reform and Manufacturing Productivity: Evidence from Indonesia.” Policy Research Working Paper 6349, World Bank, Washington, DC. Duranton, G., & Overman, H. G. (2005). “Testing for localization using micro -geographic data.” Review of economic studies, 1077-1106. Duranton, G., & Turner, M. A. (2011). The fundamental law of road congestion: Evidence from US cities. The American Economic Review, 2616-2652. Duranton, G., E. Ghani, A Grover Goswami, W. Kerr. 2015. “The Misallocation of Land and Other Factors of Production in India,” World Bank Policy Research Working Paper 7221. Ellis, P. and M. Roberts. 2016. Leveraging Urbanization in South Asia, World Bank: Washington, DC. Faber, B. 2014. “Trade Integration, Market Size, and Industrialization: Evidence from China’s National Trunk Highway System”, Review of Economic Studies Fafchamps, M. and F. Shilpi. 2005. “Cities and Specialization: Evidence from South Asia.” Economic Journal 115 Fernandes, A. and A. Pakes. 2008. “Factor Utilization in Indian Manufacturing: A Look at the World Bank Investment Climate Surveys Data,” NBER Working Paper 14178. Fernandes, A. and G. Sharma. 2012. “Together We Stand? Agglomeration in Indian Manufacturing,” World Bank Policy Research Working Paper 6062. Fernandes, A. M., and C. Paunov. (2012). “Foreign Direct Investment in Services and Manufacturing Productivity: Evidence for Chile.” Journal of Development Economics 97 (2): 305–21. cx Ferrantino 2012 Ferrantino 2013 Ferrari, A. and I. S. Dhingra. 2009. India’s Investment Climate: Voices of Indian Business. Washington, DC: World Bank. Foster, L., J. Haltiwanger, and C. Syverson. 2008. “Reallocation, firm turnover, and efficiency: Selection on productivity or profitability?” American Economic Review 98: 394-425. Ghani, E., A. Grover Goswami, and W. R. Kerr. 2012a. “Is India's Manufacturing Sector Moving Away From Cities?” NBER Working Paper 17992 Ghani, E., A. Grover Goswami, and W. R. Kerr. 2012b. “Highway to Success: The Impact of the Golden Quadrilateral Project for the Location and Performance of Indian Manufacturing,” NBER Working Paper 18524. Ghani, E., W. Kerr, and S. D. O’Connell. 2011. “Spatial Determinants of Entrepreneurship in India,” NBER Working Paper 17514. Ghani, E., W. Kerr, and S. D. O’Connell. 2013. “Input Usage and Productivity in Indian Manufacturing Pants,” World Bank Policy Research Working Paper 6656. Glaeser, E. and W. Kerr. 2009. “Local Industrial Conditions and Entrepreneurship: How Much of the Spatial Distribution Can We Explain?” Journal of Economics and Management Strategy 18(3): 623- 663. Goldberg, P., A. Khandelwal, N. Pavcnik and P Topalova, (2010). "Imported Intermediate Inputs and Domestic Product Growth: Evidence from India," The Quarterly Journal of Economics 125(4), pages 1727-1767. Gomez-Mera, L., T. Kenyon, Y. Margalit, J. Reis and G. Varela (2015). "New Voices in Investment : A Survey of Investors from Emerging Countries," World Bank Publications, The World Bank. Goni, E. and W. F. Maloney (2014). “Why Don’t Poor Countries Do R&D?, “ World Bank Policy Research Working Paper 6811. Görg, 2000 Gupta, A., I. Patnaik and A. Shah. 2013. “Learning by Exporting: Evidence from India” ADB Working Paper Series on Regional Economic Integration No 119, August. Hall, R. and C I. Jones (1999), “Why Do Some Countries Produce So Much More Output per Worker than Others?,” Quarterly Journal of Economics, 83-116. Hallward- Driemeier, M. C. 2007. “Improving the Climate for Investment and Business in South Asia,” in S. Ahmed and E. Ghani (eds), South Asia Growth and Regional Integration, Delhi: Macmillan India Ltd. Harrison et al., 2008 Harrison, A. and A. Rodríguez-Clare. 2010. Trade, Foreign Investment, and Industrial Policy for Developing Countries, Handbook of Development Economics, Vol 5: 4039-4214. Harris-Todaro, 1970 cxi Hasan, R., D. Mitra, and A. Sundaram. 2013. "What explains the high capital intensity of Indian manufacturing?", Indian Growth and Development Review 6(2): 212 – 241. Hausmann, R., J. Hwang and D. Rodrik (2006). “What you Export Matters”, NBER Working Paper 11905. Helpman, E., and P. R. Krugman (1985), Market Structure and Foreign Trade. Increasing Returns, Imperfect Competition, and the International Economy, Cambridge, MA: MIT Press. Herrendorf, B., R. Rogerson, Á. Valentinyi, 2013a. “Growth and Structural Transformation.” NBER Working Paper 18996. Herrendorf, B., R. Rogerson, and A.Valentinyi, 2013b. “Two Perspectives on Preferenc es and Structural Transformation,” American Economic Review, 2752–2789. Hicks, J. R. (1935) “Annual Survey of Economic Theory: The Theory of Monopoly”, Econometrica, , 3(1), pp. 1-20. Hopenhayn, H. (1992) “Entry, Exit, and Firm Dynamics in Long Run Equilibrium” Econometrica, 60 (1992), 1127–1150. Howitt, P. and P. Aghion. 1998. “Capital accumulation and innovation as complementary factors in long- run growth.” Journal of Economic Growth 3(2), 111–130. Hsieh, C., and P. J. Klenow (2009) “Misallocation and Manufacturing TFP in China and India” Quarterly Journal of Economics, 124(4), 1403–1448. Hsieh, C., and P. J. Klenow (2014) “The Life Cycle of Plants in India and Mexico” Quarterly Journal of Economics, 129 (3), 1035–1083. Hsieh, C.-T., and B. A. Olken. 2014. “The Missing ‘Missing Middle,’” Journal of Economic Perspectives 28(3): 89-108. Hsieh, C.-T., and P. J. Klenow. 2009. “Misallocation and Manufacturing TFP in China and India.” Quarterly Journal of Economics 124(4): 1403–48. Hummels, Rapoport and Yi 1998 Jacobs, J. 1969. The Economy of Cities. Vintage. New York Kathuria, S., A. Portugal, and S. Shahid (2015). “Barriers to regional integration in South Asia: the case of para-tariffs”, mimeo. World Bank, Washington, DC. World Bank. Krugman, Cooper and Srinivasan 1995 Krugman, P. (1994) “Peddling Prosperity: Economic Sense and Nonsense in the Age of Diminished Expectations”, Norton Publishers. Lage de Sousa, F. (2014). Does crime affect migration flows? Papers in Regional Science, 93(S1), S99- S111. Lall et al. 2011. Lall, S.V., Shalizi, Z. and Deichmann, U. 2004. Agglomeration Economies and Productivity in Indian Industry. Journal of Development Economics, vol. 73, pp. 643-673. cxii Leemput, E. 2014. ‘Intra- versus Inter-country trade barriers: Evidence from India” Levinsohn, J., and Petrin, A. 2003. Estimating production functions using inputs to control for unobservables. The Review of Economic Studies, vol. 70(2), pp. 317-341. Lewis, W (2004) “The Power of Productivity: Wealth, Poverty, and the Threat to Global Stability” Chicago: University of Chicago Press. Li, Y. and M. Rama. 2015. “Firm Dynamics, Productivity Growth, and Job Creation in Developing Countries: The Role of Micro- and Small Enterprises,” World Bank Research Observer 30: 3-38. Liu, L. and J. R. Tybout. 1996. “Productivity Growth in Colombia and Chile: Panel-Based Evidence on the Role of Entry, Exit and Learning.” In Industrial Evolution in Developing Countries, ed. M. Roberts and J. R. Tybout. New York: Oxford University Press. Lopez-Acevedo, G. and H. Tan. 2003. “Mexico: In-Firm Training for the Knowledge Economy,” World Bank Policy Research Working Paper 2957. Marshall, A. 1920. Principles of Economics. Macmillan. London. Martin, P., & Ottaviano, G. I. (2001). “Growth and agglomeration.” International Economic Review, 42(4), 947-968. Martin, P., Mayer, T. and Mayneris, F. 2011. Spatial Concentration and Plant-Level Productivity in France. Journal of Urban Economics, Elsevier, vol. 69, pp. 182-195. McKenzie, D. and C. Woodruff. 2015. “Business Practices in Small Firms in Developing Countries,” NBER Working Paper 21505. McKinsey Global Institute (1998) “Productivity: The Key to an Accelerated Development Path for Brazil” (Washington, DC: McKinsey Global Institute, 1998). Melitz, M. J. 2003. “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity”, Econometrica 71: 1695–1725. Melitz, M. J. and G. Ottaviano. 2008. “Market Size, Trade, and Productivity,” Review of Economic Studies 75: 295-316. Melitz, M. J., and S. Polanec. 2012. “Dynamic Olley-Pakes Productivity Decomposition with Entry and Exit,” NBER Working Paper 18182. National Bureau of Economic Research, Cambridge, MA. Michaels G., F. Rauch & S. J. Redding, 2012. "Urbanization and Structural Transformation," The Quarterly Journal of Economics 127(2): 535‐586. Mukim, M. 2011. “Industry and the Urge to Cluster: A Study of the Informal Sector in India,” Spatial Economics Research Centre (SERC) Discussion Paper 72. Munshi,K. and M. Rosenzweig, 2013. "Networks and Misallocation: Insurance, Migration, and the Rural- Urban Wage Gap." Ngai, R. and C. A. Pissarides. 2004. Structural change in a multi-sector model of growth. OECD 2003. OECD Science, Technology and Industry Scoreboard 2003. OECD Publishing. OECD, 2008 cxiii OECD. 2009. Investment Policy Review: India. OECD. 2015. The Future of Productivity, Paris: OECD. Olley, G. S., and A. Pakes. 1996. “The Dynamics of Productivity in the Telecommunications Equipment Industry.” Econometrica 64 (6): 1263–97. Ozden, C. and M. Sewadeh, 2010. “How Important is Migration?” in The Poor Half Billion in South Asia, E. Ghani (ed.) Oxford University Press Pagés, C. ed. 2010. The Age of Productivity: Transforming Economies from the Bottom Up. New York: Palgrave Macmillan. Porter, M. 1990. “The Competitive Advantage of Nations,” Harvard Business Review March – April. Rodriguez-Clare, A., 2007. "Clusters and Comparative Advantage: Implications for Industrial Policy," Journal of Development Economics 82: 43-57. Rosenthal, S. S. and Strange, W. C. 2004. Evidence on the nature and sources of agglomeration economies. In: Henderson, J.V., Thisse, J.F. (Eds.), Handbook of Regional and Urban Economics, vol. 4. Elsevier, pp. 2119-2171 (Chapter 49). Saia, A., D. Andrews and S. Albrizio (2015), “Public Policy and Spillovers from the Global Productivity Frontier: Industry Level Evidence”, OECD Economics Department Working Papers, No. 1238. Schumpeter, J. 1942. Capitalism, Socialism, and Democracy. New York: Harper. Shiferaw, A. 2007. “Firm Heterogeneity and Market Selection in Sub-Saharan Africa: Does It Spur Industrial Progress?” Economic Development and Cultural Change 55(2): 393-423. Sturgeon, T., and Memedovic. 2011. Taglioni and Winkler 2014 Topalova, P. and A. Khandelwal, (2011). "Trade Liberalization and Firm Productivity: The Case of India," The Review of Economics and Statistics, MIT Press, vol. 93(3), pages 995-1009. Tybout, J. R. 1996. “Heterogeneity and Productivity Growth: Assessing the Evidence.” In M. J. Roberts, and J. R. Tybout, eds. Industrial Evolution in Developing Countries: Micro Patterns of Turnover, Productivity, and Market Structure. Oxford: Oxford University Press. Van Beveren, I. 2012. Total factor productivity estimation: A practical review. Journal of Economic Surveys, vol. 26(1), pp. 98-128. Verma, R. 2012. “Can total factor productivity explain value added growth in services?” Journal of Development Economics 99(1): 163–177. Wagner, 2007. World Bank (2012a). World Development Report: Jobs. World Bank (2012b). More and Better Jobs in South Asia. World Bank. 2006a. The Investment Climate in South Asia. Washington, DC: World Bank. World Bank. 2006b. The Maldives: Sustaining Growth and Improving the Investment Climate. Washington, DC: World Bank. cxiv World Bank. 2008a. Afghanistan Investment Climate Assessment. Washington, DC: World Bank. World Bank. 2008b. Harnessing Competitiveness for Stronger Inclusive Growth: Bangladesh Investment Climate Assessment. Washington, DC: World Bank. World Bank. 2009. Pakistan Investment Climate Assessment. Washington, DC: World Bank. World Bank. 2010. Bhutan Investment Climate Assessment Report: Vitalizing the Private sector, Creating Jobs. Washington, DC: World Bank. World Bank. 2015. Leveraging Urbanization in South Asia: Managing Spatial Transformation for Prosperity and Livability. Washington, DC. World Bank. 2016. “Stiches to Riches,” Washington, DC: World Bank. cxv