by
KL Krishna
Bishwanath Goldar
Deb Kusum Das
Suresh Chand Aggarwal
Abdul Azeez Erumban
Pilu Chandra Das

Profiles of Authors

K L Krishna, a Founder Member of CDE (Centre for Development Economics), at the Delhi School of Economics, has been the Principal Investigator of the RBI-funded India KLEMS Productivity Project since 2009. With post-graduate qualifications in Statistics from the University of Kerala, and the Indian Statistical Institute, Kolkata, and Ph.D. degree in Economics from the University of Chicago, he taught Econometrics and Industrial Economics for four decades and supervised around 45 Ph.D. and M.Phil. theses in a variety of fields. He served as Head of the Department of economics. and Director, Delhi School of Economics in the 1980s and 1990s. He was the Founder Managing Editor of the Journal of Quantitative Economics for one decade, and President of the Indian Econometric Society (TIES) for one term. He is a member of the TIES Trust. He was the Chairman or member of several official Committees of the Government of India. He held the position of Chairman of CESS, Hyderabad, during 2007-13, and MIDS, Chennai during 2013-20.

Bishwanath Goldar is a Retired Professor of Economics of the Institute of Economic Growth, Delhi. He has an M.A. in Economics and Ph.D. from Delhi School of Economics, University of Delhi. He has been at the Institute of Economic Growth from 1979 to 2014, except for three brief stints at a professorial or equivalent position at the National Institute of Public Finance and Policy (1988-90), the Indian Council for Research on International Economic Relations (2003-04), and the Jawaharlal Nehru University (2012-13). During 2015-2016 he was a National Fellow of the Indian Council of Social Science Research, affiliated with the Institute of Economic Growth. He is a former member of the National Statistical Commission and is currently a member of the Standing Committee on Economic Statistics, constituted by the Ministry of Statistics and Programme Implementation, Government of India. His area of specialization is industrial economics and international trade. The bulk of his research has been on productivity and employment in Indian Industries, wage share in value added, and price-cost margin in Indian manufacturing, export performance of Industrial firms, effective protection of Indian industries, impact of trade reforms on the performance of industrial firms, and foreign direct investment in India. He has also undertaken studies on rural water supply in India and Bangladesh, water supply in Delhi, pollution of river water and river water quality in India, and the environmental aspects of Indian industries including studies on energy efficiency in Indian industrial firms and the impact of environmental performance of industrial firms on their stock prices.

Deb Kusum Das, Professor at the Department of Economics, Ramjas College. He has a PhD from the Delhi School of Economics and received the EXIM Bank IEDRA Award 2004 for his doctoral dissertation, “Some Aspects of Productivity and Trade in Indian Industry”. He is also associated with ICRIER as an external researcher and researched on important issues related to Indian economy: jobs, labour intensive manufacturing, India’s global competitiveness. His research interests are empirical international trade, labour markets and productivity growth of the Indian economy. He is the co-founder of a network for South Asian undergraduate students of Economics (SAESM).

Suresh Chand Aggarwal is currently the Senior Fellow, ICSSR at Department of Business Economics. Earlier retired as Head and Professor from the Department of Business Economics (now known as Department of Finance and Business Economics), University of Delhi. He has guided a number of students for their Ph.D. and M.Phil. thesis. He does research in Labour Economics, Econometrics and Development Economics and has published a number of papers in the related areas in national and international Journals. He has been associated as a Lead researcher on Labour in the India KLEMS project from its beginning and participated in numerous national and international conferences and has contributed to a number of papers under the KLEMS project. He has earlier done projects and consultancies for many organizations, e.g., ILO, World bank, UGC, ICRIER, etc. His current projects are RBI funded 'INDIA KLEMS' Project, and ICSSR funded ‘Inclusive Growth Project’.

Abdul Erumban teaches and researches at the University of Groningen (RuG), Groningen, The Netherlands. He is also a senior research fellow at the Conference Board (TCB) and an academic member of the Productivity Institute at the Alliance Manchester Business School. During 2013-2019, he worked as a senior economist at TCB, leading their research on global productivity, global economic outlook, and emerging markets. He has an M.A. in Development Economics from John Matthai Centre, University of Calicut, an M.Phil in Applied Economics from the Centre for Development Studies, and a Ph.D. in Economics from the University of Groningen. His research centers on productivity, technological change, structural change, digital transformation and its impact, globalization, global value chain, and international comparisons of economic development. He maintains a particular interest in economic issues in emerging markets, particularly India, China, and the Middle East economies. Being an active participant in the KLEMS research initiatives aimed to understand productivity dynamics in the world's major economies, he sustains a good relationship with international networks on productivity research. He is an advisor to the Asia KLEMS, which aims to understand productivity dynamics in Asian economies, and a lead researcher on capital in the Reserve Bank of India's India KLEMS project, seeking to understand productivity and competitiveness in Indian industries. He has also actively participated in European Commission's World Input-Output Database project at the University of Groningen. His principal publications are in the Journal of Economic Perspectives, Review of Income and Wealth, Journal of Comparative Economics, Industrial and Corporate Change, and Structural Change and Economic Dynamics.

Pilu Chandra Das is an Assistant Professor in the department of economics at Kidderpore College, University of Calcutta. He has an M.A. in economics and M.Phil. from Delhi School of Economics, University of Delhi. His M.Phil. thesis was on “Total Factor Productivity in Indian Organised Manufacturing: the story of the Noughties”, dealing with productivity in Indian registered manufacturing sector. He has been associated with the India KLEMS project since 2012. His research interest is mostly in productivity and has been a co-author of several papers prepared under the India KLEMS project. He has several published papers. Earlier, he was associated with ICRIER as a research assistant in a project named ‘Estimating Domestic Value Added and Foreign Content in India’s Exports’ sponsored by the Department of Economic Affairs, Ministry of Finance, Government of India. He was also associated with National University of Singapore in constructing the variables for the Singapore KLEMS data set.

Foreword

The innovative work of the India KLEMS team has resulted in the groundbreaking India Productivity Report, which is a comprehensive report primarily using the India KLEMS database to examine various aspects of productivity dynamics in Indian industries. The India KLEMS project, which follows the standards of World KLEMS and EU KLEMS, provides a database on capital, labour, and intermediate inputs along with total factor productivity, thus facilitating a comparison of India's productivity dynamics with other global players. It is an important step in analyzing the role of factor accumulation, productivity, and structural change in the Indian economy.

The India KLEMS project came to fruition in 2009 with Prof. KL Krishna as chairperson. I have known Krishna since he received his PhD from the University of Chicago in 1967, after completing his dissertation on productivity and then returning to DSE at the University of Delhi. Krishna has been the ideal lead on this project. From the inception of the project, Prof. Bishwanath Goldar served as advisor and the late Deb Kusum Das served as the coordinator. Prof. Suresh Chand Aggarwal and Dr. Abdul Azeez Erumban were among the original project team. Most-recently, Pilu Chandra Das joined the team in 2015. They, along with numerous collaborators, have made an outstanding contribution with the publication of the India Productivity Report. The Report was more than a decade-long endeavor, during which time numerous articles have been published in international journals by the team. The India KLEMS team has attended and presented at all World KLEMS conferences, starting in 2010, and all Asia KLEMS meetings starting in 2011. The India team will host the next Asia KLEMS meeting in 2022.

As in many other Asian economies, overcoming the legacies of past policies has been a major challenge for India. India had adopted the path of an inward-oriented closed economy policy atmosphere since it embarked on its path of planned industrialization in the mid-1950s. Since the mid-1980s, there has been some rethinking in the policy circles. Liberalization of the Indian economy to private and foreign investors began in 1988, which led to a foreign exchange crisis and IMF program in 1991. Manmohan Singh was appointed Finance Minister in July 1991, during which time India undertook bold steps to reform the economy. This changed the economic development path in the country, eventually giving India a more prominent and visible presence in the global economy. The pro-market policy changes have altered the business climate for the private and external sectors, making the importance of productivity growth more significant in the Indian context. While many researchers have explored the productivity dynamics in India, their work was confined mostly to the formal manufacturing sector, which had relatively better data. The constraints on data for inputs and output at the detailed industry level made it difficult to produce a fair analysis of factor inputs and productivity in other sectors of the economy.

A country’s economic policies and investment climate are key components of factor accumulation and productivity, which drive economic growth. Therefore, understanding sources of economic growth is paramount to researchers and policymakers. In today’s market-oriented economy, productivity growth is as important to fostering economic growth in India as it is in raising the living standards of millions of people in the world's second most populated country. Moreover, analysis of changes in factor inputs and productivity at the industry level is essential in identifying India's growth path to facilitate a pro-growth structural transformation. Evidence suggests consistent decline in poverty over the years in India. However, as home to the world's young population, India has vast human capital potential, if tapped properly. Additionally, market opportunities for investors, along with its resources, provide the potential to improve India's living standards. Productivity plays a key role in this process by improving production efficiency and helping generate surplus for additional consumption and investment. Overcoming productivity growth challenges requires a better understanding of the underlying features of productivity of specific industries.

Since the 1990s, intensified global integration as well as global fragmented production chains have resulted in drastic changes in the global economy. This has created opportunities for many developing economies, including India, to participate in segments of the global value chain, where they have a comparative advantage, and further climb up the quality ladder. Increased productivity would help better position India in the global value chain and fend off the intense competition from imports which displaces domestic producers. It is this perspective that gives the India KLEMS project and the India Productivity Report added significance.

The Report also documents the changes in the employment structure of the economy, which features a fall in agriculture – as one would expect – but stagnant manufacturing. The Report has noted the recent fall in manufacturing jobs in India and the failure of the sector to absorb workers who leave the primary sector. While overall employment growth has been relatively weak, there has been a shift in the structure towards service sector jobs. Along with the need to create manufacturing jobs, the Report analyzes energy intensity to reveal the need for policy implementation that encourages energy conservation in the manufacturing sector. Policy adjustments are also recommended to enhance India's human capital. While improvements are being made, the pace must accelerate given the size of the population and labour force. Meanwhile, the agricultural sector remains the largest job-providing broad sector in India although its relative importance in job creation is declining. By developing future extensions of the KLEMS data, researchers may gain an understanding of dynamics within the sector by considering sub-sectors within the agricultural sector. Undoubtedly, the India KLEMS database and the productivity report demonstrate substantial improvements in the data and analysis of the Indian economy. The continued work on the data will lend itself to fewer assumptions. Currently, in addition to several comparability issues across various employment surveys, the researchers needed to make assumptions to fill in the gaps between survey years, as there are no comprehensive official time-series estimates available.

With the liberalization of the economy in the 1990s the country became attractive to domestic and foreign investors, but hurdles still exist, namely stimulating investments in the formal sectors such as the much-important manufacturing sector. According to the Report the investment to GDP ratio in India has recently fallen. The need for reforming the labour market, improving infrastructure – both physical and human – and investment climate, are essential to stimulate investment and productivity. Although India is an excellent exporter of IT professionals and services, previous research by the India KLEMS team indicated a tremendous untapped potential in many sectors of the economy from the use of ICT, through investing in IT equipment, software, and communication. Taking into consideration various policy reforms, the trends in productivity demonstrate that productivity benefits from outward-oriented policies rather than inward-oriented ones.

The India Productivity Report, as well as the India KLEMS data, have a profound impact not just on India, but on the global economy. The Report does a phenomenal job of unveiling productivity dynamics in India, which creates a clearer understanding of the growth process in the country. The India KLEMS data is a crucial tool for researchers to more intensely study and understand various aspects of India's economic growth in the coming years. Moreover, the database can serve as an invaluable resource for policymakers. Globally, the KLEMS research has advanced further to incorporate the roles of ICT and intangibles in the growth process. As the KLEMS research continues, focusing on the participation of several industries in the global value chain and the potential for specializing in specific tasks and sectors will impact India research. I look forward to the policies and research that will result from this report, as well as further development of the data.

This research was made possible by the generous support of the Reserve Bank of India. For almost 13 years, RBI has funded the work on India KLEMS at the Indian Council for Research on International Economic Relations, and the Centre for Development Economics at the Delhi School of Economics. This level of support has been instrumental in building India KLEMS and the data that through continued policy and research will impact the global economy.

In closing, I would like to offer my profound appreciation for the work of the late Deb Kusum Das. His role leading the India KLEMS initiative was key in bringing this publication to fruition. His focus on quantifying trade barriers in India and the impact of the economic liberalization helped drive the research of the entire group. Applause to the India KLEMS team and sincere gratitude to D.K.D. At the time of his passing, he was serving as the President of Asia KLEMS, where his passion for his work was impactful. I, along with his many friends and colleagues, will be forever grateful for his contribution to the economic community.

Dale W. Jorgenson

University Research Professor, Harvard University

Preface and Acknowledgements

This report, titled “India Productivity Report” is based on research work carried out under the India KLEMS project at the Centre for Development Economics (CDE), Delhi School of Economics (DSE) in collaboration with the Reserve Bank of India (RBI).

We, the team members of the research project, prepare this preface and acknowledgements to the report with a deep sense of grief because of the sudden and untimely passing away of Prof. Deb Kusum Das on the Christmas eve of 2021. Prof. Das has been a prominent member of the research team. He is the person who brought us into the research team. This report marking the culmination of the research project has been his dream undertaking. He played a pivotal role in the endeavour to prepare this report, from conceptualizing the content of different chapters at the early stage, to ensuring that the chapters get written in time, to getting the chapters copy edited as the work progressed, and even to attending to nitty-gritties such as the colours of lines and bars in the graphs placed in different chapters as the work had come almost to an end. The team members gratefully acknowledge the amount of work Prof Deb Kusum Das had done for the preparation of the report till the cruel hands of destiny took him away, and they have now collectively taken over the responsibility to ensure that the report be of high quality, as envisioned by Prof. Deb Kusum Das.

As said earlier, the India KLEMS research project is housed at the Centre for Development Economics (CDE), Delhi School of Economics (DSE). The research project has been there from 2014. Prof. K.L. Krishna led the research project. Prof. Bishwanath Goldar acted as research advisor. Other senior members of the team that carried out research work under the India KLEMS project are Prof. Deb Kusum Das who acted as the project coordinator, Prof. Suresh Chand Aggarwal, Dr Abdul Azeez Erumban, and Shri Pilu Chandra Das. Dr Erumban is at the Groningen Growth and Development Centre, University of Groningen, the Netherlands. The research team is thankful to the Groningen Growth and Development Centre, University of Groningen for the research time Dr Erumban has devoted to the project.

The project team has received ample guidance from a Research Advisory Committee whose members include the leading international experts on measurement of productivity, the most prominent of them being Prof. Dale Jorgenson of Harvard University, Cambridge MA, USA. Other experts in the advisory committee who provided direction to the research done in the project include Prof. Marcel Timmer, Prof. Bart van Ark, and Prof. Mary O’Mahony.

The team will fail in its duty if it does not place on record the encouragement that they have received from the international KLEMS research community. Special thanks are due to Prof. Hak Kil Pyo, Prof. Kyoji Fukao and Prof. Harry Wu for the interaction that the team members had with them in connection with Asia KLEMS, particularly in attending meetings and conferences of Asia KLEMS. Prof Wu has been very kind to the India KLEMS research team. He has provided data and has actively participated in the studies undertaken on India-China productivity comparison. The India KLEMS research team is grateful to Prof Wu for immense helpfulness and intellectual engagement.

The India KLEMS project first started at Indian Council for Research on International Economic Relations (ICRIER) in 2009, where the foundations of the subsequent research work was laid. Prior to that, a feasibility report for construction of data series on output and inputs of different sectors of Indian economy and carrying out research based on those data, similar in spirit and design to EU-KLEMS, was prepared by Prof. K.L. Krishna, Prof. Arup Mitra and Prof. Bishwanath Goldar. The feasibility report provided the basis for the development of a research proposal which was submitted to the RBI. This met with a positive response and the RBI very kindly funded the project from its inception. The first phase of the research project continued in ICRIER till 2013. Later in 2014, the new phase of the project began at CDE, DSE. The National Statistical Office (NSO) had provided a great deal of support in the form of data and technical advice when the project was at ICRIER, and they have provided such valuable support and advice also later when the new phase of the project began and continued at CDE, DSE.

A series of consultation workshops were held at ICRIER in the initial phase of the research project. This was done for developing the methodologies for the measurement of inputs, output and productivity, and further refinements were made in the meetings and annual workshops held at CDE, DSE. A large number of scholars and experts, and persons with substantial knowledge of the data particularly India’s national accounts have contributed majorly in the ten odd years’ journey of the project. It is difficult to name all of them. To take a few names out of the long list, the research team members gratefully acknowledge that they have immensely benefited in the course of last 14 years of work on the project from the intellectual inputs received from Prof. T.N. Srinivasan, Prof. Isher Judge Ahluwalia, Prof. K. Sundaram, Dr Pronab Sen, Prof. T.C.A. Anant, Prof. T.S. Papola, Prof. Ashok Gulati, Prof. Simrit Kaur, Prof. Surender Kumar, Prof. Ravindra Dholakia, Prof. Pushpa Trivedi, Mr Ramesh Kolli, Mr. G. Raveendran, Mr. Bimal Giri, Mr Nilachal Ray, Dr G.C. Manna and Shri Ashish Kumar.

The National Accounts Division of the National Statistical Office (NSO) has provided a great deal of support to the research project both when it was at ICRIER and when it got relocated at CDE, DSE. The support has been in the form of data and technical advice. Indeed, some unpublished data on the asset composition of investment in different industries for different years has been a major, invaluable help in the preparation of capital input series. For this support regarding capital series and in general about output and input series, the research team thanks Mr S.V. Ramanamurthy, Mr P.C. Mohanan, Mr. P.C. Nirala, Ms. T. Rajeshawari, and Ms. Anindita Sinharoy. Mr Ramanamurthy has been attending the coordination committee meetings and advisory committee meetings and has contributed to the project through his advice.

Interaction with officers of RBI has been of immense benefit to the research project. In the presentations made at the Mumbai Office of RBI in 2011 and 2012, considerable insights were obtained from the comments and suggestions received from Late Dr Subir Gokarn (Deputy Governor, RBI) and Shri Deepak Mohanty (Executive Director, RBI). Subsequently, in July 2018, the research team got an opportunity to make a presentation of their research findings at a workshop at the Mumbai Office of the RBI. Dr Viral Acharya (Deputy Governor, RBI) gave the introductory observations and chaired one of the sessions of the workshop, and Dr Amartya Lahiri contributed to the workshop with his comments and suggestions. The comments and suggestion received at the workshop were a great help to the research. The team wishes to thank particularly Prof. T.C.A. Anant, Prof. Pami Dua, Prof. Chetan Ghate, Prof. K. Narayanan, and Prof. Pushpa Trivedi. In the course of that workshop, several RBI officers made presentations based on their research, which provided useful intellectual inputs for the research undertaken by the India KLEMS team.

Over the last seven years, since the second phase of the project started at CDE, DSE from 2014, the India KLEMS research team had the benefit of interacting with RBI officials in the co-ordination committee meetings, advisory committee meetings, and at the annual workshops where the research undertaken during the year was presented. There is a long list of RBI officers to whom the India KLEMS team is indebted for intellectual inputs and ideas that they have provided in such meetings. The presentations made by RBI officers at the annual workshops held under the research project have been a matter of great encouragement to the India KLEMS research team. The research team wishes particularly to thank Dr. Rajeev Ranjan, Ms. Balbir Kaur, Ms. Rekha Mishra, Dr. Jai Chander, and Mr. S.V. Arunachalam for the support, encouragement and constructive suggestions in the coordination meetings and advisory committee meetings. Other RBI officers that the India KLEMS research teams wish to thank sincerely include Shri B. M. Misra, Dr Satyananda Sahoo, Ms. Rigzen Yangdol and Mr Avdhesh Kumar Shukla. The Indian KLEMS research team thanks the RBI officers who made presentations of their research at the annual workshops of the KLEMS project held at DSE. To name some of them, these include Shri Sarthak Gulati, Shri Utsav Saksena, Shri Avdhesh Shukla, Ms. V. Dhanya, Mr. Sonna Thangzason, Ms. Sonam Choudhury, Shri Rajib Das, Shri Siddhartha Nath, Dr. Harendra Behra and Shri Saurabh Sharma. In addition, for chairing the sessions of the annual workshops, the India KLEMS team would like to thank Dr M.D. Patra and Dr Rajeev Ranjan.

In the course of last two years, the India KLEMS research team has been interacting with several senior officers of RBI, Dr M. D. Patra, Dr Sitikantha Pattanaik, Dr D. P. Rath, and Dr Rajiv Ranjan for the coordination of the project. The team is sincerely thankful to them. The team would like to thank Mr. Avdhesh Kumar Shukla for the interactions it had with him in the course of meetings and annual workshops. During 2021, training programs for RBI officers were held for the purpose of skill transfer, so that after 2021, the project could be housed at the RBI, implemented by the RBI officers. The India KLEMS research team greatly appreciates the efforts made towards learning and interest shown in the training programs by Dr Sadhan Chattopadhyay, Shri Siddhartha Nath, and Dr Sreerupa Sengupta.

The annual workshops under the KLEMS project gave an opportunity to interact with other researchers working in the area of productivity in India. For making presentations at the workshops and for participation in the discussions, thanks are due to Prof. Dibyendu Maiti, Dr Jagganath Mallick, Ms. Niti Khandelwal, and Ms Rupika Khanna.

The India KLEMS research team is grateful to Dr Rakesh Mohan who was the Director of ICRIER when the research proposal and the request for grant was given to the RBI. For providing ample encouragement and for helping in administrative matters, thanks are due to the successive Directors of ICRIER during 2009-2013 when the research project was housed at ICRIER, Dr Rajeev Kumar, Dr Parthasarathi Shome, and Dr Rajat Kathuria, and to the successive Executive Directors of CDE, Prof. Sunil Kanwar Prof J.V. Meenakshi, , Prof. Rohini Somanathan, Prof. Shreekant Gupta, Prof. Aditya Bhattacharjea, and Prof. Dibyendu Maiti in the course of last eight years when the project was housed at CDE. The ICRIER staff and CDE staff have been always helpful in the implementation of the research project, for which the research team wish to record their appreciation and thanks. Among the ICRIER staff, the KLEMS research team are especially thankful to Shri Manmeet Ahuja and Mr Krishan Kumar, who along with their team members provided considerable help in organizing workshops, co-ordination committee meetings, and other meetings connected with research work.

The research team particularly wants to thank Mr Surjit Singh, Manager of CDE during the first five years of the project at CDE, 2014 to 2018, and Ms Deepika Garg, who has been the Manager of CDE since 2019. The research team wishes to thank Shri Rajesh Papnai who maintains accounts, the ICT personnel over the years, Mr. Sanjeev Sharma, Ms. Mandeep Kaur, Mr. Raju Chauhan, Mr. Sonveer Vats and Mr. Rohit Kohli, for the ICT related support they have provided, and other subordinate staff, Mr. Mritunjay Bisht and Mr. Ashok Kumar for making arrangements in connection with meetings and workshops.

The research work under the India KLEMS project involved a massive amount of work for data collection and processing. This would not have been possible without the support received from other members of the research team, which is gratefully acknowledged. In the course of the initial four years of the project at ICRIER valuable research work has been done and support provided by Ms Kuehlika De, Ms Shreerupa Sengupta, Mr. Gunajit Kalita, Ms Deepika Wadhwa, Mr Jaggannath Mallick, Mr Pilu Chandra Das, Mr Subhojit Bhattacharjya, and Mr Parth Goyal. In the course of last seven years so, from 2014 to 2021, support for research in the India KLEMS project has been provided by Ms. Sanghita Mondal, Mr Vinay Sharma, Mr. Maajid Mehaboob Chakkarathodi, Mr. Anuj Goyal, Ms. Mehak Gupta, Ms. Prachi Madan, Mr Sk Md Azharuddin, Ms. Suchetna Pahwa and Mr. Samiran Dutta. The research team members do not words enough to thank these persons for the role they have played through their hard work and dedication for the successful completion of the India KLEMS research project.

We wish to place on record our great appreciation of the review of the Report done by Prof. Hak K. Pyo, Professor Emeritus, Faculty of Economics, Seoul National University, Prof. H. Chun and Dr. K.H. Rhee, members of the Korea KLEMS team, and Prof. Minh Khuong Vu of Lee Kuan Yew School of Public Policy, National University of Singapore. We have immensely benefited from their comments and suggestions. A presentation on the Report was made at a Webinar organized by the RBI in July 2022. Valuable comments and suggestion were received from the RBI officers who were discussants at the Webinar. Detailed, excellent comments were received subsequently from the National Accounts Analysis Division of the Department of Economic and Policy Research, RBI. We express our gratitude to the RBI officers for the comments and suggestions that they have given which has helped us improve the Report.

Finally, the research team wishes to place on record their appreciation of the work done by Ms Poonam Madan for copy editing the manuscript prepared by members of the research team.

K.L. Krishna
Bishwanath Goldar
Suresh Chand Aggarwal
Abdul Azeez Erumban, and
Pilu Chandra Das

Chapter 1: Introduction to India KLEMS Approach

1.1 Introduction

Economic growth and productivity at the global and country level is of major interest to researchers and policymakers, even more during global shocks and slowdowns, where the focus of economic policy becomes to rejuvenate “growth.” This gives productivity a prime place in academic research directed at understanding why some countries are able to do much better than other similarly placed countries in terms of the rate of economic growth and raising the standards of living.

Going from the general to the specific, for India, a major emerging economy, productivity is the key force or driving engine for rapidly fostering the development and growth process.¹ India has substantial economic potentials thanks to its enormous human and natural resources. Yet, the country is unable to lift in a short period the living standards of her vast population to sufficiently high levels owing to constraints of various kinds. A swift, substantial augmentation of productivity would enable India’s goods and services production units to penetrate in a much bigger way into the global export markets. Productivity improvements will help the country get better integrated into the global value chains and fend off the strong competition from imports that tends to displace domestic producers. It is also important for making small and medium firms more competitive, especially in the manufacturing sector, and for enhancing value addition in various economic activities securing higher profits, which would encourage further investments and higher remuneration for the worker. All of these will help make the size of the economy increasingly bigger and the Indian people progressively more prosperous. An understanding of India’s achievements regarding productivity advances, how productivity advances have contributed to economic growth, and the forces underlying the trends in productivity is obviously important both from an academic and policy point of view.

While productivity can be measured using multiple approaches (Box 1), the discussion here is focused on introducing the KLEMS-based approach to productivity measurement in the Indian economy. The KLEMS approach takes into account the roles played by capital, labour, energy, materials and services as inputs in output growth. The evolution of the India KLEMS research effort is aimed at building quality as well as detailed disaggregated data for analyzing growth and productivity at the industry level in the context of the World KLEMS initiative, which aims at construction of comparable datasets for cross-country as well as country-specific detailed productivity analysis. A review of global literature on productivity research throws up some very significant studies indicating the importance of analysing the contribution of productivity to economic growth. Indeed, the research focussed on productivity as a source of growth makes an important contribution to the literature on economic growth.

Box 1: Different Measures of Productivity

This report, titled India Productivity Report (IPR) provides the detailed analysis required to understand India’s performance since the 1980s, both at the economy-wide and disaggregated industry levels. The report is based on the India KLEMS database, version 2019, which includes important inputs as variables, such as capital (K), labour (L), energy (E) material (M) and services (S), at the industry level to construct estimates of labour and total factor productivity (TFP) across a set of 27 industries from seven broad sectors (Annexure 1A mentions the full list) using the growth accounting technique for productivity measurement and analysis (Box 2 outlines some measurement issues).

This report devotes separate chapters to various aspects of the analysis undertaken, covering themes that have remained in focus and new themes added over the years, since the India KLEMS productivity study was launched in 2009. The report covers the analytical framework and detailed methodology as well as substantive empirical results for 27 industries, broad sectors and the overall economy. Alternative measures of output, including gross value of output and gross value added have been distinguished in the presentation of the results. The data sources and construction of variables are discussed in detail in the next chapter (Major point given in Annexure 1B). The period of study is from is from 1980-81(written alternatively as 1980) to 2017-18 (written as 2017) and includes periods of high growth (2003-07) as well as global slowdown (2008-17)².

Box 2: Evolving treatment of productivity in the global context

1.2 Evolution of India KLEMS Research

The India KLEMS (Capital, Labour, Energy, Materials and Services, as inputs) research on growth and productivity in the Indian economy began in 2009 at ICRIER (Indian Council for Research on International Economic Relations), a leading think-tank in Delhi, at the instance of Prof. Dale Jorgenson of Harvard University. Prof. Jorgenson, a pioneer in KLEMS research worldwide, was on a visit to India in 2007 when he suggested to late Dr. Isher Judge Ahluwalia, chairperson of ICRIER, to undertake productivity research on the lines of EU-KLEMS which had been set up in 2003 for growth and productivity research at disaggregated levels in the advanced economies, comparing the European Union, the United States and Japan. ICRIER set up a working group in 2007 consisting of Prof. K. L. Krishna, Prof. Arup Mitra and Prof. B. Goldar to prepare a feasibility report which provided a basis or foundation to develop a detailed research proposal on the lines of EU KLEMS.³

Stage 1

ICRIER accepted the working group’s report, and a request was made to RBI (Reserve Bank India) for a research grant. After due consideration, RBI consented to support the project financially, subject to appropriate terms and conditions. An advisory committee was formed with Prof K. L. Krishna as chairperson and nominees from RBI and CSO (Central Statistics Office), ICRIER external experts, Prof. Jorgenson, Prof. Bart van Ark, Prof. Marcel Timmer, and Prof. Mary O’Mahony of EU KLEMS to monitor and give advice for the research. The project was launched in 2009, with Prof Bishwanath Goldar as advisor and Dr. Deb Kusum Das as the coordinator. Prof. Suresh Chand Aggarwal and Dr. Abdul Azeez Erumban were the key members of the research team right from the beginning. The other members of the research team in the initial years were Ms Deepika Wadhwa, Mr. Gunajit Kalita, Mr. Jaggananath Mallick. Mr. Suvojit Bhattacharjya and Mr. Parth Goyal, and subsequent new members of the team were Ms Kuhelika De, Ms Sreerupa Sengupta, and Mr. Pilu Chandra Das. It is because of the excellent research support and high dedication to work of these members of the team that the India KLEMS project could make laudable progress.

The first major task in the India KLEMS project was to construct time series data on gross output, GVA (Gross Value added) and five inputs, K (capital), L (labour), E (energy), M (materials), and S (Services) and factor income shares from 1980-81 (1980) onwards for 26 disaggregate industries comprising the Indian economy (number of industries later enhanced to 27). The data sets had to be consistent with the National Accounts Statistics (NAS) and other official economic statistics of the Indian government. The 26-industry classification was adopted from EU (European Union) KLEMS project for the sake of facilitating international comparisons (see Annexure Table A.1.1). The data sets of the Indian KLEMS project covering a period of 29 years, from 1980-81 (1980) to 2008-09 (2008) were used to prepare a comprehensive report, ‘Estimates of Productivity growth for Indian Economy’ and submitted to RBI in 2013. RBI published these on its website in July 2014.

Stage 2

Since January 2015, the India KLEMS Project has been housed at the Centre for Development Economics (CDE), Delhi School of Economics, with continued financial support from RBI and the same team of researchers with Prof. K.L Krishna as the Principal Investigator and Chairman of the advisory committee. Prof. Bishwanath Goldar has continued as advisor and Dr. Deb Kusum Das as co-ordinator. Prof. Suresh Chand Aggarwal has continued to specialise in labour input data, and Dr. Abdul Azeez Erumban in Capital input data. Mr. Pilu Chandra Das has assisted the team on construction of annual time series on Output, Gross Output (GO), and Gross Value Added (GVA), and the five inputs, K, L, E, M and S, and factor income shares. Every year since then, the updated data set and the data manual have been submitted to RBI, as part of the deliverables. In addition, analytical papers using the latest data set have been prepared in accordance with the terms of the grant for the year. Members of the research team have made extensive use of the India KLEMS data sets to prepare research papers and present them at various national and international conferences (See Annexure 1C). Of special significance to the world KLEMS initiative are two papers⁴ comparing productivity performance in India and China that were undertaken in collaboration with researchers from China.

With effect from 2022, RBI takes over the India KLEMS project. The India Productivity Report (IPR) covering the period 1980-2017 has been under preparation by the research team for the past two years to be submitted to RBI.

1.3 Global Productivity Research

The advent of research on productivity dates to the works of Solow (1957) and Jorgenson and Griliches (1967). Since the World KLEMS research platform was initiated in 2010, there has been a continuous flow of research on sources of growth and productivity, with a primary focus on disaggregated sectors, incorporating the notion of a production function at the industry level.

Following the success of the EU KLEMS project, Prof. Dale Jorgenson, Prof. Marcel Timmer and Prof. Bart Van Ark established the World KLEMS Initiative in 2010 at the first World KLEMS conference held at Harvard University. This is a platform of more than 40 countries, encompassing Europe, United States, Asia and Latin America. Two major regional platforms were also developed. The Chile-based Economic Commission for Latin American and the Caribbean (ECLAC), which is an agency of the United Nations, established the Latin American regional organization, LA KLEMS in 2009. The Asia KLEMS network was set up in 2010, comprising Japan, South Korea and Taiwan alongside China and India. These networks have organized several world and regional conference and meetings⁵ and the next Asia KLEMS conference is due to be held in India in 2022. The core of these regional networks has been to construct country specific KLEMS datasets. These conferences showcase research outputs in measurement and analysis of multifactor productivity with due attention to both country studies and cross-country comparisons.

Productivity as a source of our understanding of economy-wide growth has been extensively researched both within and outside the KLEMS platform. We highlight here some of the important recent literature in this context.

Van Ark et al. (2021) provide a key recent investigation into the question of how to avoid another phase of sluggish growth, after what occurred following the global financial crisis leading to a period of global slowdown for advanced economies. The authors state that in the past 15 years, productivity growth in advanced economies has significantly slowed, giving rise to the productivity paradox of the new digital economy – that is, increased business spending on information and communication technology assets and digital services without a noticeable increase in productivity. The study provides evidence that structural changes apparently underline the slowing productivity growth rates at the macro level. In the United States, most of the positive contribution to productivity growth is coming from the digital-manufacturing sector. The Euro Area and the United Kingdom show larger productivity contributions from the digital-using sectors. The paper outlines a series of policy measures to ensure medium as well as long-term recovery.

Dabla-Norris et al (2015) observe that trend growth in advanced economies has been slowing, driven by a slowdown in human and capital accumulation and declining total factor productivity (TFP). The decline in TFP before the crisis reflected not only the reallocation of resources to sectors where productivity growth was slower, but also declining productivity growth within those sectors, which increasingly account for the bulk of employment and economic activity.

Dieppe (2020) is a comprehensive contribution to global productivity research, covering an exhaustive sample of countries with a variety of data sources, including macroeconomic, sectoral and firm-level data. The book provides an in-depth view on Emerging Market and Developing Economies (EMDEs) relative to Advanced Economies (AEs), by using a comprehensive dataset for 35 AEs and 129 EMDEs. The largely policy-oriented assessments provide several perspectives on – trends and prospects of productivity growth; global, regional, domestic and sectoral drivers of productivity; factor reallocation and technological change; inter-country productivity convergence; and in the context of the COVID-19 pandemic, the effect of natural disasters and economic disruptions on productivity growth. Several important findings emerge from these assessments, namely weak investment, weak TFP growth and fading gains from factor reallocation. A key finding is that labour productivity growth in EMDEs has been undergoing multiple surges and declines since the 1980s, each of increasing magnitude over time. However, the COVID-19 pandemic threatens to cause the largest and most broad-based decline in EMDE productivity growth, compounding a trend slowdown in labour productivity growth in these countries, which began after the 2007-09 Global Financial Crisis (GFC).

Fraumeni (2019) is a major attempt towards creating a foundational resource for the understanding of the various new and existing methodologies and their corresponding results. Most of the chapters use the KLEMS-type models to analyze sources and origins of growth and productivity for a host of nations – G7, E7, EU, Latin America, Norway, China, Taiwan, Japan, Korea, India and other South Asian countries. By going beyond traditional measurement techniques, the book explores new methodologies meant to better address the contemporary challenges of the global economy.

Several other interesting studies may be mentioned here.⁶ Using the Conference Board’s Total Economy Database (TED), Erumban and Van Ark (2018), report that global labour productivity growth has undergone a drastic slowdown since the beginning of the 2010s – with the slowdown being driven largely by declines in TFP growth. Nomura (2018) aggregates output, labour and capital input growth rates for the period 1970-2014 across a group of 21 Asian countries by taking into account factor-price differentials. The study reports a post-crisis reduction in the TFP growth rate of the group. He identifies that regional productivity slowdown in Asia has been caused mainly by the post-crisis decline in TFP growth of the Chinese economy, on which the Asian economy has become increasingly reliant, and also due to the decline in India’s TFP performance over the same period. Hofman and Aravena (2018) use the growth accounting approach to explain the Latin American and Caribbean region’s economic growth during 1990-2013. They report that Latin American countries have substantially lower labour productivity than developed countries, with a gap that appears to be increasing mainly due to the relatively sluggish contribution of capital deepening and TFP growth in this region.

1.4 Productivity research in India

1.4.1 Early Research

While there are several Indian studies on productivity, covering the aggregate economy as well as broad sectors, the latter are largely about organized manufacturing and only a few on agriculture and service sectors. Bosworth et al. (2007) presented estimates of the sources of economic growth for the sub-periods of 1960-80 and 1980-2004, for the aggregate economy as well as at the sector level for agriculture, industry and services. The productivity estimates of Bosworth and Maertens (2010) echoed the pattern observed by Bosworth et al. (2007). Other earlier studies on sources of growth include Dholakia (2002), Sivasubramonian (2004) and Virmani (2004)⁷.

In context of the manufacturing sector, such studies indicated a slow rate of TFP growth in Indian manufacturing for the period 1951 to 1979 (Reddy and Rao, 1962; Banerji, 1975; Goldar, 1986; and Ahluwalia, 1985). The literature in the 1990s started exploring the biases in productivity estimates of the manufacturing sector due to the use of the value-added function and the method of single deflation⁸ of value-added. In particular, Balakrishnan and Pushpangadan (1994) used the double-deflation method for value-added to contest the assertion of the study by Ahluwalia (1991) (which used single-deflated value-added) regarding the acceleration in manufacturing TFP growth in the 1980s.⁹ Subsequent studies mostly concluded that the post-reform improvement in TFP growth of Indian manufacturing was non-existent and rather, was slower compared with the TFP performance of the 1980s.¹⁰ The handful of earlier studies pertaining to the unorganized manufacturing sector (Prakash, 2004; Unni et al. ,2001; Kathuria et al., 2010) mostly reported a downward trend in TFP growth.

The earlier productivity research on the Indian service sector had focused on specific sub-sectors, such as – Railways (Sailaja, 1988; Alivelu, 2006), Airlines (Hashim, 2003), Insurance (Sinha, 2007), ICT industries (Mathur, 2007) and a relatively larger number of studies were about the Banking sub-sector (De, 2004; Sinha and Chatterjee, 2008). For the services sector as a whole, Goldar and Mitra (2010) and Virmani (2004) indicated an acceleration in TFP growth after 1980. Even before the inception of the India KLEMS project, Banga and Goldar (2007) had used an econometric approach with the KLEMS production function framework to conclude that the contribution of the services (S) input to output and productivity growth of the organized manufacturing sector had increased substantially in the 1990s, with trade reforms identified as a major cause.

The productivity research revolving around the agriculture sector has looked at both –the aggregate agriculture sector and sub-sectors such as crops, livestock, including crop-specific analysis of major crops such as rice and wheat. A number of studies found improvement in the TFP index of the Indian agriculture sector, following the introduction of high yielding varieties (Rosegrant and Evenson, 1992; Evenson et al., 1993; Fan et al., 1999). In particular, Fan et al. (1999) found a relatively faster TFP growth rate of 2.5 per cent per annum for the 1980s and a marginal slowdown to 2.3 per cent per annum in the 1990s. The introduction of high yielding varieties, adoption of new technologies and improvement in rural infrastructure were argued to have induced faster productivity growth. Although most of the studies estimated technological change for aggregate agriculture, micro-level farm data¹¹ were deployed by several crop-specific studies (Pinstrup et al., 1991; Sindhu and Byerlee, 1992; Kumar and Rosegrant, 1994; Joshi et al., 2003). Krishna (2006) is a useful review of the early literature on TFP growth in Indian agriculture and its determinants, and Bhalla (2006) is a relevant study of the impact of agricultural growth on rural poverty, and it draws attention to measures for raising labour productivity in agriculture and diversification to high-value crops.

1.4.2 Recent Research

Productivity analysis continues to be important in the Indian research agenda and agricultural productivity, in particular, continues to engage researchers (Nicholas Rada and David Schimmelpfennig 2015, Emerick 2018, Balaji and Suresh Babu 2020). Initiatives by the Indian government through several policy programmes such as “Make in India” attempt to boost competitiveness in the manufacturing sector, which has been characterised by low TFP growth. Several studies have sought to investigate the impacts of such initiatives by looking at the determinants of low productivity, keeping in consideration aspects of foreign involvement (Chawla, 2020), mark ups (Mukherjee and Chanda, 2020), firm size (Mitra and Sharma, 2018) and exports (Narayanan and Sahu, 2020). The use of diverse datasets pertaining to informal and formal manufacturing at firm as well as industry level enables analysis of the manufacturing dynamics prevalent in the Indian economy.¹²

The rapid expansion of the service sector and emergence of the service-driven growth has provided much scope for research, on issues related to the nature and implications of growth of this sector, including exports and employment (Bhide et al., 2021). Further, within the services-manufacturing linkage, the servification of manufacturing and the impact of reforms in the services sector¹³ on manufacturing performance remain dominant research themes in the Indian context. The importance of service-sector productivity with emphasis on both market and non-market services and its determinants is a potential research agenda that still needs to be explored.

1.4.3 Productivity research based on KLEMS database

Since the construction of KLEMS dataset with estimates of labour as well as total factor productivity growth by industries, several comprehensive research papers have been written by the KLEMS research team. These papers, using growth accounting technique as their core methodology, attempt to analyse the performance of the aggregate economy and its sub sectors during 1980-2018. Prominent amongst these are Das et al. (2016), Erumban et al. (2019), Goldar et al. (2017), Krishna et al (2020) and Das et al. (2019). Their focus has been evaluation of productivity performance within different economic policy regimes, issues of structural change and growth and assessment of sources of growth for the Indian economy. In addition, Wu, Krishna et al., (2016) undertook a comparison of India’s growth and productivity performance with China. A related work that compares TFP levels in India and China (Erumban et al., 2019) is in progress. Other studies include analysis of structural changes in employment (Krishna et al., 2016), trends in labour quality, (Krishna et al., 2018), and growth and jobs - Indian experience in the international perspective (Aggarwal and Goldar, 2019). Further, issues such as ICT investment and growth (Erumban and Das, 2019), productivity growth and levels in formal and informal manufacturing (Krishna et al., 2018), productivity dynamics in services (Krishna et al., 2016), structural change (Erumban et al., 2019) and a paper exploring India’s trade-productivity linkage using KLEMS and other datasets (Das, 2019) have also been addressed.¹⁴

The research outputs of the KLEMS group bring several novel features to the analysis of productivity growth in the Indian economy. Its period of analysis witnessed piecemeal deregulation from 1980 until 1991 and subsequently, long-term reforms in trade and industry that followed the crisis of 1991-92. Further, the high-growth phase (2003-08) of the Indian economy followed by periods of global shocks and world slowdown in productivity allows examination of several research hypotheses, including the sources of growth and the contribution of productivity enhancement to the observed growth. The construction of dataset for 27 sectors of the economy allowed disaggregated industry-level analyses, which enabled the identification of the leading sectors within groups such as market economy, market services and manufacturing. The industry perspective provides important insights when compared with the aggregate. Analysis of how growth of inputs and total factor productivity of each industry contribute to aggregate value added, thereby also determining each industry’s contribution to aggregate growth forms a part of the research agenda. The growth accounting technique allows for studying sources of overall growth as well sources of both labour productivity and total factor productivity growth. In addition, while assessing the explicit role of inputs when accounting for output growth, the contribution of the inputs to observed productivity gets captured significantly better when the construction of variables, especially labour and capital, allows for heterogeneity within inputs than cases where measurements of these primary inputs did not consider heterogeneity. Finally, the main dataset used in our analysis is the India KLEMS database on inputs, outputs and productivity trends for 27 sectors of the Indian economy from 1980 onwards within a growth accounting framework. The dataset, publicly available at the RBI website¹⁵, illustrates to the research and policy communities, the benefits of using detailed data.

1.5 Report summary and contribution of chapters

The next chapter (2) discusses the main database used in our analysis -- the India KLEMS dataset that provides harmonized statistics on outputs, inputs and productivity trends for 27 industries comprising the Indian economy; a wide range of manufacturing and services sector industries find place in the list of 27 industries. The dataset is constructed in a manner so as to maintain a close association with National Accounts Statistics (NAS), National Statistical Office, Ministry of Statistics and Programme Implementation, Government of India, with all variables benchmarked with the NAS database. It is freely downloadable and available at the RBI website.¹⁶ The chapter describes in brief how the series on various variables in the India KLEMS dataset are constructed and at the same time devotes considerable attention to the details of the data sources employed and methodology, which also receives attention in the subsequent chapters of the report where the data are analyzed. It may be pointed out here that for the construction of the database, several supplementary data sources are used in addition to National Accounts Statistics (NAS), such as the National Sample Survey Office (NSSO) data on employment and unorganized manufacturing, Periodic Labour Force Survey (PLFS), Annual Survey of Industries (ASI), Wholesale Price Index (WPI), Input-Output Transactions Tables (IOTT) and Supply-Use Tables (SUT). Several datasets not in the public domain were provided by National Statistical office (NSO) and information on several important variables had to be extracted using ASI unit level data.

We claim justifiably that an industry-level focus allows for deeper insights when compared with aggregate analysis. It is important to point out that several industries are responsible for the observed productivity and growth performance. Given the large variation across industries with respect to technological characteristics and the regulatory environment, the observing of industries within certain broad groups – manufacturing as well as services, which drive the growth or productivity performance has profound implications for policy making. In addition, the use of detailed input measures allows us to study heterogeneity of inputs. Hence, this report, based on the India KLEMS research project (2009-21), provides new methodological perspectives on the use of data in economic growth and productivity analysis at a disaggregate level.

Chapter 3 discusses the observed changes in structure of the economy, followed by growth decomposition at the industry as well as sectoral level in context of growth in value added. It goes on to examine the contribution of intermediate inputs to the observed growth, using a gross output specification of the industry level production function. It finds that value added share of agriculture & allied activities declined steadily over the past four decades and that of services sector increased from 37 per cent in 1980 to 53 per cent in 2017. A large part of this increase is on account of doubling of the share of market services from 15 to 32 per cent. Of the 6.03 per cent per annum observed average growth for the aggregate economy during 1981 to 2017, 2.03 percentage points came from market services, 1.21 from manufacturing, 1.26 from non-market services and less than 0.8 from agriculture and allied activities. Comparing growth in output with that in input, we observe a strong positive correlation between gross output and intermediate input growth. Further, it is observed that intermediate inputs share in gross output is relatively high in all manufacturing industries and within intermediate inputs, materials input’s share is high across a majority of the sectors or industries.

The next two chapters, 4 and 5, provide an in-depth review of labour and capital—two primary inputs used in assessing both labour and total factor productivity. Chapter 4 first discusses the profile of workers in the Indian economy and then an overview of the structure of employment during the period 1980-2017. It also looks at aspects of labour such as, labour quality and labour income share, from an industry perspective. The main variables used are (1) employment (2) growth rate of employment, (3) growth rate of labour inputs, (4) labour quality indices, (5) growth rate of labour quality and (6) labour income share in value added. The India KLEMS dataset provides time series and cross section evidence of labour input for 27 industry groups. It is also the only consistent time series data available on employment by industry for different categories of labour input. Using this database, a detailed insight is provided on the role of labour input in accounting for output and productivity growth at both aggregate and industry level.

Chapter 4 offers several important insights pertaining to labour used in Indian industries. Employment, as measured by workforce participation rates have been declining in India, especially among rural females. Growth in employment in India also slowed considerably in the recent period and there was an absolute decline of eight million employed persons between 2011 and 2017. The level of employment declined in agriculture, mining, and manufacturing, and the growth in employment was driven mainly by construction and market services, underlying evidence of significant changes in the structure of employment. Further, during 2011-2017, the share of more educated workers significantly increased in the workforce. Another significant observation is that alongside overall growth in employment, there had been a significant growth in the labour quality index, leading to faster growth in labour input during 1980-2017. The observed change in labour income share seems to have occurred mainly due to the ‘within industry’ effects in the sectors and structural change effects were small as compared to the ‘within industry’ effect. The real product wages had experienced a consistent increase in the average annual growth rate during the overall period of 1980-2017. Finally, growth in employment and in labour quality index had neither been uniform across industries, nor across sub-periods, and considerable variations are observed in both these growth rates.

In an annexure to this chapter, an attempt has been made investigate some important factors impacting employment in different major segments of the economy. A positive effect of real value added and a negative effect of real wage rate is found. The models estimated could predict the fall that has actually occurred in employment in agriculture and manufacturing and in aggregate employment between 2011 and 2017. This finding has important implications for the issue of comparability of data sources used in the study to build employment series for recent years.

Chapter 5 analyses the trends in deployment of physical capital in Indian industries during the period 1980-2017. It examines the trends in investment-to-GDP ratio, real investment, capital stock, and growth rates of capital services across various industries. Capital stock is measured using the standard perpetual inventory method for individual assets and capital services as a weighted aggregate individual asset-specific stock of capital. The difference between the two measures--capital stock and capital service growth rates--is the effect of changing composition of capital stock. The chapter documents the changes in capital composition in the broad sectors of the economy, along with trends in the capital-output ratio and capital intensity measured as capital stock per worker. It observes an increasing role of equipment capital in the Indian economy, which helps improve capital composition effects. Capital services growth rates were mostly higher than capital stock growth rates across the board.

In Chapter 6, we examine trends in labour productivity and possible drivers of the observed productivity performances. Measure of labour productivity is deployed to carry out an analysis for broad sectors and the 27 disaggregated industries. Analysis of factor reallocation and sectoral decomposition is also attempted. Finally, for an international perspective, it compares growth and labour productivity in India with that in a select set of countries. The chapter offers several findings--There is large variation in observed labour productivity both, over time and across industries. The observed labour productivity growth is faster in capital-intensive and ICT-based industries. Further, labour productivity growth rates fluctuate considerably during the sub periods-1994-2007 and 2008-2017. The construction industry is the only sector that shows negative growth in labour productivity for the period 1980-2017. Industry decomposition shows that aggregate labour productivity growth is explained mainly by ‘within industry’ productivity growth, and ‘between’ or ‘structural change’ explains only a small part of it. The static reallocation effect has been positive and substantial during this entire period, suggesting a positive structural change effect. The main driver of the observed labour productivity growth in a majority of the industries is growth in capital intensity, with TFP growth contributing significantly to very few industries. The contribution of TFP is evident in ICT industries, with a range of 28 to 40 per cent. Finally, despite achieving one of the highest growth rates in labour productivity at an average of 5 per cent per annum during 1980-2017, the level of labour productivity in India is still very low as compared with countries such as BRICS, and other selected developing and developed countries.

A detailed examination of industry origins of India’s growth and productivity forms the core of Chapter 7. The disaggregated industry-level production accounts are analyzed for the 27 India KLEMS industries and six broad sectors for the overall period of 1981-2017 and for two sub-periods—the pre-GFC phase of 1992-2007, pertaining to post-systematic reforms and 2009-17, the post-GFC phase of global productivity slowdown. The chapter uses growth accounting techniques based on the gross output framework to examine the industry-level sources of output growth. It deploys the methodologically superior and practically sound Production Possibility Frontier (PPF) approach to analyse the origins of aggregate value-added and TFP growth, instead of the more traditional Aggregate Production Function (APF) approach.

The findings of the chapter are three-fold: Decomposition of the sources of growth using the gross output-based production approach yeilds several observations—First, intermediate inputs were the largest contributor to the observed output growth across the 27 industry groups, driven mostly by the contribution of Material (M) input. However, the contribution from the Services (S) input was substantial for the services industries--signaling the importance of intermediate transactions within this sector; Second, capital input was the second-most important contributor after intermediate inputs, especially for the services industries. Third, the contribution from labour input and TFP growth were negligible except for a few instances. A review of the industry origins of GVA growth reveals that apart from the agriculture & allied industries sector, industries belonging to the services sector exhibited impressive contributions to aggregate value-added growth. The substantial contribution of services was a combination of high value-added shares and impressive value-added growth rates. None of the manufacturing industries made it to the list of top contributors to aggregate value-added growth -- raising questions about the numerous policies that have been aimed at enhancing the sector’s competitiveness. The largest contribution to overall productivity growth came from agriculture & allied industries and public administration, a non-market service industry. Analysis at the broad sector level offers the following observations—first, the services sector was the largest contributor to aggregate value-added growth and TFP. The contribution of this sector to aggregate value-added growth was driven mostly by the strong performance of market services, while the corresponding contributions to aggregate TFP growth was driven by the non-market sub-sector. Second, aggregate TFP growth, estimate derived from the PPF, is mostly accounted for by Domar-weighted industry-level TFP growth, however substantial reallocation effects of labour and capital do exist.

Chapter 8 is contextualised to the dualistic structure of manufacturing in India and considers the substantive differences between the formal and informal manufacturing segments in terms of growth in real GVA, employment and productivity. The dominance of the informal segment was evident in its two-third share in employment in 2017-18, and one-third share in GVA. Comparing the two segments in terms of growth in GVA and employment, we observe that faster growth in the formal segment accounted for seven million more jobs out of total 10 million in the overall manufacturing sector during the period of 1999-2017 and its share in employment had thus risen in recent years from 25 per cent in 1999 to 34 per cent in 2017. Labour productivity in the informal manufacturing segment had always been lower than that of the organised segment, though the relative difference had narrowed down and had always exerted a downward pressure, a ‘drag’ on the labour productivity of the overall manufacturing sector. Decomposition of the labour productivity growth shows a positive ‘within’ industry change and structural (or between industries) change, thus pointing to the phenomenon of movement of workers from less productive to industries that were more productive. The chapter offers some policy suggestions, such as, a greater focus could be on those industries, which could aid in accelerating the process of formalisation of Indian manufacturing, both from capital and labour point of view. In addition, unorganised sector enterprises may be incentivised to grow and become competitive by removing obstacles through the provisioning of cheap finance, regular supply of electricity and other necessary inputs, and access to new appropriate technology and markets.

Research on India’s productivity performance has highlighted that factor accumulation rather than productivity improvements are behind the observed growth of the economy. Chapters 9 and 10 contributes to this literature by reviewing factor intensity, especially energy and capital intensities.

The analysis of capital intensity in Chapter 9 reveals that the Indian economy was increasingly becoming capital intensive, amid a falling relative price of capital compared with labour. Much of the recent expansion in capital services and capital deepening was in the market services sector. The construction sector was another sector where capital intensity had been rapidly rising. While these were primarily in alliance with the fall in relative price of capital, the trends in the manufacturing sector remain a challenging question--as the relative prices of capital in this sector were among the highest across the broad sectors of the economy, yet it continues to see high growth in capital per worker.

The results of the econometric analysis of capital intensity indicate that the finding of a marked increase in the growth rate in real wages during 2003-2017 as compared with 1981-2002 fits well with (and may be treated as a major explanation for) the finding of an accelerated growth in capital intensity in the Indian economy during 2003-2017. A significant positive effect of real wages on capital intensity is found from both an analysis based on time series data and an analysis based on panel data.

Analysis of trends in energy intensity in Chapter 10 is confined to manufacturing, which accounts for a substantial part of the economy’s energy consumption, with a focus on the formal segment of manufacturing at the aggregate and industry level for the period 2004-05 to 2016-17. A distinguishing feature here is that energy intensity is computed by taking a physical measure of energy. The analysis of trends in capital intensity has a wider industrial coverage, extending to the entire economy. The study covers the period 1980-81 to 2017-18.

The analysis shows several significant findings. First, in the period from 1980 to the end of 2000s, there was a downward trend in energy intensity in manufacturing, which got subsequently halted. The energy intensity series for aggregate manufacturing shows a mild upward trend after 2010, whereas the energy intensity series for organized manufacturing based on a different data source, namely ASI, shows that after 2011 there was a marked slowdown in the downward trend as compared with the trend prevailing previously. Second, the average annual growth rate in capital intensity at the aggregate economy level during the period 2003-2017 was significantly higher than that during 1981-2002. Decomposition analysis revealed that a sizeable part of the hike in the growth rate in capital intensity in the period 2003-2017 as compared with the period 1994-2002 is explained by the effect of the inter-industry reallocation of labour.

The econometric analysis of determinants of energy use and energy intensity presented in the chapter reveals a significant negative relationship between energy price and energy demand in manufacturing. An examination of movements in prices of energy and output shows that there was a fall in the real price of energy input in manufacturing in the period 2011-12 to 2017-18 and this could be one of the important factors that halted or slowed down the previously prevailing downward trend in energy intensity in manufacturing. Further analysis of trends in energy intensity using a time-series model finds the important role played by real price of energy. The analysis brings out that movements in manufacturing sector output and real price of energy fails to explain fully the movements in energy intensity in the more recent period, and the reversal from the previously prevailing downward trends seem traceable to some factors that tend to raise energy intensity of manufacturing.

Chapter 11 examines the growth and productivity trends in the Indian services sector for the overall period of 1980-2017 and various sub-periods. The analysis has been done at the disaggregated level for the ten services industries and at the sectoral level for market and non-market services. Within the market services sub-sector, ICT-intensive and non-ICT-intensive service segments are analysed separately. Since there is substantial heterogeneity within the service sector, our detailed sectoral focus helps understand the relative position of market services versus non-market services, and ICT-using and producing services versus non-ICT services in the overall service sector performance in India. We use labour productivity as a measure of productivity performance and observe sources of labour productivity growth for the disaggregated services.

Our main observation is that labour productivity in India’s service sector had been growing substantially in the period 1980-2017 and much of this productivity gain was accruing through acceleration in non-market services productivity. In the pre-reforms phase (1980-93), non-market services showed higher labour productivity growth relative to the market services, while the converse was true in the reform phase (1994-2002) and the golden growth phase (2003-07). However, market services fell behind non-market services in terms of labour productivity growth in the global slowdown phase (2008-17). A closer examination further suggests that ICT-intensive sectors, in particular trade and postal & telecommunication had driven much of the market services productivity growth, while non-market services growth had been driven by public administration. We further observe that TFP growth accounted for much of the improvements in labour productivity in financial services, public administration, and overwhelmingly for transportation and storage. For the rest of the services industries, except for education (where the contribution of TFP growth and capital deepening was very close), the contribution of TFP growth was mostly small. For non-market services, we find very low or negligible contributions from overall productivity growth in accounting for labour productivity growth. The decomposition of gross output growth reveals that capital and intermediate inputs (services input in particular) were the significant contributors to the observed output growth rates in different service industries.

The role of TFP growth in accounting for the observed growth was negligible. In the pre-reform phase, non-market services outperformed that of market services in this regard; however, we observe that TFP growth of market services was higher in the reform phase of 1994-2002. Overall, in terms of broad sectors, the highest contribution for the period 1980-2017 was from services. In terms of industry-wise contribution to aggregate TFP growth for the period of 1980 to 2017, the top three contributors include agriculture et al, public administration and defence, and transport and storage. Financial services and ‘other services’ also contributed significantly.

This substantiates our assertion about the growing importance of services in the overall economy and offers evidence that high productivity growth in many service sectors underlies the current dynamism in service sector growth to a considerable extent.

Economic growth in India and China has been often compared in the literature, as these giant economies have lately been the major driving force of recent global growth. These two economies share the experience of a shift from a socialist past to a market economy, which happened around the early 1990s. However, their paths to economic growth have been different. While China's growth was largely driven by the manufacturing sector, India relied substantially on its services sector. One of the important features of the country KLEMS datasets, such as the India KLEMS, is that they allow cross-country comparison of growth and levels of both labour and total factor productivities. Chapter 12 addresses that aspect of KLEMS database by undertaking a comparison of China and India’s growth and productivity performance. This chapter compares the proximate sources of economic growth in the two countries, i.e., the role of capital accumulation and total factor productivity in driving labour productivity, using a comparable methodology and detailed industry-level data for the period 1980-2017. The analysis reveals that China's labour productivity had grown much faster than that of India over the last 40 years, thereby widening the gap between the two countries’ productivity levels. A massive tide in its non-service economy has driven China's labour productivity growth compared with India, where the services economy did relatively better over the years. China performed relatively better in all the nine industry groups reviewed in this chapter, except for the financial and business services sectors in the 1990s and non-market services through the entire 1981-2017 period. China's dominance in the industrial sector was also visible in its TFP performance. China performed better in the aggregate economy, agriculture, and industry, while India’s service sector performance was much better.

The analysis in the chapter tends to suggest that the Chinese manufacturing sector was significantly more successful than the Indian in creating productivity by investing in physical capital. The marginal productivity of capital was higher in Chinese manufacturing compared with India. As already well-established in the literature, India's growth and productivity have been driven by the services sector, and our results indicate that more than capital deepening, it was the skill intensity of workers that appears to have had a higher marginal impact on labour productivity in the market services sector in India. Evidently, upskilling its workers had a larger productivity impact in India, primarily in its market services, compared with China. The capital-driven growth in China's manufacturing and skill-driven growth in India's market services are clearly brought out by our analysis in terms of the marginal productivity effects of these inputs in the two economies.

Chapter 13 sums up the main findings of the study. It outlines certain data issues in the construction of the India KLEMS database. The chapter ends with a discussion on the way forward in terms of strengthening the database and in terms of the research to be undertaken based on the data.

Annexure 1A: List of Industries in India KLEMS Database

Annexure 1B: The data sources and construction of variables – key points

• The latest available dataset is India KLEMS dataset, version 2020. The period covered is 1980-81 to 2018-19. The Report is based on India KLEMS database, 2019 in which data were provided for the years, 1908-81 to 2017-18.

• Data are provided for 27 industries into which the economy has been divided.

• The industrial classification adopted and the concepts used for the measurement of output, inputs and productivity are such that the dataset is comparable (by and large) to such datasets available for other countries. This facilitates international comparative studies on productivity.

• The details of construction of data series are available in the DATA MANUAL. This document is titled: Measuring Productivity at the Industry Level – The India KLEMS Database, Data Manual 2019.

• The main source of data for constructing the KLEMS dataset (containing output, inputs, factor shares and productivity series) is the National Accounts Statistics (NSO).

• Supplementary data are obtained mainly from: Annual Survey of Industries; NSS survey results in respect of employment-unemployment surveys and unorganized manufacturing (unincorporated enterprises) surveys; Input-output tables and supply-use tables.

• Deflators for materials and energy input are formed with the help of wholesale price indices; deflators for services input make use of implicit deflators for services in National Accounts Statistics.

• Gross output series and gross value added series are mostly formed from national accounts data. The gap between the two provides intermediate input.

• The materials, energy and services input series are constructed by splitting total intermediate inputs with the help of Input-Output and Supply-use tables.

• Labour input is measured as an index of labour service flows, taking into account changes in worker composition, in terms of levels of education.

• One of the widely used methodologies to capture changes in labour composition is given by Jorgenson, Gollop and Fraumeni (JGF) (1987). This is adopted for constructing the labour input series in the India KLEMS dataset.

• The growth rate in labour input may be decomposed into growth in labour quantity (employment) and labour quality (or composition effect).

• Capital input measure is based on a flow concept instead of a stock concept – similar to labour.

• The rationale is that assets are not directly used in the production process, rather the service delivered by these assets are the inputs to the production. The empirical measurement of capital services is complicated -- due to the difficulty in quantifying the flow of capital services delivered by a unit of capital. Therefore, the usual practice is to assume proportionality between capital services and capital stock at individual asset level (Jorgenson, 1963; Jorgenson and Griliches, 1967; Hulten, 1986). The service flow from a given stock differs from asset to asset. This is taken into account in the construction of capital input series in Indian KLEMS database.

• The growth rate in capital input may be decomposed into growth in capital stock and the effect of changes in composition. The breakup is provided in the dataset. See data manual for more details.

Annexure 1C: Publications of the Research Team Based on India KLEMS Database

Chapter 2: Sources and Methodology: India KLEMS Database

2.1 Introduction

The accurate measurement of inputs and output is of prime importance in any study on productivity. While preparing the India KLEMS database, much attention has been paid to the measurement of inputs and output, considering the theory and international practices of productivity measurement. The database uses several sources of information and appropriate measurement approaches, largely consistent with the prevailing KLEMS approach (see Jorgenson et al., 1987; OECD, 2001; Timmer et al., 2010) to build a consistent data series. The India KLEMS covers 27 industries that constitute the entire Indian economy and provides annual data on primary inputs (capital and labour), intermediate inputs (energy, materials, and services), and output (gross value added and gross output) since 1980.¹⁷ This chapter discusses the methodology adopted in measuring productivity in the India KLEMS, along with detailed discussions on the construction of each of the above-mentioned variables, and factor income shares, particularly labour income share in gross value added, and the sources of data used in constructing them.

The remaining part of the chapter is structured as follows. Section 2.2 discusses the approach to measuring productivity in the India KLEMS database, which is the standard neoclassical growth accounting approach. In Section 2.3, the approaches to construct each variable along with the sources of data are discussed in individual sub-sections. It also contains discussion on construction of the series on labour income share in gross value added. Section 2.4 lays out some outstanding issues that may be considered in the future, and the last section provides a summary of the chapter.

2.2 Approach to measuring productivity in India KLEMS: The growth accounting methodology

Measuring the proximate sources of output growth¹⁸, i.e., factor accumulation and productivity, at the individual industry level is an important feature of the India KLEMS database. To accomplish this, the approach we follow is the standard growth accounting (see the discussion below), using both value-added function and gross output function frameworks. This approach requires data on inputs and output at the industry level. Therefore, the subsequent parts of this chapter also discuss the construction of variables required for implementing the growth accounting methodology.

2.2.1 The KLEMS production function

The most common approach to measuring productivity and factor input contribution to output growth is to assume a value-added production function, i.e.

where V is real value added, L is labour input, K is capital input, and A is an indicator of overall efficiency in using inputs in the production process, or total factor productivity (TFP), all for industry i (the subscript i will disappear when 2.1 is considered for the aggregate economy) and year t. In eq. (2.1), industry value added is a function of industry-level capital and labour inputs. Nonetheless, while this function is quite appropriate at the aggregate economy level, it is less appropriate at the industry level for important reasons.

Value added is essentially a nominal concept, as there is no physical quantity of value added. Therefore, in practice, one derives real value added as a volume index obtained as a residual after subtracting intermediate inputs from the gross output. In the productivity measurement using value added function, this implies that the production function links technological change exclusively to real value added and primary inputs (OECD, 2001). Therefore, a value-added production function is valid only when intermediate inputs in the production process are separable from primary inputs. The conditions of separability are quite restrictive (see Goldman and Uzawa, 1964) and are mostly rejected in empirical tests by several previous studies.¹⁹ Therefore, industry-level productivity measures using value added functions are likely to produce a biased picture of technological change if technological change considers both primary and intermediate inputs in the production process (OECD, 2001). The production of one industry may crucially be linked to the output of another industry through the use of intermediate inputs. Due to such inter-industry reliance for intermediate inputs, it is unrealistic to assume industry-level technological change to happen solely based on the primary inputs. This makes the use of a gross output function to measure industry-level productivity more appropriate. Therefore, following the standard growth accounting approach followed in the KLEMS type of growth decomposition (see Timmer et al, 2010), India KLEMS assumes a gross output production function at the industry level. The gross output in any given industry j is assumed to be a function of factor inputs (capital and labour), intermediate inputs (energy, materials, and services), and total factor productivity, i.e.

where Y is real gross output, E, M, and S are real intermediate inputs, respectively energy, materials, and services inputs.

It is important to note that both inputs and output in equations (2.1 and 2.2) are aggregates of many components. For instance, capital may consist of several types of assets (e.g., computers, machinery, or buildings) and vintages, and labour may consist of different skill groups, age groups, or gender. Following the KLEMS practice globally, in the India KLEMS we take the heterogeneity in labour and capital inputs into account while accounting for factor input contribution to output growth (to be discussed in detail later in the chapter). In the case of intermediate inputs, as is evident from equation (2.2), we distinguish between three distinct types of intermediate inputs used in the production process, which are energy, materials, and services.

2.2.2 The decomposition of value added and output growth

Using the production functions specified in (2.1) and (2.2), and assuming additivity of nominal compensation of factor inputs and nominal values of intermediate inputs (i.e., the sum of input costs, such as wages to workers, rental values to capital, and intermediate input costs, is equal to nominal value of output for any given industry), constant returns to scale, and competitive factor market, we can decompose output and value-added growth into contributions from inputs and productivity. First, industry value added growth can be decomposed into contribution of factor inputs (capital and labour) and TFP as:

Equation (2.3) has been widely used in the literature, including in studies on productivity in Indian economy (e.g., Bosworth and Collins, 2008). It decomposes value-added growth, and therefore, does not account for the contribution of intermediate inputs to the production process. As mentioned earlier, to account for these inputs, we need to use a gross output function. The difference between gross output and gross value added for any given industry is the intermediate inputs. However, at the aggregate level, as the intermediate inputs used in one industry is the output of another industry, the inter-industry transactions cancel out, and a value- added function is appropriate. To measure productivity at industry level, as is the case with the India KLEMS, a more appropriate approach would be to use a gross output function (Jorgenson et al, 1987).²⁰ By expanding equation (2.3) to include the intermediate input, we can express the gross output-based growth accounting equation as:

where P^y_j,t is the gross output price, P^x_j,t is the intermediate input price for input X (X= energy, materials, and services), and X is the volume of the respective intermediate input. The income share of capital is obtained as a residual after subtracting the income share of all other factors from nominal gross output value.

It may be noted that the value added based TFP measure in equation (2.3) A^v_j,t and the gross output based TFP in equation (2.5), A^y_j,t have an accounting relationship (Hulten, 1978).²¹ The value added TFP is the product of gross output TFP and the ratio of gross output to gross value added, therefore,

A^v_j,t is indicative of improvements in the productivity of factor inputs, labour and capital, and is more interesting from an aggregate economy and welfare perspective. A^y_j,t, on the other hand, accounts for the efficiency of all inputs – labour, capital and intermediate inputs – and is important from a technological change point of view (see Das et al, 2016).

Note that in all the decompositions discussed above, we considered both labour and capital as single inputs. However, capital consists of various types of assets of several vintages (e.g., computers (of processor vintages such as Core i3, i5, i7, etc.), trucks, buildings, etc.). Similarly, workers consist of different age groups, educational or skill levels, and gender. While accounting for the contribution of these inputs to the production process, it is imperative to account for these differences, which will be discussed when we explain the construction of data and variables.

2.3. Data and Variables: Construction and Approaches

To practically implement the growth accounting methodology discussed in the previous section, we need data on inputs (capital, labour, and intermediate inputs), factor income shares and output for each individual industry. In the India KLEMS database, we have 27 industries (see Table 2.1 and Annexure Table 2A.1), which can broadly be categorised as agriculture, manufacturing (consisting of 13 industries), non-manufacturing industries (consisting of three industries), services (consisting of the market and non-market services, with six industries in the former, and four in the latter group). All variables are constructed, and growth accounting is performed individually for these 27 industries. Therefore, to construct aggregate indicators, such as productivity growth in the aggregate economy, total manufacturing sector, total service sector, etc., we use various aggregation procedures, such as the Tornqvist aggregation, Domar aggregation, or aggregate production function. These methods are discussed in individual chapters, wherever applicable.

The time period covered in this report is 1980-2017, and all the variables are constructed for this period. Since most analyses in this study are based on growth rates, the first year for which we compute growth rate is 1981, and therefore, the analysis period for all growth analysis is 1981-2017. For comparison purposes, in most chapters, this period is divided into three sub-periods – 1981-93, 1994-2002, and 2003-17. The first sub-period is largely consistent with the first set of liberalisation policies in India, whereas the second and third periods are the full liberalisation policy period since the 1990s and the fast-growing phase of the 2000s, during which several incremental policy reforms were introduced. In some chapters, the last period may also be divided further into the pre-global financial crisis period, 2003-07, and post-global financial crisis period, 2008-17.

The main source of India KLEMS data is the annual National Accounts Statistics (NAS), published by the National Statistical Office (NSO). However, we also rely on other sources such as the Input-Output (IO) tables, various rounds of National Sample Survey Office (NSSO) surveys on employment and unemployment and unorganised manufacturing (NSSO was merged with CSO to form NSO in 2019; however, at several places references is being made to NSSO surveys to impart greater clarity in the discussion), and Annual Survey of Industries (ASI) wherever required. This section discusses the sources of data and the approach to construct each of the variables available in the 2019 version of the India KLEMS database. For elaborated discussion on individual variable construction, interested readers may also refer to the India KLEMS data manual (Das et al., 2019). In particular, data on labour input is not available from the national accounts, which necessitated compiling consistent employment series from NSSO surveys.

2.3.1 Gross Output

The gross output of an industry is defined as the value of industry production using primary factors like labour, capital, and intermediate input purchased from other industries. We obtained gross output at current and constant prices (2011-2012 base) for all industries available from the NAS for years since 2011. For earlier years, although one could take the data from the NAS back-series, the 2011-12 base back-series did not provide gross output estimates by industries except for agriculture and allied sector, mining and quarrying, and construction. Therefore, for the remaining industries, the historical series on gross output for the period 1980-2011 period is estimated by multiplying industry value added data from NAS 2011-12 back series with the respective output/value added (see the discussion of value added below) ratio obtained from back series with 2004-05 base.

These alternations, however, were not sufficient to obtain gross output series for all the 27 industries in the KLEMS database, as the number of NAS industries was smaller than the KLEMS requirement. The NAS estimates of gross output for a few industry groups are at a higher aggregate level than the India KLEMS industry groups, requiring splitting of the aggregates. In such cases, NAS estimates of output for the manufacturing sectors have been split using inter-industry gross output share from the ASI (for the registered manufacturing segments)²² and the NSSO rounds (for the unregistered segments). Moreover, for unregistered manufacturing, gross output data is available in NAS from 2004-05 onwards only. For the missing years, the output/value added ratio from NSSO unorganized survey rounds has been applied to measures of value added from national accounts estimates of unorganised manufacturing industries. Since these ratios are available only for the NSSO survey years, we linearly interpolate the ratios for intervening years.

For service sectors, national accounts do not provide any estimates of gross output prior to 2011-12. We construct estimates of output for the service sector with the help of IO tables. We take the benchmark year gross output to value added ratio for the relevant service sector from IO tables for years 1978-79, 1983-84, 1989-90, 1993-94, 1998-99, 2003-04, and 2007-08. These ratios are linearly interpolated for intervening years and applied to the gross value-added series obtained from national accounts to derive the output estimates consistent with NAS at current prices.

To create an output series at constant price, we use the NAS estimates directly wherever the NAS sector classification matches the KLEMS classifications. For industries where direct NAS estimates are not available, the nominal estimates are deflated with value added deflators for the respective industries derived from the NAS. For industries for which we create gross output estimates using ASI and NSSO industry distributions, we first deflate the nominal series using weighted wholesale price indices (WPI) for the relevant industry groups, obtained from the Office of the Economic Advisor. The deflated industry series are then normalized i.e., proportionally adjusted to match with the published NAS aggregate real output of those sectors.

2.3.2 Gross Value Added

The gross value added of an industry is the value of output less the value of all intermediate inputs used in the production process (see the next section for a discussion on the measurement of intermediate inputs). When measured at factor cost, it is the sum of compensation paid to labour and capital inputs (Berndt and Hesse, 1986). NAS provides estimates of gross value added by industry of use at current and constant (2011-12) prices since 1950 at both aggregate and disaggregate levels in its 2011-12 base back series. We use this series until 2011-12, and for years after 2011, we extend the series using the most recent national accounts data. Specifically, we use the 2011-17 data available from the 2019 annual NAS.

The KLEMS industries for which data are directly taken from NAS are agriculture & allied sector, mining & quarrying, electricity, gas & water supply, construction, trade, hotels & restaurants, transportation & storage, communication, financial services, public administration & defense, education, and health services. For the KLEMS industry business services, we combine the value-added data for three NAS sectors, which are ‘information & computer related services’, ‘professional, scientific & technical services including R&D’ and ‘administrative & support service activities & other professional services’. Similarly, we combine the NAS sectors ‘real estate’, ‘ownership of dwellings’ and ‘community & social services’ (excluding ‘education & health’) to obtain the KLEMS industry ‘other services’. For the manufacturing sector, the NAS provides separate estimates on value added in the organised (registered) and unorganised (unregistered) segments until 2011-12. These aggregate estimates are split using industry shares from the ASI for the organised segment and the NSSO unorganised manufacturing surveys for the unorganised segment. For the unregistered manufacturing sector, we have used industry distributions from six rounds of NSSO unorganised surveys, which are 40th round (1984-1985), 45th round (1989-90), 51st round (1994-95), 56th round (2000-01), 62nd round (2005-06), 67th round (2010-11) and 73rd round (2015-16).

However, NAS has not been distinguishing between organised and unorganised segments of the manufacturing industries since the introduction of the 2011-12 series. Instead, it provides a split between the corporate (consisting of both public sector and private corporations) and household sectors. As in registered manufacturing, we use the ASI industry distribution to split the value-added data in the corporate sector and the NSSO unorganised manufacturing industry distributions for splitting the household sector value added. The sum of registered and unregistered for the years prior to 2011-12 and the sum of household and corporate sector for years since 2011-12 are considered as value added data for the manufacturing industries in the India KLEMS database. It should be noted that the NAS series for the corporate segment of manufacturing is based on the Ministry of Corporate Affairs’ MCA-21 database, which follows an enterprise approach compared to the establishment approach followed in the ASI. This might lead to a slight inconsistency between the two data sets. Despite this limitation regarding using ASI data for splitting the MCA-21-based NAS manufacturing estimates,²³ we do not yet have a better alternative solution.

It is to be noted that our estimates of value added are adjusted for financial intermediation services indirectly measured (FISIM). The value of such services forms a part of the income originating in the banking and insurance sector and, as such, is deducted from the GVA.

For deflating the current price estimates of value added in the manufacturing sector, we use wholesale price deflators, while for other sectors they are deflated using implicit deflators from the national accounts.

2.3.3 Intermediate inputs

Our approach to measuring a time series of intermediate inputs (energy, materials, and services inputs) follows the methodology developed by Jorgenson et al. (1987) and extended by Jorgenson (1990). It is similar to the one elucidated by Timmer et al. (2010, Chapter 3) in the context of EU KLEMS (also see Jorgenson et al., 2005, Chapter 4). The cornerstone of this approach is a time series of input-output (IO) tables that gives the flows of all commodities in the economy and payments to primary factors (Jorgenson et al., 2005; Timmer et al., 2010). Every commodity is accounted for in the IO transaction tables, regardless of whether produced by a domestic source or imported, and every use is noted, whether purchased by industry or by a final demand element.

We first match the 27 KLEMS industries with the industry classifications in various benchmark IO tables available for India. Then, for each benchmark year, the share of the three intermediate input groups—energy, materials, and services inputs—in total intermediate inputs are calculated. These proportions are linearly interpolated for intervening years to create a time series of distribution of total intermediate inputs into the three input categories. This interpolation assumes that the input-output coefficients for each industry change progressively between two benchmark years. This approach has been adopted for all the KLEMS industries except for public administration and defense, for which the data on intermediate input consumption was not available in the IO tables. For this industry, we rely on information from NAS and system of national accounts (SNA) tables. Specifically, we used the ‘Net Purchase of Commodities and Services’ of the Administrative Department as total intermediate input. Then using the information on the distribution of ‘Government Final Consumption Expenditure’ between different categories, we distribute the total intermediate input into materials, energy, and services inputs. Once we have the time series of the distribution of the three inputs for all the 27 industries, we apply those to the nominal values of total intermediate input from the national accounts. The national account consistent total intermediate inputs are constructed as the difference between gross output and gross value added, discussed in the previous sections. We obtain the intermediate deliveries in each sector by subtracting the nominal gross value added from the nominal gross output. Thus, combining the IO and NAS we construct a consistent time series of intermediate inputs—energy, materials, and services—in current prices.

To construct the constant price series, the nominal values of intermediate inputs are deflated using a weighted deflator for each of the three intermediate inputs.²⁴ For constructing weighted deflators, we use the WPI for the industries that produce the input components of intermediate input categories energy, and materials. These wholesale price deflators are combined with relevant industry weights from input-output transaction tables. We allow the weights to change over time using varying weights for three different IO benchmarks, 1989, 1998, and 2007. The weights from the 1989 table have been used for the period 1980 to 1993, the weights from the 1998 table have been used for the period 1993 to 2003, and the weights from the 2007 table have been used for the price series for the period since 2003. Once the three individual series were formed, these were spliced to construct one single intermediate input price series used to deflate the nominal values of energy, and materials. This approach, however, was not feasible for the services inputs, as the WPI is not available for services. Therefore, we use implicit GDP deflators derived from the NAS to deflate service inputs. For individual services input, implicit deflators were formed and then these were combined using industry specific weights taken from three IO benchmarks, 1989, 1998, and 2007.

2.3.4 Labour Input

The contribution of labour input to production comes from the pure number of workers or hours worked—a quantity measure and the accumulation of worker skills—a quality measure. India KLEMS makes a distinction between the quantity and quality contributions of labour input into the production process while measuring productivity. Although the number of hours is an appropriate measure of labour quantity for productivity analysis, such data is hard to obtain for developing countries, including India. The data sources on employment in India do not provide information on exact number of hours worked by a person and measure it only as half-day and full-day. Therefore, implicitly assuming proportionality between the number of workers and the number of hours, researchers often use the number of workers, or employment, as a measure of labour quantity. This approach is followed in the India KLEMS. This section discusses the sources of employment data used in the India KLEMS and the approach followed to measure employment for the 27 KLEMS industries for 1980-2017. The section also discusses the approach we follow to measure labour quality.

2.3.4.1 Data Sources

In India, the main sources of data on the number of workers and the distribution of employment by economic activities can be categorised into three. The first set of sources cover all sectors of the economy, which are decennial Population Census (Census) and the quinquennial Employment & Unemployment Surveys (EUS) of the NSSO²⁵. The second set covers the organized segment of the economy only, which are the ASI, and the employment market information program, by the Directorate General of Employment & Training (DGE&T) of the Ministry of Labour (Labour Bureau).²⁶ The last is NSSO’s unregistered sector surveys, which provide data on employment in the unorganised manufacturing industries for selected rounds. The NSSO unregistered surveys also cover unincorporated non-agricultural enterprises (excluding construction) in two of the recent rounds (67th (2010-11) and 73rd (2015-16)). While the surveys by Census, EUS, and the Labour Bureau are household surveys, the ones by ASI and NSSO for the unregistered sector are establishment-based surveys.

The India KLEMS relies primarily on the quinquennial EUS because it covers the entire economy and uses a consistent definition of the activity status of the person since the beginning of the quinquennial rounds of the surveys. While the Population Census also covers the entire economy, it is often criticised for underestimating workforce participation rates (WPRs), leading to lower employment estimates (Sivasubramonian, 2004). Moreover, the frequency of census estimates is once in a decade, while that of the EUS is five years, providing a more accurate picture of trends in employment. Although the EUS is often assumed to be understating population (Sivasubramonian, 2004; Ghose, 2016), it provides consistent WPRs. The estimates obtained from EUS are widely used for planning and policy purposes and by academic researchers. We use the WPR estimates from the quinquennial rounds of EUS, and population estimates²⁷ to derive employment data for India KLEMS.

During 1972-2011, the NSSO conducted nine rounds of EUS, following a comparable approach over various rounds.²⁸ In addition, from 1989-90 to 2004-05, the NSSO had also conducted annual EUS based on a small sample, often called the 'thin rounds'. However, these thin round results are considered less reliable, as their reference period is shorter compared with the quinquennial surveys, thus overlooking the patterns of seasonal variations, and their sample size is also relatively small. More recently, the National Statistical Office (NSO) started conducting periodic labour force surveys (PLFS), which are aimed to bring out high frequency data on employment. The results of the PLFS are not exactly comparable with the earlier quinquennial surveys, and hence should be used with caution.

The EUS provides several concepts of employment, depending upon the reference period it uses (one year, one week, and each day of the reference week) to determine the activity status. We use the usual and subsidiary status (UPSS) definition of employment, which is widely considered a good measure of employment (Visaria, 1996; Bosworth et al., 2007; Sundaram, 2008; Rangarajan, 2009), and is also adopted by the NAS in its labour input method.²⁹ Six quinquennial rounds of EUS are used to create the KLEMS employment series for the period 1980-2017. These are the 38th round (1983), 43rd round (1987-88), 50th round (1993-94), 55th round (1999-2000), 61st round (2004-05), and 68th round (2011-12) rounds and the PLFS, 2017-18.³⁰

2.3.4.2 Measuring persons employed at the Industry Level

In all the Employment & Unemployment Surveys, information about the person’s principal and subsidiary economic activity is collected and an industry code is assigned for the principal and the subsidiary industry in which the person is employed. NSSO used National Industry Classification (NIC)-1970 for classification of persons employed by industry in its 38th and 43rd rounds of EUS. These industrial classifications changed in the subsequent rounds. While NIC 1987 is followed in the 50th round, NIC 1998 is followed in the 55th and 61st rounds, and NIC 2008 is used in the 68th round and the PLFS. Therefore, to create a consistent series over time, the starting point was to create a concordance between the 27 KLEMS sectors, and NIC-1970, NIC-1987, NIC-1998 and NIC-2008. These concordances were formed and are available in India KLEMS data manual (Das et al., 2019).

Employment numbers for each industry has been computed using the following steps:

1. We first obtain the WPRs by UPSS for four groups: Rural Male (RM), Rural Female (RF), Urban Male (UM), and Urban Female (UF) from unit level data of each round of EUS. These are then applied to the corresponding period’s population estimates from census³¹ to estimate the number of persons employed in the four segments. These four segments are added to create the total employment.³²

2. The 27 KLEMS industry distribution of employment is obtained from unit level data of each survey round for each group, rural male, rural female, urban male, and urban female.

3. The 27-industry distribution of employment from EUS is then applied to the total number of persons employed in order to obtain number of employees, H_j,s,g, in each industry by gender and rural/urban areas, where H_j,s,g is number of employed persons for any given industry ‘j’, sector ‘s’ (s=1 for rural and 2 for urban sectors), and gender ‘g’ (g=1 for male and 2 for female).

4. The estimates of employment for the intervening years between the major EUS rounds have been obtained by simple interpolation of the four segments—rural male, rural females, urban male and the urban females, ignoring the outliers³³ where the two closest rounds have then been used for interpolation.³⁴ While, the estimates from the major rounds are centred around the mid-year of the survey rounds, i.e., July for 38th round and 1st Jan for the other rounds; the interpolation has been centred around 1st October so as to coincide with the mid-year of the national income estimates which refer to the Indian financial year from April to March. So, the base is shifted to October while taking the forecasts.

5. Total persons employed for any given industry in a year are then obtained as the sum of the H_j,s,g over the two gender and two sectors, Σ_sΣ_gH_j,s,g.

6. Extrapolate the employment data backward for 1980-81 to 1982-83 period using the interpolation of the broad industrial classification of 32nd round (1977-78) and 38thround (1983-84). The broad sector data for 32nd round is used as a bench- mark, because the 27-industry classification cannot be obtained for the 32nd round due to non-availability of unit level data.

Through steps 1 to 6, we obtain a time series of employment (total persons employed) in each industry in the India KLEMS for the period 1980-2017. However, the India KLEMS also accounts for the heterogeneity of worker groups, under the assumption that the skill differences of employees will have differing impact on productivity. To account for such skill differences in the workforce, we construct a labour quality index, the approach to which is described in the next sub-section.

2.3.4.3 Measuring Labour Quality Index

The measurement of labour quality is essentially an attempt to distinguish one labour type from the other taking into account the embodied human capital in each person. The source of human capital could be through investment in education, experience, and training. The KLEMS approach to estimate labour quality distinguishes workers according to their educational attainment and experience and weights each category's growth rate using their marginal productivities, approximated by their wage shares. Nevertheless, many limitations of India’s employment statistics, especially the non-availability of information on wages/ earnings of different category of workers, make it difficult to follow this approach in a pertinent way. Regardless, we construct a labour quality index for India KLEMS industries, computed only based on education overlooking other dimensions such as experience.

The quality of labour force is of considerable importance in the context of productivity measurement, and one of the widely used methodologies to capture changes in labour quality is given by Jorgenson et al., (1987).

First, define the growth rate of aggregate labour input in industry İ (ΔIn L_j,t) as a Törnqvist index of persons worked by individual labour types ‘l’ as follows:³⁵

The first term on the right-hand side indicates the change in labour input and the second term indicates the change in total persons employed in industry ‘İ’. It can easily be seen that if proportions of each labour type in the labour force change, this will have an impact on the growth of labour input beyond any change in total persons worked. The index of aggregate labour quality thus measures the changes in the educational composition of work force of the economy.

The labour quality growth rates have been computed using equation (2.11) for each of the 27 industries using five education categories namely: (i) up to primary, (ii) primary, (iii) middle, (iv) secondary & higher secondary, and (v) above higher secondary. The estimates are generated using five EUS rounds, 38th, 50th, 55th, 61st, and 68th and the PLFS rounds of NSO. They are then indexed to 1983 (38th round) as the base. For years between the EUS rounds, the index has been interpolated, and for 1980, 1981 and 1982 they are extrapolated. Finally, the entire series of labour quality index is rebased to 1980, the first year of the India KLEMS database.

In constructing the basic data used in implementing equation 2.11, the following steps have been performed:

1. Compute the distribution of persons employed by the five educational groups for all the 27 industries for the selected major rounds of EUS³⁷ and PLFS.

2. Apply these proportions to the number of employed persons in different industries in the major rounds and PLFS to obtain the distribution of persons by education groups.

3. Estimate the earnings data from NSO which relates to mainly the regular and casual persons employed. It may however be mentioned that even for these two groups, for a large number of persons employed, the wages are either missing or given as zero. While the earnings of regular and casual workers are estimated directly from the survey data, the earnings of self-employed persons have been estimated econometrically and a Mincer wage equation is used to estimate the earnings for self-employed persons.³⁸ The Mincer wage equation is estimated after correcting for sample selection bias by using the Heckman’s³⁹ two step procedure (Heckman, 1976). The Mincer function has been applied to the earnings of casual and regular employees where the earnings have been regressed on dummy variables representing age groups, gender, education, location (rural or urban), marital status, social exclusion, and industry. The identification factors used in the first stage are age, gender, marital status, and type of household. The corresponding earnings of the self-employed are obtained as the predicted value with similar traits.⁴⁰ The average wages per day are then computed for persons employed of different type of employment, i.e., self-employed, regular, and casual combined together; whose wages are available.

4. Estimate labour quality index, following Jorgenson et al (1987)⁴¹ using the equations explained above, and the data compiled in steps 1-3.

2.3.5 Capital Input

This section outlines the methodology employed to estimate capital input for the 27 industries in the India KLEMS database used in various chapters of this report. As in the case of labour, where we distinguished between various workers, capital consists of different asset types of multiple vintages, each with a distinct productivity profile. All the productivity calculations in India KLEMS and this report use capital services as the input to production, which is estimated using the approach developed by Jorgenson and Griliches (1967), and outlined in Jorgenson et al., (1987). The main difference between conventional measures of capital stock and the measures of capital services used in India KLEMS is that the latter accounts for the differences in productivity between different asset types and vintages. While the capital stock measures, when accounted for appropriate depreciation profiles, only consider the heterogeneity of different vintages of capital, they do not account for asset heterogeneity. For instance, aggregate capital stock measures would assume that a computer's productivity is the same as that of a car. However, proper measures of capital stock would account for the decline in the efficiency of both computers and cars over their lifetime.

To measure capital services, using the Jorgenson et al (1987) approach, we need estimates of capital stock for detailed asset types and the shares of each of these assets in total capital compensation. Using the Törnqvist approximation to the continuous Divisia index under the assumption of instantaneous adjustability of capital, aggregate capital services growth rate is derived as a weighted growth rate of individual capital assets, the weights being the compensation shares of each asset, i.e.,

where P^K_k,j,t is the rental price or user cost of capital asset type k in year t. Therefore, the numerator, which is the product of user cost of asset k and the capital stock of asset k, is the capital compensation received by asset k. The numerator is the sum of compensation received by all assets, or the total capital compensation, and therefore v_k,t is the share of asset k in total capital compensation. Under the neoclassical assumption of price-marginal product equality, it effectively incorporates the productivity or qualitative differences in the contribution of various asset types to aggregate capital services, as the composition of aggregate capital stock in any industry changes (see Jorgenson, 2001). For instance, as the marginal productivity of information and communication technology (ICT) capital is higher than that of other assets, a change in the composition of capital towards ICT capital will result in higher capital services, which will be captured by a larger value of the v for ICT assets.

It is evident from (2.12) that two important components of capital service measure are the asset wise capital stock, S_k,t and the service price (rental price) of capital assets, P^K_k,t. Assuming a geometric depreciation rate, δ_k which is constant over time, but different for each asset type, capital stock in asset k in and year t can be constructed using standard Perpetual Inventory Method (PIM) as:

where, I_k,j,t is the real investment in asset type k.

The rental price of capital P^K_k,j,t reflects the price at which the investor is indifferent between buying and renting the capital good for a one-year lease in the rental market. In the absence of taxation, the rental price equation can be derived as the sum of the nominal rate of return, the nominal cost of depreciation, less capital gain (see Jorgenson and Griliches, 1967; and Christensen and Jorgenson, 1969):⁴²

where P^I_k,j,t is the investment price (or acquisition price) of asset k in year t, and i_t is the nominal rate of return. The first component is the rate of return component, which captures the opportunity cost of purchasing a capital asset. The second component is the depreciation component, and the last component, which is the difference between investment prices in year t and t-1, captures the capital gain. Thus, this formula shows that the rental price is determined by the nominal rate of return, economic depreciation costs, and asset-specific capital gains. The rate of return i in equation (2.16) can be measured either as an internal or ex-post rate of return – using the information on realised capital compensation—or an external or ex-ante rate of return—such as the prevailing long-term interest rate in the economy. There is continuing debate on which rate is more appropriate for productivity analysis. Empirical evidence suggests that the choice of a particular rate may impact final estimates of capital services and productivity (see Oulton, 2007; Erumban, 2008a). In India KLEMS, following the arguments that investors make their investment decisions considering the prevailing interest rates in the economy, we use an external rate of return (see more details in part (iv) sub-section 2.3.5.1).

Since our measure of capital input takes account of asset heterogeneity,⁴³ it was essential to obtain investment data by asset type to estimate equation (2.15). We distinguish between three different asset types — construction, transport equipment, and machinery (includes ICT and non-ICT machinery). We exploit multiple sources of information for the construction of our database on capital services. This includes NAS that provide information on broad sectors of the economy, the ASI covering the organised manufacturing sector, and the NSSO rounds for unorganised manufacturing. Even though we use multiple sources of data, our final estimates are entirely consistent with the aggregate data obtained from the NAS.

2.3.5.1 Data Sources

i) Asset-wise investment for broad sectors of the economy

Industry-level estimates of capital input require detailed asset-by-industry investment matrices. The NAS provides information on aggregate capital formation by the industry of use for nine broad sectors, which, nevertheless, was not sufficient for our purpose. Therefore, we have collected more detailed data on assets and industries from the NSO.⁴⁴ This is the data underlying the published aggregate gross fixed capital formation (GFCF) by the broad industry groups, separately for public and private sectors. For those sectors for which the investment matrices were not available from NSO, we gather information from other sources (e.g., ASI for organised manufacturing and NSSO surveys for unorganised manufacturing) and benchmark it to the aggregate investment series from the National Accounts. The data used in the current version of the India KLEMS is based on the revised NAS with 2011-12 base and the detailed asset-industry data for this benchmark is available only for the period 2011-12 to 2015-16. Therefore, we extrapolate the series using growth rates from the previous version of the data for earlier years. However, there were slight differences between the industry groups available in the current release of the NAS data and the previous ones (see Annexure Table 2A.4), which required some matching of sectors before combining the two series.

Since we did not obtain detailed asset/industry-wise data on capital formation from the NSO, after 2015-16, we relied solely on publicly available data from the National Accounts for recent years. The publicly available data contains more detailed information now than what it used to be in the past. Therefore, the reliance on the NAS public data did not create too many challenges. However, there were two outstanding issues. First, the asset transport equipment was not separately available in this data. Therefore, we assumed the share of transport equipment in the total machinery and equipment group in 2015-16 to remain constant to create nominal investment in transport equipment since 2016-17. The nominal investment was deflated using the GFCF deflator for transport equipment. Second, the published NAS does not provide us a distinction between organised and unorganised manufacturing, which was available in the previous versions. Therefore, to continue the older series and exploit the detailed asset and industry information from the ASI and NSSO unorganised manufacturing survey (see below), we have proxied the organised manufacturing by corporate and public sector manufacturing and unorganised manufacturing by household manufacturing.

ii) Asset-wise investment for non-NAS sectors

NAS provides data only for nine broad sectors, while we have 27 industries, which necessitated further splitting of some of the NAS sectors. This includes aggregate manufacturing (registered and unregistered separately) with 13 sub-sectors; other services into four sub-sectors; and real estate activities and business services into two sub-sectors. The manufacturing sector investment data was disaggregated into 13 KLEMS subsectors using industry and asset distributions calculated from ASI (for the organised sector) and NSSO (for the unorganised sector) data. Investment series in the service sector has been split into sub-sector using two alternative approaches — value added shares and capital/labour ratio in the higher aggregate industry. However, the final data used are based on value added shares, as a sensitivity analysis did not show a significant difference between the two.

a) Registered Manufacturing: In order to split the aggregate capital formation in the organised manufacturing sector into 13 study sectors, we use the ASI. Following ASI's definition of GFCF, i.e., actual additions (newly purchased, second hand and own construction) minus deductions plus depreciation adjustment for discarded assets during the year, we measure GFCF from ASI micro-level data for each of the 13-manufacturing industries.

The detailed yearly volumes beginning 1964-65 were used to derive the GFCF by asset type directly. For the years 1964-1978, the relevant data are obtained from published detailed volumes. For the period since 1983-84, ASI has generated detailed tables from Block C of the ASI schedule that contains data on fixed assets. Data for missing years are interpolated using the changes in investment using the book value method (Goldar, 1986).

Once the investment in each of these assets and industries is generated using ASI data, we apply this industry-asset distribution⁴⁵ to the published aggregate NAS GFCF series for the organised manufacturing sector. It may also be noted that from 1960-61 to 1971-72, ASI data are for the census sector, and from 1973-74 onwards, they are for the factory sector. To make these two series comparable over the years, we convert the data prior to 1972 to the factory sector using the factory/census ratio in 1973. Thus, after these adjustments, we obtain investment data for 13 manufacturing industries (according to KLEMS classification) by asset types, consistent with the NAS aggregate.

b) Unregistered Manufacturing: The data required for creating the gross investment series for the 13 sectors of the unorganized manufacturing sector are obtained from various rounds of NSSO surveys on unorganized manufacturing. We use six rounds of NSSO surveys that cover the period 1989-2016. These are 45th round (1989-90), 51st round (1994-95), 56th round (2000-01), 62nd round (2005-06), 67th round (2010-11), and 73rd round (2015-16). Unit level data has been aggregated to 13 industries using the appropriate concordance tables. NSSO provides net addition to owned assets during the reference year within the block of fixed assets, and we use this as a measure of our investment. The investment series constructed for six rounds were interpolated to obtain the annual time series of unorganized sector gross fixed capital formation by asset type. As in the registered sector's case, once the investment by asset types across industries is constructed, the asset-industry distribution is applied to the published NAS aggregate GFCF in unregistered manufacturing to obtain NAS consistent GFCF by asset type and industries.

iii) Investment Prices by Asset Types

In order to compute asset-wise capital stock using PIM (equation 2.15) and rental price (equation 2.16), we require asset-wise investment price deflators. Since NSO has provided us with investment data by industries and assets both in current and constant prices, we could derive the price deflators with base 2011-12. For years before 2011, prices were spliced using 2004-05 base investment deflators. These deflators are directly used for all the three asset categories we have. Since there was no separate asset-wise data available for transport equipment since 2016-17, the investment deflator for transport equipment from 2016-17 was extrapolated using the trend in the investment price of total machinery & equipment obtained from NAS. For industries for which NSO data were not available, the deflator from the nearest aggregate sector is applied (e.g., for all the 13 registered manufacturing industries, we use the aggregate registered manufacturing deflator from the NAS).

iv) Initial Stock, Depreciation Rates, and Rate of Return

As is evident from equations 2.12 to 2.16, our capital input estimates require time-series data on asset-wise capital stock. Capital stock has been constructed using the PIM, where the capital stock (S) is defined as a weighted sum of past investments with weights given by the relative efficiencies of capital goods at different ages, which requires data on current investment by asset types, investment prices by asset types and depreciation rate. Also, for the practical implementation of PIM to estimate asset-wise capital stock, we require an estimate of the initial benchmark stock. NAS provides estimates of net capital stock since 1950 for all the broad sectors in its Statement 17: Net Fixed Capital Stock by industry of use. We take the NAS estimate of real net capital stock in 1950 (in 2011-12 prices) as our benchmark stock for all non-manufacturing sectors, and for manufacturing sectors, the same is taken for the year 1964.⁴⁶ However, since the NAS estimate is available only for broad sectors and for aggregate capital, we use our industry-asset distribution of GFCF for the benchmark year in order to create net fixed capital stock estimates by asset type for all 27 industries.

NAS also provides detailed tables on the assumed life of assets used for computing capital stock, for private units, administrative units as well as departmental and non-departmental units by asset types.⁴⁷ We use these lifetime estimates to derive appropriate depreciation rates for non-ICT assets, using a double declining balance rate. Following the NAS, we assume 80 years of lifetime for buildings, 20 years for transport equipment, and 25 years for machinery and equipment (see Appendix 26.2; NSO, 2007). The corresponding depreciation rates are 2.5 percent for buildings and construction, 10 percent for transport equipment, and 8 percent for machinery.⁴⁸ Subsequently, we build our capital stock series by asset types for all the 27 industries using our GFCF series from 1950 (1964) onwards for the non-manufacturing (manufacturing) sectors.

Our measure of capital input is arrived at using equation (2.12), for which we also require estimates of rental prices (see equation 2.16). Assuming that the flow of capital services is proportional to the capital stock at the individual asset level, aggregate capital flows can be obtained using a translog quantity index by weighting growth in the stock of each asset by the average shares of each asset in the value of capital compensation, as in (2.12). The rate of return (i) in equation (2.16) represents the opportunity cost of capital. As stated earlier, in the India KLEMS, it is measured as an external rate of return, proxied by the average of return on government securities and prime lending rate obtained from the Reserve Bank of India. Therefore, we use a real rate, which is net of capital gain. Hence, the capital gain component in equation (2.16) is excluded while estimating rental price using the external rate of return, obtaining

where r is the real rate of return, nominal interest rate adjusted for consumer price inflation rate. The consumer price indices (CPI) are obtained from IMF and World Bank.

2.3.6 Labour Income Share

In the previous sections, we discussed the measurement of output, primary inputs, and intermediate inputs. However, as is evident from equations 2.3 and 2.5, we also require data on factor income shares to measure productivity using growth accounting. While the income shares of intermediate inputs can be directly measured as the ratio of nominal values of energy, materials, and services inputs to nominal values of gross output, the income share of capital and labour needs to be calculated separately. This is because the income of capital is not directly observable, and the income of labour involves the imputation of self-employed workers' wages, which is usually not available in published official statistics. As we discussed earlier, once we have the estimate of labour income share, the income share of capital can be computed as a residual, under the assumption of constant returns to scale. In this section, we discuss the measurement of labour income share in value added.

The basic data sources used for the computation of labour income share are the NAS (covering the entire economy), the ASI (covering organised manufacturing industries) and unit level data of NSSO survey of unorganised manufacturing enterprises. The NAS publishes data on net domestic product (NDP) comprising of compensation of employees (CE), operating surplus (OS) and mixed income (MI) for major industry groups—agriculture, forestry, logging & fishing; mining & quarrying; manufacturing—registered & unregistered; electricity, gas & water supply; construction; trade; hotels & restaurants; transport & storage; communication; banking & insurance; real estate, ownership of dwelling & professional services; public administration; and other services. As in the case of GFCF (see section 2.3.5.1), NAS does not provide estimates of factor incomes separately for registered and unregistered manufacturing since 2011-12. Instead, it provides a split between corporate sector and household sector. The MI component consists of income accruing to capital and the wage income of the self-employed workers and is an important part of labour income in the unorganised segment of the economy. Therefore, to accurately calculate total labour income in any given industry (consisting of both registered and unregistered segments), we need to combine the CE (which is the total income earned by workers other than self-employed) and the portion of MI that can be attributed to wages of self-employed workers.

To compute labour income share, we first collect the data on CE and MI for all the NAS industry groups mentioned earlier. However, only nine industry groups—(i) agriculture, forestry, logging & fishing, (ii) mining & quarrying, (iii) electricity, gas, and water supply and (iv) construction, (v) trade, (vi) hotels and restaurants, (vii) post and telecommunication, (viii) financial services and (ix) public administration and defence—are directly comparable to India KLEMS industry classification. For the other sectors, we had to split the NAS aggregates using industry distributions from other sources.

We split the NAS estimates of factor incomes for registered and corporate sector manufacturing into 13 KLEMS manufacturing industries using the distribution of ASI data on total emoluments. For the unorganised manufacturing and household sector, we use the distribution of wage payments (to hired workers) in different industries from the NSO unorganised surveys. Similarly, the MI in the unorganised and household manufacturing sector are split across industries using the distribution of mixed income in various industries. For this purpose, the mixed income component has been obtained after subtracting the wage payments (to hired workers) from the total value added in any given industry.

Unlike the ASI data for organised manufacturing, the data for unorganised manufacturing enterprises are available for only the NSO survey years. The distributions for the non-survey years have been linearly interpolated, except for years before 1989 and years after 2016. For years before 1989, we use a constant distribution of 1989-90, and for years after 2016, we use the distribution of 2015-16, as the surveys used in this study starts in 1989-90 (45th round) and ends in 2015-16 (73rd round).

For the remaining service industries, the aggregate estimates of relevant industry groups from NAS are distributed across the KLEMS industries using their gross value-added share. For instance, the estimate of factor incomes for ‘other services’ in NAS has been split into KLEMS industries (i) education, (ii) health & social work, and (iii) other services. Once we construct the series of CE and ME for all the KLEMS industries, we measure the labour income share in value added as (also see equation 2.4):

An important step in the implementation of equation (2.18) is the estimation of η_j. We estimate η_j using industry-wise data on the number of self-employed workers, wage rate of self-employed and the number of days of work per week from unit records of NSO employment-unemployment survey.⁴⁹ We use two approaches to get an estimate of η_j, and the final series on labour income shares for each industry are arrived at using an average of the two estimates of η_j.

In the first approach, we estimate labour income of self-employed workers for each industry for the NSO survey years, 1983-84, 1993-94, 1999-00, 2004-05, 2009-10, 2011-12 and the PLFS, 2017-18 using the number of self-employed, the number of days of work per week, and the estimated wage rate of self-employed using a Mincer equation (see section 2.3.3). The estimated self-employed income is divided by the mixed income derived from NAS to get an estimate of η. We do not estimate the η for all the KLEMS industries due to data reliability issues. Rather, in some cases, the estimates are made for a broad industry group, and these group estimates are applied to all the constituent industries.⁵⁰

In the second approach, we directly estimate the ratio of income of the self-employed workers to the income of regular and casual workers. This ratio is multiplied with the NAS estimate of CE to derive the labour income component of the MI.⁵¹ η in this case is obtained as a ratio of the labour income component of MI and total MI. If the estimated labour component of MI exceeds the NAS reported estimate of MI, which was the case in a number of industries/years, the estimate of η has been taken as unity.

Examining the estimates obtained by the first approach, it was found that the estimated labour income shares out of mixed income varied significantly among the estimates for the eight years for which the ratio in question has been estimated. Therefore, the average value of η has been computed for each industry or industry group. This has been done separately for the estimates based on approach-1 and approach-2. The value of η used in equation (2.18) is an average of the average estimates from approach 1 and 2.

2.4 Data Challenges & Outstanding Issues

2.4.1 Gross Output and Intermediate Inputs

For years prior to 2011, NAS do not provide estimates of gross output of services sector industries and hence we have to rely on IO tables which are available at an interval of about 5 years. This necessitates interpolation and assumption of linearly interpolated (steadily changing at a constant rate) shares of value added in output (for each industry in the intervening years) for measuring output of services sector industries. Similar interpolation was necessary for constructing intermediate inputs, which are broken down into energy, materials and services using the IO tables.

The weaknesses of the estimates of gross output of services industries and the estimated break up of intermediate inputs, noted above, can be eased to a large extent by constructing a timeseries of IO tables using the available IO tables and applying RAS technique. Such tables have been made and are available for India in the World Input-Output database (WIOD). The number of sectors into which the economy is divided in the WIOD tables is much smaller than the number of sectors available in the IO tables, which have been used for constructing India KLEMS database. Constructing detailed IO tables for India for each year would involve a great deal of effort, but this is needed to improve estimates of gross output of services industries and estimates of energy, materials, and services inputs.

It needs to be pointed out that the estimated time series on intermediate inputs at constant price will not be consistent with the intermediate input at constant price implied by NAS data, i.e., the deflated value of intermediate input would not match the gap between value added and gross output at constant price from NAS. This is so because NAS uses a single deflation method to estimate Gross Value Added and Gross Output at constant price.⁵²

2.4.2 Labour Input

The major data challenge that arises in constructing a continuous time series on employment is the lack of reliable estimates of annual employment. Therefore, we had to interpolate employment between major rounds of NSSO linearly to create a yearly series, which seriously compromises the annual fluctuations in employment. As a consequence, it may affect the annual estimates of labour productivity and total factor productivity. Therefore, this approach warrants further improvement to incorporate the annual fluctuations in the series between NSSO survey years. However, it is unclear yet, what sources of information can help improve the annual employment series, and therefore, needs further inquiry.

Lately, the National Statistical Office (NSO) conducts annual and quarterly PLFS surveys and provides data on employment and wages. However, these are not entirely consistent with the earlier surveys in the way the respondents are selected, and survey is administered.⁵³ Once we have more rounds of annual data from PLFS, we may be able to make the time series on employment more consistent and reliable.

As discussed earlier, the labour quantity measure used in the India KLEMS database is employment. However, it may be noted that the most appropriate labour quantity measure productivity measurement is hours worked. We had no option but to proxy it with employment, implicitly assuming that the changes in hours are proportional to changes in employment, due to lack of adequate information on hours worked in Indian industries. The problem with such an approach is that it can overstate employment if people are only working part-time, which may have important implications for productivity. The ups and downs in activity in the economy may impact the average hours of work done by the workers in India in a year, and basing the quantity measure of labour input on head count thus fails to capture this aspect which may affect the measurement of productivity.

2.4.3 Labour Income Share

A major limitation of the estimates of labour income share in value added is that the parameter representing the labour income component out of the total income of the self-employed (i.e., mixed income reported in NAS) has been taken as time invariant to facilitate computations. This parameter varies across industries but taken to the same for different years. This fails to consider the possibility that the relevant proportion may change over time at least in some of the industries.

Another limitation of the estimates of labour income share is connected with the method employed for estimating the wage rates of self-employed workers in different industries. As explained earlier, this is based on an econometrically estimated Mincer equation. One equation is estimated for each benchmark year, and the wage rates estimated for benchmark years have been interpolated, which involves some degree of approximation. The estimated equation takes into account differences in industry affiliation by introducing dummy variables for industry affiliation. This is a second-best approach as estimating separate equations for major sectors of the economy, for example, separate estimates for workers engaged in agriculture, industry, and services, might provide a better result. It is needless to say that any econometric issues in the estimated wage rate of self-employed based on the estimated Mincer equation along with approximations being made by using interpolation between benchmark year estimates will affect the estimated wage rate of self-employed income.

2.4.4 Capital Input

The measures of capital input available in India KLEMS are based on only three asset categories, construction, machinery, and transport equipment. Therefore, it does not consider the enormous and distinct role of ICT capital in contributing to productivity and growth. The absence of ICT estimates in the database is driven by insufficient data on investment in ICT assets, such as hardware, software, and communication equipment in the National Accounts. However, during recent years, NAS has extended its asset coverage to include software and ICT equipment. Erumban and Das (2016) have made some estimates for the aggregate economy using input-output tables for ICT equipment and relying on NAS for software investment. Similarly, Erumban and Das (2020) have made some initial estimates of ICT capital by industries, which still remains a challenging task. Nevertheless, now that more data is available from NAS, it is possible to explore this further, and future extensions of the database may explore this. Similarly, following the SNA 2008 guidelines, the NAS now provides estimates of intangible capital such as intellectual property and research and development (R&D), which may be a useful extension to the capital input database.

Two additional challenges are arising since the new system of national accounts. The first is the lack of separate data on registered and unregistered manufacturing, which poses challenges in using ASI and NSO data to split aggregate manufacturing data into 13 industry groups. Instead of the registered and unregistered split, NSO now provides a division between the household and corporate sectors. We currently address this challenge by considering the household sector as unregistered manufacturing and corporate sectors (the sum of public sector and private corporations) as a registered sector.⁵⁴

The second challenge is the lack of separate data on transport equipment and machinery assets in the new series of national accounts, which follows United Nation's System of National Accounts (SNA)-2008. Although, for the time being, we adopt ad-hoc approaches to fix this, for the long-term, it might be necessary to combine transport equipment and machinery into one single asset. Finally, as mentioned earlier, the debate on whether it is appropriate to use an external or an internal rate of return, is unsettled in the literature. It may be worth exploring estimates of capital using the internal rate of return. Some early estimates of the internal rate of return and the impact of using those in the capital service estimation in India (Erumban, 2021) suggest some effect, although not large enough to significantly impact any conclusions on productivity growth.

A last but not the least point is the inconsistency between official statistics on GFCF between various sources. For instance, the reported GFCF series for the registered manufacturing differs substantially between ASI and NAS, as the two sources followed different approaches to gather data. This, however, is beyond the reach of India KLEMS database, although it is important to acknowledge the existence of such differences between the various sources.

2.4.5 Methodology of TFP Measurement

As mentioned in the beginning of the chapter, the India KLEMS database uses the growth accounting methodology to compute contributions of TFP and factor inputs to output growth. However, the growth accounting approach entails certain restrictive assumptions such as constant returns to scale and the equality of factor incomes and their marginal products. Although these assumptions may be questioned, especially in the context of emerging markets, we follow the growth accounting approach to ensure consistency and comparability in KLEMS databases across countries. The KLEMS databased of other countries, including the emerging economies, follow the growth accounting methodology despite its known limitations. Nonetheless, the KLEMS database may be used to produce alternate estimates of TFP growth, relaxing the neoclassical assumptions, as the database includes all relevant information on inputs and output. Attention to this is drawn in the way forward section in Chapter 13.

2.5 Summary

India KLEMS database consists of industry-level estimates of output and inputs—primary inputs and intermediate inputs, since 1980 for 27 industries. This chapter discussed the approach we take in constructing these variables and the main data sources used. The database provides estimates of gross output and value-added measures of output, and estimates of employment, capital, and intermediate inputs as inputs.

The availability of value added and output measures at the industry level allows one to estimate productivity using gross output function, incorporating the important role of intermediate inputs in the production process at the industry level. At the same time, at the aggregate level, one can still use the value-added function. While the measures of value added are largely taken from published NAS, complemented by Annual Survey of Industries (ASI) and National Sample Survey Office (NSSO) surveys on the unorganised manufacturing sector, measures of output involved combining data from these sources with input-output (IO) tables. Additionally, the construction of intermediate input prices required combining intermediate input weights from IO tables with wholesale price indices of two intermediate input categories—energy and materials—and the use of implicit GDP deflators from NAS for services input.

The labour input is further distinguished between pure measures of quantity—number of employed workers and measures of labour quality—the skill composition of the workforce. For both the measures, the primary data source is the NSSO's employment and unemployment surveys and Periodic Labour Force Surveys (PLFS). The measurement of labour quality involved estimation of self-employed wages, which was accomplished using an econometric approach, using a Mincer regression (1974). Further, the NSSO and PLFS surveys in combination with the National Accounts (NAS) have also been used to derive the income share of workers in value added and output, or the labour compensation shares.

Similarly, capital input, measurement of which is subject to severe discussion in the economics profession, has been measured after accounting for asset and vintage heterogeneities. By distinguishing between three asset types—machinery, construction, and transport equipment—and weighing the growth rate of each asset's capital stock by their relative compensation (or rental value) shares, we treat them as distinct assets in the aggregate capital input. For each individual asset, we also allow efficiency decline for older vintages by allowing appropriate depreciation of individual assets while calculating capital stock measures, using the standard perpetual inventory method. This way, as in the case of labour input, one can also distinguish between pure quantity measures—aggregate capital stock and quality measures—capital composition effect, or the share of high and low productive assets in the total capital stock. The basic data required for the construction of capital input, which included nominal investment, investment prices, depreciation rates, and interest rates, are mainly taken from NAS, ASI, NSSO unorganised manufacturing surveys, and the Reserve Bank of India.

Following the discussion of the methodology and data sources, this chapter also identifies some outstanding issues that could still help further improve the KLEMS database. This includes improving labour input measures and extending capital input measurement to include information and communication technology assets and intangible assets. The data and variables discussed in this chapter and the industry aggregation groups provided in Annexure Table 2A.1 will be used in various chapters of this report.

Annexure 2A: Additional Tables

Chapter 3: Economic Growth in India: An Industry Perspective

3.1 Introduction

During the first 30 years after Independence, the Indian economy was able to transition from being virtually stagnant into one growing at a sustained, albeit modest pace of growth (3.5 per cent per annum) [(Acharya et al. (2003)]. Subsequently, in the decade of the 1980s, its annual average GDP growth rate surged to 5 per cent and subsequently to 6 per cent in the 1990s [(Kotwal et al. (2011)]. The growth rate picked up further in 2000. However, after an impressive average annual growth rate of 7.8 per cent during 2001-08, the global financial crisis and the Great Recession led to India’s economic slowdown starting 2009 and the growth rate declined to 6.8 per cent during 2009-16 [Nayyar (2019)].

The disappointing growth between 1955 and 1978, often marked as a failure of Nehru-Mahalanobis growth strategy, was due to the deepening of import substitution and industrial regulation, a reduced role of the market and some exogenous shocks such as oil prices and wars with neighbouring countries. The subsequent gradual acceleration in growth, especially since the early 1980s, has attracted a lot of attention from economists and policy analysts and a significant amount of research has been carried out on this theme. India’s economic growth since the 1980s has been attributed to different sets of factors by different authors. For instance, Subramanian (2008) emphasises that the skilled human capital created in the pre-1980s contributed to faster growth of both industry and services in the 1980s when pro-business policies were implemented. On the other hand, Bhattacharjea (2008) argues that the recovery of public investment in the 1980s could have influenced private investment favourably, leading to faster growth. Acharya et al. (2003) provides a more demand-side perspective, arguing that strong domestic demand, significant export growth along with liberal supply-side policies led to a high growth in the 1980s.

Growth performance in the decades of 1990s and 2000s attracted even greater attention because of the series of economic reforms carried out by India in the early 1990s. These reforms led to deregulation and liberalisation of the Indian economy, which is argued to have increased the growth rate (Basu and Maertens, 2007; Panagariya, 2008; Virmani, 2004, 2006). Nayyar (2019) pointed out that rapid investment growth coincided with rapid export growth since 1991, leading to rapid GDP growth. However, the Indian economy lost this growth momentum in the latter half of the 1990s due to setbacks to the fiscal correction process and some slackening in the pace of structural reforms, which coincided with the onset of the East Asian financial crisis [Mohan (2008)]. With a recovery in its industrial sector and a sustained growth in services, India, however, regained its growth momentum during 2003-07 and became the world’s second fastest growing economy after China (Nagaraj, 2008; Balakrishnan and Parameswaran, 2007). Ghose (2019) emphasised that surging exports of information-technology related services and growing inflow of foreign investment and remittances fuelled the growth acceleration and a decline in foreign finance and remittances in the post-2012 period led to growth deceleration.

However, the euphoria about the impressive growth performance for most part of the period from 1980 to 2017 is somewhat dampened by the fact that the sector-wise contributions to this performance do not present a very impressive picture. Mohan (2008) highlights that India’s agricultural growth has been subject to large variation over the decades. A marked recovery of the agriculture sector in the 1980s has been followed by a slowdown in the subsequent decades. On the other hand, he adds that, it is the continuing and consistent acceleration in growth in services over the decades that really accounts for the continuous acceleration in overall GDP growth of the country since the 1980s. Nagaraj (2008) observes that while the services-sector boom since 1991-92 has been dominated by communications and business services; the communications boom seems largely domestic-demand-led growth and business services seem entirely export-driven. Balakrishnan and Parameswaran (2007) too, infer that the acceleration in the growth of the Indian economy over the last quarter century has consistently been led by services. On the other hand, they highlight that, the contribution of manufacturing to these transitions has been relatively small, despite most of the policy changes being targeted towards the manufacturing sector. Ghosh (2019) pointed out that manufacturing had led the growth process during the first 30 years after independence and only in the early 1990s, services become the lead sector. Through the subsequent reforms and growth accelerations since 1980s, the position of services in the growth process only got strengthened. Basu (2019) said that a boost from information technology sector triggered the overall services sector growth in 2000s. He highlighted that a policy shift in the computing sector in the late 1970s, shrinking government bureaucracy after 1991 and tax exemption for IT products played a significant role in the success of India’s IT sector. Nayyar (2019) found that India has followed the non-traditional pattern of structural change, where unskilled agriculture labour was drawn to the urban informal services sector and contributed to economic growth through this labour transfer between sectors.

Two distinct features of the past analyses on economic growth in India are: 1) they are confined mainly to the aggregate economy or three sectors (agriculture, industry, and services) and hardly trace the differing roles of various industries within the economy in driving growth. 2) Economic growth is analysed using value added, which is indeed appropriate for the aggregate economy, but fails to acknowledge properly the role of intermediate inputs in the production process, while analysing industry-level growth dynamics. The India KLEMS database facilitates industry-level analysis of trends in and proximate sources of economic growth using gross-output and value-added functions. The data on gross output, gross value added, and intermediate input, along with primary inputs labour and capital (see Chapter 2) are available for 27 industries that constitute the aggregate Indian economy. Such detailed analysis will help one understand the impact of various policies on economic growth in general and specific industries within the economy. The remainder of this chapter is organised as follows. Section 3.2 presents the structure of the Indian economy during the 1980-2017 period. We discuss the trends in value added growth of the aggregate economy, and contribution of 27 KLEMS industries and broad sectors to aggregate growth in Section 3.3. The growth rates of gross output and intermediate inputs for the 27 industries are discussed in Section 3.4. Conclusions are presented in Section 3.5.

3.2 Changes in the structure of the economy

In this section, we document industry shares of value added in Indian economy since 1980. In Figure 3.1, we depict the time-series of the changing structure of Indian economy in terms of value-added shares since 1980 for select industry aggregates. In addition, shares of value added for 27 industries are provided in Table 3.1. We have divided the whole period 1980-2017 into four sub-periods, 1980-93; 1994-2002; 2003-07 and 2008-17.

We observe that the value-added share of agriculture & allied activities has declined steadily over the past four decades. It has declined from 36 per cent in 1980 to 18 per cent in 2017. On the contrary, the share of services sector increased from 37 per cent in 1980 to 53 per cent in 2017. The share of non-market services, which consists of public administration, education and health services, remained stagnant around 20 per cent over the entire period whereas the share of market services, which constitute trade, transportation, business and financial services, has increased from 15 per cent to 32 per cent. The increase in share of market services mirrored in an increase in the overall services share. Except for a slight increase during the mid-1990s, the share of manufacturing sector in the total GDP has remained stagnant since 1980. There is an improvement in the share of construction — it increased from 5 per cent in 1980 to 10 per cent in 2008 and then fell to 8 per cent in 2017. The share of utilities sector (electricity, gas & water supply) and mining sector remained stable over the entire period.

In terms of the percentage distribution of value-added shares across 27 India KLEMS industries, agriculture and allied activities had the highest share of more than 17 per cent of GDP in 2017, followed by trade (10.8 per cent), other services (9.4 per cent), business services (8.4 per cent), and construction (7.8 per cent). However, the share of agriculture and allied activities has been declining sharply since 1980 — it fell by 19 percentage points in close to 4 decades. Within manufacturing industries, chemicals & chemical products, petroleum & nuclear fuel, non-metallic mineral products, electrical & optical equipment, and transport equipment have shown an increasing share. However, increases in the share of these industries were not sufficient to offset the decline in the share of other sectors, resulting in no improvement in the share of manufacturing in the overall economy. Chemicals & chemical products observed highest increase, from 1.2 per cent in 1980 to 2.5 per cent in 2017, whereas textiles & leather has witnessed the highest decline, from 3.7 per cent in 1980 to 2.1 per cent in 2017.

Clearly, services sector has been the single largest contributor to value added in the post-1980 period. In terms of value-added share in 2017, out of the top 10 industries, 7 industries are from services sector. The steep increase in share of services between 1994 and 2003 is due to expansion of the market services, of which trade, business services, financial services and transport & storage had high value-added shares in all five years. The share of business services had risen sharply from 1 per cent in 1980 to 8.4 per cent in 2017, followed by trade (increased from 6.9 per cent to 10.8 per cent) and financial services (increased from 2.7 per cent to 5.4 per cent). Value added share of non-market services remained constant. Within non-market services, education, health, and public administration observed an increase in value-added shares between 1980 and 2017, while other services recorded a decline.

3.3 Aggregate value-added growth and industry and broad sector contributions to aggregate growth

This section presents the growth rate of value added in the aggregate economy, broad sectors of the economy, and detailed 27 KLEMS industries. The first two—the aggregate economy and broad sector growth rates for this chapter—are obtained as Tornqvist aggregates of industry value-added growth rates. In a Tornqvist aggregation approach, the growth rate of each industry value added is weighted by the nominal value-added share of that industry while calculating the aggregate value-added growth rate. Another commonly used approach is to simply sum the real value added across industries to obtain aggregate value added and compute the growth rate—an approach often called the aggregate production function approach (see Jorgenson et al., 2007). The Tornquvist approach, which is consistent with the aggregate production possibility frontier approach described in Jorgenson et al. (2007), relaxes several assumptions in the aggregate production function approach, such as a common production function across all industries.

Figure 3.2 provides the annual growth rate of aggregate value added during 1980-2017, and Table 3.2 provides the average growth rates for the aggregate economy and broad sectors for the four sub-periods and the entire period. On average, the Indian economy grew at 6 per cent per annum during 1981 to 2017. The annual economic growth rate fell sharply in 1991 due to a balance of payment crisis. It subsequently recovered and it grew steadily till 1996 at the time of the East Asian financial crisis—when it slowed until 2002. After that, India witnessed a high growth phase until the global financial crisis in 2007-08.

The average annual value-added growth rate of the Indian economy was 5.3 per cent during 1981-93. The average annual growth rate in Gross Value Added (GVA) for manufacturing, mining & quarrying, utility, market services and non-market services was between 6.1 and 8.1 per cent, whereas it was 3.2 per cent for agriculture & allied activities and 3.9 per cent for construction. During 1994-2003, GVA growth improved slightly, with construction and market services seeing a rapid rise, even though agriculture & allied activities, mining & quarrying, utilities, and non-market services registered a fall. During the 2003-07 period, when the aggregate economy witnessed a growth acceleration (7.7 per cent per annum), two sectors that registered very impressive growth were manufacturing and construction. The most marked growth acceleration occurred in construction with a trend growth of 12 per cent during 2002-07, while manufacturing observed a jump in growth rate from 6.1 per cent to 9.2 per cent. The growth rate of agriculture & allied activities also accelerated during this period. The overall growth rate declined to 6.4 per cent in the post-financial crisis period and the growth performance of all the broad sectors except non-market services appears similar to the aggregate economy trend.

Figure 3.4 shows value added growth rates for 27 industries during 1981-2017. We observe that 19 out of 27 India KLEMS industries grew at a faster rate than the economy (6 per cent per annum). Business services grew the fastest at 11.9 per cent per annum. The other well-performing industries are post & telecommunications, electrical & optical equipment, rubber & plastic, coke & petroleum products, transport equipment, financial services, health & social work, chemicals and chemical products—each growing at more than 8 per cent per annum. At the other end, GVA growth has been quite slow in industries such as, wood & wood products, agriculture & allied activities, mining & quarrying, and other services. The performance of agriculture & allied activities, which is essential for India’s economic growth, was relatively very poor. It is also evident (Annexure Table 3A.1) that there is wide variation in value-added growth rates across the 27 KLEMS industries and across the four sub periods. More industries (10 of 13 that slowed) recorded slower growth in the 1994-2002 sub-period compared to 1981-93, while in the subsequent phase of 2003-07 the slowing was more on account of services—five were services of the seven industries that slowed their pace of growth.

A further comparison between the periods 2003-07 and 2008-16 (Figure 3.5) allows us to identify the industries primarily responsible for the growth slowdown in the 2010s. The industries have been ranked by growth in the 2003-07 period—the highest growth took place in manufacturing industries such as, electrical & optical equipment, pulp & paper products, and machinery n.e.c. Service industries such as, business services, post & telecommunications, hotels & restaurants, health, and construction, witnessed more than 10 per cent per annum growth. At the other end, there are industries such as, wood & wood products, rubber & plastic products, other services, agriculture and allied activities and public administration, which registered less than 5 per cent per annum growth rate during this phase. Value-added growth rates declined over these two sub-periods in most industries. Out of 27 KLEMS industries, only nine recorded increased growth. The decline in growth rate is much more for the industrial sector. So, the improved growth performance of the Indian economy during 2003-07 and economic slowdown after the 2008 financial crisis are largely on account of the manufacturing sector.

The extent to which an industry can explain differences in aggregate growth does not only depend on its growth rate, but also on its share in value added. Another way to observe this divergence between value-added share and growth in individual industries is to calculate the contribution of each industry to total economic growth. This is done by multiplying individual industry growth with gross value-added share of that industry. In Figures 3.6 and 3.7, we have plotted the contribution of broad sectors and the 27 industries to the aggregate economy GVA growth. As we have seen, services have been the best performing sector in the Indian economy, both in terms of its value-added share and its growth rate. The sector’s contribution to aggregate economic growth more than doubled from 2.1 per cent in 1981 to 4.2 per cent in 2017. The contribution of market services increased significantly from less than 1 per cent in 1981 to 2.4 per cent in 2017. The contribution of manufacturing sector for the entire period has been around 1.2 per cent. Agriculture has contributed less than 0.8 per cent. It is evident from Figure 3.6 that the Indian economy had crossed the barrier of growth of 9.5 per cent in 1988 when agriculture contributed almost 4.4 per cent of the share. Clearly, for the economy to grow at such a rapid pace, such substantive contribution of the agriculture sector was vital.

Table 3.3 depicts the sectoral contributions to aggregate value-added growth of the economy. During 1981-93, the service sector contributed about 2.63 percentage points towards the aggregate value-added growth rate of 5.4 per cent. Market and non-market services contributed 1.19 and 1.44 percentage points, respectively and agriculture contributed 1 percentage point. The contribution of industry was 1.72 percentage points, with 1 percentage point coming from the manufacturing sector. The picture changed during the period 1994-2002, when contribution of agriculture and non-market services declined significantly, whereas industry’s contribution remained stagnant. The market services sector did see a significant improvement in its value-added growth contribution—from 1.13 percentage point in the previous period to 2.3 percentage point. Indian economy achieved impressive growth during the 2003-2007 period with improvement from 5.7 per cent to 7.7 per cent. This improvement in GVA growth was reflected in almost all segments of the economy, except in the non-market services. Contribution of manufacturing and construction sector to aggregate value-added growth improved significantly. During 2008-2017, aggregate value-added growth rate declined. In absolute term, the contribution to aggregate GVA growth of all broad sectors except non-market services also declined.

Even though we have observed that Indian agriculture sector grew at 3.1 per cent per annum during the period 1981-2017, in terms of contribution to total value-added growth it has made highest contribution (Figure 3.7), followed by trade. Five important service industries that made substantive contributions are trade, other services, business services, financial intermediation, public administration, and defence—their combined share in value-added was around 40 per cent in 2017. Business services grew at the fastest pace (11.86 per cent per annum) during 1980-2017, while trade and financial services grew at a moderate rate of 7 per cent during the same period.

We have seen that service sector contributed around half of the India’s economic growth. But the rise in the overall growth rate during 2003-07 or the fall in the post-financial crisis period stemmed mainly from the industrial sector—specifically from construction and manufacturing. The contribution of agriculture & allied activities and market services to the rise in economic growth rate was much lower. In the post-financial crisis period, all the broad sectors except non-market services observed a decline in the contribution to aggregate economic growth. A further analysis of select industries from the manufacturing and services sectors will help us to identify the industries with substantive potential for growth as well as examine their performance and contribution in terms of economic growth.

Veeramani and Dhir (2017) have emphasised that traditional labour-intensive products such as, food products, beverages and tobacco, wood & wood products, pulp & paper products etc., and assembly-related network products such as, electronics and electrical machinery have strong potential for growth. However, we found that the labour-intensive industries such as, wood & wood product and pulp & paper products were the least contributing industries over the last four decades. Based on their shares in value-added, their rates of growth and the potential for growth, we have chosen eight key industries—agriculture; food products, beverages and tobacco; chemicals & chemical products; basic metals & metal products; electrical & optical equipment; business services; financial services and trade. The combined share in value-added of these 8 industries is around 50 per cent. It is to be noted that the decline in value-added share for agriculture & allied activities during the last four decades has been compensated by the other key industries, particularly by the three services industries – trade, financial services & business services. Except for agriculture & allied activities, all other key industries’ growth rate is higher than the aggregate economic growth rate. However, all these industries experienced a slowdown in value-added growth—more so in the manufacturing sector than services in the post-financial crisis period. Electrical & optical equipment recorded the biggest fall in growth rate. Between 1981-93 and 2008-17, the contribution of services increased significantly, albeit not that of manufacturing, even as it decreased for agriculture & allied activities.

The above data analysis establishes that while agriculture & allied activities and the industrial sector contributed significantly towards high economic growth during 2003-07, it is the market services sector which grew steadily over the last four decades. Within the market services, business services and trade contributed the most.

3.4 Growth of industry output and intermediate inputs

This section presents an analysis of gross output and intermediate input series for the 27 KLEMS industries, in terms of growth of gross output and intermediate input and contribution of intermediate input to gross output growth.

Figure 3.7 and Table 3.5 present the average annual growth rates of gross output of the 27 KLEMS industries. For all 27 industries, the average growth rate of output was 7.39 per cent per annum and of these, 12 industries grew faster than the industry average.

Output growth is found to be most rapid in business services (12.40 per cent). Of the five fastest growing industries, two are producers of services (business services and post & telecommunication) and three are in manufacturing (electrical & optical equipment, manufacturing n.e.c., rubber and plastic products). All five posted growth rates of over 9.97 per cent during the period 1981-2017. At the other end, output growth has been quite slow for wood & wood products and agriculture and allied activities.

Table 3.5 depicts annual average growth rates of output for the four sub periods under consideration. During 1981-93, the best performing industry was rubber and plastic products, growing at the average annual rate of approximately 12 per cent per annum. Five other industries, two of which were from manufacturing—electrical & optical equipment and other manufacturing—grew at the average annual rate of more than 9 per cent per annum. Business services was the only service sector industry which grew at more than 10 per cent per annum, followed by financial services (9.9 per cent per annum). During 1994-2002, all the services except other services and financial services registered higher growth rates compared with the previous sub-period. However, within manufacturing, output growth of six industries improved, while that of seven industries declined. During 2003-07, as many as half the industries achieved their highest output growth, reinforcing the rapid growth achieved by the Indian economy during this period. Among these, electrical & optical equipment, manufacturing n.e.c., pulp, paper & paper products, basic metal & fabricated metal products, construction, machinery, n.e.c., food products, beverages & tobacco, hotels & restaurants, and health & social work registered above 10 per cent growth rates. During 2008-17, growth in gross output accelerated for transport equipment, rubber & plastic product, education, wood & wood products, public administration, and trade. At the other end, the lowest growth rates were observed for hotels & restaurants, coke & petroleum products, chemicals & chemicals products, post & telecommunication, food products, beverages & tobacco, and transport & storage.

Table 3.6 shows the annual average growth of gross output, intermediate inputs and three categories of intermediate inputs—material, energy, and service—for all 27 industries. There was wide variation across industries with the growth rate of input ranging from 15.7 per cent in business services to 2.1 per cent in agriculture and allied sector. The growth in intermediate input was relatively faster in business services, manufacturing n.e.c., post & telecommunications, financial services, electrical & optical equipment and relatively very slow in agriculture and allied, wood & wood products, health & social work, coke & petroleum products, and hotels & restaurants. Comparing output growth with input growth, we observe a strong positive correlation between gross output and intermediate input growth (Figure 3.8)—the fastest-growing industries in terms of input growth were also among the fastest in gross output growth.

The growth rates of material, energy and service inputs show considerable variation across industries. The fastest-growing industries in terms of material input were business services, post & telecommunications, manufacturing, n.e.c., financial services, other non-metallic mineral products, mining & quarrying, rubber & plastic products, and electrical & optical equipment. At the other end, material input growth was slow in agriculture and allied, wood and wood products, electricity, gas & water supply, education, and hotels & restaurants. The growth in energy input was highest in business services, education, financial services, post and telecommunications, and trade, was negative in wood & wood products, and almost stagnant in chemicals & chemicals products, hotels & restaurants, and agriculture. The growth rate of services input was high in business services, other services, education, manufacturing n.e.c. and financial services, and very low for wood & wood products, health and social work, agriculture, and chemicals & chemicals products.

The contribution of an input is measured as the product of an input’s share in value-added and its growth rate. Thus, each input contributes to output in proportion to its value-added share. The estimated factor income shares in gross output for 27 industries are presented in Table 3.7. While estimates have been made for all the years from 1981 to 2017, the table reflects only the period averages.

We find that the share of intermediate inputs in gross output is relatively high in all manufacturing industries, electricity, gas & water supply, construction and two services industries viz., hotels & restaurants and post & telecommunication. Among intermediate inputs, material input share is high for all these industries including health & social work, followed by services input share and energy input share.

Figures 3.9, 3.10 and 3.11 rank the industries in terms of the contributions of material, energy and services input. The mean contributions of these inputs are 2.63, 0.35 and 1.18 percentage points respectively for the period 1981 to 2017. While comparing the contribution of factor inputs—material, energy, services, labour and capital, we found that material input is the dominant source of growth for a majority of the industries. The contribution of energy and services input to output growth has been relatively low as compared to that of material input. Except for agriculture and allied sector and some services (trade, financial services, business services, public administration, education, health, and other services) more than 50 per cent output growth was achieved because of growth in intermediate inputs

Material input contributed significantly to gross output growth in certain industries such as, manufacturing n.e.c., coke & petroleum products, rubber & plastic products, electrical & optical products, chemical products. The industries where the contribution of material input to output growth was less than 0.5 percentage points are agriculture, education, trade, public administration, and financial services. The contribution of energy input was the highest in transport & storage, electricity, gas & water supply and other non-metallic mineral products and services input contribution to output growth was the highest in business services, transport & storage, manufacturing n.e.c., electricity, gas & water supply, and financial services.

3.5 Summary and Conclusion

The main objective of the India KLEMS project was to construct gross value-added, gross output, labour input, capital service input, intermediate inputs (energy, material, and services) and total factor productivity series of 27 India KLEMS industries from 1980 to 2017. Chapter 2 presented the basic data sources and methodology for building a time series of gross value added, gross output, material, energy, and services input from 1980 to 2017 for 27 industries comprising the Indian economy by using NAS and Input Output Transactions Table (IOTT) data. Based on the constructed time series, this chapter has analysed the trend growth of GVA, GVO and intermediate inputs—materials, energy and services input for 27 industries of the Indian economy for the period 1980-2017. We have done this in two parts—first, we analyse the structure and trends at the level of disaggregated 27 industries as well as at the level of the broad sectors of the economy. In the second part, we have examined the growth rate of gross output and intermediate inputs—materials, energy and services inputs, and analysed the contribution of intermediate inputs in output growth.

One of the most interesting features of the Indian economy since 1980s is the emergence of services as the dominant sector and as the main driver of GDP growth. Among fast-growing developing countries, India is distinctive for the role played by the service sector. Whereas earlier developers grew on the basis of exports of labour-intensive manufactures, India has concentrated on services. Although there are other emerging markets where the share of services in GDP exceeds the share of manufacturing, India stands out for the size and dynamism of its service sector. India’s service sector observed an increase in its share in aggregate GVA from nearly 37 percent to more than 53 percent between 1980 and 2017. The share of agriculture in aggregate GVA declined by 19 percentage points and that of manufacturing remained at around 17 per cent during the period. In terms of GVA shares of 27 industries in 2017, although agriculture and allied activities share declined by 19.1 percentage points, it was still the largest sector with 17.2 per cent share in 2017. Further, seven of the top 10 industries in terms of GVA share were from the service sector.

We observe that the Indian economy grew at 6.03 per cent per annum during 1981 to 2017 and the growth improved over different sub-periods until the 2008 global financial crisis. While growth in aggregate value-added the Indian economy was between 5 to 6 per cent during 1981-2002, it improved to 7.65 per cent during 2003-07, and then declined slightly to 6.43 per cent in the subsequent sub-period. In terms of broad sectors, services grew at a faster rate (7.31 per cent per annum) than others, while agriculture grew at a much slower rate (3.06 per cent per annum). Market services such as business services, post & telecommunication, and financial services registered much faster growth than non-market services industries such as, health & social work and education & public administration. Within the manufacturing sector, electrical & optical equipment, rubber & plastic products, coke & petroleum products, and transport equipment grew faster than other industries. Of the average growth of 6.03 per cent per annum for the aggregate economy during 1981-2017, 2.03 percentage points came from market services, 1.21 and 1.26 percentage points from manufacturing and non-market services respectively, and less than 0.8 percentage points from agriculture and allied sector. Within the market services, business services and trade contributed most.

Our analysis reveals a more prominent effect of the global financial crisis at the disaggregate industry level. For as many as 19 of the 27 KLEMS industries, growth rates declined after 2007-08.

The annual growth rate of gross output varied across industries ranging from 1.16 per cent in wood & wood products to 12.4 per cent in business services. Out of 27 industries, 12 observed higher growth rates compared with the industry mean— among them, eight industries are from the manufacturing sector. Agriculture and mining grew at much slower rates than the industry mean of 7.39 per cent per annum, respectively 2.81 and 5.50 per cent. Positive correlation between gross output and intermediate input growth is also observed.

In context of the contributions of intermediate inputs to growth in gross output, we observe that, in general, contribution of materials input is higher than the other factors of production. It is as high as 40 to 50 per cent in the case of manufacturing industries. Except for transportation & storage, electricity, gas & water supply, non-metallic mineral and metal products, the contribution of energy input is negligible (less than 3 per cent), while the contribution of service inputs to gross output lies between 10 to 20 per cent. Evidently, the contribution of intermediate input, particularly material input, remained the dominant source of output growth for most industries. Therefore, intermediate inputs play a crucial role in explaining economic growth at the industry level, which is why, we have argued for the use of gross output instead of value-added in productivity analysis.

Annexure 3A: Additional Table

Chapter 4: Employment, Labour Quality and Labour Income Share

4.1 Introduction

Providing gainful employment to its workforce has always been an integral aspect of economic development in any country, be it a developed or a developing one. However, the problem of job creation has been worsening in several countries owing to multiple factors, including changes in the structure of the labour market. According to ILO (2014), the process of globalisation has also aided the ‘precarisation’ of labour in many countries. India has not remained unaffected by labour market changes and faces similar employment challenges.

The Indian economy has experienced some structural transformations in both GDP and employment after the 1990s economic reforms. However, the structural transformation in employment has been slow as compared to that in GDP (Kochhar et al., 2006; Islam, 2008; Papola and Sahu, 2012; Krishna, 2015). The past few years have also created a situation of inadequate job creation despite reasonably strong economic growth. India’s GDP grew at an average rate of 7.3 per cent over the of the period 2014-18, but this growth did not create enough jobs--the increase in employment has not been commensurate with the increase in the workforce. The Indian economy has been slowing since 2016-17 with a continuous fall in its GDP growth rate from 8.26 per cent to 6.8 per cent per annum during 2017-18 and the total unemployment rate by usual status has gone up significantly from 2.2 per cent in 2011-12 to 6.1 per cent in 2017-18. During the same period, the youth unemployment rate (15-29 years) also shot up across all segments--rural and urban, and males and females.⁵⁵ Some of the reasons pointed out for the recent slowdown in the Indian economy (Nagraj, 2020) are: (i) a decline in the rate of savings and gross capital formation, (ii) unfavourable macro-economic situation with rise in international oil prices, and (iii) a slowdown in GDP of many other countries-post financial crisis of 2008-09, leading to a fall in India’s exports.

The challenge then is to bring the Indian economy back to a faster-growth trajectory and accelerate the process of structural transformation. One of the important routes to achieve that objective is through increase in the growth rate of productivity. But an insight into any analysis of productivity requires data on output and inputs--both labour and capital--for the aggregate economy and at the disaggregated industry level. Due to the lack of availability of any such data in a continuous time series, the India KLEMS project, took on the challenge to build a credible time series for the required data, including an employment series.⁵⁶

Because of the limitations of reliable official data on employment, there are differences of opinion on the actual trend in employment in the recent past. Based on the available data sources, different authors have estimated employment in India for different time periods, but some of these estimates differ from each other (Table 4A.1 in Annexure) mainly due to (i) different estimates of projected population, (ii) use of different employment status and age groups, (iii) different reference periods.⁵⁷ The sources and problems of employment statistics in India have been widely discussed in the literature (Sivasubramonian; 2004, Himanshu; 2011, Ghose; 2016) and also in chapter 2 of the current report. Based on the employment series constructed for India KLEMS (2019), this chapter discusses trends in employment and the causes underlying slow employment growth in India. Trends in the growth of the labour quality index and labour income share have also been analysed in it.

4.2 Profile of the Workforce since 1980-81

The workforce participation rate (WFPR) of male and female workers employed in Usual Status (usual principal and subsidiary status or UPSS) in rural and urban sectors since 1983 in all major rounds and PLFS (2017-18) are summarised in Table 4.1. It can be observed that the total WFPR was largely stagnant at around 42 per cent between 1983 and 2004-05. There has been a decline since then, and by 2017-18, the figure had declined to as low as 34.7 per cent.

On comparing, we observe that the WFPR in urban areas has always been lower than in rural India, but the gap has narrowed down over the period from 10 percentage points in 1983 to just 4 points in 2011-12, and subsequently to only one percentage point in 2017-18. This convergence between the rural and urban sectors has come mainly because of a fall in the rural WFPR by 9 percentage points-from 44.5 per cent to 35 per cent. In general, the WFPR has been much lower among females, both in rural and urban sectors, than among their male counterparts. It was only 16.5 per cent for females against 52 per cent for males in 2017-18. However, while the urban female WFPR has remained stable around 15 per cent, the rural female workforce participation rate has declined by almost half from 34 per cent to 17.5 per cent from 1983 to 2017.

The figures are to be read along with the explanatory note for comparability (as cautioned by the Government of India in its PLFS Report)

A number of explanations have been put forward in the literature for the low WFPR among females. Predominant among them is the ‘distressed employment’ hypothesis, which argues that rural females take up work only during distress and if the economy is doing well, they tend to withdraw from the workforce (Ghose, 2016; Mehrotra and Sinha, 2017). Another plausible explanation is that an increasing number of females are joining and staying in schools and colleges (Mehrotra et al., 2014; Thomas, 2015, Ghose, 2016) or attending to domestic duties (Thomas, 2015). As incomes rise further, a larger proportion of women are seen to work again. Thus, a U-shaped relationship is found between female WFPR and economic development (Tam, 2011). Another explanation focuses on increasing measurement errors in work participation data from the National Sample Surveys (Himanshu 2011; Hirway 2012). Whatever the reasons, the WFPR of females in India is relatively very low-- a mere 27 per cent in 2016, as compared with 80 per cent in Nepal, 67 per cent in Bhutan, 64 per cent in China, 59 per cent in Brazil (Aggarwal and Goldar, 2019).

The significant decline in WFPR for females and for all persons in 2017-18 seems to be driven by a significant fall in the rural figures--for both males and females, possibly due to lack of proper employment opportunities, --especially in agriculture and related activities. The reasons for a decline in women participation in the labour market⁵⁸ also include ‘social restrictions’ on women participation in the labour force, discrimination at workplaces in terms of wages (Srivastava & Srivastava, 2010), getting regular jobs (Goldar & Aggarwal, 2015), and the absence of ample job opportunities for women⁵⁹ (Thomas, 2015) who are heavily employed in the informal sector without any legal protection. The withdrawal of rural women from the labour force is also prompted by decline in poverty (Ghose, 2016; Mehrotra and Sinha, 2017), and increased migration by men of the households to urban areas in search of jobs that puts family responsibility of childcare and elder care on women. There is thus, a great potential to increase the female WFPR in India by providing employment opportunities in the non-agriculture sectors where, however, the gender disparity is even higher.⁶⁰ Policy interventions are therefore necessary to bring about the parity. This could happen if improvements are made in women’s access to quality education, suitable training and skill development, improved access to childcare and elder care, improved maternity protection for balancing work-life responsibilities, parental leave, increased job-flexibility, public safety and improved safe and accessible public transport, encouragement to women friendly policies in the private sector, and an awareness about the importance of work by women to their family. These all are to be part of a coordinated policy framework which promotes a pattern of growth that creates job opportunities for women (Aggarwal and Goldar, 2019). The efforts to increase the women WFPR are also imperative so that a sufficient supply of labour is adequately available to meet the growing demand due to a sustained growth in GDP.

The WFPR among males is not significantly different between rural and urban sectors and after convergence between the two for the period 1983 to 2005, the urban rate exceeded that in rural sectors. This could be due to faster growth in GDP, especially in the services sector, giving rise to more urban employment opportunities and making urban centres more attractive to migrants.

It is not only the level and the structure of employment, which is key to understand the growth process, but its quality as well. Rodgers (2020, p13) points out that the two dimensions of quantity and quality of work need to be considered together as in the ILO’s Decent Work agenda. The concept of quality of employment is multidimensional (Rodgers, 2020) and has been defined on the basis of the nature of employment, the availability of a written job contract, and the social security benefits available in the job (Nayyar, 2012; Papola & Sharma, 2015, Ghose, 2015). Based on these parameters it is perceived that the ‘best’ jobs are regular and salaried ones especially in the formal sector, and the worst ones are the casual jobs in the informal sectors. There are many consequences of increase in ‘informal’⁶¹ economy and ‘informal’ employment. Generally, the ‘informal’ activities lack social protection and insurance; have low productivity and low wages; have small-sized enterprises; and are a small contributor to national exchequer because most of these are out of the tax net.

With a high share of the employment in the casual and self-employment categories (76.2 percent in 2018), one may infer that the quality of employment in India is still very poor, though there has been a slight improvement in recent years due to an increase in the share of regular employment from 15 percent in 1999 to 23.8 percent in 2018 (Ghose and Kumar, 2021, Table 8). Based on an employment quality index, Ghose and Kumar (2021, p14) also present evidence of a steady improvement in the average quality of employment during 1999-2018, aided by a decline of employment share of the casual labour, and an increase in the employment share of the regular and salary employees.

4.3 Structure and growth in employment

4.3.1 Structure of employment in India during 1980-2017

In this section, we briefly discuss the structural change in employment in India from 1980 to 2017.⁶² The model for structural change was first developed by Lewis (1954), under which as a country develops, more resources (labour) shift from less productive activities of agriculture sector into activities with higher productivity in the non-agricultural sector in a two-sector economy. Subsequent literature (Kuznets, 1966; Chenery and Syrquin, 1975) extended the structural shift from agriculture to industry and to services and has been discussed by many researchers (Felipe and Estrada, 2008; Felipe et al., 2007; Krishna, et al., 2018; Erumban et.al. 2019). In Table 4.2, we provide the sectoral distribution of employment in India from 1980-81 to 2017-18 to observe how the employment structure has been changing over time.

Agriculture now employs just two-fifth of the workforce as compared with 70 per cent in 1980. The size of the agriculture workforce also reduced in absolute numbers for the first time in 2005-06 and has continued to fall since then. However, we find that manufacturing sector experienced a slow growth in employment and while its share increased marginally from 10.4 per cent in 1980 to 11.5 per cent in 2017--the share has remained stagnant since 2003. The combined share of mining and electricity has remained stagnant at around 0.8-0.9 per cent and these two broad sectors do not play any significant role in employment generation in the Indian economy. The maximum growth in employment comes from the construction sector where it increased from just 2 per cent in 1980 to 11.6 per cent in 2017, a sixfold increase with a compound annual growth rate (CAGR) of 6 per cent. Employment in the services sector grew continuously over the period and its share in total employment has doubled from 17 per cent in 1980 to 34 per cent in 2017 with a CAGR of 3.2 per cent (Table 4.3), with the pace of growth being faster in market services (3.7 per cent) than in nonmarket services (2.4 per cent). While the share of employment in market services has increased from 9 per cent to 22 per cent (an increase of 150 per cent), in non-market services it increased from 8 to 12 per cent, i.e., by just 50 per cent. So, the structural change has been such that employment has shifted from agriculture sector to non-agriculture sector, especially to construction and the services sectors. The structural change in employment in India has bypassed the traditional route of agriculture to industry to services and has been a direct shift from agriculture to services.

4.3.2 Growth in employment

In order to have a more detailed understanding of the structural change in employment, we now turn to the analysis of the trend in employment growth in the broad sectors and in 27 India KLEMS industries during the period 1981-2017. The growth in employment for this entire period, along with four sub-periods 1981-93, 1994-2002, 2003-07, and 2008-17 are depicted in Table 4.3 for the seven broad sectors of India’s economy.

4.3.2.1 Growth in Employment in the Broad sectors

Table 4.3 shows that over the period 1981-2017, the growth in employment took place at an average rate of just 1.33 per cent per annuum. The share of agriculture sector, which still retains the largest share of employment, has been a laggard with almost zero per cent growth. Among the other broad sectors, growth is driven mainly by the construction sector and the services sector where the average annual growth in employment has been 6 and 3.2 per cent respectively. Within the service sector, it is the market services which have consistently grown faster than the non-market services. Employment growth in construction has been high and it has been able to absorb most of the labour which displaced from agriculture, mainly due to the matching up of the skills in the two sectors. The average annual growth in employment in the manufacturing sector at 1.6 per cent has been quite slow though is still relatively faster than the growth rate in employment in agriculture. It seems manufacturing could not play the role of being the engine of employment growth in the economy.

However, the growth in employment has not been uniform across all the sub-periods. Employment growth in agriculture was the highest in the pre-reform period of 1980-93 as the sector was then the major source of employment in the country. However, with the opening up of the economy, opportunities opened up in both manufacturing and in the services sector. This induced labour to move away from agriculture to other greener pastures with higher wages, and perhaps, higher productivity. As a result, job growth in agriculture decelerated after 1993 and in fact, employment declined in absolute numbers leading to a negative employment growth during the sub-periods 2003-07 and 2008-17. We find a similar pattern of employment growth in the mining and quarrying sector. The immediate benefit of economic reforms with removal or relaxation of many licensing rules and lowering of custom duties, was enjoyed by the manufacturing sector and employment during 1994 -2002 grew by 2.4 per cent as compared with 2 per cent during the pre-reform period of 1981-93. Although the growth in employment in manufacturing decelerated a bit during 2003-08, it was still quite impressive at almost 2 per cent. However, the 2008-17 sub-period brought a substantive slowdown to the pace of employment growth in manufacturing, that may be attributed to the impact of the global financial crisis (GFC), increasing international competitiveness and the emergence of strong global value chains (GVC).

The reforms opened the gates for substantive growth in many service sector industries, such as business services, financial services, telecommunication services, other services, etc. but employment generation has not been commensurate. Employment in services grew at a rate of 3.2 per cent during 1981-2017 but the growth has gradually decelerated since 2003. The deceleration is faster for market services (around one percentage point in both the sub-periods of 2003-07 and 2008-17) than non-market services. The construction sector, however, got a boost from expansion of the manufacturing and the services sectors through increasing demand for urban real estate--both housing and commercial, and infrastructure. The construction sector witnessed an expansion of job opportunities and an increase in its share in total employment. Growth in employment in the sector has been very high, especially during the sub-period of 2003-07 due to expansionary fiscal policy of the government, but some slowdown was experienced after the GFC.

The overall employment situation has been worsening since 2003, with a deceleration in employment growth. Ghose and Kumar (2021) finds that the deceleration is more for the less-educated (educated only up to primary), indicating a ‘skill-biased’ technological change in the process of production. Post -1999, they find evidence of rapidly declining employment opportunities for less educated workers in both agriculture and non-agriculture sectors. This finding is also supported by what is reflected in Figure 4.3 on the educational distribution of the employed. It highlights that while the share of the ‘less educated’ among total employed persons, has reduced from more than 80 per cent to just around 45 per cent, that of the more educated-- above secondary (or more than 10 years of education) has increased almost four-fold from just 9 per cent to 33 per cent and this process started in the 1990s with the onset of economic reforms. However, Ghose and Kumar (2021) also point out that the increase in employment of more-educated persons was slower than the increase in their working age population. The authors highlight that since 1999, employment elasticity has declined in all the broad sectors, possibly due to increasing mechanisation of agriculture and ‘from any or all of three possible developments: technological advances and associated increases in capital intensity in the constituent (narrowly defined) sub-sectors, changes in the structure of output involving increases in the shares of more technology-and-capital-intensive products, and increase in the share of large enterprises (which generally employ more technology-and-capital-intensive methods of production than small enterprises) in the sector’s output,’⁶³ indicating evidence of rapid technological advances and rising capital intensity in the manufacturing and services sectors. It seems that the 2008 global financial crisis considerably slowed employment growth in the Indian economy because of the domestic and spill over effects of slowdowns observed in many other globally integrated economies. Further discussion on employment growth in India especially in the 2011-2017 period based on an econometric analysis is presented in Annexure 4B.

In order to better understand the underlying changes in employment growth within the broad sectors, it will be helpful to next focus on the 27 individual KLEMS industries.

4.3.2.2 Growth in employment in the 27 India KLEMS industries

In this section, we discuss the trends in employment growth in the 27 India KLEMS industries during the period 1980-2017. The growth in persons employed for the entire period, 1981-2017, along with four sub-periods 1981-1993, 1994-2002, 2003-2007, and 2008-2017 are depicted in Table 4.4, while Figure 4.1 provides a ranking of the industries according to employment growth for the entire period 1981-2017.

From the Table 4.4, we observe that the growth rate of employment varies significantly across industries reflecting a change in the speed at which jobs are created in different sectors during the period. Employment has shifted from farm to the non-farm sector, particularly towards construction and some non-market services such as trade, transport, business and financial services. For the full period of 1980-2017, total employment has increased from 283.6 million to 464.7 million with an average annual growth rate of 1.33 per cent. Meanwhile, agriculture and mining & quarrying, which together employed the largest share of 42.7 per cent of the total employment in 2017, experienced a stagnant growth in employment with almost no annual growth in agriculture and a modest growth of 0.76 per cent in mining. In fact, the agriculture sector employed 1.5 million less persons in 2017 as compared with 1980, and the mining sector provided jobs to only 4.8 lacs extra persons over the 38-year period.

Among the 13 KLEMS industries withing manufacturing, very few industries show fast trend growth in employment. As a result, though manufacturing added 24 million jobs between 1980 and 2017, its share in total employment remained relatively stagnant between 10.5-11.5 per cent during the period. Within manufacturing, the industries which have grown at a fast rate, more than 4 per cent per annum, are electrical and optical equipment; machinery, nec; rubber and plastic products; and coke, refined petroleum products (Figure 4.1). At the other extreme, there are industries, mostly the traditional labour-intensive manufacturing industries, where employment growth has been quite slow (at less than 2 per cent) i.e., textiles and leather products; wood and products of wood; food products; other non-metallic mineral products; chemicals; etc. If we wish to increase the share of manufacturing in total employment then some of the slow-growing sectors need policy initiatives and incentives to promote them.

Construction is one important sector that has consistently experienced fast employment growth at an average of 6 per cent over the entire period, adding 48 million jobs, which is more than one-fourth of all new jobs added during the period. Though construction has low labour productivity, it has absorbed a major proportion of the displaced labour from agriculture. It now contributes 11.6 per cent to total employment in 2017 as compared with just 2 per cent in 1980. Growth in employment in the construction sector, which has a large share of casual labour, initially accelerated due to expanding government programmes--rural employment guarantee schemes, rural housing schemes and rural roads schemes, but these programmes did not progress much between 2011 and 2018 (Ghose and Kumar, 2021). However, construction still has vast employment potential and can be tapped in the future for more employment opportunities.

Most service sub-sectors, except public administration, grew at more than 3 per cent during the 38-year period, resulting in addition of 109 million jobs. Of these, 77 million jobs were added by market services alone. While trade added the maximum number of jobs at 34 million, 17 million jobs were added by transport and storage industry, 14 million by ‘other services’, 13 million by education and 12 million by business services. The service sector thereby contributed 60 per cent to the total additional employment between 1980 and 2017. All activities in the services sector such as trade; hotels; transport; post & telecommunication; financial services; education; and health have registered fast employment growth. Most of this growth can be credited to liberalisation and opening up of the sector after the economic reforms were initiated in 1991.⁶⁴ However, attention needs to be paid to further tap the employment potential in the services sector, in order to absorb the shift of labour from agriculture.

While total employment grew at a substantial rate of 2.1 per cent and 1.6 per cent in the first two sub-periods of 1981-93 and 1994-2002, the rate of growth declined after 2002 -- ending up in an absolute decline in employment of 8 million persons between 2011-12 and 2017-18. In fact, deceleration in the growth in employment could be seen even during the earlier NSS-EUS rounds between 1999-2004 and 2004-11.⁶⁵ Agriculture and mining also have a negative average employment growth of -1.6 and -1.9 per cent respectively during 2003 to 2017, due to an absolute decline in employment of 57 million in agriculture and 0.6 million in mining. Consequently, their share in total employment has reduced significantly from 58 per cent in 2003 to just 43 per cent in 2017. The deceleration in growth in employment after 2002 is also experienced by many other industries across both, manufacturing and the services sector. Notable among these are: Food products, textiles, wood products, paper products, chemicals, basic metals, trade, hotels, transport, post and telecommunication, public admin., and education. We thus find a widespread phenomenon of falling employment growth.

However, despite low employment growth rate, the unemployment rate has not been high, because of absorption of workers as casual wage labour or as self-employed⁶⁶, till PLFS (2017-18) when it showed a sharp increase in the open unemployment rate perhaps due to the limit on the absorption capacity of the informal sector (Rodgers, 2020, p.4). Ghose and Kumar (2021), however, argue that the sharp increase in the unemployment rate during 2011-18 was essentially the effect of the sharp decline in the employment rate of the educated, who in the absence of any suitable job opportunities waited for the right employment and joined the ranks of unemployed while the less-educated just opted out of the labour force after losing all hopes of getting work.

4.3.2.3 Some plausible explanations for the slowdown in employment growth

India experienced a severe slowing of growth in employment after the global financial crisis (GFC) of 2008, with the average annual growth rate for employment falling from 1.3 per cent for the period 2003-07, to just 0.6 per cent for 2008-17. However, an international comparison shows that India was not alone in experiencing such deceleration (Figure 4.2), with many other countries also facing significant adverse impacts. It is evident that all BRICS countries, and most of the other selected countries of Philippines, Thailand, Vietnam, US and Japan observed a significantly lower growth in employment during 2008-17 as compared with the pre-GFC period of 2003-07. Only few countries, e.g., Bangladesh, Indonesia and Malaysia were able to face the GFC and sustain the pace of growth in employment.

Several explanations have been provided for the low employment growth in India, which include low work-force participation rates (WFPR), low labour intensity in several sectors (Panagariya, 2008), rigid labour laws (Fallon & Lucas, 1991; Besley & Burgess, 2004), increased use of capital- intensive technology (Das et al., 2015; Jha, 2015), rapid advancements in technology, low employment elasticity⁶⁷, decline in poverty of the households that induces women not to work in low-productive jobs (MGI, 2017) and non-availability of suitable jobs.

India is now facing a serious problem with its relatively slow employment growth since 2008. Most sectors have registered a decline in the rate of employment growth since the global financial crisis of 2008.⁶⁸ The industries of coke, refined petroleum products and nuclear fuel, machinery, nec., electrical and optical equipment, are the few exceptions that have experienced a real surge in employment growth. Dhawan and Sengupta (2020) have identified 11 manufacturing industries, where integration with the global value chains⁶⁹ could accelerate the growth of output and employment in India. These include auto-vehicles and vehicle components, capital goods, pharma, chemicals, agriculture and food, metals, textiles and apparel, leather and rubber, renewable energy, etc. A meaningful employment policy thus has to be designed which can understand the dynamics of the employment potential of each industry and take necessary sector specific measures to tap their potential. India has to identify the strongest growing industries of the economy, both in manufacturing and modern services and also the skills that are in demand. Skill enhancement and required training for the acquisition of these skills should be the focus of the educational and employment progammes. Since India has not invested enough in the human infrastructure to reap the benefit of its demographic transition, so investment in human and physical infrastructure will need to be scaled up to promote entrepreneurship and create jobs. The next generation of economic reforms necessary to promote and maintain international competitiveness must also be undertaken at the earliest so as to reap the benefits of GVC and the integration of the different economies.

4.4 Growth in Labour Quality Index

The discussion about labour quality emanates from the importance of human capital which helps improve the quality of labour input and thus the growth of output through increase in labour productivity. According to the Human Capital Index (World Bank, 2020, pp. ix), ‘human capital consists of the knowledge, skills, and health that people accumulate over their lives. People’s health and education have undeniable intrinsic value, and human capital also enables people to realise their potential as productive members of society. More human capital is associated with higher earnings for people, and higher income for countries.’⁷⁰

However, growth in labour quality index is mainly a measure of change in the skill composition of the total workforce and total compensation over a period of time. Labour quality though is intertwined with human capital, but it does not directly embody health of the people and is a narrower construct than human capital. The labour quality index is sensitive to the measurement of earnings and the distribution of labour across different categories. Any change in labour quality in India can be partially explained by inter-industry shifts of labour and partially by a change in the composition of labour. The changes in the composition of labour in India are summarised in Figure 4.3, which indicates that there has been an increase in the share of high skilled⁷¹ labour (with education above Hr. Sec.) from mere 2.4 per cent to almost 13 per cent between 1983 and 2017-18. It is also evident that the proportion of employed persons with low level of education, i.e., either with below primary level of education or with primary level of education, has consistently decreased over the period along with the increase in high skilled labour.

4.4.1 Growth in Labour Quality index in Broad sectors

Generally, the index of labour quality improves over time with improvements in the skill level (i.e., with education and experience) of labour. While measuring labour quality is a data-intensive and data accuracy-sensitive exercise, it is also an important exercise given that the quality of labour significantly affects availability of labour input for growth. From Table 4.5, it is evident that during 1981 to 2017, the index of labour quality in India increased at an annual growth rate of 1.2 per cent, which is marginally slower than the growth in employment at 1.3 per cent (Table 4.3). It thus indicates that the growth in labour quality index is almost as important as the growth in employment for growth in labour input. The growth in labour quality index has been relatively fast in mining and manufacturing sectors and has been slow in agriculture and construction sectors. Growth in the index has been moderate for the services sector at an average of 0.9 per cent per year, but the growth rate, as expected, is slightly higher in market services as compared with non-market services. It reflects the skill requirements and the nature of activity of the sectors. Since the growth in labour quality is affected by both the changes in the skill composition (education) and the growth in labour compensation (wages paid to labour), we hardly notice any significant changes in the skill composition of the labour who works in agriculture and construction. It is also clear that growth in labour quality index in most of the broad sectors has been faster in the sub-period 2003-07 as compared with the period 2008-17, possibly due to faster growth in wages during this sub-period.

4.4.2 Growth in labour quality index for the 27 India KLEMS industries

The growth in labour quality index for the 27 KLEMS industries is presented in Table 4.6 for all the four sub-periods as well as for the entire period of 1981-2017. It shows that the growth in labour quality index over 1981-2017 has wide variations both between industries and also over time within the same industry. While the agriculture sector observed an annual growth of only 0.3 per cent, mining experienced among the highest annual rates of growth of 1.32 per cent across all 27 industries.

Within the manufacturing sector, we find that eight out of 13 industries have grown at a rate of 0.7 per cent or higher, e.g., food products, paper products, chemicals, rubber, machinery, transport, while the remaining five have shown a lower growth rate. While the labour quality index in the utility industry grew at an average rate of 0.7 per cent per annum, the growth has been much slower in the construction industry mainly due to the predominance of the low-skilled labour in the construction industry—that is mostly casual labour without any job security, earning very low wages, often even below the statutory minimum wage rates. The industries within the services sector did not experience the similar wide variation in the growth of index of labour quality, mainly because of the similar nature of the skills- mainly medium to high-skill, of the person employed in these industries. On one extreme of the variation, we have ‘other’ services and health with more than or equal to 0.7 per cent annual growth, and on the other extreme the industries of ‘business’ services and financial services observed an annual growth of around 0.4 per cent in labour quality index.

On the whole, we find that not only skill quality of each industry is different, but it has changed differently over the whole period and during the sub-periods (Table 4.6) and there is no discernible common pattern during the four sub-periods.

4.5 Growth in Labour Input

The growth rate of labour input, measured as the growth rate in number of persons employed plus growth rate of labour quality index is briefly analysed in this section.

4.5.1 Growth in Labour Input in Broad sectors

It is observed that the growth rates of labour input are higher than those of persons employed because of contribution of labour quality index, but the trend in growth in labour input is similar to that of the growth in employment. Like growth in employment, the fastest growth in labour input has been observed by the construction sector, followed by the services sector- induced by acceleration in the market services, and the electricity sector (Table 4.7).

4.5.2 Growth in Labour Input for the 27 KLEMS industries

The average annual growth rate of labour input for the period 1980-2017 in Figure 4.4 shows that the combined push of the two growth rates led the growth in labour input to be the fastest in business services; construction; electrical and optical equipment; machinery, nec; financial services; rubber and plastic products; and post & telecommunication. On the other hand, there are industries where growth in labour input has been quite slow i.e., agriculture; wood and products of wood; public administration and defence; textiles and leather products; and food products, beverages and tobacco. The growth in labour input is an important indicator for policy makers to pay special attention. to those industries which contribute significantly to labour input growth.

4.6 Labour Income Share

The distribution of income between capital, labour and intermediate inputs, is an important element in growth accounting because income shares, under conditions of competitive markets, can be used to measure the contributions that each factor makes towards output growth. So, measurement of labour income share plays an important role in the growth accounting framework to compute total factor productivity (TFP).⁷² In production analysis, with only two factors of production -- labour and capital, the labour income share is defined as the ratio of labour income to gross value added (GVA). Then, under the assumption of constant returns to scale, the sum of the labour income share and the capital income share is equal to one and the capital income share can be obtained as the residual of one minus the labour income share.

The importance of labour income share also lies in understanding the changes in the distribution of national income. A downward trend in the labour income share is an indication that a higher share of national income is going to the capital, thus implying an unequal distribution of national income. However, it is to be noted that since labour income share (SL) is inversely related to labour productivity (LP) (equation 4.3); therefore, its distribution would also help in explaining the pattern of labour productivity.⁷³ So, if product wage rises faster than labour productivity, the labour income share increases. If they move in tandem, the labour income share remains unchanged. Thus, unless the growth in product wages (nominal wages deflated by price index of output) is faster than the growth in labour productivity, the growth in labour income share is bound to be negative.

4.6.1 Trends in labour income share in India

For an understanding of the trend in labour income share for the broad sectors of the Indian economy, the results for selected years are presented⁷⁴ in Table 4.8. While labour income share for each industry is obtained directly from equation 4.1; that for each sector can be defined as the weighted average of each industry’s labour income share (equation 4.4). The weight for each industry is the share of the industry’s value added in the total value added of the sector. Therefore, the trend of total labour income share in a sector (or in the economy) can be decomposed by the change in the labour income share within each industry over the period and the change in value added share of each industry within the sector (or economy). Thus,

where S_st and S_it are the labour income share in sector ‘s’ and industry ‘i’ at period t and v_it is the share of value added in the i^th industry at period t.

From Table 4.8 it is clear that the total economy and all the broad sectors have a long-term tendency of a declining trend in labour income share. A closer analysis reveals that in 1980, the labour income share varied from a maximum of 79 per cent in construction to the minimum of 35 per cent in electricity, followed by manufacturing, at 39 per cent and services and agriculture at around 57 per cent each. The non-market services had a higher labour income share, at 61 per cent as compared to 52 per cent for the market services sector. However, by 2017, labour income shares of the total economy and of the services sector had each declined by around four percentage points because of a change in the structure of the economy, by more than half in mining and by eight percentage points in manufacturing. The decline in the labour income share has been sharper for market services, by more than 11 percentage points, as compared to non-market services, which witnessed a decline of just one percentage point.

An inter-temporal comparison shows that between 1980 and 2003, the labour income share has declined in the intermediate years for the all the broad sectors and the total economy except the construction sector where it is almost stagnant at around 79 per cent. After 2003, while there is a continuous decline in the share for mining and construction, there was a marginal increase in labour income share for manufacturing and services--both market and non-market. But the share has remained stagnant for agriculture. Though an increase in labour income share of 2.6 percentage points is observed in the total economy between 2008 and 2017, the figure for 2017-18 is still lower than that what it was in 1980-81. The increase between 2008 and 2017 seems to have been induced by an increase in the share for the manufacturing and services sectors. One of the likely reasons for this is the fast increase in real product wages in manufacturing and market services during this period (Table 4.12).

Much attention has been paid lately to the falling labour income share in many countries across the world. Dao et al. (2017), point out that there has been a downward trend in labour share in global income since the early 1990s and this downward trend is observed for advanced economies as well as emerging market and developing economies (EMDEs). For India, a decline in the labour income share in manufacturing has been found by a few scholars (Goldar and Aggarwal, 2005; Goldar, 2013; Abraham and Sasikumar, 2017; Goldar, 2022).

4.6.2 The volatility of the labour income share

Based on the notion of constant labour income share advanced by Kaldor (1956, 1961), Solow (1958) proposed two ways to evaluate the stability of the labour income share. The first one is absolute stability, which means to examine the changing extent of total labour income share (expressed by serial variance) in a certain time span. The other one is relative stability, which requires that the variance of total labour income share should be less than the variance in each sector or industry (Zhou, 2016, p.68). We have used the same methodology to explain the volatility in the labour income share through absolute and relative stability. While Table 4.9 shows the results of absolute stability in labour income share for the economy during 1980 to 2017, the results for relative stability are presented in Table 4.10. The results in Table 4.9 do not satisfy the condition of absolute stability and show that the volatility value for the combined period lies, contrary to the theory, outside the volatility values for the sub-periods.

Table 4.10 shows the standard deviations of labour income shares in the total economy and the broad sectors. It is evident that the condition of relative stability is satisfied as it is found that the volatility of the labour income share for the total economy is generally lower than ‘within’ sector volatility, except agriculture.⁷⁵ The two results together show that although the labour income share in each sector has a relatively larger change, the change of total labour income share can be relatively stable due to the sectoral structure change, but the stability is missing over time.

4.6.3 Reallocation of change in the labour income share

Since the labour income share for the economy is computed by the weighted labour income share of its constituent industries (equation 4.4), the change in the labour income share could be explained either by change in the labour income share of each industry or by the change in the value-added share; i.e., a structural change or by change in both the components.⁷⁶ Zhou (2016) has used the methodology proposed earlier by Ruiz (2005) and Young (2006) in which the total change is decomposed into three effects--the ‘within’ industry effect; the ‘structural change’ effect (‘between’ effect) and the ‘covariance’ effect. The present chapter has used the equation proposed by Zhou (2016, p.70) for the KLEMS 27 industries and is as follows:

The three effects obtained by using equation 4.5 for the 27 KLEMS industries are depicted in Table 4.11. It shows that the changes in labour income share over the period 1981-2017 have not been uniform. While there has been a general decline of 0.12 percentage points per year over the entire period (Table 4.11), it varies between the different sub-periods of the study. The decline in wage share is observed for the first three sub-periods and is relatively large in sub-period of 2003 to 2007, but the share increased in the fourth sub-period, 2008-2017. For the entire period of 1981-2017 and the first three sub-periods, both within and between industry effects are responsible for the decline in the wage share but within industry effect is greater than the structural effect. In the period of post-GFC, 2008-17, both the effects have actually contributed to the increase in wage share. The result of a stronger within-industry effect than the between-industry effect to explain the change in aggregate labour income share is also supported by Abraham and Sasikumar (2017)⁷⁷ for Indian organized manufacturing sector, and OECD (2012) for 26 OECD countries and 20 industries in the private sector since 1990.

Since it has been noticed from equation 4.3 and the discussion on labour income share that the behaviour of labour income share is induced by the behaviour of real product wages, we present a brief overview of the average growth rates of real product wages for the broad sectors for 1981-2017 and the sub-periods in Table 4.12 and the trends for broad sectors in Figure 4.5, in order to complete the discussion on labour income share.

It is evident from Table 4.12 that the total economy has experienced a consistent acceleration in average growth rate of real product wages in all the sub-periods. The same trend is visible in all the broad sectors except electricity and construction where a deceleration in growth in real product wages is observed in the sub-period of 2008-17. The growth in real product wages is almost same for both market and non-market services at four per cent per annum. Another feature is the decline in the level of real product wages in the sub-period 1980-93 in both mining and construction, resulting in a negative average growth rate during 1981-93 and also an overall negative average growth rate for the entire period 1981-2017 for the construction sector, and only a small positive increase in the mining sector. A similar narrative is visible from Figure 4.5 showing a trend in the index of real product wages (the level of real product wages in 1980-81 is taken to be 100) for all the broad sectors.

The total economy has observed an increase of 360 per cent in level of real product wages induced by more than 425 per cent increase in the level of wages in the manufacturing sector (Panel A). The slowest increase is evident in the mining sector, and the construction sector- which has actually witnessed a decline in the index of real product wages (panel B).

4.7 Summary and Conclusion

The chapter presents an analysis of the trend in time series of persons employed, labour quality index and labour input from 1980 to 2017 for the broad sector of the economy and the 27 KLEMS industries comprising the Indian economy.

The main findings of the chapter can be summarised as: (a) The overall WFPR in India has been falling, especially for rural females; (b) There has been a significant change in the structure of employment in the Indian economy between 1980 and 2017-- while the share of agriculture has reduced in total employment from 70 per cent to 42 per cent, the share of the services sector has doubled from 17 per cent to 34 per cent, pushed by a two and a half times increase in the share of market services from 9 per cent to 22 per cent during the same period-- thus establishing that growth has been mainly service-led; (c) The growth in employment has slowed down considerably in the recent period--there has been an absolute decline of 8 million employed persons between 2011-12 and 2017-18--and the unemployment rate has increased significantly during the same period. The employment level declined between 2011 and 2017 in agriculture, mining, as well as in manufacturing and while employment growth has been driven mainly by construction (a sector that had substantially helped absorb labour from agriculture till 2011) and the services sectors, especially market services, this growth has also slowed; (d) A change is noticed in the education profile of the work force as the proportion of more-educated workers has significantly increased; (e) Along with overall growth in employment, there has been a significant growth, at 1.2 per cent per annum, in the labour quality index leading to a faster growth in labour input during 1981-2017. Growth in employment and growth in labour quality index have not been uniform, with considerable variations being observed across industries and across sub-periods in both; (f) Though there is a long-term trend of a fall in labour income share for the economy, no unique trend could be discerned at the broad sector level. The change in labour income share seems to have occurred mainly due to ‘within industry’ effect in the sectors and ‘within industry’ effect is larger as compared with the structural change effect. Real product wages have experienced a consistent increase, with an annual average growth rate of four per cent for the period of 1981-2017.

We thus find that India’s employment situation has been quite precarious with an absolute decline in total employment and shooting unemployment rates between 2011-17. However, with a fall in employment elasticity from 0.3-0.4 in 1999 to 0.1 - 0.2 in 2017, unless India achieves a growth rate 8-9 per cent in real GDP that is job-rich, it will not be able to meet the challenge of providing employment to the backlog and the additions to labour force.⁷⁸ Rather than a macro approach, the key to policymaking appears to lie in a focus on sectoral understanding and microeconomics.

Annexure 4A: Additional Tables

Annexure 4B: Trends in employment in the period 2011-2017 – assessing the impact of growth in real wages

While discussing the structure and growth in employment in Section 4.3 it was noted that the growth rate in aggregate employment during 2008-2017 was lower than that during 1981-1993, 1994-2002 and 2003-2007. During 2008-2017, there was a fall in employment in agriculture and allied activities and in mining and quarrying. Employment in manufacturing was almost stagnant during 2008-2017 whereas the growth rate was about two percent per annum or higher during 1981-1993, 1994-2002 and 2003-2007. A more detailed analysis reveals that between 2011 and 2017, there was an absolute fall in employment in manufacturing.

This annexure is devoted to an analysis of employment trends during 2011 to 2017, complementing the analysis presented in Section 4.3 of this chapter. Between 2011 and 2017, there was a fall in aggregate employment and employment in the agriculture and allied sector fell by about 29 million between these two years. Employment in construction increased by about 5 million, but this gain was partly neutralized by the fall in employment in manufacturing by about 2 million. The fall in employment in agriculture was compensated mostly by the increase in employment that took place in services, where the increase between 2011 and 2012 was by about 20 million. Thus, the total employment in the Indian economy fell by about 7 million between 2011 and 2017.

In this annexure, an econometric analysis is undertaken of the trends in employment in the recent period. An employment function is estimated by applying the ARDL model using time series data for the period 1980-2011. In the next step, using the estimated model parameters, employment is predicted for the years 2012-2017, and for earlier years. What is of interest is to find out if the model-predicted employment for the period 2012-2017 is much higher than the actual employment in this period.

For the purpose of this analysis, the economy is divided into four parts: (i) agriculture and allied, (ii) construction, (iii) industry excluding construction, i.e., manufacturing, mining, and electricity, gas and water supply taken together, and (iv) services. A log-linear specification is used. Logarithm of employment is the dependent variable. Logarithm of real GVA and logarithm of real wage rate are taken as the explanatory variables. In the equations estimated for industry (excluding construction) and for the services sector, an additional variable has been introduced – this is the ratio of ICT capital stock to total capital stock. The ICT series was a part of the India KLEMS database, version 2018, which has been used for this analysis. In that series, data on ICT capital stock are provided from 1990-91 to 2016-17. The series has been extrapolated to 2017-18 by applying growth rate in 2016-17. To normalize the ICT capital stock series, this has been divided by the total capital stock of the economy after excluding that for agriculture and allied sector. Since investment in ICT assets have presumably been made mostly in the industries other than agriculture and allied sector, the denominator has been formed in this manner.

The estimated long-run coefficients of the employment function for the four segments of the Indian economy are presented in Table 4B.1. The results indicate a negative effect of real wages on employment. This impact is found to be relatively bigger in the agriculture and allied sector. A positive effect of real GVA growth on employment is found for all four segments of the economy, as expected.

There is need for caution in the interpretation of the coefficient of the ICT intensity variable because it is highly correlated with the trend term, and may thus pick up the trend effect, i.e., it might be picking up the effects of technological change and other such developments taking place in the economy over time, along with the effect of ICT capital on employment. Yet, based on the results obtained, it would perhaps not be wrong to claim that there is some indication that investment in ICT had a negative effect on employment generation in the services sector. No such negative effect is found for the industry sector (excluding construction). Rather, ICT investments made in the economy seem to have helped in job creation in manufacturing.

As mentioned above, the models for explaining employment level have been estimated for the period up to 2011 with the objective of predicting employment for subsequent years and then assessing how well the actual employment during 2012-2017 compares with that predicted by the model. This is depicted in Figure 4B.1.

The estimated models for different segments of the economy predict a fall in aggregate employment between 2011 and 2017, which matches in terms of direction with the data on actual employment although the data for the recent period has not been used in the process of model estimation. The fact that the actual employment for 2017 is fairly close to that predicted by the model implies the observed fall in employment between 2011 and 2017 is attributable mostly to the changes that have taken place in the explanatory variables used in the models. Arguably, hikes in real wages have played a role in causing a fall in employment at the aggregate level.

Turning now to the estimates for the four segments of the economy (see Fig. 4B.2), in the industry sector (excluding construction), the average growth rate in real wages was 5.4 percent per annum during 2003-2011 which increased to 8.2 percent per annum during 2012-2017. The fact that there has been a fall in employment in this sector between 2011 and 2017 appears to be explainable mostly by this factor. By contrast, in the agriculture and allied sector, there has been only a moderate increase in the growth rate of real wages, up from 5.1 percent per annum during 2003-2011 to 5.6 percent per annum during 2012-2017. The fall in employment that took place in this sector between 2011 and 2017 thus cannot be attributed only or mostly to increase in the growth rate in real wages. The explanation lies elsewhere too. Another important factor that seems to have played a role is the fall in the growth rate in real GVA – it fell from 3.9 percent per annum during 2003-2011 to 3.1 percent per annum during 2012-2017.⁷⁹

It is interesting to observe that the fall in aggregate employment in the Indian economy between 2011 and 2017 has been smaller than that predicted by employment estimates generated by the model. According to the model-based estimates, employment should have decreased by about 13 million whereas the actual fall was by about 7 million. The source of this difference lies primarily in the industry sector. Going by the employment estimates based on the estimated model, employment in industry (excluding construction) should have decreased by about 7.2 million between 2011 and 2017 – the actual fall was much lower at 2.4 million. It appears that the negative effect of an acceleration in the growth rate in real wages in manufacturing during 2012-2017 was offset to a considerable extent by certain other developments in the manufacturing sector.

An interesting academic question is if the models were estimated using data up to 2004 (EUS 61st round) and then the model were used to predict employment for 2011 (EUS 68th round) and 2018 (PLFS-I), would the predicted values match the actual, particularly would the prediction be of an increase in aggregate employment between 2004 and 2011 and a fall between 2011 and 2018. This is also a question about checking the robustness of the methodology employed. When this exercise is done, the methodology and the associated predictions are found to be meeting squarely the requirements of robustness. For the period 2004-2011, a fall in employment in agriculture and allied sector and a rise in employment in the other three segments of the economy and at the aggregate level is predicted which tallies with the actual. For the period 2011-2018, a fall in employment in agriculture and allied sector and industry sector (excluding construction) is predicted which tallies with the actual. For construction and services, an increase in employment is predicted for 2011-2017, which tallies with the actual.

Figure 4B.3 shows the prediction based on the model estimated using data up to 2011 (called estimate-1) and the prediction based on the model estimated using data up to 2004 (called estimate-2). A comparison is made with actual employment. It is evident that the two sets of predictions are close to each other, and also close to the actual employment. Although the model is estimated using data up to 2004, the prediction for 2011 differs from actual only by 1%, and the prediction for 2017 differs from actual by 0.1%.

The analysis presented above has a bearing on an issue concerning the employment series used for the analysis presented in the report. This is the issue of data comparability since in the process of construction of the employment series for the years beyond 2011 the employment estimates based on PLFS have been used in conjunction with the estimates based on EUS 68th round. This issue is taken up for a detailed discussion below.

The trends in employment in the 2010s indicated by the PLFS in combination with the EUS 68th round for 2011 are quite different, in several respects, from the trends for the 1990s and 2000s revealed by EUS of 2011 along with that for earlier years. Thus, a fall is seen in aggregate employment between 2011 and 2017, and such a fall has occurred in the agriculture and allied sector and in manufacturing. While a fall in employment in agriculture had occurred also between 2004 and 2011, a fall in employment in manufacturing between successive employment surveys did not occur in the employment estimates of earlier rounds of EUS. The fact that manufacturing achieved a relatively high average annual growth rate in real GVA of about 7.2 percent during 2012-2017 whereas employment in manufacturing decline at the rate of (-)0.7 percent per annum (a fall by about 2.3 million persons) in this period appears puzzling, and contrary to expectations. A negative employment elasticity in manufacturing over a period of seven year has not occurred for Indian manufacturing ever in the period 1980-2011 going by India KLEMS data (perhaps true also for the 1970s, 1960s and 1950s). As regards the fall in employment in agriculture, this is not unrealistic or altogether unexpected, and it should be treated as a mark of the achievements made in terms of structural transformation in the development process in India. While the growth rate of employment in construction during 2012-2017 is positive at 1.6 percent per annum, it has fallen sharply from the growth rate of about 9 percent per annum achieved during 2005-2011. Another interesting difference between the survey findings of PLFS and those of EUS 68th (for 2011) is that the unemployment rate was 2.2 percent in 2011 (previously, 2.3 percent in 1999 and 2.4 percent in 2004) and it was much higher at 6.1 percent in 2017.

In several recent studies, the EUS and PLFS data have been combined to analyse the post-2011 trend in employment in India. These include Mehrotra and Parida (2019), Agarwal and Goldar (2020), Ghose and Kumar (2021) and Padhi and Motkuri (2021). In this literature, the issue of comparability of EUS 68th round and PLFS has received attention (see, for example, Jajoria and Jatav, 2020; Padmaja et al., 2020; Mehrotra and Parida, 2019; Padhi and Motkuri, 2021). The general impression that seems to be emerging from the literature is that there is no serious problem of data incomparability in using PLFS data with EUS 68th round data (although, in some articles on this topic, serious concerns have been expressed). It should be noted in this context that in the econometric analysis presented in this annexure, looking into the impact of real wages and real value added on employment, the model estimates predict a fall in employment between 2011 and 2017 in agriculture, industry (excluding construction) and in aggregate economy which is in line with the actual fall as indicated by EUS 68th round and the PLFS. Thus, the econometric results provide some reasons to feel confident about the employment series used for analysis in the Report. In other words, the finding of a fall in aggregate employment and employment in manufacturing between 2011 and 2017 is depicting the reality, and not arising from some issues of data incomparability.⁸⁰

Based on the studies that have been undertaken on the recent trends in the employment situation in India using PLFS data in conjunction with EUS 68th round data, some key points that emerge may be briefly stated here (some of these points were already noted above in this chapter, but worth repeating here): (1) The rate of unemployment is relatively high (6.1 percent in 2017 and 5.8 percent in 2018). It is still higher among the youth. The rate of unemployment among persons of age 15-29 years was 17.8 percent in 2017 (up from 6.1 percent in 2011). (2) There has been a significant fall in women work participation rate in rural areas, from 37 percent in 2011 to 25.5 percent in 2017. Among women of age 15-29 years in rural areas, the work participation rate has fallen from 41 percent in 2004 to 26 percent in 2011 and then to 14 percent in 2017. (3) The labour force participation rate among youth bore a negative correlation with the economic status of the household in 2011 as revealed by the pattern observed different quintile classes of households according to monthly per capital consumer expenditure (MPCE). This has now changed to a mild positive correlation – the labour force participation rate in the lowest quintile has come down from 48 percent in 2011 to 38 percent in 2017 (Padhi and Motkuri, 2021). This indicates that the youth in rural areas belonging to relatively poorer households is now less inclined and less under compulsion to join labour force than was the situation in 2011. They would rather stay unemployed than take up a low paying low quality job. These facts along with other related facts (such as youth of lower income strata continuing to pursue education in a greater scale) bring out that there has been rationalization of labour in agriculture and this is reflected in the fall in agricultural employment that took place between 2011 and 2017 (also in the period 2004 to 2011). The rationalization of workforce and the removal of surplus labour in agriculture got reflected in the growth in labour productivity and total factor productivity. This, by itself, should not be treated as a matter of serious concern, because in the process of economic development labour is expected to shift from agriculture to more productive jobs in the non-agricultural sector.

A much more important point to note regarding the trends in employment in India in the recent period is that the employment in manufacturing has fallen, and thus manufacturing has not been able to absorb the workers that moved out of agriculture. The analysis presented later in Chapter 8 of the Report brings out that the growth rate in employment in the formal segment of manufacturing had fallen from 4.8 percent per annum during 2000-2007 to 1.1 percent per annum during 2008-2017. In the informal segment, the growth rates in employment were 1.6 percent per annum during 2000-2007 and it came down to (-) 0.5 percent per annum during 2008-2017. Evidently, the fall in employment in manufacturing in the recent period is traceable to the failure of the informal manufacturing sector in sufficiently enhancing job opportunities.

Chapter 5: Investment and Capital Input

5.1 Introduction

Capital is an essential input in the production process, and its importance in understanding the sources of growth is widely acknowledged in the literature. Empirical analysis of sources of economic growth requires pertinent measures of capital input to quantify these sources properly. However, measuring capital is the most complex of all input measurements in productivity and growth analysis. The India KLEMS database, which is primarily aiming to understand sources of growth using a growth accounting framework, provides estimates of capital input at the industry level. These measures are based on the concept of capital services rather than conventional capital stock measures. A measure of capital services is most appropriate from the perspective of productivity analysis, as it helps account for the contribution of capital into production more accurately (Jorgenson, 1963; Jorgenson and Griliches, 1967; Hall and Jorgenson, 1967).

This chapter analyses the trends in physical capital in Indian industries during the period 1980-81 (financial year, written as 1980) to 2017-18 (2017). Specifically, we examine the trends in investment/GDP ratio, real investment, capital stock, and growth rates of capital services at the industry level. As in the case of other chapters in this report, all the data used in the analysis are based on India KLEMS dataset version 2019 (hereafter IKD-2019). The analysis in this Chapter will be primarily confined to broad sectors of the economy, while more detailed tables are provided in the annexure. The analysis covers the period 1980-81 to 2017-18, subdivided into three sub-periods, 1980-81 to 1993-94, 1994-95 to 2002-03, and 2003-04 to 2017-18.

The methodology used to construct the capital service growth rates is based on Jorgenson (1967) and Christensen and Jorgenson (1969).⁸¹ This approach, rooted in neoclassical theory, required detailed asset-wise data on investment, investment prices, rate of return of capital, and the user cost (or rental prices) of capital for the 27 KLEMS industries. While all the final estimates of gross capital formation (referred to as investment throughout this paper) are consistent with the National Accounts Statistics (NAS) with base year 2011-12, several imputations have been made to generate consistent series of capital input for individual industries and assets. For this purpose, we have relied heavily on the annual survey of industries (ASI) for the formal manufacturing sector and National Sample Survey (NSS) for the unorganised manufacturing sector. Moreover, we have obtained detailed asset-industry data on GFCF from NAS, which are not usually published.

The aggregate capital services growth rates are measured using estimates of asset-specific capital stocks and user costs (or rental prices) for three individual assets. The three assets considered are construction, machinery, and transport equipment. Rental prices or the user costs of these individual assets are calculated using distinct asset prices and depreciation rates, for each of the 27 KLEMS industries, which includes 13 manufacturing industries and 14 non-manufacturing industries. This basically means that the India KLEMS consists of data on GFCF, investment price deflators, rental prices, and capital stock for three individual assets for all the 27 industries. All this information is combined to produce a measure of aggregate capital service growth rates or the rental weighted growth rates of three individual asset-wise capital stock, which forms the capital input used in the underlying growth accounting exercise in the India KLEMS. The chapter will not discuss the details of the underlying methodology used in calculating capital stock and capital services, nor the construction of basic investment data series. For a detailed explanation of the methodology used in constructing the capital input series in IKD 2019, please refer to the India KLEMS Data Manual.

We observe an increasing role of equipment capital (i.e., machinery and transport equipment) in the Indian economy, which helps improve capital composition effects. Capital services growth rates are mostly higher than capital stock growth rates across the board. Much of the recent expansion in capital services and capital deepening is in the market services sector.

The chapter is organised in five sections, and the rest of it is structured as follows. In the second section, we focus on the composition and growth in real investment across broad sectors of the economy. We discuss the distribution of investment in the Indian economy across sectors by two broadly-defined asset categories—equipment and non-equipment, and trends in real investment growth across these two categories. It also documents the investment-to-GDP ratios in different sectors of the economy. Section 5.3 seeks to analyse the changing composition and growth rate of estimated capital stock by broad sectors of the economy over the years. In Section 5.4, we discuss the capital service growth rates, along with the contribution of equipment and non-equipment assets to capital service growth rates in each of the broad sectors of the economy. Section 5.5 looks at the composition effects, where we compare the growth rates of capital stock with that of capital services. The final section concludes.

5.2 Composition and Growth of Investment

In this section we examine the changes in the composition of investment across sectors and assets, along with the growth rate of real investment. Detailed industry tables are provided in the annexure.

5.2.1 Distribution of Investment across Broad Sectors

Figures 5.1 and 5.2 depict the distribution of total gross fixed capital formation, along with equipment and non-equipment assets (throughout this report, non-equipment refers to building and construction assets) across the seven broad sectors of the economy. Among these, the non-market services sector constituted the highest share in aggregate investment in the Indian economy, on average, for the entire period, followed by market services. Together, these two sectors, which constitute the economy's service sector, comprised at least half of the total investment in the economy across the three time periods we have chosen. Manufacturing came third, except for the period 1994-95 to 2002-03, during which it had a slightly higher share than market services. This was likely a reflection of the post-reform investment uptick in the manufacturing sector.

The manufacturing sector registered the biggest improvement in investment share in the 1990s, followed by non-market services and construction. All other sectors dropped their shares. However, in the post-2000 period, manufacturing share dropped substantially by more than 5 per cent, which was almost entirely offset by market services. Other sectors that lost shares include utilities, agriculture, and non-market services, whereas construction share improved.

There are, however, compositional differences among sectors regarding their relative share in aggregate investment. The manufacturing sector maintained an edge over other sectors in terms of equipment investment in the 1990s and 2000s. Industries in the market services sector and utility industry are the others that had high equipment share, whereas agriculture, mining, and construction industries contributed relatively less. In the case of non-equipment, however, non-market services constituted the highest portion of the investment. For instance, non-market services occupy more than 40 per cent of all construction investment in the economy. Although manufacturing had about 16 per cent in the 1980s and 1990s, its share fell to 12 per cent in the 2000s, which was somewhat captured by market services that improved from 14 per cent in the 1980s to 27 per cent in the 2000s. Agriculture nearly halved its relative presence in construction investment from 19 per cent in the 1980s to 10 per cent in the 2000s.

However, the investment to GDP ratio measured in real terms was the highest in the utility industry, which is not unexpected, given the high capital intensity in the energy sector (Figure 5.3). Among the other sectors, the highest investment to GDP ratio was in the non-market services, which includes the government sector. Agriculture and construction industries had the lowest investment intensity. This pattern holds across the three time periods. However, investment intensity increased in all the sectors in the 2000s compared with the 1990s, except manufacturing. The highest increase in the non-utility sectors was in construction, followed by non-market services. Agriculture also saw an increase, in fact, a consistent increase from the 1980s to the 1990s and further in the 2000s. On the other hand, the investment/GDP ratio in manufacturing, after a remarkable rise in the 1990s, dropped marginally in the 2000s. In fact, in the 1990s, investment intensity fell substantially in mining, utilities, and market services. Simultaneously, it improved in manufacturing, non-market services, and construction, leading to a nearly stable overall investment/GDP ratio. In the 2000s, with the substantial increase in the utilities and most sectors except for manufacturing, the overall investment intensity improved.

5.2.2 Growth rate of real investment by broad sectors

In the India KLEMS database, the capital stock and capital services are constructed using real investment, measured as nominal investment deflated by relevant asset-wise investment deflators. The growth rate of real investment for the broad sectors are plotted in Figures 5.4 to 5.6. These are the growth rates of simple additions of individual industry investments and are not weighted growth rates. Figure 5.4 shows the growth rate of total investment across all assets, and in Figures 5.5 and 5.6, we divide the total investment into equipment and non-equipment assets, respectively.

Real investment growth was the highest in the aggregate economy during the 2003-18 period. This was driven by faster growth in equipment investment. During the other two sub-periods, the aggregate investment growth hovered around 5 to 6 per cent, whereas it was about 8 per cent during the 2000s.

Aggregate economy real investment growth was 5.3 per cent in the 1980s, when investment grew between 5 and 7 per cent on average in mining, manufacturing, utilities, and services—both market and non-market, whereas it was dismal at 1.3 per cent in agriculture, and moderate at 3.4 in construction. In the 1990s, investment growth improved in the economy, with construction and non-market services, both seeing a rapid rise from the previous period. While manufacturing still registered an impressive 5.5 per cent growth, it was lower than the previous period. In the 2000s, the agriculture sector, which had already improved its growth in the 1990s over the 1980s, further sustained a growth rate above 4.5 per cent. Among the other sectors, mining, construction, and market services grew at a substantial rate, although the construction sector saw an erosion in its growth rate from the previous period. Investment growth in manufacturing improved, registering an 8 per cent growth rate, while it fell in the non-market services sector.

Among the two asset categories, investment in construction grew at about 4 per cent in the economy in the 1980s, which increased to 8.5 per cent in the 1990s. The rise was primarily a result of a massive increase in construction and market services, but other sectors such as agriculture, manufacturing, and non-market services also showed an uptick. While the Indian economy seems to have seen a construction boom in the 1990s, the overall growth rate of construction investment fell to 5.4 per cent in the 2000s. So, while construction investment increased in all sectors in the 2000s, with the highest increase in construction, mining and quarrying, and market services, the rate of increase in most sectors was lower than in the 1990s. While agriculture, manufacturing, and non-market services saw a growth rate of about 3-5 per cent, other sectors had a 7 to 9 per cent growth rate, resulting in overall growth of 5.4 per cent.

Equipment investment had its best performance in the 2000s compared with the previous periods. While it grew at 8 per cent in the 1980s, its growth performance was very dismal at 2 per cent in the 1990s—mainly on account of market services. The 2000s experienced a significant boost with nearly 12 per cent growth—with equipment investment showing an even higher increase in general, and double-digit growth in most sectors. Agriculture was the only exception. Even there, the growth rate was an impressive 8.5 per cent. Market services were the best performer among all the seven broad sectors in terms of equipment investment growth.

Source: India KLEMS dataset 2019 version

5.3 Composition and growth rates of capital stock

Using real investment by different asset types,⁸² we estimate capital stock for each asset type by industry, using the standard perpetual inventory method. In doing this, we use asset-specific depreciation rates (see the data manual). This section documents the trends in and composition of capital stock across broad sectors of the economy. First, we examine the changes in the composition of capital stock, by looking at the share of equipment and non-equipment assets in total capital stock. Then, we look at the growth rates of capital stock over the three time periods across broad sectors of the economy.

5.3.1 Changing composition of capital stock

Figure 5.7 depicts the composition of aggregate capital stock over the last 37 years, which shows a clear shift from non-equipment capital to equipment capital. Yet the overall capital stock in the economy is dominated by construction capital stock. For instance, the share of equipment capital stock increased from 15 per cent in 1980 to 32 per cent in 2017. Indeed, this has resulted in a drop of non-equipment share from 85 per cent to 68 per cent, yet close to 70 per cent of all capital stock in the economy consists of buildings and structures.

Among the broad sectors of the economy, the highest share of equipment capital in the overall capital stock is in the mining sector (Figure 5.8). Among the non-mining sectors, manufacturing, utilities, and construction sectors maintain dominance over other sectors, and non-market services and agriculture are the ones with the lowest equipment capital stock among the broad sectors. The share of equipment in total capital stock increased over the years in all sectors of the economy, except for market services where it declined.

The overall manufacturing sector clearly shows an increase in equipment share but a closer look at the disaggregated industry level suggests that the high equipment share is primarily in petroleum, paper, transport equipment, and food products industries (Annexure Table 5A.3). Many industries, such as textiles, non-metallic mineral products, and machinery, have seen a fall in equipment share.

5.3.2 Growth rates of capital stock

Figure 5.9 shows the annual growth rate of capital stock in each of the three assets in the aggregate economy, and in Figure 5.10, we have the average growth rate of total capital stock in each broad sector. As shown in Figure 5.9, growth in equipment capital was much more volatile⁸³ compared with non-equipment and aggregate capital stock. In the 1980s through early 2000s, equipment capital grew somewhat stable, but after a drop in 2002-03, it accelerated substantively until the global financial crisis. Ever since then, it started falling, along with the construction capital stock. But again, during 2015-17 it shows more resilience. On the other hand, construction capital increased consistently until the global financial crisis and started falling since then.

Total capital stock in the aggregate economy grew at around 5 per cent in the first two sub-periods, which accelerated to 7 per cent in the 2000s (see Figure 5.10). Interestingly, in terms of the asset composition, while the growth rate of construction capital stock improved from 4 per cent in the 1980s to 5.3 per cent in the 1990s, equipment stock growth fell from 7.4 per cent to 5.3 per cent, thus both assets registered the same growth rate in the 1990s. The uptick in the total capital stock growth in the 2000s is supported by both assets, although much rapid improvement is seen in the equipment capital stock.

5.4 Capital service growth rates

Aggregate capital service growth rates, obtained as weighted growth rates of individual asset-specific capital stock, with user-cost as weights, are provided in Figure 5.13. The user costs were obtained using the methodology developed by Jorgenson (1967) and Christensen and Jorgenson (1969) but deploy an external rate of return.⁸⁴ The aggregate capital service growth rates in the chart are divided into contributions from equipment assets and non-equipment assets. First, aggregate capital services in the economy have increased from about 5 per cent in the 1980s and 1990s to close to 8 per cent in the 2000s. Second, most of these increases in aggregate economy capital services come from construction and services sectors, with market services dominating the latter. The manufacturing sector has seen somewhat stable growth rate of capital service over the entire period of analysis, although it was relatively better in the 1990s.

Looking at the contribution of the two asset types—equipment, and non-equipment (construction), in absolute terms, the contribution of construction and building investment fell in the 1990s in agriculture, mining, manufacturing, and utility sectors. At the same time, it improved in the construction and services sectors (both market and non-market service). The volume of increases was much higher than the volume of declines, resulting in an overall improvement in the construction's contribution to capital services growth in the aggregate economy. In the 2000s, the contribution of construction capital to capital service growth rate further increased across the board, except in the manufacturing sector, where it fell marginally.

The contribution of equipment capital— often argued to have a higher role in boosting economic growth (De Long and Summers, 1991)⁸⁵ —increased only in agriculture in both periods. Among the other sectors, it fell substantially in mining, utility, and market services, and slightly in non-market services in the 1990s, while it improved in the mining and manufacturing sectors. However, the improvement in manufacturing has been minor and was confined primarily to petroleum, chemicals, electrical & optical equipment, and transport equipment sectors. In all other sectors, it fell (Table 5.1). In the 2000s, construction and manufacturing saw declines in equipment capital service growth rate, while all other broad sectors saw improving contribution from equipment capital. The decline in manufacturing was modest, as the substantial fall in several sectors including petroleum, electrical & optical equipment, and non-metallic minerals, was somewhat offset by improvements in basic metals and machinery sectors (Table 5.1).

In relative terms, in the 1980s, more than 80 per cent of the overall capital services growth in the agricultural sector and more than 60 per cent in the non-market services came from investments in new buildings. In the 1990s, although the relative importance of non-equipment capital fell in the agriculture sector, it still contributed more than 60 per cent of the sector's overall capital service growth rate. Simultaneously, the relative contribution of construction assets increased to above 80 per cent in the non-market services. Moreover, market services also joined this group, with almost 3/4th of its overall services coming from construction assets. The high relative contribution of construction in these three sectors led to more than half of the overall capital service growth in the entire economy being driven by construction investment accumulation. However, in the 2000s, construction investment's relative importance fell across the board except in the construction sector. Yet, the non-market services and agriculture maintained relatively high contributions from investment in buildings—50 to 65 per cent.

Interestingly, in the 1980s, 60 per cent of overall capital service growth in the aggregate economy was driven by equipment capital accumulation. This originated primarily from mining, manufacturing, utilities, and construction, where 60 to 90 per cent of capital service growth was from equipment. However, the relative importance of equipment capital fell in the 1990s. Although it improved in the 2000s, it was still less than 60 per cent of the overall capital service growth rate observed in the 1980s. The drop in the 1990s and the improvement in the 2000s were both driven primarily by services.

In relative terms, the contribution of equipment capital improved in the agriculture sector the most, from just 14 per cent of overall capital service growth in the 1980s to nearly 40 per cent in the 1990s. At the same time, it improved from 68 per cent to 74 per cent in manufacturing and fell everywhere else, with two services segments of the economy seeing rapid declines. Overall, with a substantial drop in equipment contributions in the market and non-market services, the contribution from equipment in the economy's capital service growth rate dropped to less than half in the 1990s.

In the 2000s, however, the contribution of equipment capital relative to non-equipment increased, everywhere except in construction, with considerable improvement in utilities and services and marginal in other sectors. In the construction sector, the absolute contribution of equipment capital growth fell from 9.7 per cent to 8.3 per cent, while that of construction improved from 3.6 per cent to 4.4 per cent, thus lowering its relative importance and leading to a fall in the overall capital service growth in the sector.

5.5 Capital composition effects

This section looks at the difference between the growth rates of conventional capital stock measures and our measures of capital services. This difference is often considered an indicator of the quality of capital, as it is a direct function of the change in capital composition towards higher quality equipment (Jorgenson, 2001). For instance, increasing the share of ICT equipment in the total capital increases the overall quality of capital, as they have relatively high marginal products. In this paper, following Harper et al. (1989), we call it a ‘composition effect’, as it is primarily a result of accounting for asset composition changes.

Capital stock and capital service growth rates for the 27 industries are given in the annexure Tables 5A.5 and 5A.6. The results suggest an acceleration in capital service growth rates over the three time periods in the aggregate economy and more than half of the industries. Table 5.2 provides the capital composition effects for three time periods, for each of the 27 industries and the aggregate economy. Faster growth in capital services compared with capital stock would indicate a positive composition effect, while the opposite would suggest a negative composition effect. Interestingly, the overall composition effect was high in the 1980-93 period, while it was negative in the 1990s, and turned positive again in the 2000s. These aggregate pictures, however, are reflections of wide-spread positive composition effect across sectors in the 1980s and 2000s, while the negative effects in the 1990s were mainly emanating from a few sectors. In the 1980s, out of the 27 sectors, 22 showed a higher capital service growth rate than capital stock growth. Among those with negative composition-effects are sectors, trade, construction, and petroleum. However, they registered a positive composition effect in the 1990s, despite many more sectors showing negative composition effect during this period. Only 17 of the 27 sectors had a positive composition effect in the 1990s, while many sectors, including non-metallic minerals, transport services, machinery, and public administration, had a negative effect. The aggregate economy-wide composition effect was also negative during this period, indicating a shift away from high marginal productive assets. In the 2000s, capital services grew faster than capital stock at the aggregate level, indicating a positive composition effect. As in the 1980s, 22 of the 27 sectors had positive composition effect, with the dominant ones being post and telecommunication services, business services, other manufacturing, paper products & publishing, and transport services. On the other extreme, with negative composition effects, were hotels and restaurants, textiles, electric and optical equipment, and public administration.

5.6 Summary and conclusions

The role of capital in economic growth is widely acknowledged in the economic growth literature. However, capital remains one of the most challenging concepts when it comes to measurement. Barring several controversies around the measurement of aggregate capital for understanding the contribution of factor inputs and productivity to economic growth (for instance, see Cohen, and Harcourt, 2003 for a review of the famous Cambridge Controversy), many economists now use capital stock and capital services to understand sources of growth. From a productivity and sources of growth perspective, what is quite important is to know how much of the services of capital—just as in the case of labour—has been used in the production process so that it can be accounted for accurately. This has led to a wider consensus that an appropriate measure of capital input in the productivity analysis is capital services, commonly measured as a weighted index of asset-wise capital stocks, with each asset being adjusted for its marginal productivity (Schreyer et al, 2003). This paper documents the trends in capital stock and capital services for 27 industries, available in the India KLEMS database. The chapter starts with examining the trends in investment and then analyses trends in capital stock, capital services, and capital composition effects. The chapter is largely descriptive in nature but opens several avenues for further analytical research.

We observe that the aggregate investment growth has improved over the three-time periods we have considered—1980-81 to 1993-94, 1994-95 to 2002-03, and 2003-04 to 2017-18, with the fastest growth being registered in the 2000s. Construction and market services sectors were the two main drivers of the high growth in the 2000s, but all other broad sectors of the economy, including manufacturing, also grew faster in the 2000s. A rapid increase in equipment investment facilitated the faster growth in the total investment in the 2000s, the growth rate of which accelerated substantially across the board, while that of non-equipment investment growth eased compared with the previous period.

In context of capital stock and capital services, we observe that, in general, the composition of capital stock is moving towards an increased share of equipment capital in the overall economy. However, there are differences across sectors. For instance, non-market services and agriculture sectors, despite seeing a consistent increase in the equipment share, maintain a relatively low presence of equipment capital compared with manufacturing construction and market services. The general rise in the share of equipment capital also leads to a positive capital composition effect, resulting in a relatively higher capital service growth rate compared to the capital stock growth rate. Overall, the aggregate capital services in the economy have increased from about 5 per cent in the 1980s and 1990s to close to 8 per cent in the 2000s.

Annexure 5A: Additional Tables

Chapter 6: Trends and drivers of labour productivity in Indian industries

6.1 Introduction

India’s pattern of economic growth and subsequent structural changes during different policy regimes has been summarised in many studies (e.g., Krishna, 2015; Jha, 2015, Krishna, et al., 2017, Erumban, et. al., 2019). The GDP growth rate has been quite remarkable during the period 2001-2011 at 7.9 per cent as compared with 5.6 per cent during the earlier decade of 1991-2001 (Krishna, 2015). The acceleration of growth in the past two decades-1991-2011 has been attributed to, among others, substantial policy changes initiated in the early 1990s and that continued through the 2000s. The period of 2001-11 has been defined as period of service-led growth in India by some economists (Ghose, 2015; Das, et al, 2015), as the industrial sector grew at 7.8 per cent whereas services registered an impressive annual average growth of 9.4 per cent. Consequently, this resulted in a structural transformation marked by a larger share of services sector in the overall GDP, and the several aspects of this structural transformation has been analysed in the literature (Kochhar et al, 2006; Islam, 2008; Papola and Sahu, 2012; Krishna, 2015; Chandrasekhar, 2015; Erumban et al., 2015, Krishna, et al., 2017). The primary sector’s share in GDP⁸⁶ has declined to just 17.9 per cent in 2017-18 (at 2011-12 prices) from 38.5 per cent in 1990-91. The share of the services sector, on the other hand, has increased from 37.6 per cent to 53.9 per cent and the share of the secondary sector just increased from 24 per cent to 28.2 per cent, with the share of manufacturing increasing from 14.9 per cent to 18 per cent. Agriculture contributes only 15 per cent to GDP in 2017 and still employs 42 per cent of the workforce. This structural transformation has thus been from the primary to the tertiary sector, having bypassed the secondary sector. Simultaneously, there has also been a structural shift of the labour force from primary sector (where its share of employment declined from 64.6 per cent in 1990-91 to 42.3 per cent in 2017-18), to services (where the employment share increased from 20 per cent to 33.8 per cent during the same period). This kind of duality cannot happen unless there is a simultaneous structural change happening in the sphere of labour productivity.

Such a structural change is part of the economic development process of developing economies that are characterised by large productivity gaps among their different sectors and industries. An economy’s overall productivity depends not only on what’s happening within industries, but also on the reallocation of resources across sectors. So, the structural change in labour productivity at the economy level will take place whenever there is reallocation of labour from low-productivity to high-productivity sectors.

Measuring labour productivity is important—as the time profile of labour productivity not only reflects how productively labour is used to generate output but also the personal capacities of persons employed or the intensity of their effort. According to OECD (2008) ‘labour productivity is a revealing indicator of several economic indicators as it offers a dynamic measure of economic growth, competitiveness, and living standards within an economy’. It is in this context that the chapter has examined India’s labour productivity growth at the industry level to understand the proximate sources of growth in the Indian economy for the period 1980-2017. We compute and analyse the estimates of labour productivity (LP) growth for each of the 27 industries, using the India KLEMS database.

The chapter is organised as follows: in Section 6.2 we briefly discuss the literature on labour productivity and the methodology and dataset in Section 6.3. Section 6.4 looks at trends in labour productivity for the 27 industries. Analysis of sources of labour productivity and the ‘virtuous’ pattern of growth is carried out in Section 6.5 and Section 6.6 analyses the results for the broad sectors of the economy and the sectoral decomposition for disaggregated industries. A cross-country comparison of labour productivity is covered in Section 6.7 and the last section 6.8 offers a conclusion.

6.2: Literature on Labour Productivity

APO (2018) considers labour productivity and its drivers important because of the close links to GDP per capita. Productivity accurately measures the performance of an economy, an industry and a firm in terms of the quantity and the efficiency with which inputs, especially labour, are used to produce goods and services. The efficiency of labour, through its impact on unit cost of production, would also have bearing on the competitiveness of the economy and on its participation in global supply chains. The importance of labour productivity is also emphasised by Krugman (1994), who said that ‘Productivity isn’t everything, but in the long run it is almost everything’. The writings earlier by OECD (2016) and later by Dieppe(ed.2020) has brought labour productivity in focus once again.

According to Dieppe (ed. 2020), the experience of the emerging market and developing economies (EMDEs) reveals that growth in labour productivity (LP) also helps in reducing poverty through its impact on the long-term growth in per capita income. The report shows (Fig 1, p.i5) that countries with faster labour productivity growth during 1981-2015, are also the ones which succeeded in reducing their extreme poverty at a quicker pace. Dieppe (ed. 2020) also observes that any external shock, e.g., the global financial crisis (GFC) or Covid 19,⁸⁷ is likely to have serious impact on labour productivity, resulting in loss in per capita income and increase in poverty and larger is the impact with more severe downturns. Surges or declines in labour productivity generally follow cyclical upswings and declines in GDP and can be persistent if the countries have weak fiscal positions and necessary safeguards are not taken to protect labour.

Labour productivity is, however, a partial measure of productivity and reflects the joint influence of a host of factors depending on the growth accounting framework being considered. In a gross output framework, the contribution to labour productivity growth comes from four sources: namely capital deepening with more or better capital (technology); labour quality, or changes in the composition of labour, which comes from investment in human capital; contribution of intermediate input deepening which reflects the impact of more intermediate-intensive production on labour productivity; and finally from TFP growth which contributes to labour productivity point-for point (Jorgenson, 2005; p299). However, in a gross value-added framework there are only three contributors to labour productivity growth; namely - changes in labour composition or labour quality, contribution of capital service per person employed or capital intensity, and total factor productivity (TFP) growth. Based on global productivity trends, Dieppe and Kindberg-Hanlon (2020, p. 316) also find that the drivers of labour productivity growth are “capital-deepening (growth of capital per unit of labour input) or technological and organisational changes, including the adoption of more efficient methods of production, in some cases through capital investment (Hulten, 1992)”. The report also finds that ‘technological’ drivers of productivity explain a large percentage of the variation in labour productivity during 1980-2018 and the technological improvements during the period were responsible for the decline in employment by 70 per cent in the EMDE countries and 90 per cent in the advanced economies. The largest negative effects on employment occurred in those industries, which tended to be more amenable to labour-saving innovation than other industries.

International experience indicates that labour productivity of an economy also increases due to reallocation of labour from low-productivity sectors/industries to high-productivity sectors/industries, and Dieppe and Kindberg-Hanlon (2020, p.i8) notes that “labour reallocation towards higher-productive sectors has historically accounted for about two-fifths of overall productivity growth in EMDEs”. They observe that shocks such as the global financial crisis and Covid-19 pandemic may weaken the mechanism of structural change by restricting the mobility of people, thereby slowing the geographical and sectoral labour reallocation.

Labour productivity reflects the volume of output produced per unit of labour input (OECD, 2016) and is described as one of the key indicators of economic performance and as an essential driver of change in living standards. It however, also depends on the use of other inputs, such as capital—both physical and intangible; and technical and organisational change. OECD defines labour productivity as GDP per hour worked⁸⁸ and its growth is measured in GDP at constant prices in the national currency of the country. Dieppe (ed. 2020) also pointed out that GDP per hour worked may be a better proposition to measure labour productivity, but for international comparisons, found it convenient to use output per worker because of wider availability of data for workers than hours. He, however, mentions doing this has the disadvantages of ‘failure to account for the quality and intensity of labour input (pp. 15, Box 1)’.

Though labour input may be preferably measured in terms of hours worked, in the case of India, the data on hours worked are either not available or are of low quality and hence labour input is measured in terms of the number of persons.⁸⁹

6.3. Methodology and Dataset

In this section, we discuss the methodology for computing labour productivity in the 27 individual industries in Indian economy. We find the trend in labour productivity over the period 1980-2017 and estimate its sources of growth. We also use the decomposition method (Fabricant, 1942) to find the within-industry productivity growth component and the between-industry component to understand the reallocation or structural change between the industries.

In the present chapter, labour productivity is defined as real gross value added per unit of person employed. From the growth accounting framework, we find that there is a strong and direct relationship of labour productivity growth with growth in total factor productivity (TFP). It is because growth in output basically depends on the growth in inputs and the efficiency with which these inputs are used, known as TFP and the inputs used are not only labour and capital but also intermediate inputs. However, the gross value-added version is used and labour productivity, its sources and the decomposition are all based on the value added⁹⁰ concept following OECD.

6.3.1 Measurement of labour productivity

While at the industry level ‘i’, labour productivity (LP) is obtained directly by taking the ratio of real GVA (V_i) to employment (L_i), for the broad sectors it may be obtained as the sum of real GVA divided by the sum of employment for all the industries in the broad sector. Thus,

where LP_i and LP_s are labour productivity in the i^th industry and the s^th broad sector; V_i and L_i are the real gross value added and employment for the i^th industry and s is the number of industries in a broad sector. However, labour productivity growth is computed simply by the difference of growth rate of value added and growth rate of persons employed by UPSS. It may be mentioned that although labour productivity is a partial productivity measure, it is closely related to wages and, hence, to living standards (Murray, 2016).⁹¹

6.3.2 Sources of Labour productivity growth

The sources of labour productivity growth are therefore change in capital intensity, change in labour composition, change in intermediate inputs and TFP growth, as shown in (6.3). However, as mentioned earlier, in a gross value-added framework, which is used in the current chapter, the only contributors to labour productivity growth are three; namely - changes in labour composition, contribution of capital service per person employed or capital intensity, and TFP growth, and the growth accounting equation then is:

Dieppe, et.al. (2020, p17, Box 1) however, points out to the limitation in the measurement of both output and inputs (K and L). He argues that “Capital services is a measure of the total physical capital available for production without necessarily considering how much of the existing capital is used actively in the production process (capital utilisation). Similarly, labour input, even if it is finely measured as total working hours, does not account for variable labour effort. This may lead to an overly cyclical measure of productivity.” In such a situation, TFP, which is a measure of residual does not only estimate technological change but also any error in the measurement of capital and labour. The neglect to measure and include ‘natural capital’ as an input is also likely to over-estimate the TFP growth effect. According to Dieppe, et.al. (2020, p18), the measurement of new IT products and changes in the quality of ICT has not been accurately done in many national accounts. However, based on many studies on the impact of mismeasurement on the slowdown in productivity after the GFC, the author concludes (p.19) that “despite some evidence for mismeasurement it is unlikely that a significant part of the broad-based slowdown in productivity growth since the global financial crisis can be explained by it alone (Cerra and Saxena 2017; Syverson 2016).”

6.3.3 Decomposition of labour productivity- within and between effects

In this section we study the contribution of the structural change to growth in aggregate labour productivity for the 27 sectors of the Indian economy for the period 1980 to 2017. We use decomposition methods to understand the contributions of structural change to growth in aggregate factor productivity. The decomposition method applied for this purpose can be traced to the methodology in the study undertaken by Baily, Hulten, and Campbell (1992) in the context of manufacturing plants and which was subsequently modified by Haltiwanger (1997).

The same decomposition method is now widely used to decompose the growth rate of aggregate factor productivity growth in to a ‘within’ industry productivity growth component and a ‘between’ effect, also known as reallocation effect. It is the ‘between effect’ which is used as a measure of structural change by various researchers (e.g., Ghani, 2013; Goldar and Aggarwal, 2015; Goldar, et al., 2017). The same method is applied for reallocation of labour input in the present paper and is discussed below.

The equation 6.5 is similar to equation 6.2, except while equation 6.5 defines labour productivity at the aggregate level of an economy, 6.2 defines it at the sector level. Using the decomposition methodology used by Goldar et al. (2017), the annual change in labour productivity levels is then decomposed into within-industry productivity change and a reallocation effect:

The first term in equation 6.6 is the within industry effect, or the effect of productivity change within each industry. It denotes for each industry, the sum of change in productivity levels weighted by its respective employment share in the previous period. The second term is between industry effect—a measure of worker reallocation across industries—and is calculated by a change in employment weighted by levels of productivity in the previous period. If it is positive, it suggests that workers move to sectors with above-average productivity levels, and this term is often considered as a static reallocation (or between) effect and is also described as the structural change term measuring the movement of workers from low-productivity to high-productivity industries. The last term in the equation is an interaction of change in employment share and change in productivity, and thus it measures the combined effect of changes in employment shares and sectoral productivity in a given period – a co-variance or dynamic reallocation term (Ghani, 2013; Erumban, et.al., 2015). A positive dynamic reallocation term would suggest that the factor input is moving to sectors that see positive productivity growth. We obtain the decomposition of labour productivity growth rates by dividing both sides of equation 6.6 by aggregate labour productivity in the previous period.

6.3.4 Data and variables

The source of data used in the current chapter is India KLEMS database, 2019 version.⁹⁴ The data covers the period 1980-2017, and all the growth rates analysed in the study are calculated for three sub-periods during this period, 1981-93, 1994-2002, 2003-17, with the last sub-period splitting further into pre-and post-global financial crisis, i.e., 2003-07 and 2008-17. All data in India KLEMS refers to the financial year, as with national accounts data. However, the periods in this report will be expressed in terms of the calendar year. For instance, the period 1980-2017 refers to 1980-81 to 2017-18 and financial year 1980-81 is expressed as 1980.

6.4: Labour Productivity – Trends and behaviour

In this section we first explain the behaviour of labour productivity growth in the 27 KLEMS industries in section 6.4.1 followed by a discussion about the changes in the growth rates of labour productivity in section 6.4.2.

6.4.1 Labour Productivity growth in the disaggregate industries

Trends in growth of labour productivity for 27 industries that constitute the KLEMS framework are presented in Table 6.1 and Figure 6.1. It is evident that the agriculture sector observed a consistent growth of around 2 per cent in labour productivity till 2002 and thereafter, a surge between 2003 and 2017, mainly because of deceleration in growth in employment. As both the level of and growth in employment declined at an average annual rate of 2 per cent during 2008-17, labour moved out from agriculture to other labour-intensive industries⁹⁵ with similar skill requirements, mainly construction, which experienced an acceleration of 5 per cent in its growth in employment during the same period. Both agriculture and construction not only had and still have similar skill composition, with 65-68 per cent of unskilled labour and only 2.6-3.4 per cent of high-skill labour respectively,⁹⁶ but both also employ a large proportion of the total employed persons in the economy- 49 and 11 per cent respectively. The other labour-intensive industries are all relatively small in terms of level of employment and do not have much labour absorption capacity. The only other labour-intensive industry with a sizable share in total employment in 2011-12, around 10 per cent, is trade but in 2011, it had only one-third of its employees with low skill and 15 per cent with high skill. Hence, the skill composition and requirement of this industry was different from what was available from the displacement of labour from agriculture. Thus, the trade industry had very limited capacity to absorb the labour that moved out of agriculture and employment in the industry only accelerated at an average annual growth rate of 1.2 per cent during 2008-17. So, while the decline in employment in agriculture boosted its labour productivity, this also led to a decline in the growth in labour productivity of its labour-absorbing industry, namely construction. However, for the overall period of 1981-2017, labour productivity growth in the agriculture sector, although is lower than for the total economy, is still reasonable at 3 per cent per annum.

Within the manufacturing sector (comprising 13 industries starting from ‘food products’ at number 3 to ‘manufacturing, nec.; recycling’ at number 15 in Table 6.1) traditional industries such as machinery, nec.; wood and products of wood; pulp & paper; and basic metals and fabricated metal products have low labour productivity. While wood industry is labour intensive, the other industries also employ substantial labour relative to capital and are thus expected to be low in labour productivity. Further, those manufacturing industries, which lately (2008-17) experienced a surge in the growth in labour productivity are textiles, textile products, leather and footwear, wood and products of wood; rubber and plastic products; other non-metallic mineral products; transport equipment; and manufacturing, n.e.c; recycling.

Three industries, that are also often clubbed for analysis with the ‘non-manufacturing’ broad sector, are mining and quarrying; electricity, gas and water supply; and construction. A close look at the individual behaviour of these three industries reveals that construction not only has the lowest labour productivity among the three, but also experienced a small negative average growth in labour productivity during 1981-2017. However, the negative growth is restricted only to the first sub-period of 1981-93. Despite remaining a low-productivity sector, the construction sector has boomed in the post-reforms period—with employment growing rapidly. The reasons for low productivity could be the massive deployment of labour in the sector especially after 2004-05 when it absorbed most of the labour displaced from agriculture without any significant improvement in technology. So, growth in the sector’s output could not keep pace with growth in employment. On the other hand, the growth in labour productivity in the electricity industry was quite high at 5.5 per cent during 1981-93 but it subsequently became stagnant and later declined after the 2008 GFC. The mining industry however, consistently experienced an increasing rate of growth in labour productivity in all the sub-periods. One of the main reasons for its relative success is the increase in capital intensity of the industry and the opening of some segments of the industry to the private sector, which would have promoted competition and efficiency.

The six industries from number 18 to 23 that are defined as market-services, show much variation among them (Table 6.1). While trade, and post and telecommunication record high growth in labour productivity, the other four industries—hotel, transport, financial services, and business services—display a relatively low growth. The trend in growth in labour productivity in these six industries is also not uniform. While ‘Hotels and Restaurants’, ‘Financial Services’, and ‘Business Services’ show a slower growth after 2002; the other three industries have shown a tendency of faster growth after 2002. In the non-market service sector consisting of four industries (number 24 to 27), variation in labour productivity is not substantive. These industries, except health have grown faster during 2008-17 as a result of gradual reforms in the sector and a fast pace of overall economic growth.

Figure 6.1 showing growth in labour productivity over the entire period of 1981-2017 reveals that the top five industries in terms of growth in labour productivity are chemicals, other non-metallic mineral products, public administration, post and telecommunication, and textiles. The five industries that experienced the slowest growth are construction, wood and products of wood, machinery, n.e.c., other services, and transport & storage. Therefore, large differences in levels and growth of labour productivity across the 27 industries are observed. These results are supported by two earlier studies—by Aggarwal (2018) for 27 KLEMS industries and Goldar and Sengupta (2016) for 2-digit organised and unorganised manufacturing industries. Both these studies also find different growth rates of labour productivity in different industries within the same period and across different sub-periods.

A further investigation of the underlying factors which could explain the differences in the growth of labour productivity is attempted separately for manufacturing industries (Figure 6.2) and the services industries (Figure 6.3). The analysis shows (Figure 6.2) that generally, as expected, the average labour productivity during 1981-2017 in the 13 manufacturing industries has accelerated the fastest in the capital-intensive industries followed by labour-intensive industries and less capital-intensive industries. The same trend is also discernible in the growth in GVA as well as in employment. It indicates that high capital-intensive industries generally observe a fast growth in both GVA and employment as compared with other manufacturing industries. It is, however, observed that growth in labour productivity in the labour-intensive manufacturing industries has not followed a constant trend as it varies between different sub-periods—and it was quite high during the period 2008-17 due to a deceleration of 1.8 per cent per annum in the employment growth during that period.

Detailed analysis of the 10 service sector industries (Figure 6.3) empirically indicates that over the entire period of 1981-2017, growth in labour productivity has been the highest in the high capital-intensive service industries, but the difference in the average annual labour productivity growth between these and the labour-intensive service industries s only marginal. The same is true about the growth in GVA and employment, where we observe that the high capital-intensive service industries show higher growth in both GVA and employment, but the growth in labour productivity is almost the same because of the uniform difference in the two growth rates. Inter-temporarily, we do not find a clear trend in the labour productivity growth across service industries with varying capital intensity. The labour-intensive service industries show an impressive acceleration in labour productivity growth during all the sub-periods due to a very high growth in capital intensity in the economy at 6.9 per cent per annum during 2008-17 compared with just 2.5 per cent during 1981-93.⁹⁷ The growth in labour productivity also accelerated quite fast for the capital-intensive service industries during the sub-periods till 2002, but in 2003-07 these industries experienced a marginal slowdown due to slowing growth in employment from 5.1 per cent during 1994-2002 to 3.4 per cent during 2003-07 and in growth in GVA from 10.6 per cent to 8.2 per cent.

Figure 6.4 depicts an analysis of the growth of labour productivity in manufacturing and service industries based on ICT. The 27 KLEMS industries are divided into ICT-producing, ICT-using and non-ICT industries (see Annexure Table 6.A). The figure reveals that (i) during 1981-2017, within the manufacturing industries the non-ICT industries have experienced the highest labour productivity growth due to a low growth in employment, just 2 per cent per annum as compared with 5.2 for ICT-producing and 3.2 per cent for ICT-using manufacturing industries with a quite similar growth in GVA; (ii) in the service industries it is the ICT-producing industries, which show the highest growth in labour productivity due to high growth in both employment and GVA; (iii) the growth in labour productivity in the ICT-using industries has been in between the ICT-producing and non-ICT industries both in manufacturing and service industries; (iv) inter-temporarily there is no clear trend in growth of labour productivity, thus clearly reflecting the uniqueness of each of the manufacturing and the service industries.

6.4.2 Changes in the growth rate of labour productivity

In order to examine the industry level change in labour productivity over time to understand volatility in the growth rates, we consider two sample periods—the period 1994-2007 is taken as the sub-period starting from economic reforms (including the period of high growth) to the beginning of global financial crisis, and 2008-2017 is taken as the sub-period post-global financial crisis (including tapering of high growth).

Figure 6.5 ranks the industries in terms of change in productivity growth between these two sub-periods. The industries that experienced a sharp growth in the second period are wood; manufacturing, nec; rubber; textiles; and education. The majority of the industries here is labour intensive in nature. The industries that showed the largest decline are electrical equipment; basic metals; coke & refined petroleum; hotels; and business services. The observed volatility in the growth rate of labour productivity is evident from the sharp increases in the region of 1-14 per cent per annum and the sharp declines ranging between 1 and 9 per cent per annum post-2008. The trend calls for more attention to be paid to the specific sectors to minimise the sharp fluctuation in the growth rates for better planning by the industry.

6.5 Sources of Growth in Labour Productivity and the ‘virtuous’ pattern of growth

This section deals with the sources of growth in labour productivity in sub-section 6.5.1 and analysis of the ‘virtuous’ pattern of growth is undertaken in 6.5.2.

6.5.1 Sources of Growth in Labour Productivity

The moot question that arises and could help in understanding the underlying trend in the growth of labour productivity is: what are the drivers of labour productivity? This section seeks to address this question.

Figure 6.6 captures the contribution of the three sources of growth in labour productivity⁹⁸—growth in labour quality, capital intensity (capital deepening) and TFP growth for the 27 industries—for the period 1981 to 2017. The figure shows that across the 27 industries, the mean contribution of growth in capital deepening is 3 percentage points out of the mean aggregate labour productivity of 3.98 percentage points; and is followed by just 0.65 percentage points by TFP growth; and only by 0.27 percentage points by growth in labour composition index (or labour quality index). In most of the industries, the contribution of labour composition growth has been quite marginal— less than 10 per cent of its aggregate labour productivity growth (mean being only 7 per cent). It contributes more than 10 per cent to the aggregate labour productivity growth only in the three industries of transport & storage; other services; and health & social work. In 23 out of 27 industries, capital deepening growth contributed the maximum to labour productivity growth, and in the remaining four industries, it is growth in TFP that contributed the maximum to labour productivity growth. These four industries with substantial contribution by TFP growth are manufacturing, n.e.c.; transport & storage; financial services; and public administration. The contribution by TFP growth is negative in eight industries, which does not auger well for the growth of these industries or for the economy.

Figure 6.7 reflects the findings of an investigation of the drivers of growth based on ICT use. It reveals that during 1981-2017, the average labour productivity growth rate is the highest in the ICT-producing industries and is the lowest among the non-ICT industries. It also shows that the average contribution of capital deepening is higher than TFP growth and labour quality composition in all the three types of industries—ICT-producing, ICT-using and non-ICT industries. However, while the ICT-producing industries have the highest contribution from TFP growth, capital deepening has the highest contribution in the average growth of labour productivity of the non-ICT industries. Evidently, ICT plays an important role in accelerating the growth of labour productivity and positively impacts TFP growth.

Sectorally, capital-deepening growth has contributed substantially to labour productivity growth in all manufacturing industries except manufacturing, n.e.c. In the services sector, it is TFP growth that has done this (with 30 per cent or more contribution) except in trade and business services, where capital deepening has contributed more than 85 per cent of the total growth in labour productivity. In fact, capital deepening has been the primary driver in 23 of the 27 industries.

Some reasons for capital deepening in Indian industry⁹⁹ over the period are (i) reduced cost of capital (Unni, 2015), (ii) relatively low price of capital due to interest subsidies and cheap electricity (Chandrasekhar, 2008), (iii) use of new imported capital intensive technology (Felipe & Hasan, 2006), (iv) market distortions in the form of subsidies to capital and high wages in modern industries (Felipe, 2010), and (v) the emergence and growth of the ‘new service economy’ which is more capital intensive. The adoption of new technology leading to automation and increase in capital intensity of firms and industries, especially in the organised sector in India, is also confirmed by Kapoor (2016), Das et al. (2015) and Goldar (2000). But from a long-term growth prospect, it is necessary that growth in labour productivity is driven more by TFP growth rather than by factor inputs.

However, the Indian experience is similar to the other Asian countries where capital deepening contributes almost 50 per cent to labour productivity growth and is thus an important engine of labour productivity growth (APO, 2018, pp.78). Dana, et.al. (2020, p.270) has identified some challenges in South Asia including India that are an obstacle to the growth of labour productivity—and these are low rates of female labour force participation, limited integration in global value chains, low R & D development, lack of adequate infrastructure, and the large size of India’s informal sector. Overcoming these obstacles can help in the growth of labour productivity.

6.5.2 The virtuous pattern of growth of labour productivity

Once the track record of the 27 industries on their performance on growth in labour productivity is analysed, it is desirable, especially from the policy perspective, to find out which of these industries have followed the ‘virtuous’ path. Mas and Benages (2016) have used two criteria—(i) creating employment, and (ii) achieving growth in labour productivity—to assess if a sector/industry is following a ‘virtuous path’. Figure 6.8 shows there are very few industries in India, which satisfy these two criteria to generate fast growth in employment (above-average growth indicated by vertical blue line) and in productivity (yellow horizontal line) so that these industries could be promoted for creating productive employment. The figure shows that these are the industries of post and telecommunication, rubber and plastic, coke & petroleum, transport equipment, health, manufacturing; n.e.c., and chemical, all of which have experienced higher-than-average labour productivity and employment growth. These industries may be carefully monitored and incentivised to bear the fruits of the virtuous path of high growth in labour productivity and high growth in employment. The industries in other quadrants, especially close to the mean lines, e.g., education, trade, electricity, electrical equipment, food, non-metallic industries, and business services may also be suitably incentivised to move them towards the virtuous path.

6.6 Labour productivity at the broad sectoral level, reallocation and sectoral decomposition

6.6.1 Level of labour productivity in the broad sectors

As the level of labour productivity plays a crucial role in determining the level of earnings and hence the living standard of people, its growth helps in the growth of GDP of an economy. We, therefore, briefly discuss trends in the level of labour productivity in the broad sectors of the economy to understand the changes that have taken place in the structure of the Indian economy. As indicated by standard economic theory and proven by the economic history of several countries, labour would always have the incentive and the tendency to move from less-productive sectors to more-productive sectors to improve its income level. It is, therefore, expected that with economic growth, labour shifts from agriculture, which typically has the lowest relative labour productivity, to manufacturing and then to services. However, this has not happened in India, where manufacturing has been bypassed. It is, therefore, interesting to analyse the behaviour of labour productivity in India’s broad sectors. Table 6.2 reflects the average labour productivity for the broad sectors over the period 1980 to 2017 and its four sub-periods. It reveals that (i) the level of labour productivity has increased for all the broad sectors over all the sub-periods, and for the entire period, (ii) labour productivity varies across sectors, with agriculture recording the lowest level of average labour productivity, at just Rs.48,000 during 1980-2017, as compared with Rs.165,000 in manufacturing, Rs176,000 in the market services, and Rs.240,000 in non-market services sector for the same period; (iii) while the index of labour productivity increased between 1980 and 2017 from 100 to more than 300 for agriculture (Figure 6.9), the index increased to 672, 508 and 375 for manufacturing, market services and non-market services, respectively— indicating that both the level and the growth in labour productivity for agriculture has been the lowest, (iv) the segments of the non-manufacturing sector—mining and quarrying and electricity, gas and water supply recorded very high levels of average labour productivity (Rs 795,000 and Rs 760,000 respectively) during 1980-2017, but (v) the construction sector, which is quite labour intensive and has over the years provided huge employment, suffers from low level of labour productivity (an average of Rs. 150,000 during 1980-2017), and both the level and index have declined during the recent period. So, unless suitable measures including improved techniques are adopted, it would not be a huge draw for labour especially that which is migrating from agriculture.

It is worth emphasising that it is the three high capital-intensity industries—coke, chemicals, and transport equipment—where labour productivity has grown faster than the other manufacturing industries and in these three, capital deepening has been the major source of this growth. Among the service industries too, growth in labour productivity has been fastest in capital-intensive industries and these industries are (Table 6.1) post & telecommunication, business services, public administration, and other services. In all these four high capital-intensive service industries, capital deepening has been the important source of growth in labour productivity and only public administration shows evidence of a significant contribution from TFP growth. Clearly, therefore, in the labour-surplus economy of India, it is not labour but capital-intensity that has accelerated the growth in labour productivity.

6.6.2 Growth of labour productivity in the broad sectors

The trends in growth of labour productivity at broad sector level for the period 1981 to 2017 are summarised in Table 6.3, which shows that the average growth rate of labour productivity for the total economy has an upward trend during the four sub-periods. While the growth was less than 4 per cent before 2002, it was more than 6 per cent during the subsequent period of 2003-17. It is also seen from table that labour productivity in the four broad sectors —agriculture, mining, manufacturing and services (market and non-market) accelerating during the period 2003-17 in comparison with the period 1981-2002. The electricity sector also recorded a high rate of growth in labour productivity at more than 4 per cent, although the rate slowed in the later sub-periods. The construction sector, on the other hand, experienced much volatility and an overall negative growth in labour productivity.

Two recent studies also observed acceleration in growth in labour productivity for the economy and for the manufacturing sector in the post-2000 period. Aggarwal (2018) finds that for the 27 India KLEMS industries, the median growth of labour productivity is 6.7 per cent per annum during 2003-11 as compared with 5 per cent per annum during 1994-2002. Goldar and Sengupta (2016) noted that labour productivity growth for total Indian manufacturing during 1999-2010 was 5.4 per cent per annum as compared with 3.8 per cent per annum during 1990-99.

Analysing labour productivity in India, Dieppe (ed. 2020) has pointed out that labour productivity in India remained low during 1980s due to India’s controlled economy with heavy regulation (the “license raj”) and widespread corruption that stifled manufacturing, investment, and technology adoption. But, after the reforms of 1991, many restrictions on product and factor markets were reduced and trade was made freer with reduced restrictions, spurring a surge in productivity growth (Rodrik and Subramanian 2005; Virmani and Hashim 2011). However, the level of labour productivity in India remains very low, despite some recent convergence, as compared with advanced countries and many EMDEs.