1 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY MOBILITY AND TRANSPORT CONNECTIVITY SERIES THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Matías Herrera Dappe, Muneeza Mehmood Alam, and Luis Andres 2 MOBILITY AND TRANSPORT CONNECTIVITY SERIES © 2021 The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work 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. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: “Herrera Dappe, Matías, Muneeza Mehmood Alam, and Luis Andres. 2021. The Road to Opportunities in Rural India: The Economic and Social Impacts of PMGSY. Washington, DC: Word Bank. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be consid- ered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opin- ions expressed in the adaptation are the sole responsibility of the author or authors of the adapta- tion and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third- party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of compo- nents can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. Image credits All images used in this report created by PMGSY Project Team, World Bank THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Contents Foreword................................................................................................................................................................................................................................6 Acknowledgments...............................................................................................................................................................................................................8 Abbreviations........................................................................................................................................................................................................................8 About the Authors...............................................................................................................................................................................................................9 Executive Summary...........................................................................................................................................................................................................10 Introduction.........................................................................................................................................................................................................................12 Evaluating the Impact of PMGSY................................................................................................................................................. 13 Organization of the Report.......................................................................................................................................................... 15 References...................................................................................................................................................................................... 16 Chapter 1. Overview of PMGSY.......................................................................................................................................................................................18 ...................................................................................................................................... 19 History of Rural Road Building in India. What Is PMGSY?............................................................................................................................................................................. 20 The Online Management, Monitoring, and Accounting System (OMMAS)............................................................................ 22 References...................................................................................................................................................................................... 23 Chapter 2. Methodology and Data................................................................................................................................................................................ 24 Empirical Approach....................................................................................................................................................................... 25 Data . .............................................................................................................................................................................................. 27 ....................................................................................................................................................................... 33 Robustness Checks. References...................................................................................................................................................................................... 34 Chapter 3. Impacts of PMGSY on Travel Patterns.................................................................................................................................................... 35 Review of the Literature .............................................................................................................................................................. 36 Results of This Study..................................................................................................................................................................... 36 Concluding Remarks..................................................................................................................................................................... 44 References...................................................................................................................................................................................... 45 Annex 3A Summary Statistics...................................................................................................................................................... 46 Chapter 4. Impact of PMGSY on Economic Opportunities and Well-Being.......................................................................................................48 Review of the Literature .............................................................................................................................................................. 49 Results of This Study..................................................................................................................................................................... 50 Concluding Remarks..................................................................................................................................................................... 57 References...................................................................................................................................................................................... 58 Annex 4A Summary Statistics...................................................................................................................................................... 60 Chapter 5. Impact of PMGSY on Wealth and Human Capital Accumulation..................................................................................................... 62 Review of the Literature .............................................................................................................................................................. 63 Results of This Study..................................................................................................................................................................... 64 Concluding Remarks..................................................................................................................................................................... 72 References...................................................................................................................................................................................... 73 Annex 5A Summary Statistics...................................................................................................................................................... 75 4 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Tables Table 1.1 Modules of the Online Management, Monitoring, and Accounting System (OMMAS)System (OMMAS)................. 23 Table 2.1 Number of habitations and households included in the panel .................................................................................... 28 Table 2.2 Summary statistics for village-level control variables from the 2011 Census of India . ............................................. 29 Table 2.3 Summary statistics for household-level control variables ............................................................................................. 30 Table 2.4 Summary statistics for village-level control variables from the 2001 Census of India................................................ 30 Table 2.5 Test of parallel trends in treatment and control groups, 1991–2001........................................................................... 32 Table 3.1 Impact of PMGSY roads on travel time and distance, by destination .......................................................................... 37 Table 3.2 Differential impacts of PMGSY roads on travel time and distance based on terrain ................................................. 39 Table 3.3 Impact of PMGSY roads on travel cost, by destination. ................................................................................................... 40 Table 3.4 Impact of PMGSY roads on trip frequency, by destination and mode.......................................................................... 41 Table 3.5 Differential impact of PMGSY roads on trip frequency based on terrain ruggedness............................................... 42 Table 3.6 Impact of PMGSY roads on crop transport patterns and costs. ..................................................................................... 43 Table 3.7 Differential impact of PMGSY roads on crop transport patterns and costs based on terrain ruggedness............. 43 Table 3A.1 Summary statistics on outcome variables. ..................................................................................................................... 46 Table 3A.2 Summary statistics of agriculture-related outcome variables..................................................................................... 47 Table 4.1 Impact of PMGSY roads on employment.......................................................................................................................... 50 Table 4.2 Differential impact of PMGSY roads on employment based on distance .................................................................... 51 Table 4.3 Differential impact of PMGSY roads on employment based on terrain. ....................................................................... 53 Table 4.4 Impact of PMGSY roads on agriculture............................................................................................................................. 54 Table 4.5 Differential impact of PMGSY roads on agriculture based on distance and terrain................................................... 55 Table 4A.1 Summary statistics on employment outcomes............................................................................................................. 60 Table 4A.2 Summary statistics on agriculture-related outcome variables.................................................................................... 61 Table 5.1 Wealth index based on principal component analysis.................................................................................................... 64 Table 5.2 Impact of PMGSY roads on household wealth................................................................................................................. 66 Table 5.3 Differential impact of PMGSY roads on household wealth based on distance and terrain....................................... 67 Table 5.4 Impact of PMGSY roads on years of completed schooling............................................................................................. 68 Table 5.5 Differential impact of PMGSY roads on years of schooling of girls............................................................................... 68 Table 5.6 Differential impact of PMGSY roads on years of completed schooling based on distance and terrain................... 69 Table 5.7 Impact of PMGSY roads on health-seeking behavior...................................................................................................... 70 5 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 5.8 Differential impact of PMGSY roads on health-seeking behavior based on distance and terrain............................ 71 Table 5.9 Differential impact of PMGSY roads on immunization of girls. ...................................................................................... 72 Table 5A.1 Summary statistics on wealth indexes, 2009 and 2017................................................................................................ 75 Table 5A.2 Summary statistics on years of completed schooling, 2009 and 2017. ...................................................................... 75 Table 5A.3 Summary statistics on health outcomes, 2009 and 2017............................................................................................. 76 Table 5A.4 Summary statistics on immunization of children under four...................................................................................... 76 Figures Figure I.1 Results chain of rural road interventions......................................................................................................................... 14 Figure 2.1 Trends in nightlights.......................................................................................................................................................... 31 Figure 2.2 Number of control and treatment habitations connected, 2003–17........................................................................... 33 6 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Foreword At the turn of the millennium, about 330,000 habita- tions in India lacked a connection to an all-weather road. The 300 million affected residents were severely constrained in their access to economic opportuni- ties, including opportunities outside agriculture, and basic services, such as education and health. In response to limited rural connectivity, in 2000, the government of India launched the Pradhan Mantri Gram Sadak Yojana (PMGSY) rural roads program. The scheme targeted the provision of all-weather roads to about 178,000 habitations. By 2018, it had achieved 90 percent of its goal, having connected more than 159,000 habitations through about 562,000 kilometers of new and improved roads, at a cost of about $27 billion. Since the mid-2000s, India has become a mid- dle-income country. Average GDP per capita nearly quadrupled between 2000 and 2016, rising from US$440 to US$1,700. One of the forces behind this transformation was the movement of people out of agriculture and into manufacturing and services. Rising labor earnings caused by the movement to non-farm work and the unprecedented rise in wages for unskilled labor propelled a rapid decline in pov- erty between 2005 and 2012. Improved rural connectivity played an important role in supporting this progress. This report presents an evaluation of the impact of PMGSY on economic and social outcomes in the states of Himachal Pradesh, Madhya Pradesh, and Rajasthan. The evaluation finds that improved accessibility had only a limited impact on agriculture, but it caused an important shift in the employment structure in newly connected 7 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY habitations: A significant share of male farmers The World Bank has been engaged in PMGSY since switched to the non-agricultural sector, and women its inception and this impact evaluation is part of the took over their agricultural activities. Farmers were engagement. The Bank provided technical support to also able to sell their products in more distant mar- develop a rural roads policy framework and a mainte- kets on more favorable terms, thanks to the newly nance policy for rural roads; built the capacity of rural built or upgraded roads. Road connectivity also road agencies; undertook systematic assessments of improved access to education and health, with both PMGSY to enhance its design and implementation; boys and girls attending school longer and the share and developed asset management plans, environ- of babies delivered at home decreasing sharply. mentally optimized design guidelines, and the train- Earlier research has documented some positive ing framework for PMGSY. It has provided PMGSY effects of PMGSY. This impact evaluation builds on with more than US$2 billion in loans in concessional the literature by making use of a novel panel dataset terms to build rural roads in nine states (Bihar, based on surveys carried out exclusively to evaluate Himachal Pradesh, Jharkhand, Meghalaya, Punjab, the program. Use of these purposefully conducted Rajasthan, Tripura, Uttar Pradesh, and Uttarakhand). surveys at the household- and habitation-level allows for a fine-grained analysis covering a broad range of The results of this impact evaluation underscore the relevant questions, including the program’s impact benefits of providing access to reliable rural roads. on mobility, employment, agricultural production and Policymakers in India and beyond may draw inspira- distribution, wealth, access to education and health tion from these findings to promote rural connectivity services, and differential impacts across habitations in a manner that yields benefits to rural communities and genders. and the country as a whole. Guangzhe Chen Regional Director South Asia Infrastructure World Bank Acknowledgments This report was prepared by a team led by Matias Herrera Dappe under the overall direction of Karla González Carvajal, and Shomik Raj Mehndiratta. The core team included Muneeza Mehmood Alam, Luis Andres, Reenu Aneja, Aphichoke Kotikula, Stephanie Gimenez Stahlberg, Akib Khan, and Basheer Saeed. The extended team included, Sonali David, Tonmoy Islam, and Yiran Xin. Administrative support was provided by Aruna Aysha Das and Tema Alawari Kio-Michael. Barbara Karni edited the report. The team thanks the following colleagues for their helpful contributions, comments, and suggestions: Arnab Bandyopadhyay, Vincenzo Di Maro, Mesfin Wadajo Jijo, Ashok Kumar, Dhushyanth Raju, and Jennifer Solotaroff. The team thanks the following officials from the government of India for their helpful contributions, comments, and support: Rajesh Bhushan (Additional Secretary, Cabinet Secretariat, and former Joint Secretary, Ministry of Rural Development, and Director General, National Rural Roads Development Agency [NRIDA]); Uttam Kumar (Director, NRIDA); I.K. Pateriya (Director, NRIDA); A.V. Rajesh (Joint Director, NRIDA); and Alka Upadhyay (Joint Secretary, Ministry of Rural Development, and Director General, NRIDA). Peer reviewers Sam Asher, Simon David Ellis, George Joseph, Rinku Murgai, and Fernanda Ruiz Nuñez pro- vided insightful and constructive comments. Financial support from the Australian government is gratefully acknowledged. Abbreviations CNCPL Comprehensive New Connectivity Priority List DPIU District Program Implementation Unit DRRP District Rural Roads Plan MoRD Ministry of Rural Development NRIDA National Rural Infrastructure Development Agency OMMAS Online Management, Monitoring, and Accounting System PMGSY Pradhan Mantri Gram Sadak Yojana SRRDA State and Rural Roads Development Agency About the Authors Muneeza Mehmood Alam is a Senior Transport Economist in the Middle East and North Africa Transport Unit of the World Bank. Muneeza brings with her more than 15 years of experience working on development issues and joined the World Bank in 2015. During her time at the Bank, Muneeza has worked on a myriad of topics relating to transport and economic policy, particularly economic corridors and regional connectivity, urban transport, logistics, gender and spatial inclusion, and electric mobility. She has a keen interest in under- standing the mechanisms through which the economic and social benefits of transport investments can be maximized and more equitably distributed. Muneeza has previously worked in the Global, as well as, South Asia Transport Units of the World Bank. Before joining the World Bank, she worked in eco- nomic consulting. She holds a PhD in economics from Yale University. Luis Andrés is the Infrastructure Sector Leader and Lead Economist for Brazil at the World Bank. He oversees operations and analytical work in the transport, energy, water, and digital development portfolios implemented by the World Bank. Before joining the World Bank, he was the Chief of Staff for the Secretary of Fiscal and Social Equity for the government of Argentina and held other positions at the Chief of Cabinet of Ministries and the Ministry of Economy, Works, and Public Services in Argentina. He has written more than 90 books, book chapters, monographs, and articles on development policy issues, with a focus on infrastructure. He holds an engineering degree from the University of Buenos Aires and a PhD in economics from the University of Chicago. Matías Herrera Dappe is a Senior Economist in the Transport Global team in the Infrastructure Practice Group of the World Bank, where he leads policy research programs on infrastructure. He has published extensively on the links between transport and trade and transport and economic development, the efficiency of ports and logistics, infrastructure investment needs and access, private participation in infrastructure, competition, and auctions. Before join- ing the World Bank, he worked for consulting firms and think tanks, advising governments and companies in Latin America, North America, and Europe. He holds a PhD in economics from the University of Maryland, College Park. 10 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Executive Summary At the end of the 20th century, about 300 million PMGSY INCREASED ACCESS TO ECONOMIC people in rural India had limited connectivity with OPPORTUNITIES, TRIGGERING A CHANGE IN THE the rest of India and the world because their villages STRUCTURE OF EMPLOYMENT IN RURAL INDIA and habitations lacked all-weather road access. The improved accessibility provided by PMGSY roads In response to this poor connectivity and limited triggered a shift from farm to nonfarm employment, opportunities, in 2000 the prime minister announced particularly nonfarm employment outside the the Pradhan Mantri Gram Sadak Yojana (PMGSY) habitation. As a consequence of PMGSY roads, the rural roads program. The program seeks to establish rate of primary employment in the nonfarm sector farm-to-market connectivity by providing access to increased by about 12 percentage points in the all-weather roads to about 178,000 habitations across habitations studied. This increase represents a 33 India. By December 2017 it had built more than percent increase over the average share of nonfarm 550,000 kilometers of rural roads, connecting more primary employment in 2009 in habitations that were than 159,000 habitations, at a cost of $27 billion. connected after 2009. The share of people with pri- mary employment outside their habitation increased This report presents the results of an impact evalu- by 8 percentage points. This increase represents a ation of PMGSY that uses a difference-in-difference 35 percent increase relative to the average share of approach and panel data from the states of Himachal primary employment outside the habitation in 2009 Pradesh, Madhya Pradesh, and Rajasthan collected in in habitations that were connected after 2009. Most 2009 and 2017. workers who switched to nonfarm employment were men. Women stepped in to take care of the farm PMGSY IMPROVED ACCESSIBILITY, after road connectivity was improved. The entrance PARTICULARLY IN HILLY AREAS of women into the workforce was the main force On average, people travelled to their destinations, behind the 5.5 percent increase in employment in particularly work, in shorter time, thanks to improved connected habitations. connectivity, but they did not change the distance travelled. Reductions in travel time were greater in The impact on employment was correlated with dis- hilly areas. Households’ transport costs did not seem tance from urban areas. In habitations 10 kilometers to have changed after PMGSY roads were built, possi- farther away from the nearest urban agglomeration bly because people switched from walking and public than the average habitation, employment increased transport to bicycles and private motorized vehicles. by 6 percentage points more than in the average habitation. More isolated habitations also experi- Thanks to improved connectivity, people in hilly areas enced larger increases in the shares of students made more trips to work per week; in flat areas they and housewives getting part-time jobs. The shift made fewer weekly trips to work. Patterns on trips to to employment outside the habitation, particularly local markets were similar. nonfarm employment, was more pronounced in hilly areas than in flat land, a finding that is consistent with people making more weekly trips to work in those areas. 11 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY PMGSY improved farm-to-market connectivity, PMGSY roads had a positive impact on human but it had limited impact on farming practices capital formation in rural India, with boys and PMGSY roads yielded an 8 percentage point increase girls benefiting equally in the share of crops transported to markets for sale, Improved rural connectivity provides a long-term a tripling over levels observed before PMGSY roads and sustained boost in the living standards of rural were built. The increase in the share of crops sold at populations if it allows households to accumulate market was larger in hillier areas than in flatter ones. wealth and human capital. In the habitations studied, Farmers selling food grains traveled 8.9 kilometers rural roads had a positive but small effect on the farther after the PMGSY roads were built, which sug- average wealth of households, equivalent to adding gests that farmers were travelling to locations where small appliances (like a pressure cooker and radio) prices for their crops were higher. The cost to carry to the household’s assets. The estimated impact on the crops did not seem to have changed as a result wealth is statistically significant only under certain of improved connectivity. Hence the results suggest specifications. improvement in rural road connectivity allowed farmers to take advantage of more favorable market On average, children who were in middle or high conditions. school at the time their habitation was connected had 0.7 more years of schooling in 2017 as a result Increased access to markets had some effects of PMGSY roads that were built in the previous three on farming practices, but they were smaller than years. The analysis found no overall impact on pri- expected, potentially hinting at the need for comple- mary schooling, although years of primary schooling mentary interventions to support the development of rose in hilly areas. agriculture value chains. The average land area under cultivation did not change after road construction, The share of babies delivered at home decreased by except in hilly areas, where it decreased. Rural roads 30 percent in connected habitations, and the reduc- had a positive impact on food grain yields in habita- tion was even larger in habitations farther away from tions farther away from urban agglomerations and in urban agglomerations. Young children in connected hilly areas, which could be driven by less productive habitations were also less likely to fall sick, possibly land being taken out of cultivation. because vaccination take-up among children under the age of four increased by 15 percentage points, with boys and girls benefiting equally. 12 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Introduction 13 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY At the end of the 20th century, about 330,000 of and van Ommere 2010) or to open a business at or India’s 825,000 villages and rural habitations (hamlets closer to home (Rosenthal and Strange 2012). It also or subvillages) lacked all-weather road access.1 As a lowers female educational attainment and access to result, about 300 million people—28 percent of the health care services. population—had limited connectivity with the rest of India and the world. It was against this background of poor connectivity and limited opportunities that India’s prime minister Limited connectivity means that people have to travel announced the Pradhan Mantri Gram Sadak Yojana long distances to reach certain places and to pay more (PMGSY) rural road program, in 2000. PMGSY set a to do so. It thus limits access to economic opportu- target of connecting every habitation with a popula- nities (employment, product markets) and to basic tion of more than 500 with all-weather roads.2 It was services (education and health). Limited connectivity estimated that about 178,000 habitations would be also hinders interactions between productive activities provided with connectivity under the program. By holding down economic efficiency and growth. 2018 PMGSY had delivered more than 550,000 kilo- meters of all-weather rural roads, connecting more Limited access to economic opportunities and human than 159,000 habitations, at a cost of $27 billion. capital accumulation hinders improvements in the liv- ing standards of the rural poor. High transport costs are an impediment to higher income (Escobal and Ponce 2002, Jacoby and Minten 2009, Cuong 2011); Evaluating the per capita consumption and better livelihoods (Emran and Hou 2013); and poverty reduction (Dercon and Impact of PMGSY others 2008; Khandker, Bakht, and Koolwal 2009). Households with limited connectivity are also more A wealth of anecdotal evidence suggests that PMGSY likely to be poor in multiple dimensions (Ali and oth- has had positive impacts on rural communities. ers 2015). According to one beneficiary, “It would take nearly two to three days to reach the nearest hospital as Limited connectivity affects men and women differ- we had only camel carts to transport our sick and ently, because they have different mobility patterns pregnant women and children. Many of them would (Hanson 2010) and face different restrictions and die on the way to hospital. Now, the nearest hospital challenges. In many contexts gender-related restric- … is just two hours away” (World Bank 2014). Another tions and challenges to mobility translate into limited beneficiary noted, “My husband can now cycle to the access to and utilization of economic and social local market to sell the farm produce. My children opportunities. Poor transportation can lead women can now go to the English-language school in town” to limit their job search radius (Gutiérrez-i-Puigarnau (World Bank 2014). 1 An all-weather road is one that is negotiable every season of the year, implying that the road bed is drained effectively (by adequate cross-drainage structures, such as culverts, minor bridges and causeways). It need not be paved, surfaced, or black-topped. 2 The cut-off is 250 people in the hill states (the North-Eastern states, Sikkim, Himachal Pradesh, Jammu and Kashmir, and Uttarakhand); desert areas; and tribal areas. 14 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Figure I.1 Results chain of rural road interventions Improved Economic Capital Rural roads accessibility opportunities accumulation • Employment • Wealth • Agriculture production • Human Capital Several studies document the impacts of PMGSY findings in the literature on the impacts of PMGSY. It (Mukherjee 2012; Bell and van Dillen 2014, 2015; answers eight questions, based on the result chain Banerjee and Sachdeva 2015; Asher, Garg, and presented in figure I.1: Novosad 2020; Aggarwal 2018; Asher and Novosad 2020; Adukia, Asher, and Novosad 2020). Overall, they 1. What are the impacts of PMGSY roads on find that PMGSY had positive albeit limited impact mobility? on the economic structure of rural areas and on education and health outcomes. This report adds to 2. What are the impacts of PMGSY roads on trans- this body of research by using a novel panel dataset portation of farm products to markets? based on household and habitation-level surveys conducted in 2009 and 2017 with the purpose of 3. Did PMGSY roads improve access to economic evaluating the program. opportunities through changes in employment? All previous research on PMGSY except Bell and van 4. How did agriculture production respond to Dillen (2014, 2015) used census and surveys designed improved farm-to-market access? for purposes other than evaluating the impact of roads, limiting the research questions that could 5. Did households in connected habitations increase be tackled. Only one previous study examined the their wealth? impacts of PMGSY roads on mobility (Bell and van Dillen 2015), who examine accessibility to medical 6. Are PMGSY roads providing the foundations for facilities. The use of census and surveys designed for sustained poverty reduction, by increasing access purposes other than evaluating the impact of roads to education and health services? also constrains the analysis to the village or a higher level of aggregation, which attenuates the estimated 7. Did some types of habitations benefit more from effect of rural roads, as the intervention is at the PMGSY roads than others? habitation level. 8. Did men and women benefit differently from This report presents the results of an impact evalu- improved connectivity? ation of PMGSY and compares the results with the 15 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Organization of the Report This report is organized as follows. provides an overview of the PMGSY program. It discusses Chapter 1 its design, its salient features, and progress so far. presents the methodological framework and describes Chapter 2 the data used in the analysis. assesses the transport-related impacts. Chapter 3 describes the impacts on economic opportunities and Chapter 4 examines differential impacts. looks at the impacts on wealth and human capital. Chapter 5 16 MOBILITY AND TRANSPORT CONNECTIVITY SERIES References Adukia, A., S. Asher, and P. Novosad. 2020. “Educational Investment Responses to Economic Opportunity: Evidence from Indian Road Construction.” American Economic Journal: Applied Economics 12(1): 348–76. Aggarwal, S. 2018. “Do Rural Roads Create Pathways out of Poverty? Evidence from India.” Journal of Economic Development 133: 375–95. Ali, R., A.F. Barra, C.N. Berg, R. Damania, J. Nash, and J. Russ. 2015. “Transport Infrastructure and Welfare: An Application to Nigeria.” Policy Research Working Paper 7271, World Bank, Washington, DC. Asher, S., T. Garg, and P. Novosad. 2020. “The Ecological Footprint of Transportation Infrastructure.” The Economic Journal 130(629): 1173–1199. Asher, S., and P. Novosad. 2020. “Rural Roads and Local Economic Development.” American Economic Review 110(3): 797–823. Banerjee, R., and A. Sachdeva. 2015. “Pathways to Preventive Health: Evidence from India’s Rural Road Program.” University of Southern California Dornsife Institute for New Economic Thinking, Research Paper 15-19. Bell, Clive, and Susanne van Dillen. 2014. “How Does India’s Rural Roads Program Affect the Grassroots? Findings from a Survey in Orissa.” Land Economics 90 (2): 372–94. ———. 2015. “The Ways to Good Health? Rural Roads, Illness and Treatment in Upland Orissa.” Discussion Paper, University of Heidelberg, Department of Economics, Heidelberg, Germany. Cuong, N.V. 2011. “Estimation of the Impact of Rural Roads on Household Welfare in Viet Nam.” Asia-Pacific Development Journal 18 (2): 105–35. Dercon, S., D.O. Gilligan, J. Hoddinott, and T. Woldehanna. 2007. “The Impact of Roads and Agricultural Extension on Consumption Growth and Poverty in Fifteen Ethiopian Villages.” CSAE Working Paper 2007–01, Centre for the Study of African Economies, Oxford University, Oxford. Emran, M.S., and Z. Hou. 2013. “Access to Markets and Rural Poverty: Evidence from Household Consumption in China.” Review of Economics and Statistics 95 (2): 682–97. Escobal, J., and C. Ponce. 2002. “The Benefits of Rural Roads: Enhancing Income Opportunities for the Rural Poor.” Grupo de Análisis para el Desarrollo Working Paper 40–1, Lima. Gutiérrez-i-Puigarnau, E., and Jos van Ommere. 2010. “Labour Supply and Commuting.” Journal of Urban Economics 68 (1): 82–89. Hanson, Susan. 2010. “Gender and Mobility: New Approaches for Informing Sustainability.” Gender, Place and Culture 17 (1): 5–23. 17 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Jacoby, H.G., and B. Minten. 2009. “On Measuring the Benefits of Lower Transport Costs.” Journal of Development Economics 89 (1): 28–38. Khandker, S.R., Z. Bakht, and G. B. Koolwal. 2009. “The Poverty Impact of Rural Roads: Evidence from Bangladesh.” Economic Development and Cultural Change 57 (4): 685–722. Mu, R., and D. van de Walle. 2011. “Rural Roads and Local Market Development in Vietnam.” Journal of Development Studies 47 (5): 709–34. Mukherjee, Mukta. 2012. “Do Better Roads Increase School Enrollment? Evidence from a Unique Road Policy in India.” https://ssrn.com/abstract=2207761. Rosenthal, Stuart S., and William C. Strange. 2012. “Female Entrepreneurship, Agglomeration, and a New Spatial Mismatch.” Review of Economics and Statistics 94 (3): 764–88. World Bank. 2014. “Improving Connectivity across Rural India.” http://www.worldbank.org/en/ results/2014/04/10/improving-connectivity-roads-rural-india. 18 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Chapter 1. Overview of PMGSY 19 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY History of Rural Road Building in India In 1929 the government of India appointed the National Rural Employment Programme, the Rural Jayakar Committee, the first organized effort at road Landless Employment Guarantee Programme, and building at the national level. As a follow-up to a the Employment Assurance Scheme. In some states, recommendation of the committee, the central gov- market committees and sugar cane societies also ernment set up the Indian Roads Congress in 1934. It built rural roads. facilitated the formulation of a sequence of plans: the Nagpur Plan (1943–62), the Bombay Plan (1961–81), Although road length increased under these and the Lucknow Plan (1981–2001). schemes, these efforts were not enough to achieve full connectivity of villages. The specifications for Each plan was more ambitious than the previous these roads were probably good enough for ani- one, setting standards, norms, and targets for road mal-drawn carts and pedestrians, but the advent development of various categories. The Bombay of fast and heavy vehicles, coupled with the lack of Plan envisaged that no village should be more than a adequate maintenance, caused rapid deterioration mile and a half from a road in developed agricultural of these roads, sometimes rendering them impass- areas, three miles from a road in semi-developed able. At the end of the 20th century, about 330,000 areas, and five miles from a road in underdeveloped of India’s 825,000 villages and habitations (hamlets and uncultivable areas. or subvillages) lacked any all-weather road access (Government of India 2006).1 As a result, about 300 The Lucknow Plan set accessibility targets and pro- million people—28 percent of India’s population— vided direction on how states could prepare their had limited access to education, health services, and own perspective plans for road development, bearing economic and social opportunities. in mind differences in the land-use patterns, popu- lation, terrain, stage of and potential for economic In response to the huge gap in rural connectivity, in development, and social infrastructure needed to October 1999, the president of India announced a achieve a balanced road network. Several states new program to build all-weather roads to connect all followed through by formulating their own plans for villages and habitations. The government constituted rural roads. the National Rural Road Development Committee on January 2000 to make specific recommendations on Before 2000, rural roads were built largely under the way forward. Subsequently, the prime minister rural development programs. The Minimum Need announced the Prime Minister’s Rural Roads Program Programme included rural roads for the first time in (Pradhan Mantri Gram Sadak Yojana [PMGSY]), which the Fifth Five-Year Plan (1974–79). In later years, rural was launched on December 2000. For the first time, roads were also constructed under several schemes the government focused on rural connectivity. funded by the central government, including the 1 A majority of the poorly connected rural communities are in 10 states: Assam, Bihar, Chhattisgarh, Himachal Pradesh, Jharkhand, Madhya Pradesh, Orissa, Rajasthan, Uttar Pradesh, and West Bengal 20 MOBILITY AND TRANSPORT CONNECTIVITY SERIES What Is PMGSY? PMGSY’s objective is to establish farm-to-market defects resulting from poor workmanship. At the end connectivity by providing access to all-weather roads of the five-year period, the road is transferred to local to eligible unconnected habitations.2 The focus of government (Panchayati Raj) institutions or other the program is a habitation, not a revenue village or owner institutions for maintenance. Panchayat. A habitation is a cluster of people living in the area whose location does not change over time, SELECTION AND SEQUENCING OF ROADS according to the program rules (Government of India PMGSY established high management standards and 2015). The PMGSY administrative database shows operating procedure that are applied nationwide. It that most villages have more than one habitation. has a well-structured framework for delivery of rural roads based on detailed guidelines. PMGSY set a target of providing all-weather road access to every habitation with a population of more The District Rural Roads Plan (DRRP) is the starting than 1,000 by 2003 and to every habitation with a point of the exercise, the basic instrument for proj- population of more than 500 by the end of 2007.3 The ect selection and prioritization of construction and population thresholds were set based on the 2001 upgradation and the basis for allocation of funds for census. It was estimated that about 178,000 habita- maintenance in the core network. Selection of road tions would be provided with connectivity under the works follows a bottom-up approach. Through a program. consultative process involving Panchayati Raj institu- tions, district Panchayats and elected representatives A PMGSY intervention includes both construction prepare the DRRP. This plan lists the complete road and maintenance. Eligible unconnected habitations network in the district (village roads, major district are connected to nearby habitations already con- roads, state roads, and national highways) and “the nected by an all-weather road or to another existing proposed roads for providing connectivity to eligible all-weather road, in order to make services (educa- unconnected habitations in an economic and efficient tional, health, and marketing facilities) available to manner in terms of cost and utility” (Government residents.4 Connectivity is provided through con- of India 2015). The district core network, consisting struction of a new all-weather road or upgrading of of the set of roads required to provide connectivity an intermediate link that cannot provide all-weather to all eligible habitations, is the key element of the connectivity. During the five years after construction DRRP. Once the district Panchayat approves the core of the road, the contractor is responsible for routine network, the plan is sent to the state-level agency maintenance of the road and all its components. and the National Rural Infrastructure Development This practice seeks to encourage contractors to Agency (NRIDA), the central implementing agency place more importance on the quality of the initial under the Ministry of Rural Development (MoRD). construction works, thereby reducing failures and 2 An all-weather road is one that is negotiable every season of the year, implying that the road bed is drained effectively (by adequate cross-drainage structures, such as culverts, minor bridges and causeways). The road need not be paved, surfaced, or black-topped. 3 The cut-off is 250 people in the hill states (the North-Eastern states, Sikkim, Himachal Pradesh, Jammu and Kashmir, and Uttarakhand); desert areas; and tribal areas 4 An eligible habitation is considered as connected if it is no more than 500 meters (1.5 kilometers of path distance in case of hills) from an all-weather road or a connected habitation. 21 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY The list of roads works to be taken up under PMGSY fully funded by the central government until 2014–15. each year is prepared by the district and intermediate Since 2015–16, the central and state governments Panchayats, in accordance with the funds made avail- have shared construction costs, with the central able for the district. Once the core network is ready, government funding 60 percent of the costs in plain the states are required to prepare a Comprehensive areas and 90 percent in special category states. New Connectivity Priority List (CNCPL), at the block Maintenance is the responsibility of the states. and district level, of all proposed road links under PMGSY. Unconnected habitations are prioritized The inherent strength of PMGSY is its strong national based on their population in the 2001 census. focus on rural road development through the Habitations with population of 1,000 or more have central implementing agency, the National Rural the highest priority, followed by habitations with pop- Infrastructure Development Agency (NRIDA) under ulations of 500–999 and habitations with populations MoRD. NRIDA is responsible for overseeing and of 250–499 (where eligible). coordinating all technical aspects and facilitating systematic monitoring of program implementation in Once the CNCPL has been firmed up, road works of the states/union territories. State rural roads devel- lower-order priority are not taken up in the district opment agencies execute the program. where road works of higher-order priority remain to be taken up. Lower-order roads can be considered PROGRESS SO FAR only when it is not feasible to execute the higher PMGSY has been identified as one of India’s 60 order of work (because, for example, land is not success stories since independence, according to available). In districts where no new connectivity is a survey conducted by India Today (an Indian news required, a Comprehensive Upgradation Priority List magazine and television channel). Implementation is prepared based on a road condition survey, with has brought a sea change in the rural roads sector, higher priority given to roads in worse condition. thanks to rigorous planning, technical specifications and standards, procurement and contracting require- ROLE OF CENTRAL AND STATE GOVERNMENTS ments, and attention to quality. As provision of all-weather road connectivity was conceived as part of the larger poverty reduction PMGSY has already met about 90 percent of its strategy, MoRD was entrusted with the task of original target. Of the 178,184 habitations intended administering and managing the program. However, to benefit from the program, 159,759 habitations as rural roads are a state responsibility under the (90 percent) have been connected, through 562,047 Constitution of India, PMGSY is executed by state/ kilometers of new and improved rural road network, union territory governments. Road construction was at a cost of about ₹1.88 trillion ($27 billion). 22 MOBILITY AND TRANSPORT CONNECTIVITY SERIES The real story is not how many kilometers were built, however, but where these roads were built. The Online Management, The states that recorded the most road construction by December 2017—Madhya Pradesh (68,796 kilo- Monitoring, and Accounting meters), Rajasthan (63,465), Uttar Pradesh (51,999), System (OMMAS) Bihar (44,705), and Odisha (43,117)—are the ones that were least connected at the turn of this century. Bihar alone had 34,586 habitations originally eligible To monitor the program and increase efficiency, under PMGSY; 27,590 (80 percent) now have road accountability, and transparency in implementation, connectivity, and work on another 6,040 habitations the government developed a web-enabled system was ongoing by December 2017. Connectivity has with a centralized database. The Online Management, been impressive in Madhya Pradesh (17,826 out of Monitoring, and Accounting System (OMMAS) facil- 18,404 eligible habitations), Rajasthan (16,165 out of itates the operational requirements of planning, 16,694), Chhattisgarh (9,368 out of 10,191) and West scheduling, monitoring, tracking, and execution Bengal (12,557 out of 18,641). Implementation capac- in implementing PMGSY. Most data are entered at ity also greatly enhanced over time, with the number the program implementation unit level. Exceptions of kilometers completed rising from 15,500 in the include technical clearance by state technical agen- program’s first year (2001) to 52,400 kilometers in its cies, the sanction of proposals by NRIDA, and moni- last full year (2017). toring by state rural roads development agencies and NRIDA. Execution of the program is decentralized, PMGSY achieved 90 percent of its initial targets 17 but monitoring is centralized. years after its launch, mainly as a result of lack of resources and limited implementation capacity. The Modules in OMMAS exist for every process in PMGSY; uniform structure of the program led to constraints, data are captured at the relevant agency level. The with many states lacking the capacity to implement process flow starts with preparation of the DRRP, the program in a timely manner. The largest number followed by identification of the core network and of unconnected habitations come from five states: preparation of proposals from the core network, Bihar (6,996 habitations), Assam (5,276), Odisha which are cleared by the state technical agencies and (2,497), Jharkhand (2,015), and West Bengal (1,632). sanctioned by MoRD. For the sanctioned proposals, tenders are published, works are awarded to the selected contractor, and agreement is executed. While works are being executed, the quality of work is monitored, and expenditures are recorded. Upon completion of work, maintenance is planned. Table 1.1 describes the system’s main modules. 23 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 1.1 Modules of the Online Management, Monitoring, and Accounting System (OMMAS) Module Description Master data Includes master data on districts, constituents, blocks, villages, habitations, Panchayats, roads, contractors, and so forth. Core network Includes District Rural Road Plans (DRRPs), which identify national highways, state high- (rural road plan) ways, major district roads, rural roads, link routes, and through-routes in each district. Proposals Includes proposals for selection of road links from the core network. Tendering Includes tendering data and contractor award details. Execution Documents progress of works (physical/financial). Online fund processing Processes requests for funds from the State Rural Roads Development Agency (SRRDA) to MoRD. States initiate proposals and forward the request to MoRD, submitting all required and relevant information. After approvals from the project and finance depart- ments of MoRD, a letter specifying the amount sanctioned and released is issued to the state. Quality monitoring Includes data on quality control inspection carried out by national quality monitors. Receipts and payments Includes accounting data on classified expenditures against each road work. Maintenance Includes physical and financial data for five years. Security and administration Used to create users, define roles, and map menus to the roles and assignment of roles to users. Analysis of rate for rural Maintains schedule of rates, based on analysis of different items of work derived from roads the “Specification for Rural Roads” published by the Indian Roads Congress, for different items. Receipts and payments bank Used by personnel of bank at which the SRRDA has account for payment of PMGSY- related bills. When bank clears checks or e-payments related to a voucher, the bank authority logs ins and reconciles the payment, which is reflected in the District Program Implementation Unit and SRRDA reports. Data gap Allows viewing of data gaps in entry of proposals. Updating of user manual Annex provides latest enhancements to OMMAS. Source: http://omms.nic.in/. References Government of India. 2006. Working Group Report of the Eleventh Five Year Plan. Planning Commission. Delhi. ———. 2015. Pradhan Mantri Gram Sadak Yojana. Program Guidelines. Delhi. 24 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Chapter 2. Methodology and Data 25 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY This chapter outlines the analytical framework for An ideal impact evaluation would randomly choose evaluating the impacts of PMGSY and describes the the habitations that would be connected through data used in the analysis. Although it is aimed for new roads from a list of eligible habitations. This both technical and nontechnical audience, some randomization would ensure that the treated and readers may wish to skip it. untreated sites were, on average, identical in terms of both observable and unobservable characteristics. In this context, the comparison of outcomes between Empirical Approach groups yields an estimate of the causal effect of the program. The aim of an impact evaluation is to go beyond Randomizing the construction of roads is not feasi- simple correlations between outcomes and program ble, because of the huge investment cost of these participation, to detect causal relationships between roads. Instead, roads that were built first can be the relevant variables. The causal effect of a program compared with roads that were built later to estimate or treatment is defined as the difference between causal impact. the observed outcome of a treated unit and the result that this same unit would have obtained in To statistically compare treated habitations with the absence of the treatment (the counterfactual untreated habitations, the analysis uses a differ- outcome). In the case of PMGSY, the treatment is the ence-in-differences approach (Gujarati 2003). This construction of an all-weather road, and the treated methodology consists of measuring the average unit is a habitation connected through that road. As change in an outcome before and after the inter- the counterfactual outcome is unobservable by defi- vention and then comparing the changes between nition, to assess the average impact of a program it the control and treatment groups. The before-after is necessary to compare the treated units to a pool of difference corrects for any time-invariant difference nonparticipants (a comparison or control group). between treatment and control; the difference between groups deals with external factors that The main challenge of an impact evaluation is to affect the target population during the interval of construct a control group. It should be as similar to analysis. Assuming that those factors affect the the test group as possible, to ensure that differences treatment and control groups equally (parallel trends in outcomes can be attributed exclusively to the pro- assumption), the double difference successfully gram. The key to the construction of a valid control isolates the true causal effect of the intervention. group are the identification assumptions, which state Even in an experimental context, where there should the conditions under which the untreated group can be no baseline differences between groups, differ- be considered comparable to the treated group. ence-in-differences may help account for some “con- tamination” of the data, especially when sample sizes are small. This approach requires the existence of baseline and postintervention (endline) information for treatment and control groups. 26 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Empirically, the difference-in-differences model can be specified as Yst=β0+β1 Treats+β2 Postt+β3 (Treats × Postt )+β4 Xst+εst where • Xst are the control variables in group s and period t. • Yst is the observed outcome in group s and period t. • β3 is the difference-in-differences estimate of the treatment effect. • Treats is a dummy variable that takes a value of 1 if the observation is from the treatment group β3 captures whether there is a significant difference in and a value of 0 if the observation is from the the outcome of interest between treatment and con- control group. trol groups as a result of the building of PMGSY roads. The difference-in-differences estimate is given by • Postt is a dummy variable that takes a value of 1 if the observation is from the posttreatment period and a value of 0 if the observation is from the pretreatment period. β3 = (Y̅treatment endline – Y̅treatment baseline ) – (Y̅control endline – Y̅control baseline ). Baseline data were collected in 2009; endline data a period of time. β3 captures the difference between were collected in 2017. The identification strategy the sum of the marginal effects in the first few years exploits the phasing of the PMGSY investment minus the marginal effects in the later years. The (the fact that some habitations were connected estimator thus underestimates the effect of PMGSY, before others). Habitations connected before which is the sum of all of the annual marginal effects. 2009 by a PMGSY road serve as the control group; most of these habitations were connected in 2007. This difference-in-differences framework is the main Habitations that were connected between 2009 and analytical tool used in this report; some variations 2017 (meaning they were unconnected in 2009) serve are applied in some sections. The study is based on as the treatment group; most of these habitations household survey data for a sample of habitations were connected in 2014. Rural roads are expected in three states: Himachal Pradesh, Madhya Pradesh, to have annual marginal effects on habitations over and Rajasthan. 27 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY The difference-in-differences analysis includes three models with state fixed-effects. Thus, across all three Data specifications, the analysis accounts for differences across states: The impact evaluation combines two main sources of data to construct a panel of habitations with two data • T he first model examines outcomes without points: baseline (2009) and endline (2017). MoRD including control variables (beyond state fixed collected data between 2009 and 2011 to understand effects). the magnitude and distribution of impacts of PMGSY. The 2009 round is used as the baseline. In 2017 the • T he second model controls for several village-level study team revisited the habitations in Himachal variables, including variables from the 2011 Pradesh, Madhya Pradesh, and Rajasthan that were census: population; whether the habitation had a surveyed in 2009. public primary school, a public middle school, or a bank; and whether power supply for domestic The baseline data lacked habitation names but use was available.1 It also includes distance to the included the end nodes of the road connecting each nearest statutory or census town and a measure habitation and the names of the state, district, block, of the ruggedness of the terrain. and village. The team merged these data with the OMMAS data to find the name of each habitation. • T he third model controls for several house- A team was sent to the identified habitations with hold-level variables, including the social group; the information on all households surveyed in 2009. It number of household members; and the gender, confirmed 80 percent of the habitations. age, marital status, and number of schooling years of the head of household. Only the habitations confirmed were surveyed in 2017 to create the panel of habitations. It consisted of 127 Heterogeneous impacts are calculated by extending habitations in the control group and 26 in the treat- the difference-in-differences model by including ment group (table 2.1). Treatment and control habi- interactions with some key characteristics, including tations are from different villages but from the same distance from an urban agglomeration, the rugged- nine districts and three states.2 On average the sam- ness of the terrain, and gender (one at a time). This ple contains 18 households per habitation at baseline part of the analysis was conducted to determine and 15 at endline.3 Because of the eight-year span whether distance from an urban agglomeration or between the surveys, about a third of the households the ruggedness of the terrain affected the impact of surveyed in 2009 could not be found in 2017. Random PMGSY roads and whether there were gender differ- replacements were selected in the corresponding ences in the program’s impact. habitations. As the dataset is a panel of habitations and a repeated cross-section of households, attrition at the household level is not a concern. 1 The 2011 Census of India is used because most of its variables relevant to the analysis were recorded in 2009, the same year as the baseline. 2 The nine districts are Kangra and Shimla in Himachal Pradesh; Dhar, Sagar, Seoni and Sidhi in Madhya Pradesh; and Barmer, Bhilwara, and Nagaur in Rajasthan. 3 The difference in the number of households between baseline and endline is a survey design feature because of budget constraints. 28 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 2.1 Number of habitations and households included in the panel 2009 2017 Item Control group Treatment group Control group Treatment group Habitations Total 127 26 127 26 Himachal Pradesh 21 3 21 3 Madhya Pradesh 43 14 43 14 Rajasthan 63 9 63 9 Households Total 2,301 425 1,914 391 Himachal Pradesh 430 51 315 45 Madhya Pradesh 697 221 644 211 Rajasthan 1,174 153 955 135 A household survey and a habitation survey were informants considered to be the habitation leader, conducted at baseline and endline. For the endline representative, or well-informed elder. It covers ame- survey, in 2017, the questionnaires were streamlined nities and general characteristics of the habitations. to remove questions for which the response rates were too low in the baseline; a few questions were Spatial data were also used as control variables in the also added, in order to collect more information analysis. Information about the presence of a public for quality-check purposes and for a potential third primary school, public middle school, and a commer- round of data collection in a few years. A section on cial bank; domestic power supply; and the population gender was also added, including questions about were obtained from the 2011 census (Government of agency, economic empowerment, and physical mobil- India 2011). The ruggedness of the terrain of the habi- ity for the main woman in the household. The rest of tation was extracted from the dataset of terrain rug- the household questionnaire collected information gedness created by Nunn and Puga (2012).4 Distance about household characteristics, profiles of house- to the nearest statutory or census town is the linear hold members, employment, travel, medical treat- distance between a habitation and the nearest statu- ment, immunization and health, cropping patterns, ary or census town in the 2011 census (the centroid of household amenities, and household assets. The the polygon marked as a statuary town). habitation survey collected data by interviewing key 4 Ruggedness is measured in hundreds of meters of elevation difference for grid points 30 arc-seconds (926 meters on a meridian) apart. 29 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Tables 2.2 and 2.3 present the mean and sample size 40 percent have a public middle school. The distance for all control variables used and the t-test for the to the nearest statutory or census town is 16 kilo- difference in the sample means. The population in meters in the control group and 13 kilometers in the the average village was 941 in the control group and treatment group. The average terrain ruggedness 916 in the treatment group. The average village has is higher in treatment habitations than in control power supply for domestic use. Only about 5 percent habitations. The treatment and control groups are of villages have a bank. More than 85 percent of not statistically different in terms of the village/habi- villages have a public primary school, but less than tation-level controls. Table 2.2 Summary statistics for village-level control variables from the 2011 Census of India Control group Treatment group Control variable Mean Na Mean Na T-test Village population 941 126 919 25 0.179 Distance to nearest statutory or census town 16 126 13 25 1.054 (kilometers)b Terrain ruggednessb 0.64 126 1.08 25 –1.218 Availability of power supply for domestic use 0.92 126 1.00 25 –1.458 Village has public primary school 0.87 126 0.80 25 –0.839 Village has public middle school 0.39 126 0.28 25 –1.026 Village has bank (commercial or cooperative) 0.06 126 0.04 25 0.315 Note: a. Data on some of village/habitation-level controls are missing for one habitation in each group. b. Variable is at the habitation level. The average households in the control and treat- percent of households in the control group and 86 ment groups were similar. At baseline the average percent of households in the treatment group were household in the control group had 5.4 members, from a scheduled caste, scheduled tribe, or other and the average household in the treatment group backward class. The small difference in household had 5.1 members. Up to 7 percent of households size and social group were the only statistically signif- were headed by a woman at baseline, which slightly icant differences. Even in randomly selected samples, increased by endline. In both groups, the average there can be some differences across treatment and household head was 47 years old at baseline, with control groups, pointing to the importance of adding almost four years of schooling; about 90 percent of control variables to the regression model. household heads were married. At baseline, about 75 30 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 2.3 Summary statistics for household-level control variables 2009 2017 Control group Treatment group Control group Treatment group Control variable Mean N Mean N T-test Mean N Mean N T-test Social group a 0.25 2,301 0.14 425 4.923 *** 0.20 1,914 0.14 391 3.103** Household size 5.4 2,300 5.1 424 2.403** 5.0 1,914 4.8 391 1.582 Female-headed household 0.06 2,286 0.07 419 –1.137 0.09 1,914 0.11 391 –0.850 Age of head of household 47 2,283 47 417 –0.008 51 1,914 50 391 1.209 Head of household is married 0.91 2,286 0.90 419 0.205 0.88 1,914 0.87 391 0.565 Years of schooling of head of 3.7 2,277 3.6 419 0.737 4.1 1,914 4.3 391 –0.745 household Note: a. Social group is the share of households that are not scheduled castes, scheduled tribes, or other backward classes, as defined by the government of India. *** p < 0.01 , ** p < 0.05. The main challenge of an impact evaluation is to using data from the 2001 Census of India and the define a control group that differs from the treatment test for differences in means. The test for differences group in only one respect: not receiving the program. in means indicates that the treatment and control Table 2.4 presents the means for the village-level con- groups were not statistically different in 2001, before trol variables for the control and treatment groups any of the habitations was connected. Table 2.4 Summary statistics for village-level control variables from the 2001 Census of India Control group Treatment group Control variable Mean N Mean N T-test Village population 817.83 127 792.64 25 0.170 Power supply for domestic use is available 0.79 127 0.92 25 –1.530 Village has a primary school 0.94 127 0.96 25 –0.435 Village has a middle school 0.25 127 0.20 25 0.531 Village has a bank (commercial or cooperative) 0.03 127 0.04 25 –0.224 31 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY The assumption of parallel trends in outcomes in a difference-in-differences model at the habitation the absence of treatment was tested using two data level was estimated. It confirmed that there is no sources: nightlights and data from the 1991 and significant difference between treatment and con- 2001 Censuses of India. Annual nightlights data trol groups over time. A difference-in-differences were used to compare economic activity (for which regression model on 27 variables from the 1991 and these data are proxies) by control and treatment 2001 Censuses of India was used to assess whether habitations between 1992 and 1998, one year before the treatment and control habituations followed the launch of PMGSY (the one-year cutoff was used significantly different trends between 1991 and 2001. to avoid picking up any impact of the expectation For all 27 variables, the null hypothesis of equal of road connectivity). The treatment and control trends could not be rejected (table 2.5). These results habitations followed similar trends (figure 2.1). To indicate that there were no discernable differences statistically test for differences in nightlights trends, across treatment and control groups over time.5 Figure 2.1 Trends in nightlights 5 4 3 2 1 0 1992 1993 1994 1995 1996 1997 1998 Control Treatment 5 Because of the small sample size (26 treatment habitations), the tests are underpowered. Both sets of results still give confidence about the comparability of treatment and control groups. 32 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 2.5 Test of parallel trends in treatment and control groups, 1991–2001 Variable P-value Variable P-value Total population 0.80 Number of post offices 0.56 Total scheduled caste population 0.94 Number of telephone connections 0.45 Total literate population 0.86 Bus services (yes/no) 0.38 Total literate population (male) 0.95 Paved approach road 0.71 Total main workers 0.96 Mud approach road 0.63 Main worker (male) 0.89 Power supply (yes/no) 0.54 Educational facilities (yes/no) 0.37 Total irrigated area 0.84 Number of primary schools 0.46 Unirrigated area 0.77 Number of middle schools 0.59 Scheduled caste population as share of 0.75 total population Number of secondary schools 0.64 Literate population as a share of total 0.20 population Medical facilities (yes/no) 0.93 Literate male population as share of 0.17 total literate population Number of hospitals 0.95 Main workers as share of total 0.43 population Number of health centers 0.68 Main male worker as share of total 0.57 main workers Tap water (yes/no) 0.71 33 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Robustness Checks A set of robustness checks was performed, including of the roads and implementation of the survey was testing the models with no fixed effects and district only a few months for these habitations, inclusion of fixed effects. For binary outcomes, logistic regres- these habitations in the control group could atten- sions were estimated. The results are discussed in the uate the impacts of PMGSY. Therefore, as a robust- following chapters. ness check, all regressions were estimated after dropping all habitations connected in 2009 from the A few habitations were connected with a PMGSY sample. The results were similar to those presented road in 2009 (figure 2.2), the year the baseline survey in chapters 3–5. was implemented. As the time between completion Figure 2.2 Number of control and treatment habitations connected, 2003–17 120 Number of habitations connected 100 80 60 40 20 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Control Treatment 34 MOBILITY AND TRANSPORT CONNECTIVITY SERIES References Government of India. 2011. “Census 2011.” http://www.censusindia.gov.in/2011census/dchb/DCHB.html. Gujarati, D.N. 2003. Basic Econometrics. New York: McGraw Hill. Lechner, M. 2010. “The Estimation of Causal Effects by Difference-in-Difference Methods.” Foundations and Trends in Econometrics 4 (3): 165–22. Nunn, N., and D. Puga. 2012. “Ruggedness: The Blessing of Bad Geography in Africa” Review of Economics and Statistics 94 (1): 20–36. 35 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Chapter 3. Impacts of PMGSY on Travel Patterns 36 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Spatial isolation is an important contributing factor Kanuganti and others (2015) find evidence of better to sustained poverty in rural areas (Chambers 1984). accessibility to medical facilities in a group of villages Better transportation infrastructure provides the in Rajasthan, India connected by an all-weather road. means to link spatially isolated rural population Aparna and others (2017) examine three geograph- to market areas and to vital social and economic ical areas in the West Midnapore district of West services. The role of transport in stimulating rural Bengal. They use simple t-tests to examine the num- development is therefore critical. ber of trips before and after these areas were served by PMGSY roads. They find that an increase in the This chapter examines the effect of PMGSY on travel number of trips for work, education, and other pur- patterns. It begins by reviewing the literature on the poses is correlated with the construction of PMGSY effect of rural roads on travel patterns. roads in West Bengal. Review of the Literature Results of This Study Investment in rural roads in developing countries IMPACT ON MOBILITY PATTERNS AND COSTS improved accessibility of rural populations, by cutting This section examines the travel patterns of rural travel times, reducing transport costs, and increasing populations in habitations connected by all-weather trip frequency. Tian, Li, and Chen (2009) investigate roads through the PMGSY program. It uses a trip as the impact of rural road investments in a group of the basic unit of measurement. Trips are character- villages in Fujian Province, China, using a propensity ized by various social and economic purposes (desti- score–matched difference-in-differences method. nations) inside and outside the habitation, as well as, They find evidence of travel time savings and an by various modes of transport. increase in trip frequency in villages in which roads had been rehabilitated. A comparative (ex ante ver- Improvement in road connectivity can affect various sus ex post) analysis by Asomani-Boateng, Fricano, aspects of a trip. It can result in people traveling and Adarkwa (2015) finds that provision of rural farther, faster, and/or cheaper. roads in rural Ghana was associated with reductions in trip duration to schools and health facilities. Did people in PMGSY-connected habitations expe- rience reductions in average trip duration and cost, Limited empirical evidence exists, however, on increases in travel speed and distance, and increases the causal impact of rural road provision on travel in trip frequency? In the sample at baseline, an patterns in India. Using a with-and-without study average trip took about 37 minutes, covered 8.2 design, Bell and van Dillen (2014) find evidence of kilometers, and cost about 19.3 rupees (see table better accessibility to primary and secondary schools 3A.1 in the annex for summary statistics). Trips to and health services and reduction in the cost of school and work tended to be shorter, took less time, transporting crops in villages connected by PMGSY in and cost less. Trips to hospitals/medical centers and Orissa. Using multiple-criteria decision-making tools, local markets took longer and cost more. On average 37 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY at baseline a household made about 8.4 trips a week. distance: People did not travel farther, but they got to Households that travelled for work or school made their destinations faster. The estimated coefficients in 6.1 and 6.3 trips a week on average, respectively. In table 3.1 show that the average travel time for people contrast households who travel to local market or connected after 2009 (treated habitations) signifi- hospital/medical center make fewer trips on average cantly decreased. This decrease appears to have per week, 1.6 and 0.7, respectively. Regarding the been driven by trips to work and the local market. modes of transport at baseline, on average house- The estimated coefficient (–13.63) of model 3 in table holds make half of their trips walking, followed by 3.1 suggests an overall travel time saving of nearly public transport (37 percent of trips), bicycle (8 per- 14 minutes for trips to the local market. For trips to cent), and motorized vehicles (4 percent). work, the time saving was about nine minutes. For distance travelled by destination, it cannot be ruled The results of the difference-in-differences analysis out that the average effect is statistically zero. suggest that on average, rural roads improved acces- sibility by reducing travel time without reducing travel Table 3.1 Impact of PMGSY roads on travel time and distance, by destination Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Travel time by destination (minutes) All destinations –10.999*** –10.259*** –10.819*** (3.94) (3.81) (3.87) Work –8.290** –8.626** –8.530** (3.98) (4.11) (3.96) School 2.108 3.198 3.031 (5.41) (4.37) (5.05) Local market –13.981** –14.388** –13.627** (5.70) (5.98) (5.61) Hospital/medical center –14.358** –10.416* –14.155** (7.01) (5.67) (7.00) Travel distance by destination (kilometers) All destinations 0.837 0.865 0.930 (1.20) (1.25) (1.20) Work –0.664 –0.597 –0.701 (2.02) (2.08) (2.02) School 0.974 1.380 1.481 (1.96) (1.85) (1.86) Local market 2.611 2.438 2.608 (1.99) (2.08) (1.98) 38 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Hospital/medical center 1.527 1.918 1.554 (2.48) (2.52) (2.48) Travel time per kilometer by destination (minutes) All destinations –2.462 –2.413 –2.557* (1.50) (1.55) (1.51) Work –5.753** –6.218** –5.809** (2.51) (2.55) (2.54) School 1.011 0.861 –0.049 (3.36) (3.43) (3.31) Local market –2.041* –1.700 –2.000* (1.09) (1.04) (1.08) Hospital/medical center –1.750 –1.787 –1.815 (1.98) (2.09) (2.01) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1. People in connected habitations enjoyed savings in it cannot be ruled out that the average effect is travel time as a result of faster travel speed (travel statistically zero (table 3.1). When interacted with time per kilometer). Travel time per kilometer vari- the ruggedness of the terrain in and around the ables have mostly negative values, suggesting an habitation, the effect becomes significant (table 3.2). increase in travel speed, with the impact on travel For habitations in hillier areas, the distance declines; time per kilometer to the workplace, local market for habitations in flatter areas, it increases. The and all destinations being statistically significant.1 On coefficient for the terrain interaction effect on travel average, the increase in travel speed to the workplace distance to hospitals and medical centers of –2.387 yielded a six-minute reduction in travel time for every (model 3 in table 3.2) implies that in habitations with kilometer traveled (the estimated coefficients range terrain that is two standard deviations more rugged between –5.753 and –6.218 across the three models). than the mean, travel distance decreased by 8 kilo- meters more than in habitations with average terrain The impact of the all-weather roads built under ruggedness.2 The impact on travel time to all destina- PMGSY, depends on the terrain around the habita- tions, the local market, and the hospital is similar with tion. For distance to hospitals and medical centers, respect to terrain. 1 The coefficients for all destinations in models 1 and 2 and local market in model 2 are significant at just below the 90 percent. 2 The mean terrain ruggedness for the sample is 0.71, the standard deviation is 1.67, and the minimum is 0.02. 39 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 3.2 Differential impacts of PMGSY roads on travel time and distance based on terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Travel time by destination (minutes) All destinations –4.478** –4.568** –4.498** (1.89) (1.89) (1.89) Work –1.246 –1.175 –1.309 (1.09) (1.10) (1.09) School –2.023 –1.781 –2.176 (1.57) (1.45) (1.57) Local market –7.493*** –7.751*** –7.512*** (2.44) (2.43) (2.45) Hospital/medical center –8.994*** –8.991*** –9.058*** (2.66) (2.61) (2.63) Travel distance by destination (kilometers) All destinations –0.626 –0.620 –0.647 (0.56) (0.56) (0.58) Work –0.543 –0.543 –0.575 (0.54) (0.54) (0.55) School –0.779 –0.647 –0.780 (0.92) (0.93) (0.88) Local market –0.643 –0.705 –0.693 (0.89) (0.88) (0.89) Hospital/medical center –2.339** –2.231** –2.387** (1.01) (1.01) (1.02) Travel time per kilometers by destination (minutes) All destinations –0.349 –0.348 –0.345 (0.34) (0.34) (0.35) Work –0.114 –0.048 –0.137 (0.65) (0.66) (0.64) School –1.224 –1.428* –1.071 (0.74) (0.80) (0.73) Local market –0.166 –0.181 –0.156 (0.32) (0.31) (0.32) Hospital/medical center –0.060 –0.051 –0.034 (0.32) (0.35) (0.31) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1. 40 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Rural roads have no statistically significant impact The same results arise when looking at travel cost per on transport costs. The estimated coefficients for kilometer. Total travel costs could increase if people travel cost to a local market range between 5.5 and switch to more expensive transport modes. The share 7.7 across the three models (table 3.3), suggesting a of motorized vehicles rose, but so did the share of five- to eight-rupee increase in travel cost. However, bicycles as mode of transport, which may explain the it cannot be ruled out that these estimates are zero. lack of impact on transport costs (table 3.4).3 Table 3.3 Impact of PMGSY roads on travel cost, by destination Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Travel cost by destination (rupees) All destinations 4.702 3.321 4.984 (3.40) (3.52) (3.33) Work 2.674 2.296 2.978 (3.27) (3.75) (3.28) School 4.556 2.960 5.181 (3.55) (3.62) (3.37) Local market 7.710 5.479 7.724 (4.97) (4.71) (4.94) Hospital/medical center 4.624 2.888 4.406 (5.86) (6.07) (5.79) Travel cost per kilometers by destination (rupees) All destinations –0.121 –0.022 –0.099 –0.57 (0.59) (0.58) Work 0.630 1.056 0.737 (0.65) (0.93) (0.72) School 0.207 0.288 0.304 (0.63) (0.72) (0.66) Local market –0.219 –0.105 –0.229 (0.47) (0.46) (0.47) Hospital/medical center –0.481 –0.370 –0.518 (1.05) (1.01) (1.07) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. 3 The share of trips by bicycle loses significance when habitations connected in 2009 are dropped. 41 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 3.4 Impact of PMGSY roads on trip frequency, by destination and mode Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Number of trips per week All destinations –0.116 0.057 -0.136 (1.02) (1.03) (1.02) Work –0.557** –0.493* –0.557** (0.26) (0.26) (0.26) School –0.065 –0.235 –0.106 (0.34) (0.32) (0.32) Local market –0.120 –0.121 –0.133 (0.28) (0.29) (0.28) Hospital/medical center –0.036 0.061 –0.044 (0.15) (0.12) (0.16) Share of trips by mode Walking –0.023 –0.015 –0.028 (0.07) (0.07) (0.07) Bicycle 0.033* 0.030* 0.031* (0.02) (0.02) (0.02) Motorized vehicles 0.073** 0.068** 0.074** (0.03) (0.03) (0.03) Public transport –0.083 –0.086 –0.079 (0.08) (0.08) (0.08) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Public transport includes buses, minibuses, auto-rickshaws, and jeeps/vans. ** p < 0.05, * p < 0.1. Rural roads have a stronger impact on the frequency two standard deviations above the mean, households of trips to work and local markets in the hardest to made 0.77 more trips to work a week than in the hab- reach habitations. The average effect of rural roads itation with average terrain ruggedness; in the flat- on the number of weekly trips for all destinations test habitation in the sample, the number of weekly except work is not statistically significant (see table trips to work decreased by 0.66. Patterns on trips to 3.4). For trips to work, the average household in hab- local markets were similar, suggesting that in hilly itations connected after 2009 traveled about half a areas, construction of all-weather roads increased trip a week less than households in the control group. access to economic opportunities. However, relative This decrease was less pronounced in habitations in to flatter areas, PMGSY roads in hillier areas did not hillier areas. The coefficient for the terrain interaction have a statistically significant impact on the number effect on trips to work of 0.231 (model 3 in table 3.5) of trips to school or hospital. implies that in habitations with terrain ruggedness 42 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 3.5 Differential impact of PMGSY roads on trip frequency based on terrain ruggedness Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Number of trips per week All destinations 0.128 0.120 0.125 (0.09) (0.09) (0.09) Work 0.234*** 0.220*** 0.231*** (0.07) (0.07) (0.07) School –0.126 –0.115 –0.128 (0.09) (0.09) (0.09) Local market 0.318*** 0.327*** 0.321*** (0.10) (0.10) (0.10) Hospital/medical center 0.064 0.040 0.065 (0.07) (0.07) (0.07) Share of trips by mode Walking –0.016 –0.017 –0.016 (0.02) (0.02) (0.02) Bicycle –0.010*** –0.009*** –0.009*** (0.00) (0.00) (0.00) Motorized vehicles –0.007 –0.006 –0.007 (0.01) (0.01) (0.01) Public transport 0.033* 0.034* 0.033* (0.02) (0.02) (0.02) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Public transport includes buses, minibuses, auto-rickshaws, and jeeps/vans. *** p < 0.01, * p < 0.1. IMPACTS ON CROP TRANSPORT PMGSY improved farm-to-market connectivity in the PATTERNS AND COSTS sample of habitations surveyed. In 2009 farmers in Improvements in road connectivity could affect sev- habitations that were not connected with a PMGSY eral aspects of crops and trips. It could, for example, road travelled to markets to sell only 4 percent of result in farmers traveling to markets to sell their their food grain crops, the most common crop in crops instead of selling the crops to middlemen that the surveyed habitations. PMGSY roads triggered come to the farm (or not selling their crops at all). an 8 percentage point increase in the share of crops It could also result in farmers travelling to markets transported to markets for sale (table 3.6), tripling the farther away and/or incurring lower transport costs. average share of crops sold in market. The impact of The objective of this section is to determine whether all-weather roads on the share of crops transported farmers in PMGSY-connected habitations enjoyed to markets was more pronounced in hillier areas than these benefits. in flatter ones, as indicated by the positive and statis- tically significant coefficient in table 3.7. 43 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 3.6 Impact of PMGSY roads on crop transport patterns and costs Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Food grains transported to market (share) 0.078 *** 0.076 ** 0.080*** (0.03) (0.03) (0.03) Distance to market for food grains (kilometers) 9.769** 7.239* 8.909** (4.20) (3.94) (3.99) Transport cost to carry food grains to market –132.742 –151.520 –171.048 (rupees) (349.54) (347.74) (334.52) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01, ** p < 0.05. Table 3.7 Differential impact of PMGSY roads on crop transport patterns and costs based on terrain ruggedness Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Food grains transported to market (share) 0.049*** 0.049*** 0.049*** (0.01) (0.01) (0.01) Distance to market for food grains (kilometers) –0.346 –0.422 –0.400 (0.95) (0.93) (0.93) Transport cost to carry food grains to market 19.959 22.399 13.396 (rupees) (59.41) (59.33) (60.03) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01. The improvement in rural road connectivity seem to ROBUSTNESS CHECKS have led farmers to take advantage of more favor- The results on mobility, crop transport patterns, and able market conditions. Farmers selling food grains costs are robust to different specifications. When travelled 8.9 kilometers farther to sell their crops using district fixed effects and no fixed effects for in response to the improved connectivity provided each outcome variable, the results from the three by PMGSY roads, which represents an 88 percent models remained largely the same as the results increase in the distance travelled. The cost to carry presented in this chapter. The only exception is that crops to markets does not seem to have increased as the average effect on distance to transport food a result of improved connectivity. The results suggest grains becomes statistically insignificant with district that farmers are travelling to locations where they fixed effects. can get better prices for their crops. 44 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Concluding Remarks The analysis presented in this chapter is the first hilly areas and decreased in flat areas, possibly indi- comprehensive assessment of the impacts of cating that people travelled to towns farther away PMGSY on households’ travel patterns and costs and and stayed overnight instead of commuting daily. farmers’ transportation patterns and costs using Shedding light on every link of the result chain would a difference-in-differences approach. Bell and van allow policy makers to design interventions that Dillen (2014) look at travel patterns to school and maximize net benefits. health facilities and costs for transporting crops in habitations connected by PMGSY. Because of data Households’ transport costs did not seem to change limitations, they used a with-and-without design or after PMGSY roads were built. This finding might single-difference approach, which raises concerns reflect the fact that people switched from walking about the attribution of the impacts to PMGSY. Asher and public transport to bicycles and private motor- and Novosad (2020) used a regression discontinuity ized vehicles. Lack of competition in the provision approach and census data. They found an increase of transport services could also explain the lack of in the availability of transport services as a result of change in transport costs. If the availability of trans- PMGSY roads. Because of data limitations, they did port services increased in the studied habitations, as not look at impacts on the demand side. Asher and Novosad (2020) found for the habitations in their sample, the poor quality of transport services The analysis presented in this chapter is one of the could explain people switching away from public main contributions of this report to understanding transport. The collection and analysis of data on how people respond to improvements in connec- market structure, competition, quality of service, and tivity. Thanks to PMGSY roads, on average people people’s choices of public transport could help policy reached their destinations, particularly work, faster. makers make decisions. The number of weekly trips to work increased in 45 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY References Aparna, B.S., P. Prasad, M.F. Cheranchery, and B. Maitra. 2017. “Effect of Development of Rural Roads and Feeder Service on Trip Characteristics of Households: A Case Study.” Transportation in Developing Economies: A Journal of the Transportation Research Group of India 3 (2): 1–10. Asher, S., and P. Novosad. 2020. “Rural Roads and Local Economic Development.” American Economic Review 110(3): 797–823. Asomani-Boateng, R., R.J. Fricano, and F. Adarkwa. 2015. “Assessing the Socio-Economic Impacts of Rural Road Improvements in Ghana: A Case Study of Transport Sector Program Support (II).” Case Studies on Transport Policy 3 (4): 355–66. Bell, C., and S. van Dillen. 2014. “How Does India’s Rural Roads Program Affect the Grassroots? Findings from a Survey in Orissa.” Land Economics 90 (2): 372–94. Chambers, R.. 1984. Rural Development: Putting the Last First. London: Longman. Escobal, J., and C. Ponce. 2002. “The Benefits of Rural Roads: Enhancing Income Opportunities for the Rural Poor.” Grupo de Análisis para el Desarrollo Working Paper 40–1, Lima. Kanuganti, S., A.K. Sarkar, A.P. Singh, and S.S. Arkatkar. 2015. “Quantification of Accessibility to Health Facilities in Rural Areas.” Case Studies on Transport Policy 3 (3): 311–20. Tian, Y., Y. Li, and X. Chen. 2009. “Evaluating the Impact of Rural Road Investment Projects on Household Welfare in Rural Area: Evidence from Fujian Province.” In ICCTP 2009: Critical Issues In Transportation Systems Planning, Development, and Management, 3602–08. Reston, VA; American Society of Civil Engineers. 46 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Annex 3A Summary Statistics Table 3A.1 Summary statistics on outcome variables 2009 2017 Control habitations Treatment habitations Control habitations Treatment habitations Outcome variable Mean N Mean N Mean N Mean N Travel time by destination (minutes) All destinations 35.2 5,583 47.4 983 36.5 3,946 37.2 749 Work 23.9 1,643 29.0 301 25.3 1,215 21.9 249 School 23.2 786 24.5 118 27.9 759 31.3 131 Local market 43.8 1,795 60.7 326 48.4 1,113 50.5 204 Hospital/medical center 44.7 1,359 64.0 238 44.5 859 48.4 165 Travel time per kilometers by destination (minutes) All destinations 11.1 5,583 13.9 983 4.7 3,946 4.8 749 Work 16.0 1,643 22.6 301 6.1 1,215 6.4 249 School 19.6 786 19.5 118 6.8 759 7.0 131 Local market 6.9 1,795 8.4 326 3.2 1,113 2.7 204 Hospital/medical center 6.1 1,359 7.6 238 2.8 859 2.9 165 Travel distance by destination (kilometers) All destinations 8.3 5,583 7.9 983 14.8 3,946 15.2 749 Work 2.9 1,643 2.2 301 8.2 1,215 6.9 249 School 3.8 786 3.7 118 8.9 759 9.8 131 Local market 11.5 1,795 10.6 326 20.7 1,113 22.2 204 Hospital/medical center 13.1 1,359 13.7 238 21.8 859 23.3 165 Travel cost by destination (rupees) All destinations 20.1 3,502 15.4 727 25.1 3,941 25.7 748 Work 8.8 666 1.8 166 12.3 1,215 8.2 249 School 8.1 421 5.2 88 11.3 758 12.5 131 Local market 23.4 1,351 19.8 261 33.8 1,112 38.9 204 Hospital/medical center 27.8 1,064 24.7 212 44.3 856 46.6 164 Travel cost per kilometers by destination (rupees) All destinations 1.9 3,502 1.9 727 1.6 3,941 1.5 748 Work 1.5 666 0.4 166 1.1 1,215 0.7 249 School 1.4 421 1.0 88 1.0 758 0.8 131 Local market 2.0 1,351 2.2 261 2.0 1,112 2.1 204 Hospital/medical center 2.3 1,064 2.9 212 2.4 856 2.8 164 47 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY 2009 2017 Control habitations Treatment habitations Control habitations Treatment habitations Outcome variable Mean N Mean N Mean N Mean N Number of trips per week To all destinations 8.3 2,255 8.4 412 8.4 1,689 8.4 324 To work 6.1 1,638 6.6 299 6.4 1,215 6.4 249 To school 6.3 785 6.5 117 6.7 759 6.8 131 To local market 1.7 1,790 1.7 325 1.1 1,113 1.0 204 To hospital/medical center 0.7 1,356 0.7 237 0.2 859 0.1 165 Share of trips by mode Walking 0.50 5,583 0.52 983 0.41 3,946 0.41 749 Bicycle 0.08 5,583 0.05 983 0.03 3,946 0.03 749 Motorized vehicle 0.05 5,583 0.03 983 0.15 3,946 0.21 749 Public transport 0.37 5,583 0.39 983 0.41 3,946 0.36 749 Note: An observation is a trip made by household members for specific purposes. N represents the total number of trips made by all households in the data set. Public transport includes bus, minibus, auto-rickshaw, and jeep/van. Table 3A.2 Summary statistics of agriculture-related outcome variables 2009 2017 Control habitations Treatment habitations Control habitations Treatment habitations Outcome variable Mean N Mean N Mean N Mean N Crops transported to market (share) Food grains 0.08 2,974 0.04 626 0.95 1,519 0.99 413 Distance to market (kilometers) Food grains 12.1 242 10.1 24 16.8 1,448 22.4 408 Cash crops 11.9 18 — 0 18.8 266 14.8 32 Horticulture 12.0 41 15.0 2 39.7 69 3.0 1 Transport cost to market (rupees) Food grains 594 223 726 22 344 1,261 322 348 Cash crops 173 14 — 0 425 224 435 25 Horticulture 1,630 40 1,457 2 4,739 65 175 2 Note: — Not available. 48 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Chapter 4. Impact of PMGSY on Economic Opportunities and Well-Being 49 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY This chapter presents the results of the differ- rural connectivity also leads farmers to specialize in ence-in-differences analysis on employment and fewer types of crops (Qin and Zhang 2016). Transport agriculture outcomes. It begins by reviewing the improvements reduce production costs by lowering literature on the effect of connectivity on economic the delivered price of inputs, including capital and opportunities and well-being of the rural poor. information (the latter by facilitating increased speed of know-how and technological diffusion). Transport improvements reduce the cost of shipping agricul- tural products to market and extend the distance to Review of the Literature break-even locations, thereby expanding the area of land where cultivation is profitable. Consequently, The rural poor often have extremely limited mobility rural connectivity increases net farm gate prices, beyond their immediate settlement, because of raises farmer incomes (Kyeyamwa and others 2008), geographical isolation and the high cost of motor- yields more stable incomes, and enables the poor to ized transport. As a result, they are not able to take improve their management of risks. advantage of employment opportunities, such as seasonal work, beyond their settlements. Reducing Studies of PMGSY find positive effects on employ- transport costs allows workers to shift from the farm ment, agriculture, and the wellbeing of rural popula- sector to the nonfarm sector (Khandker and Koolwal tion. Asher and Novosad (2020) find that a new road 2011; Gertler and others 2014; Gachassin, Najman, causes a 9 percentage points decline in the share of and Raballand 2015) and to work more (Rand 2012). agricultural workers and an equivalent rise in wage Construction of new roads also supports the emer- labor. They conclude that the changes are driven gence of new nonfarm activities (Mu and van de Walle by work outside the village. Aggarwal (2018) finds 2011). Better access to the outside world improves an increase in the labor force participation rate of access to economic opportunities and increases wel- prime-age women in districts in which more people fare (Jacoby 2000, Fafchamps and Shilpi 2009). have access to a PMGSY road. She finds evidence of greater market integration through lower prices Agriculture is the backbone of the rural economy. and increased availability of nonlocal goods in the Rural connectivity plays a pivotal role in promoting consumption basket, as well as higher adoption rates agricultural production and commercialization. of fertilizer and hybrid seeds among farmers in dis- Improved rural transport can accelerate the intro- tricts with PMGSY roads. The World Bank (2014) finds duction of improved farming practices and the some evidence of a shift in cropping patterns from transition from subsistence farming to cash crops food grains to cash crops in Jharkhand and Himachal and a market economy (Omamo 1998; Minten, Koru, Pradesh following connection to PMGSY roads. and Stifel 2013; Damania and others 2017). Improved 50 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Results of This Study IMPACT ON EMPLOYMENT Employed people include people 21- to 60-years-old The analysis shows a robust increase in employ- who report wage work or business as their primary ment as a result of PMGSY roads. The employment occupation and people 21- to 60-years-old who rate increased by 5.5 percentage points between report being a student or housewife as their primary 2009 and 2017 in habitations connected after 2009 occupation but report wage work or business as thanks to PMGSY roads. The increase represents a secondary occupation.1 Based on these definitions, 9.5 percent change relative to the mean in 2009 in at baseline the employment rate was 62 percent habitations that were unconnected at that time. The (see annex table 4A.1). Fifty-five percent of people increase in employment reflects an increase in part- reported primary employment, with 22 percent of time employment, particularly among housewives them reporting secondary employment as well; 13 who started to work, as indicated by a 12 percentage percent of students and housewives reporting having point increase in part-time employment. There was part-time employment. no increase in primary employment and secondary employment (table 4.1). Table 4.1 Impact of PMGSY roads on employment Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls 0.055* 0.054* 0.056** Employment (0.03) (0.03) (0.03) 0.013 0.012 0.013 Primary employment (0.02) (0.02) (0.02) 0.038 0.048 0.040 Secondary employment (0.05) (0.06) (0.05) 0.116*** 0.116*** 0.120*** Part-time employment (0.04) (0.04) (0.04) 0.129*** 0.130*** 0.132*** Part-time employment (housewives only) (0.04) (0.05) (0.04) 0.095 0.093 0.089 Nonfarm employment (0.06) (0.06) (0.06) 0.127** 0.126** 0.122** Nonfarm primary employment (0.06) (0.06) (0.06) 0.136 0.125 0.126 Nonfarm secondary employment (0.12) (0.12) (0.12) –0.082 –0.051 –0.089 Nonfarm part-time employment (0.16) (0.16) (0.16) 0.087* 0.082* 0.079* Primary employment outside habitation (0.05) (0.05) (0.05) 0.076* 0.074* 0.069* Primary nonfarm employment outside habitation (0.04) (0.04) (0.04) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1. 1 A person is considered employed if he or she reported any of the following occupations: farmer, agricultural labor, construction/other labor, artisan, service (private/government), trade/business, or household worker. 51 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY PMGSY roads triggered a change in the structure employment as a result of PMGSY roads was driven of employment in rural India. The rate of primary largely by people working outside their habitations, employment in the nonfarm sector increased by as it became easier to commute to other villages about 12 percentage points as a consequence of and urban agglomerations. The share of people with PMGSY roads. In 2009 on average 36 percent of nonfarm primary employment outside their habita- people working in unconnected habitations reported tion increased by 7 percentage points as a result of a primary occupation in the nonfarm sector. PMGSY PMGSY roads. roads thus increased the nonfarm primary employ- ment rate by a third. The analysis seems to indicate The effect on employment was stronger in habita- that many students and housewives stepped in tions farther from urban agglomerations. Distance to to take care of the farm after road connectivity urban agglomerations is defined as the distance from improved (as indicated by an increase in part-time a habitation to the nearest census or statutory town. employment and no observed change in the share of The increase in employment, particularly part-time nonfarm part-time employment). The analysis does employment was higher (and statistically significant) not show statistically significant changes in second- in more remote habitations. PMGSY roads caused a ary employment in the nonfarm sector as a result of 5.5 percentage point increase in the employment rate PMGSY roads. in the average habitation (see table 4.1). The coeffi- cient for the distance interaction effect on employ- PMGSY roads allow people to travel and access mar- ment of 0.006 (table 4.2) implies that in habitations kets, including labor markets, outside their habita- 10 kilometers farther away from the nearest urban tions. The share of people with primary employment agglomeration than the average habitation, the effect outside their habitation increased by 8 percentage of PMGSY roads on employment rate was 6 percent- points as a result of PMGSY roads. This increase rep- age points higher than in the average habitation. resents a 35 percent increase relative to the average The effect of PMGSY roads on part-time employment share of primary employment outside the habitation rate was 11 percentage points higher in habitations in 2009 in habitations that were connected after 10 kilometers farther away from the nearest urban 2009. The observed increase in nonfarm primary agglomeration than in the average habitation. 52 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 4.2 Differential impact of PMGSY roads on employment based on distance Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Employment 0.006 * 0.006 * 0.006* (0.00) (0.00) (0.00) Primary employment 0.002 0.002 0.002 (0.00) (0.00) (0.00) Secondary employment 0.006 0.005 0.006 (0.01) (0.01) (0.01) Part-time employment 0.011** 0.011** 0.011** (0.00) (0.00) (0.00) Part-time employment (housewives only) 0.011** 0.011** 0.011** (0.00) (0.00) (0.00) Nonfarm employment –0.004 –0.004 –0.004 (0.01) (0.01) (0.01) Nonfarm primary employment –0.003 –0.003 –0.003 (0.01) (0.01) (0.01) Nonfarm secondary employment 0.002 0.001 0.002 (0.01) (0.01) (0.01) Nonfarm part-time employment –0.020 –0.023 –0.018 (0.01) (0.01) (0.01) Primary employment outside habitation –0.001 –0.000 –0.001 (0.00) (0.00) (0.00) Primary nonfarm employment outside habitation –0.002 –0.001 –0.002 (0.01) (0.01) (0.01) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. ** p < 0.05, * p < 0.1. In hilly areas, the increase in nonfarm primary ruggedness. As there was no change in the rate of employment outside habitations was stronger nonfarm primary employment related to terrain, than in the average habitation, as indicated by the the stronger increase in nonfarm primary employ- positive coefficient in table 4.3. In habitations with ment outside the habitations indicates that people terrain ruggedness two standard deviations above switched the location of their employment without the mean, the rate of nonfarm primary employment changing the sector thanks to PMGSY roads. In hilly outside habitations increased by 9 percentage points areas, PMGSY roads led to an increase in part-time more than in the habitation with average terrain nonfarm employment (table 4.3). 53 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 4.3 Differential impact of PMGSY roads on employment based on terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Employment 0.002 0.003 0.003 (0.01) (0.01) (0.01) Primary employment –0.006 –0.006 –0.006 (0.01) (0.01) (0.01) Secondary employment –0.002 –0.004 –0.003 (0.01) (0.01) (0.01) Part-time employment 0.011 0.011 0.011 (0.02) (0.02) (0.02) Part-time employment (housewives only) 0.016 0.015 0.016 (0.02) (0.02) (0.02) Nonfarm employment 0.017 0.018* 0.017 (0.01) (0.01) (0.01) Nonfarm primary employment 0.019 0.019 0.018 (0.01) (0.01) (0.01) Nonfarm secondary employment –0.018 –0.019 –0.027 (0.05) (0.05) (0.04) Nonfarm part-time employment 0.098* 0.081 0.103* (0.06) (0.06) (0.06) Primary employment outside habitation 0.015 0.015 0.013 (0.02) (0.02) (0.02) Primary nonfarm employment outside habitation 0.030*** 0.030*** 0.028*** (0.01) (0.01) (0.01) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01, * p < 0.1. IMPACT ON AGRICULTURE yield for each type of crop increased in both groups By improving farm-to-market connectivity, PMGSY of habitations between 2009 and 2017. roads are expected to expand the area of land under cultivation, ease the introduction of improved The impact of PMGSY roads on agriculture outcomes farming practices, and/or facilitate the transition was weaker than expected, possibly suggesting the from subsistence farming to cash crops. At baseline need for complementary interventions to support the the average farmer cultivated about 10 acres of development of agriculture value chains. The analysis land. Between 2009 and 2017, the average area found no effect of rural roads on the average land of cultivated land decreased in both groups of under cultivation when considering all crops (table habitations, with habitations connected after 2009 4.4). Farmers in habitations connected after 2009 seeing a smaller decrease (see annex table 4A.2 for reduced the average land area cultivated for cash summary statistics). The average area of land farmers crops by 0.3 acres less than farmers in habitations cultivated for cash crops also decreased. The average that were already connected by 2009. 54 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 4.4 Impact of PMGSY roads on agriculture Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Area of cultivated land (acres) All crops 0.832 0.771 0.621 (1.17) (1.10) (1.19) Food grains 0.379 0.375 –0.001 (1.18) (1.11) (1.21) Cash crops 0.301** 0.300** 0.349** (0.15) (0.15) (0.16) Crop yield (quintals per acre) All crops 0.216 0.874 –0.209 (1.28) (1.28) (1.28) Food grains –0.759 –0.159 –1.022 (1.25) (1.21) (1.28) Cash crops 0.699 1.088 0.815 (2.37) (2.44) (2.28) Crop diversification (Herfindahl Index) –0.010 0.002 –0.012 (0.05) (0.05) (0.04) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Horticulture is not included because of the small number of observations in the treatment group. ** p < 0.05. The impact of PMGSY roads on agriculture outcomes between both groups of habitations cannot be varied with the level of accessibility of the habitation. attributed to PMGSY roads. However, when consider- The area of land under cultivation for food grains ing food grains, yields increased in habitations that decreased in habitations in hilly areas. In habitations were farther away from urban agglomerations and in with terrain ruggedness two standard deviations hilly areas as a result of improved connectivity (table above the mean, the area of cultivated land for food 4.5).2 In habitations with terrain ruggedness two stan- grains decreased by 1.4 acres per crop more than dard deviations above the mean, the yield of food in the habitation with average terrain ruggedness grains increased by 5 quintals per acre more than in after PMGSY roads were built. The average changes the habitation with average terrain ruggedness as a in yield between 2009 and 2017 and the differences result of PMGSY roads. 2 The heterogeneous effects on cash crops was not analyzed, because of the small sample size. 55 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 4.5 Differential impact of PMGSY roads on agriculture based on distance and terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Area of land cultivated with food grains (acres) Distance interaction effect 0.054 0.047 0.043 (0.05) (0.05) (0.05) Terrain ruggedness interaction effect –0.457** –0.453** –0.434*** (0.19) (0.19) (0.16) Yield on food grains (quintals per acre) Distance interaction effect 0.080* 0.065 0.080* (0.04) (0.04) (0.04) Terrain ruggedness interaction effect 1.568** 1.599** 1.541** (0.68) (0.67) (0.69) Crop diversification (Herfindahl Index) Distance interaction effect –0.001 –0.001 –0.001 (0.00) (0.00) (0.00) Terrain ruggedness interaction effect 0.004 0.004 0.004 (0.01) (0.01) (0.01) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Horticulture is not included because of the small number of observations in the treatment group. *** p < 0.01, ** p < 0.05, * p < 0.1. Crop diversification offers opportunities to mitigate To assess the impact of PMGSY roads, a crop diver- risks related to the production and price of crops sification index is calculated (box 4.1). PMGSY roads (Mukherjee 2010; Chhatre, Devalkar, and Seshadri do not seem to have any significant average impact 2016). Using the Herfindahl Index to examine crop on the diversification of crops across habitations (the diversification in 14 Indian states, Mukherjee (2010) coefficients for crop diversification in table 4.4 are not finds that Indian farmers increasingly adopt crop statistically significant). diversification to mitigate the risks associated with the production of a single crop. 56 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Box 4.1 Crop diversification index Crop diversification is calculated for each farmer (household) using the Herfindahl Index, which is the sum of the square shares of cultivated land for each crop on the farm: where values range between 0 and 1, and p is the share of each crop per farm, defined as where Ai is the acreage of cultivated land for each crop within a farm, and ∑n A is the total acreage of i=1 i cultivated land for all crops. An H equal to 1 indicates complete specialization on a single crop; an H close to 0 indicates high crop diversification. ROBUSTNESS CHECKS The results on employment and agriculture are statistical significance. For binary outcomes, results robust to different specifications. When using district were also checked using a binary logistic regression; fixed effects and no fixed effects for each outcome they were consistent with the results presented in variable, the results from the three models (without this chapter. The only exception is the heterogeneous controls, with village-level controls, and with house- effect of terrain ruggedness on nonfarm part-time hold-level controls) remained the same in terms of employment, which becomes insignificant with dis- the magnitude of coefficients, their sign, and their trict fixed effects. 57 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Concluding Remarks The improved mobility provided by PMGSY roads changes in farming practices. The average land area increased access to economic opportunities, trigger- under cultivation did not change as a result of road ing a change in the structure of employment in rural construction, except in habitations in hilly areas, India. Asher and Novosad (2020) find a 9 percentage where it decreased. For food grains, the main crop points decline in the share of agricultural workers in the habitations studied, rural roads had a positive and an equivalent rise in wage labor. They conclude impact only in habitations farther away from urban that work outside the village is driving the changes, agglomerations and in hilly areas. Aggarwal (2020), as they do not find commensurate changes in non- whose unit of observation is the district, not the hab- farm employment in the villages. itation, finds higher adoption of fertilizer and hybrid seeds among farmers in districts with more PMGSY The analysis presented in this chapter finds a similar roads. The results presented in this chapter are shift from farm to nonfarm employment, with the consistent with those of Asher and Novosad (2020), rate of primary employment in the nonfarm sector who find no evidence of farmers moving away from increasing by about 12 percentage points. The house- subsistence crops, land intensification, or increases hold surveys used for the analysis confirm that the in ownership of mechanized farm and irrigation shift to nonfarm employment was largely to nonfarm equipment. employment outside the habitation: The share of people with primary employment outside their habi- The results on employment and agriculture indicate tation increased by 8 percentage points as a result of that improved connectivity in rural areas mainly leads PMGSY roads. people to look for employment opportunities outside their habitations instead of improving and expanding The analysis finds that men are switching primarily their farming to take full advantage of improved to nonfarm employment and that their wives step access to input and output markets. This outcome is in to take care of the farm after road connectivity not bad per se, as villagers may be earning higher improves. The entrance of married women into the incomes by moving to more productive and higher workforce is the main force behind the increase in paying activities. However, if India wants to make employment in connected habitations. This result is more efficient use of agriculture land, it needs com- in line with Aggarwal (2018), who finds an increase plementary programs to support the development in the labor force participation rate of prime-age of agriculture value chains. Such programs should women in districts with access to a PMGSY road. look at the entire value chain, in order to remove constraints on agriculture logistics, such as lack of PMGSY improved farm-to-market connectivity, as aggregation and marketing services, take-up of more indicated by the large and significant increase in the efficient farming practices, and factors limiting econ- share of crops transported by farmers to markets. omies of scale. But the increase did not translate into significant 58 MOBILITY AND TRANSPORT CONNECTIVITY SERIES References Aggarwal, S. 2018. “Do Rural Roads Create Pathways out of Poverty? Evidence from India.” Journal of Economic Development 133: 375–95. Asher, S., and P. Novosad. 2020. “Rural Roads and Local Economic Development.” American Economic Review 110(3): 797–823. Chhatre, A., S. Devalkar, and S. Seshadri. 2016. “Crop Diversification and Risk Management in Indian Agriculture.” Decision: Official Journal of Indian Institute of Management Calcutta 43 (2): 167–79. Damania, R., C. Berg, J. Russ, A. F. Barra, J. Nash, and R. Ali. 2017. “Agricultural Technology Choice and Transport.” American Journal of Agricultural Economics 99 (1): 265–84. Fafchamps, M., and F. Shilpi. 2009. “Isolation and Subjective Welfare: Evidence from South Asia.” Economic Development and Cultural Change 57 (4): 641–83. Gachassin, M., B. Najman, B., and G. Raballand. 2015. “Roads and Diversification of Activities in Rural Areas: A Cameroon Case Study.” Development Policy Review 33 (3): 355–72. Gertler, P. J., M. Gonzalez-Navarro, T. Gracner, and A.D. Rothenberg. 2014. “The Role of Road Quality Investments on Economic Activity and Welfare: Evidence from Indonesia’s Highways.” Unpublished manu- script. http://sites.bu.edu/neudc/files/2014/10/paper_250.pdf. Jacoby, H.G. 2000. “Access to Markets and the Benefits of Rural Roads.” Economic Journal 110: 713–37. Khandker, S., R. Koolwal, and B. Gayatri. 2011. “Estimating the Long-Term Impacts of Rural Roads: A Dynamic Panel Approach.” Policy Research Working Paper 5867, World Bank, Washington, DC. Kyeyamwa, H., S. Speelman, G.V. Huylenbroeck, J. Opuda-Asibo, and W. Verbeke. 2008. “Raising Offtake from Cattle Grazed on Natural Rangelands in Sub-Saharan Africa: A Transaction Cost Economics Approach.” Agricultural Economics 39: 63–72. Minten, B., B. Koru, and D. Stifel. 2013. “The Last Mile(s) in Modern Input Distribution: Pricing, Profitability, and Adoption.” Agricultural Economics 44: 629–46. Mu, R., and D. van de Walle. 2011. “Rural Roads and Local Market Development in Vietnam.” Journal of Development Studies 47 (5): 709–34. Mukherjee, S. 2010. “Crop Diversification and Risk: An Empirical Analysis of Indian States.” MPRA Paper 35947, University Library of Munich, Germany. 59 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Omamo, S.W. 1998. “Transport Costs and Smallholder Cropping Choices: An Application to Siaya District, Kenya.” American Journal of Agricultural Economics 80 (1): 116–23. Qin, Y., and X. Zhang. 2016. “The Road to Specialization in Agricultural Production: Evidence from Rural China.” World Development 77: 1–16. Rand, J. 2011. “Evaluating the Employment-Generating Impact of Rural Roads in Nicaragua.” Journal of Development Effectiveness 3 (1): 28–43. World Bank. 2014. Rural Road Development in India: An Assessment of Distribution of PMGSY Project Benefits in Three States by Gender and Ascribed Social Groups. Washington, DC. 60 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Annex 4A Summary Statistics Table 4A.1 Summary statistics on employment outcomes 2009 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Employment 0.62 5,794 0.58 1,024 0.59 4,851 0.60 976 Primary employment 0.56 5,794 0.55 1,024 0.53 4,851 0.54 976 Secondary employment 0.22 3,061 0.19 546 0.22 2,561 0.23 522 Part-time employment 0.14 2,506 0.07 456 0.14 1,931 0.18 360 (students and housewives) Part-time employment 0.15 2,396 0.07 441 0.14 1,813 0.19 332 (housewives only) Nonfarm employment 0.42 3,570 0.36 591 0.44 2,854 0.48 588 Nonfarm primary 0.45 3,219 0.36 561 0.44 2,587 0.48 525 employment Nonfarm secondary 0.28 684 0.44 106 0.40 565 0.63 118 employment Nonfarm part-time 0.15 351 0.40 30 0.44 267 0.51 63 employment Primary employment 0.26 3,210 0.23 557 0.30 2,587 0.37 525 outside habitation Primary nonfarm employ- 0.22 3,210 0.15 557 0.26 2,587 0.28 525 ment outside habitation 61 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 4A.2 Summary statistics on agriculture-related outcome variables 2009 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Area of cultivated land (acres) All crops 10.2 1,744 9.0 313 4.7 469 4.8 117 Food grains 8.8 1,744 8.4 313 4.1 469 4.6 117 Cash crops 1.0 1,744 0.5 313 0.4 469 0.3 117 Horticulture 0.3 1,744 0.1 313 0.2 469 0.00 117 Crop yield (quintals per acre) All crops 7.3 3,451 6.9 672 9.8 2311 9.4 549 Food grains 6.4 2,946 7.1 617 9.7 1944 9.4 508 Cash crops 3.9 342 4.4 49 8.1 311 8.7 39 Horticulture 30.7 163 13.7 6 21.8 56 10.2 2 Crop diversification 0.63 1,744 0.57 313 0.63 469 0.56 117 (Herfindahl Index) Note: For area of cultivated land and crop diversification, an observation is a farmer. For crop yield, an observation is a crop cultivated by a farmer. Food grains include arhar, barley, gram, horse-gram, jowar, maize, pulses, masoor, millet, moong, moth, pulse, ragi, rice/paddy, soybeans, urad, and wheat. Cash crops include cotton; dry chilies; oilseeds (sesamum, rapeseed and mustard, linseed, and groundnut); guar seed; sugarcane; and tobacco. Horticulture includes apples, arecanut, onions, potatoes, and vegetables. 62 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Chapter 5. Impact of PMGSY on Wealth and Human Capital Accumulation 63 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Improved rural connectivity provides long-term By reducing the time it takes to travel to school, and sustained boost in the living standards of rural enhanced road access can potentially improve people. Improved access to economic opportuni- schooling outcomes. Khandker, Bakht, and Koolwal ties increases income, allowing people to increase (2009) and Khandker and Koolwal (2011) find positive consumption and wealth accumulation. Improved short- and long-run impacts on schooling—for both connectivity facilitates access to schools and health boys and girls—of a rural road construction program services, which contributes to human capital accumu- in Bangladesh, mainly because the connections to lation. These benefits of improved rural connectivity new markets increased schooling investment by can translate into long-term poverty reduction. increasing both household income and the returns to education. This chapter presents the results of the differ- ence-in-differences analysis on the impact of PMGSY Adukia, Asher, and Novosad (2020) find a positive roads on wealth, education, and health outcomes. causal impact of PMGSY all-weather roads on both It begins with a review of the literature on these the enrollment and the educational performance of outcomes. middle-school children. Consistent with the standard human capital investment model, the effects are larger for roads that are more likely to raise the returns to education and smaller for roads that are Review of the Literature more likely to raise the opportunity cost of schooling. However, roads can also increase access to outside Lower transport costs improve access to labor, inputs job opportunities, which may increase the opportu- and outputs markets, increasing income (Escobal nity cost of schooling, discouraging human capital and Ponce 2002, Jacoby and Minten 2009, Cuong investment. Aggarwal (2018) and Mukherjee (2012) 2011); raising per capita consumption and improving find that the presence of PMGSY roads increased the livelihoods of rural households (Emran and Hou middle-school enrollment as well as dropout rates for 2013); and reducing poverty (Dercon and others high school students. 2008; Khandker, Bakht, and Koolwal 2009). These effects increase households’ asset holdings, which Access to roads can also improve health-seeking reduces their vulnerability (Krishna and others 2004, behavior and health outcomes through multiple Barrett and Swallow 2006, Carter and Barrett 2006). channels. Reductions in transportation cost and Reducing vulnerability is especially important in travel time (Adhvaryu and Nyshadham 2012), environments in which credit and insurance markets improvements in health care supply, and increases do not work for the rural poor, and households rely in household income and awareness promote on their assets to smooth consumption and ensure health-seeking behavior. survival in the face of repeated shocks (Carter and Barrett 2006). 64 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Using a fuzzy regression discontinuity design, Banerjee and Sachdeva (2015) show a positive Results of This Study impact of road access on the use of preventative healthcare services (antenatal care, trained health IMPACT ON WEALTH personnel–assisted delivery, modern contraception, Two wealth indexes were constructed to assess health insurance, and water treatment) by women the impact of PMGSY roads on the socioeconomic and households in villages connected by PMGSY. status of households. The first uses principal Both demand- and supply-side factors were at play. component analysis, the methodology adopted The authors find improvements in awareness about by Gwatkin and others (2000), Filmer and Pritchett public healthcare programs, health care supply, and (2001), and McKenzie (2005). The second uses the social interaction, both within and between villages. equal weights method, as adopted by Montgomery and others (2000). Consistent with increases in the inputs to the health production function, Bell and van Dillen (2015) find The indexes are based on four types of variables: positive effects on health outcomes for a sample small asset ownership, large asset ownership, water of villages in upland Orissa. The provision of an source characteristics, and dwelling characteristics all-weather road led to reductions in the likelihood (table 5.1). Several of these variables were also used of an individual in a typical village connected by in the socioeconomic index Filmer and Pritchett PMGSY falling sick and the duration of disabling (2001) created for India. The data were standardized, illness. Improved market access had a positive impact in order to be able to use the same unit for all index on household nutrition in rural Ethiopia (Stifel and questions. For each of the assets, water sources, or Minten 2015). In rural Georgia, road rehabilitation dwelling characteristics shown in table 5.1, a value of improved access to emergency medical services by 1 was assigned if the household had it; a value of 0 cutting the time required for an ambulance to arrive was assigned if it did not.1 by one third (Lokshin and Yemtsov 2005). Table 5.1 Wealth index based on principal component analysis Standard Bottom Variable Mean deviation Weight 40% Middle 40% Top 20% Ownership of small assets Mattress 0.76 0.43 0.16 0.58 0.81 1.00 Pressure cooker 0.44 0.50 0.37 0.03 0.57 0.98 Chair 0.59 0.49 0.35 0.15 0.82 1.00 Table 0.44 0.50 0.37 0.03 0.57 0.98 1 In the principal component analysis, the covariance matrix was used and only the first component generated retained, a common practice for creating socioeconomic indexes. The first component had an eigenvalue of 1.1 and explained 31 percent of the variation. 65 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Standard Bottom Variable Mean deviation Weight 40% Middle 40% Top 20% Watch or clock 0.65 0.48 0.17 0.47 0.71 0.90 Bicycle 0.30 0.46 -0.06 0.37 0.29 0.19 Radio 0.11 0.31 0.04 0.06 0.14 0.15 Sewing machine 0.28 0.45 0.26 0.03 0.31 0.73 Mobile telephone 0.70 0.46 0.21 0.47 0.79 0.98 Ownership of large assets Motorcycle 0.28 0.45 0.13 0.15 0.32 0.45 Refrigerator 0.17 0.38 0.21 0.01 0.11 0.64 Car 0.04 0.19 0.05 0.00 0.01 0.16 Water source Improved water source 0.79 0.41 0.11 0.68 0.81 0.97 Five minutes or less to fetch 0.41 0.49 0.28 0.13 0.45 0.89 water Dwelling characteristics Latrine inside house 0.52 0.50 0.30 0.22 0.60 0.98 Kitchen in separate room 0.50 0.50 0.31 0.16 0.60 0.98 Main cooking fuel not 0.15 0.36 0.17 0.02 0.09 0.55 biomass Dwelling of mixed or 0.65 0.48 0.25 0.37 0.78 0.97 high-quality materials Ownership of some assets is more widespread than The weights from the principal component analysis others. Seventy percent of households own a mobile in table 5.1 show the importance of each variable in phone and almost 80 percent have improved water constructing the index. For example, owning a sew- source.2 But only 21 percent own a refrigerator and ing machine increases the household wealth index by 17 percent use a fuel other than biomass for cooking. a larger margin than owning a mobile phone. Richer households have higher rates of asset ownership and lower rates of “negative” assets (using biomass as the main cooking fuel, for example). 2 The definition of improved water source comes from the Millennium Development Goals (MDGs). It includes water piped into a dwelling, piped to a yard or plot, a public tap or stand pipe, a handpump, a tube well or bore well, a protected well, a protecting spring, and rainwater. 66 MOBILITY AND TRANSPORT CONNECTIVITY SERIES The richest 20 percent of households (as defined by Table 5.2 illustrates the impact of PMGSY roads the index) have higher rates of all but one of the out- on the wealth indexes. The coefficients show that comes, bicycle ownership, which is negatively related they had a positive effect on the average wealth of with ownership of other assets, dwelling characteris- households. The average wealth indexes increased in tics, and water source. Other studies have also found both groups of habitations over time, but households a negative weight for bicycle ownership (see Vyas and in habitations connected after 2009 experienced Kumaranayake 2006). larger increases than households connected before 2009. Using the equal weights wealth index as the The equal weights index is a much simpler measure dependent variable reveals that households con- of wealth. It assigns a weight of 1 to each asset, water nected after 2009 added about one more asset than source, and dwelling characteristic listed in table 5.1; households that were already connected by 2009—an it then adds up all the factors to create the index. increase of 0.24 standard deviations or a 12.4 percent This index ranges from 0 to 18, where 0 means that with respect to the median wealth in the sample. the household has none of the factors listed in table Using the principal component analysis, the average 5.1 and 18 means that it has all of them. Annex table increase in wealth was equivalent to adding a pres- 5A.1 presents the mean of the two indexes for each sure cooker and a radio to the household’s assets.3 group of habitation by year. However, the coefficients in model 3 for the principal component analysis is just below the standard 90 percent level of statistical significance. Table 5.2 Impact of PMGSY roads on household wealth Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Wealth index, principal component analysis 0.226 * 0.193 0.196 (0.13) (0.14) (0.13) Wealth index, equal weights 1.011** 0.831 0.868* (0.51) (0.52) (0.49) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. ** p < 0.05, * p < 0.1. 3 The sum of the weight times the standard deviation for a pressure cooker and a radio is almost 0.196 (see table 5.1). 67 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY The terrain of the locality affected the impact of roads deviations above the mean, household wealth index on household wealth (table 5.3). Households in hillier decreased by 0.36, which represents a 25 percent habitations experienced decreases in their average decrease with respect to the median wealth in the wealth; households in flatter habitations experienced sample.4 In the flattest habitation in the sample, increases. The coefficient for the terrain interaction the household wealth index increased by 0.3, which effect on wealth index calculated using principal represents a 21 percent increase with respect to the component analysis of –0.162 (model 3) implies that median wealth in the sample.5 in habitations with terrain ruggedness two standard Table 5.3 Differential impact of PMGSY roads on household wealth based on distance and terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Wealth index based on principal component analysis Distance interaction effect –0.001 0.002 –0.002 (0.01) (0.01) (0.01) Terrain ruggedness interaction effect –0.154*** –0.151*** –0.162*** (0.06) (0.06) (0.05) Wealth index based on equal weights method Distance interaction effect 0.002 0.010 –0.006 (0.05) (0.05) (0.04) Terrain ruggedness interaction effect –0.622*** –0.609*** –0.658*** (0.22) (0.22) (0.20) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. *** p < 0.01. IMPACT ON EDUCATION Years of completed schooling increased across all lev- connected to PMGSY roads. Ages 6–11 correspond to els of education in the habitations surveyed between primary school, 11–16 to middle school, and 16–20 to 2009 and 2017. The analysis stratifies students in high school.6 For the middle and high school samples, the sample into three groups based on their level the analysis considers only students with some school- of schooling, age, and when their habitations were ing. Annex table 5A.2 presents the summary statistics. 4 The decrease is equivalent to 0.37 standard deviations. 5 The increase is equivalent to 0.31 standard deviations. 6 Among control (treatment) habitations, 87 percent (80 percent) got connected in the two (three) years preceding the baseline (endline) survey. Children 6–11 years old in the two (three) years before the baseline (endline) survey are considered treated while in primary school. People 11–16 and 16–20 are considered treated while in middle and high school, respectively. 68 MOBILITY AND TRANSPORT CONNECTIVITY SERIES PMGSY roads had a positive impact on schooling. baseline for habitations connected after 2009. The On average, children who were in middle or high difference-in-difference analysis does not find any school at the time their habitation was connected had statistically significant average effect on years of about 0.7 more years of schooling in 2017 as a result schooling for children in primary school. There was of PMGSY roads that were built about three years no significant differential impact on girls. Both girls earlier (table 5.4). The additional years of schooling and boys benefited equally from the construction of represent about a 9 percent increase in the years PMGSY roads (table 5.5). of middle and high school relative to the average at Table 5.4 Impact of PMGSY roads on years of completed schooling Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Primary school –0.021 0.053 –0.066 (0.27) (0.29) (0.27) Middle school 0.641** 0.653** 0.681*** (0.25) (0.26) (0.25) High school 0.697** 0.735** 0.748*** (0.31) (0.30) (0.28) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Middle and high school include the intensive margin only. *** p < 0.01, ** p < 0.05. Table 5.5 Differential impact of PMGSY roads on years of schooling of girls Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Primary school –0.604 –0.468 –0.615 (0.48) (0.51) (0.50) Middle school 0.451 0.500 0.401 (0.35) (0.36) (0.33) High school 0.419 0.387 0.454 (0.54) (0.54) (0.53) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Middle and high school include the intensive margin only. 69 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY PMGSY roads had a positive and significant effect on deviations above the mean, completed schooling years of completed schooling for children in primary for children in primary school increased by 1.2 school in hill areas, with the effect stronger the more years more than in habitations with average terrain rugged the terrain (table 5.6). The coefficient for ruggedness. The strong and significant average the terrain interaction effect on years of completed effects on children in middle and high school were primary school of 0.367 (model 3) implies that in unaffected by terrain ruggedness or distance to the habitations with terrain ruggedness two standard nearest urban agglomeration (table 5.6). Table 5.6 Differential impact of PMGSY roads on years of completed schooling based on distance and terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Primary school Terrain ruggedness interaction 0.412*** 0.433*** 0.367*** (0.08) (0.07) (0.09) Distance interaction –0.014 –0.033 –0.008 (0.03) (0.03) (0.03) Middle school Terrain ruggedness interaction 0.079 0.152 0.067 (0.12) (0.12) (0.12) Distance interaction –0.000 –0.019 0.008 (0.04) (0.04) (0.04) High school Terrain ruggedness interaction –0.063 –0.069 –0.098 (0.13) (0.14) (0.13) Distance interaction –0.033 –0.037 –0.006 (0.03) (0.03) (0.03) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Middle and high school include the intensive margin only. *** p < 0.01. 70 MOBILITY AND TRANSPORT CONNECTIVITY SERIES IMPACT ON HEALTH Health-seeking behavior improved between 2009 of household members to seek medical treatment in and 2017 across all the habitations surveyed. In 2017 town; however, the coefficients are not statistically the average share of male and female household significant (see table 5.7). The analysis finds a strong members that went to the local or regional town for and statistically significant decrease in the propensity medical treatment was higher than in 2009 (see table for at-home child delivery of more than 14 percent- 5A.3 for summary statistics). The average share of age points. The reduction represents a 30 percent babies delivered at home also decreased between decrease in the share of babies delivered at home 2009 and 2017, and the average number of children relative to the average at baseline for habitations that in the household that received OPV-BCG-polio-DPT- were connected after 2009. The reduction in delivery measles vaccines increased. at home was greater in habitations farther away from urban agglomerations than in the average habitation PMGSY roads had a positive impact on health-seeking (table 5.8). The effect was weaker in habitations in behavior. The difference-in-difference analysis indi- hillier areas than in flatter areas. cates a potentially positive effect on the propensity Table 5.7 Impact of PMGSY roads on health-seeking behavior Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Went to town for treatment (share) Male 0.100 0.119 0.109 (0.08) (0.09) (0.08) Female 0.029 0.047 0.038 (0.08) (0.08) (0.08) Baby delivered at home (share) –0.147** –0.170** –0.144** (0.07) (0.07) (0.06) Number of children in household immunized 0.161 0.058 0.187 (0.18) (0.20) (0.18) Share of children in household under four 0.161** 0.191** 0.155** immunized (0.08) (0.09) (0.08) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Immunization refers to OPV-BCG-polio-DPT-measles vaccines. ** p < 0.05. 71 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Table 5.8 Differential impact of PMGSY roads on health-seeking behavior based on distance and terrain Model 1 Model 2 Model 3 Outcome variable No controls Village-level controls Household-level controls Baby delivered at home (share) Distance interaction effect –0.019** –0.017** –0.018** (0.01) (0.01) (0.01) Terrain ruggedness interaction effect 0.046** 0.049** 0.044** (0.02) (0.02) (0.02) Share of children in household under four immunized Distance interaction effect 0.015* 0.012 0.015* (0.01) (0.01) (0.01) Terrain ruggedness interaction effect –0.020 –0.020 –0.019 (0.01) (0.02) (0.01) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Immunization refers to OPV-BCG-polio-DPT-measles vaccines. ** p < 0.05, * p < 0.1. The impact of PMGSY roads on child immunization period were more likely to benefit than children was insignificant when considering all children. The born just before 2014 (in habitations connected after analysis finds a positive, albeit not statistically signif- 2009). Therefore, differences in immunization take-up icant, impact of rural roads on immunization, mea- between children under four and older children in sured by the number of all children in the household 2017 in the habitations connected after 2009 provide receiving OPV-BCG-polio-DPT-measles vaccines. It is an alternative estimate of the treatment effect, after possible that children who do not receive vaccines netting out the same differences estimated in the at the proper age do not receive them later. The habitations connected before 2009 to account for any availability of vaccination history (whether the child cohort effect. Thanks to PMGSY, the share of children was vaccinated) for all children in the household as under the age of four that received immunization recorded at endline (this information is not available increased by 15.5–19.0 percentage points, depending at baseline) allows a different identification strategy. on the specification. The improvement in immuni- zation was greater in habitations farther away from PMGSY roads had a strong and significant impact urban agglomerations than in the average habitation on immunization of children under age four. Given (see table 5.8). Among children under four, there was that 80 percent of the treatment habitations were no significant differential impact of PMGSY roads on connected in or after 2014, children born in this immunizations for girls versus boys (table 5.9). 72 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 5.9 Differential impact of PMGSY roads on immunization of girls Model 1 Model 2 Model 3 Village-level Household-level Outcome variable No controls controls controls Share of children in household under four immunized –0.189 –0.181 –0.189 (0.15) (0.15) (0.15) Note: All regressions are ordinary least squares. Standard errors are in parentheses, clustered at the habitation level. All models include state fixed effects. Immunization refers to OPV-BCG-polio-DPT-measles vaccines. ROBUSTNESS CHECKS The results on wealth are robust to different spec- in the study did not collect data on educational per- ifications. Using district fixed effects and no fixed formance. Using data on educational performance, effects for each outcome variable had little effect on Adukia, Asher, and Novosad (2020) find a positive the results from the three models (without controls, causal impact of PMGSY all-weather roads on both with village-level controls, and with household-level enrollment and educational performance for mid- controls). The statistical significance of the coefficient dle-school children. Aggarwal (2018) and Mukherjee for the principal component analysis index was just (2012) find that the presence of roads can increase below 90 percent when tested with district fixed both middle-school enrollment and dropout rates effects. The results on education and health out- for high school students. The analysis presented in comes are also robust to different specifications. chapters 4 and 5 indicate that a very small proportion of students got part-time jobs after their habitations were connected, but doing so did not translate into dropping out of school. Overall, these results are in Concluding Remarks line with the results of other research. The analysis presented in this chapter finds a small Women were more likely to travel to medical facil- but positive effect on wealth, but it is statistically ities to deliver their babies after their habitations significant only under some specifications. Asher and were connected with all-weather roads. Roads also Novosad (2020) do not find a significant effect on increased the share of young children receiving vac- asset ownership. Further research in this area would cinations. Other studies also show positive impacts help understand the link between improved employ- of PMGSY on health-seeking behavior and outcomes ment outcomes and asset accumulation, helping (see Bell and van Dillen 2015 and Banerjee and policy makers design complementary interventions to Sachdeva 2015). increase the benefits of PMGSY. The impacts on wealth and human capital accu- PMGSY roads had a positive impact on schooling in mulation could potentially be the most important rural areas, with boys and girls benefiting equally. benefits of the program, as they set the foundations School attendance is a necessary but not sufficient for long-lasting poverty reduction in rural India. Only condition for capital accumulation. The survey used time will tell if that is the case. 73 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY References Adhvaryu, A. R., and A. Nyshadham. 2012. “Schooling, Child Labor, and the Returns to Healthcare in Tanzania.” Journal of Human Resources 47 (2): 364–96. Adukia, A., S. Asher, and P. Novosad. 2020. “Educational Investment Responses to Economic Opportunity: Evidence from Indian Road Construction.” American Economic Journal: Applied Economics 12(1): 348–76. Aggarwal, S. 2018. “Do Rural Roads Create Pathways out of Poverty? Evidence from India.” Journal of Economic Development 133: 375–95. Banerjee, R., and A. Sachdeva. 2015. “Pathways to Preventive Health: Evidence from India’s Rural Road Program.” University of Southern California Dornsife Institute for New Economic Thinking, Research Paper 15-19. Barrett, C.B., and B.M. Swallow. 2006. “Fractal Poverty Traps.” World Development 34 (1): 1–15. Bell, C., and S. van Dillen. 2015. “The Ways to Good Health? Rural Roads, Illness and Treatment in Upland Orissa.” Discussion Paper, University of Heidelberg, Department of Economics, Heidelberg. Carter, M.R., and C.B. Barrett. 2006. “The Economics of Poverty Traps and Persistent Poverty: An Asset-Based Approach.” Journal of Development Studies 42 (2): 178–99. Cuong, N.V. 2011. “Estimation of the Impact of Rural Roads on Household Welfare in Viet Nam.” Asia-Pacific Development Journal 18 (2): 105–35. Dercon, S., D.O. Gilligan, J. Hoddinott, and T. Woldehanna. 2007. “The Impact of Roads and Agricultural Extension on Consumption Growth and Poverty in Fifteen Ethiopian Villages.” CSAE Working Paper 2007–01, Centre for the Study of African Economies, Oxford University, Oxford. Emran, M.S., and Z. Hou. 2013. “Access to Markets and Rural Poverty: Evidence from Household Consumption in China.” Review of Economics and Statistics 95 (2): 682–97. Escobal, J., and C. Ponce. 2002. “The Benefits of Rural Roads: Enhancing Income Opportunities for the Rural Poor.” Grupo de Análisis para el Desarrollo Working Paper 40–1, Lima. Filmer, D., and L. Pritchett. 2001. “Estimating Wealth Effects without Expenditure Data—or Tears: An Application to Educational Enrollments in States of India.” Demography 38 (1): 115–32. Gwatkin, D.R., S. Rutstein, K. Johnson, E. Suliman, A. Wagstaff, and A. Amouzou. 2000. “Socio-Economic Differences in Health, Nutrition, and Population in Ethiopia.” Working Paper 39419, World Bank, Washington, DC. 74 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Houweling T.A.J., A.E. Kunst, and J.P. Mackenbach. 2003. “Measuring Health Inequality among Children in Developing Countries: Does the Choice of the Indicator of Economic Status Matter?” International Journal for Equity in Health 2: 8. Jacoby, H.G., and B. Minten. 2009. “On Measuring the Benefits of Lower Transport Costs.” Journal of Development Economics 89 (1): 28–38. Khandker, S.R., Z. Bakht, and G.B. Koolwal. 2009. “The Poverty Impact of Rural Roads: Evidence from Bangladesh.” Economic Development and Cultural Change 57 (4): 685–722. Khandker, S.R., and G.B. Koolwal. 2011. “Estimating the Long-Term Impacts of Rural Roads: A Dynamic Panel Approach.” Policy Research Working Paper, World Bank 5867, Washington, DC. Krishna, A., P. Kristjanson, M. Radeny, and W. Nindo. 2004. “Escaping Poverty and Becoming Poor in Twenty Kenyan Villages.” Journal of Human Development 5 (2): 211–26. Lokshin, M., and R. Yemtsov. 2005. “Has Rural Infrastructure Rehabilitation in Georgia Helped the Poor?” World Bank Economic Review 19 (2): 311–33. McKenzie, D.J. 2005. “Measure Inequality with Asset Indicators.” Journal of Population Economics 18 (2): 229–60. Montgomery M.R., M. Gragnolati, K.A. Burke, and E. Paredes. 2000. “Measuring Living Standards with Proxy Variables.” Demography 37 (2): 155–74. Mukherjee, M. 2012. “Do Better Roads Increase School Enrollment? Evidence from a Unique Road Policy in India.” https://ssrn.com/abstract=2207761. Stifel, D., and B. Minten. 2015. “Market Access, Welfare, and Nutrition: Evidence from Ethiopia.” ESSP Working Paper 77, Washington, DC: International Food Policy Research Institute. Vyas, S., and L. Kumaranayake. 2006. “Constructing Socio-Economic Status Indices: How to Use Principal Component Analysis.” Health Policy Plan 21 (6): 459–68. 75 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY Annex 5A Summary Statistics Table 5A.1 Summary statistics on wealth indexes, 2009 and 2017 2009 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Wealth index based on principal 1.49 1,862 1.04 315 1.68 1,914 1.53 391 component analysis Wealth index based on equal 6.90 1,862 5.23 315 7.43 1,914 6.98 391 weights approach Table 5A.2 Summary statistics on years of completed schooling, 2009 and 2017 2009 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Boys and girls Primary school 4.0 2,134 4.3 341 4.8 1,470 5.1 250 Middle school 7.4 1,920 7.2 383 8.5 1,571 8.9 354 High school 9.1 1,237 8.6 224 10.1 1,193 10.4 292 Girls Primary school 3.9 918 4.1 156 4.8 738 4.7 117 Middle school 7.2 713 7.2 176 8.3 771 9.2 160 High school 8.8 400 8.1 84 9.9 506 10.3 132 76 MOBILITY AND TRANSPORT CONNECTIVITY SERIES Table 5A.3 Summary statistics on health outcomes, 2009 and 2017 2009 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Traveled to town for treatment (share) Male 0.49 1,223 0.34 197 0.64 949 0.62 196 Female 0.49 1,222 0.34 197 0.67 1,054 0.58 228 Girls Baby delivered at home (share) 0.33 2,226 0.47 414 0.26 764 0.25 128 Number of children in household 0.9 714 1.0 123 1.4 533 1.6 87 immunized Note: Immunization refers to OPV-BCG-polio-DPT-measles vaccines. Table 5A.4 Summary statistics on immunization of children under four Four or older in 2017 Under four in 2017 Control Treatment Control Treatment habitations habitations habitations habitations Outcome variable Mean N Mean N Mean N Mean N Share of children in household under 0.84 337 0.72 57 0.91 585 0.95 113 four immunized Share of girls in household under four 0.82 153 0.82 27 0.913 276 0.963 54 immunized Note: Immunization refers to OPV-BCG-polio-DPT-measles vaccines. 77 THE ROAD TO OPPORTUNITIES IN RURAL INDIA: THE ECONOMIC AND SOCIAL IMPACTS OF PMGSY