PRODUCTIVITY, TECHNICAL PROGRESS AND SCALE EFFICIENCY IN INDIAN MANUFACTURING: NEW EVIDENCE USING NON-PARAMETRIC APPROACH

This article studies the effects of the economic reforms on the Indian manufacturing industries. Data Envelopment Analysis was used to estimate technical and scale efficiency changes after the 1991 reform initiatives. The estimates suggest, however, that the efficiency of manufacturing industries declined during the post-reform era. The variations are different across industries, and the findings demonstrate the importance of technological progress for improving manufacturing efficiency and productivity in India.


Introduction
The Indian manufacturing sector has traversed a diversified path to industrial development since the independence in 1947.Over the last three decades, the sector has undergone significant changes through various reform programs.Along with China, India is fast emerging as a global manufacturing hub.Manufacturing sector plays a significant role in domestic economy in providing employment and its' significant contributions towards growth.The key reasons for increasing growth are cheap skilled manpower, recent investment in public infrastructure, gradual reduction in government controls, and higher inflow of private investment in 1990s.In terms of manufacturing competence, India ranked second following 2010 Global Manufacturing Competitiveness Index; an index developed by the Deloitte Touche Tohmatsu in collaboration with the US Council on Competitiveness.† † The New Industrial Policy (NIP) introduced in 1991 being outward-oriented abolished licensing of capital goods, reduced number of industries in public sector, increased foreign ownerships in domestic industries, introduced deregulation in small-scale industrial units, reduced trade barriers and induced private investment in infrastructure.These elements of reform program along with others are introduced to enhance productivity and efficiency in Indian industries.‡ ‡ The competition and new technologies generally enhance the productivity and reduce the production costs of industries with comparative advantages.
While there is a growing volume of literature undertaking an explicit comparison of reform processes, the effect of economic reforms on productivity of Indian industries † † Source: Indian Brand Equity Foundation at www.ibef.org/‡ ‡ Refer Pedersen (2000) and Ahluwalia (2002) for comprehensive discussion on the economic reform in India.
remained a matter of considerable debate.The traditional industry argument maintains that the removal of protection may result in large number of industries becoming bankrupt.
Alternatively, advocates of liberalization claim that the effect should be marginal, as only the inefficient industries exit, providing opportunities to the remaining industries to improve their performance.§ § The concept of technical efficiency indicates the degree of success in the utilization of productive resources.Technical efficiency is considered to be an important determinant of productivity growth and international competitiveness in any economy (Taymaz and Saatci, 1997).There are different schools of thought in estimating the technical efficiency.Technical efficiency consists of maximizing the level of production that can be obtained from a given combination of factors.In the Indian context, number of studies examined the technical efficiency of the manufacturing industry, e.g., Page (1984), Little et al. (1987), Patibandla (1998), Mitra (1999), Agarwal (2001), and Mitra et al. (2002), Bhandari et al. (2007aBhandari et al. ( , 2007b) and many others.Krishna and Mitra (1998) investigate the effects on competition and productivity on the dramatic 1991 trade liberalization in case of Indian manufacturing.Using firm-level data from a variety of industries, they find some evidence of an increase in the growth rate of productivity.Driffield and Kambhampati (2003) estimate frontier production functions for six manufacturing industries.Their findings suggest an increase in overall efficiency in five out of the six manufacturing industries in the post-reform period.Mukherjee and Ray (2005) examine the efficiency dynamics of a 'typical' firm in individual states during the pre and post-reform years.Their findings establish no major change in the efficiency ranking for different states after the reforms was initiated.Using a panel dataset of 121 Indian manufacturing industries from § § Refer Golder and Kumari (2002) and Hasan (2002) 1981 to 1998, Pattnayak and Thangavelu (2005) find evidence of total factor productivity improvements for most of the industries after the reform period.
While the 1991 economic reform was radical, India adopted a gradualist approach to reform, meaning a frustratingly slow pace of implementation (Ahluwalia, 2002).It suggests that it is more appropriate to examine the effect of liberalization on manufacturing sectors' efficiency using a longer time span for both pre and post-reform.How did this economic reform program shifted Indian manufacturing into global stage and influencing technical and scale economies of major industries?
In answering this question, we employ a nonparametric approach in explaining productivity changes, technical progress and scale efficiencies of industries within the sector.In this paper, we examine the impact of liberalization on the technical efficiency of Indian manufacturing sector by comparing pre and post economic reform periods.We employ an industry-level panel dataset from 1980-2003 in examining the estimates of efficiency and productivity changes.The rest of the paper is organized as follows.The methodology of the DEA technique along with Malmquist Index is described in Section 2.
Section 3 describes our dataset and variables while Section 4 reports the empirical findings.
The last section adds some concluding remarks.

Methodology of DEA and the Malmquist Index
Analysis of technical efficiency of manufacturing industries in developing countries has received considerable attention in the economic literature in recent years.Recent literature includes Onder et al. (2003) for Turkey, Pham et al. (2009) for Vietnam, Margono et al. (2010) for Indonesia, and Mastromarco (2008) for less-developed countries among others.
Technical efficiency is concerned with how closely the production unit operates to the frontier for the production possibility set.The historical roots of a rigorous approach to efficiency measurement can be traced to the works of Debreu (1951) and Farrell (1957).*** Over the past three decades, a variety of approaches, parametric and non-parametric, have been developed to investigate the failure of producers to achieve the same level of efficiency ,for a detailed survey on such methodologies, see Kalirajan and Shand (1999).In parametric models, one specifies an explicit functional form for the frontier and econometrically estimates the parameters using sample data for inputs and output, and hence the accuracy of the derived technical efficiency estimates is sensitive to the nature of the functional form specified.In contrast, the method of Data Envelopment Analysis (DEA) introduced by Charnes et al. (1978) and further generalized by Banker et al. (1984) offers a non-parametric alternative to parametric frontier production function analysis.A production frontier is empirically constructed using linear programming methods from observed inputoutput data of sample decision making units (DMUs).In this study, we adopt the outputoriented (OO) DEA that seeks the maximum proportional increase in output production, with input levels held fixed.† † † The non-parametric approach entails constructing an envelope of the most productive groups to serve as the frontier for the productive performance of all manufacturing industry groups.Thus, there will be one production frontier for each year of the sample, with differences between the frontiers of any two years representing the technical change between those years.By exploiting the computational *** It can be noted in this connection that Debreu-Farrell measure of efficiency follows radial approach.
Another imported measure, based on non-radial approach, is also used in the literature due to Koopmans (1951).For its obvious similarity with the concept of Pareto improvement it is widely known as the Pareto-Koopmans measure of efficiency.However, we have used the earlier measure in our study.† † † The output and input oriented measures provide equivalent measures of technical efficiency when constant returns to scale exist (Färe and Lovell, 1978) strength of DEA, the Malmquist productivity-change index may be decomposed into multiplicative factors that can be attributed to technical change (TC), technical efficiency change (TEC) and scale efficiency change (SEC).Lovell (1996) gives a clear description of how the DEA based Malmquist approach implements such decomposition.
The conventional setup of Färe et al. (1992) is adopted in modeling the problem as transformation of a vector of inputs x t ∈ R N + into a vector of output y t ∈ R M + .The production technology at each time period t, denoted S t , is identified as the set of all technologically feasible input-output combinations at time t (Lovell, 1996).It is constructed from the data as: where each firm produces M number of outputs using N number of inputs and K is the sample size.
When referred to a constant return to scale (CRS) technology as we have denoted here, we mean specifically that there are no further restrictions on the so-called intensity variables k t λ other than those implied by Equation 1, which also includes strong disposability of inputs and outputs (Färe et al., 1994).The output distance function (and the corresponding Malmquist index) can be computed relative to any type of technology -including variable returns to scale (VRS) -as the case may require.Relaxation of CRS to VRS entails the additional restriction to Equation 1.The output distance function arises when a feasible combination is measured against prevailing technology by ‡ ‡ ‡ D t o (x t , y t ) = inf{θ: (x t ,y t /θ) ∈S t } Such output distance functions are simply Debreu-Farrell's output-oriented technical efficiency measures.In the non-parametric approach, these technical efficiency measures can be readily computed from the output-oriented DEA models of Charnes et al.(1978) (hereafter CCR), and Banker et al. (1984)

(hereafter BCC). § § §
To measure the productivity change of industry group p from year t to year (t+1) and decompose it into some economically meaningful components, we have to calculate few distance functions, two of which are shown below and the others can be formed in the similar way.We solve the linear programming problem The distance function can also be defined when measurement is carried out against a technology of a different period, at time (t + 1) say, by ‡ ‡ ‡ We distinguish the output-oriented distance function measured relative to a production frontier based on VRS by t VRS o D ., § § § CCR and BCC are the seminal applications of DEA in the literature based on constant returns to scale and variable returns to scale formulations respectively.

D t+1
o (x t , y t ) = inf{θ: (x t ,y t /θ) ∈S t+1 } This is obtained as the solution of the linear programming problem: From such distance functions, the (output-oriented) Malmquist index for productivity change is defined as: Following Färe et al. (1994), we also decompose the Malmquist index into components representing the effects of pure efficiency change, changes in scale economies and technical change.Technical progress (TP) and technical efficiency change (TEC) are defined respectively as:  ,  , , and, thus, we can easily decompose TEC as TEC = PTEC×SEC.Therefore, four separate linear programming problems have to be solved for each sample observation in order to calculate the Malmquist index (Equation 4) for it.From Equations ( 7) and (8), it is clear that additional two, based on the VRS technology, are to be solved in order to obtain the decomposed components, namely TC, PTEC and SEC, of Malmquist index of a sample observation.

Data and Variables
The main data source is the Annual Survey of Industries (ASI), which is published by the Central Statistical Organisation of India.The ASI includes only registered manufacturing sectors.We employ data from 1980 to 2003 at the 2-digit level of manufacturing industries classified according to ISIC Rev 3.1.**** The data set is widely used in analyzing Indian industrial sector (see Besley and Burgess, 2003;Lall and Chakravorty, 2005; Pandey and **** For details on the ISIC Rev 3.1 classification see http://unstats.un.org.Dong, 2009).In this study we are analyzing the behavior of total factor productivity and its components during pre and post-reform periods, for that the data is divided into pre (1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991) and post (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004))-reform periods.† † † † The output variable we measure is gross value added figures.Using gross value added at constant prices is a common practice in the Indian empirical literature (Unel, 2003;Goldar, 2004;Rajesh and Duraisamy, 2008).The data on value added are deflated by using appropriate price deflators with 1993-94 as the base.‡ ‡ ‡ ‡ As for the input variables, it includes labor and capital.The ASI provides three measures of labor input.The first is number of workers, which can be approximated as paid manual workers.The second is number of paid employees in the organization, which includes both paid manual as well as non-manual workers.Third is total persons engaged, which relates to all persons engaged by the industry, whether for wages or not, in work connected directly or indirectly with the manufacturing process.The ASI also provides data on the number of days worked by workers and employees in a particular year.In this paper we have used 'total persons engaged' as it is the most suitable measure of labor input in our production function framework.§ § § § Gross fixed capital is used to represent capital.In many studies capital stock is measured by the book value of fixed assets while in others its flow is measured by summing † † † † To compare the performance prior to the July 1991 reforms and that following them, the conventional practice is to draw the line at 1990-91 and thus to divide the time period into the decades of the 1980s and 1990s.However, this division does not accurately reflect the division into periods prior to and following the July 1991 reforms.Indeed, because 1991-92 was a crisis year and the 1991 reforms were a response to, rather than the cause of, the crisis, the conventional practice creates a profound distortion by including the year 1991-92 in the post-1991-reform period.The July 1991 reforms and subsequent changes could not have begun to bear fruit prior to 1992-93.Therefore, 1991-92 is taken as the dividing line between the two periods.The start of the post-1991-reform period is 1992-93 (Panagariya, 2004).
‡ ‡ ‡ ‡ Industry specific wholesale price index obtained from the Office of the Economic Advisor, Ministry of Commerce and Industry.§ § § § It could be easily understood that if one follow a dual cost function framework, paid employee, the second one out of these three, would be the most suitable labor input.
rent, repairs and depreciation expenses or perpetual inventory created from annual investment data.This measure has its own shortcomings.The book value and perpetual inventory methods do not address the question of capacity utilization, whereas the flow measure may be questioned on the ground that the depreciation charges in the financial accounts may be unrelated to actual depreciation of hardware.Thus, following Kumar (2006) gross fixed capital was deflated by using the Wholesale Price Index (WPI) of machinery and machine products (base 1993-94 = 100), therefore we include the real gross fixed capital was included in the function.

Empirical Findings
This section presents and compares the empirical results pertaining to technical efficiency in Indian manufacturing industries during pre and post-reform periods.
Considering each industry group as a proponent of its characteristic type of industrial activity, we are involved in assessing the capacity of each group to use its possessions in generating its output as efficiently as possible.Such an evaluation carries with it the suggestion that industry groups are judged against each other on a common basis of outputoriented efficiency.Thus, it is considered that all industry groups are operating with respect to a common production frontier for manufacturing in India.
The key findings from computing the output-based Malmquist index (MALM) of productivity changes are summed up in Table 1 and Table 2. ***** Table 1 summarizes the performance of all industries during the pre-reform and postreform periods.
All figures in Tables 1-3 are (geometric) average annual changes measured in percentage points.
[Insert Table 1 approximately here] The MALM for the Indian manufacturing during pre-reform recorded an average annual increase of 1.7%.† † † † † Of this, pure technical efficiency change (PTEC) is 0.5% and scale efficiencies change (SEC) is 1.2%.The results imply that productivity growth in Indian manufacturing is mainly efficiency driven during the pre-reform period.TP indicates that it is constant during pre-reform period.This suggests that, in the industries studied, constant technical progress has been the main barrier to achieve high level of TFP during the period under consideration.The efficiency effects were positive and provide some evidence in support of the thinking that investment spending had not been excessive.
The MALM for the Indian manufacturing during post-reform recorded an average annual decrease of 0.4% despite some evidence of technical progress achieved by the industries unlike the pre-reform period.Of this, pure technical efficiency change for -1.9% and change in scale efficiency is -0.4%.Technical progress witnessed a growth of 2 per cent during the post-reform period.Deteriorating efficiency might be attributable to a failure to achieve technology mastery, or might be due to short run cost minimizing behavior in the face of quasi-fixed vintage capital.The decline in the level of efficiency happened in the context of higher technical progress, identified as the upward shift of the best practice technology in all industries.The growth in technical progress failed to contribute to the productivity growth in Indian manufacturing industries, owning principally to failure to improve † † † † † The MALM value in pre-reform period is 1.017 (mean ).Here, (1.017-1*100 ) provides the percentage as 1.7%.By following the same method,, we get TP as 0; PTEC as 0.5 and SEC as 1.2.
efficiency in the post-reform period.This indicates that majority of the industries failed to catch up with the shifting frontier technology, resulting in an increase in their inefficiency.
Distinguishing between pure efficiency change and technical efficiency change permits one to observe the importance of the role that non-constant returns to scale have played over time.The fairly rapid rate of structural transformation has taken place with government performing the job of giving guidelines for development.Such guidelines may take the form of new lines of manufactured output, or new methods for the same manufactured outputboth of which have the effect of replacing plant and capacity.Thus, with the new capacity being installed for new directions and an accelerated pace of technological obsolescence for old lines of manufactured output, one would expect technology to be characterized by increasing returns to scale.At the same time, large variations in the scale economies change could also be seen as a result of the sustained pace of change.
Table 2 presents results of the decomposition of productivity growth for all industries for pre and post-reform periods.It focuses on the overall picture gathered from taking averages over both the pre and post-reform periods.This shows nine out of eighteen industry groups experiencing average annual gains in pure efficiency (PTEC) over the pre-reform period.
The ten industry groups that experienced average annual technical progress were 17 (textile products), 20 (wood and wood products), 22 (publishing), 23 (coke), 26 (non-metallic mineral products), 28 (fabricated metal products), 29 (machinery and equipment N.E.C), 30 (office, accounting and computing machinery), 32 (radio) and 34 (motor vehicles).The remaining eight industry groups experienced technical regress.Technical regress refers to negative rate of technical change.It is not unusual to find technical regress in the manufacturing sector.While estimating TFPG in the organized manufacturing sector of major Indian states, Kumar (2006) observed technical regress in some states during prereform and post-reform.In a similar study by Ray (2002), technical regress was established in case of a number of states.
[Insert Table 2 approximately here] The improvement in pure technical efficiency over the period 1980 to 1991 suggests that there was a learning process, as predicted by theories of intra-firm diffusion (Kalirajan and Shand 2001).While analyzing the trend of technical efficiency and its components, pure technical efficiency and scale efficiency, during pre-reform period, in Figure 1 the periodto-period fluctuations in pure technical efficiency has apparently contributed to the fluctuations in overall technical efficiency.
[Insert Figure 1 approximately here] For post-reform period, four out of eighteen industry groups experiencing average annual gains in pure efficiency.The two industry groups that experienced average annual technical regress were 20 (wood and wood products) and 22 (publishing).The remaining sixteen industry groups experienced technical progress during post-reform period.It is generally perceived that technological progress is the main driving force behind productivity growth, especially in manufacturing industries.In fact, TFP has often been considered synonymous with TP.The performance of the manufacturing sector in this respect have been fairly satisfactory during the post-reform period with an average annual rate of 2 per cent (refer, Table 1).The highest TP was exhibited by coke, petroleum products and nuclear fuel industry followed by office, accounting and computing machinery industry.Finally, the decomposition of technical efficiency for each industry is depicted in Figure 2.
[ Insert Figure 2 approximately here] The detail findings in Table 3 reveal that three industry groups -23 (Coke, petroleum products and natural fuel), 24 (Chemical and chemical products) and 30 (Office, accounting and computing machinery) recorded no change in pure efficiency over the entire pre-reform period.This implies that they were operating at the frontier and thus, were deemed efficient.
During the pre-reform period the top five industries in terms of their efficiency levels include 32 (Radio, television and communication equipments), 20 and 21 (wood and wood products; paper and paper products), 26 and 25 (non-metallic mineral products and rubber and plastic products).The least efficient industries turn out to be basic metals and textile products.The textile products industry is one of the largest industrial sectors in India.
However, efficiency of this sector has been constrained by a variety of factors including low technological base, lack of adequately trained manpower and little research and development to improve product and process technologies.In fact, Table 3 shows that out of 18 industry groups, 11 industry groups enjoyed improvements in pure efficiency at some point during the pre-reform period.
[Insert Table 3 approximately here] The detail findings in Table 3 also reveal that three industry groups -23 (Coke, petroleum products and natural fuel), 24 (Chemical and chemical products) and 30 (Office, accounting and computing machinery) recorded no change in pure efficiency in the postreform period also.The top five efficient industries are leather and related products, basic metals, rubber and plastic products, motor vehicles and machinery and equipment.The least efficient industries are radio, television and communication equipments and wood and wood products.The results shows that out of 18 industry groups, 8 industry groups enjoyed improvements in pure efficiency at some point during the post-reform period.
The two columns (in bold ) of Table 4 showing means for the entire pre-and post-reform periods can be used to get a comparison of efficiency behavior during these periods.In case of most of the manufacturing groups, the sector witnessed a decline in efficiency behavior during post-reform period.A similar finding for the Indian manufacturing sector reported in Mukherjee and Majumder (2007) implies that the Indian manufacturing sector has registered decreasing efficiency in the reform period.During pre-reform period Radio, television and communication equipments industry (32) was the most technical efficient one when compared to other industry groups, in contrast to this, this industry has recorded the lowest efficiency during post-reform period.In the competitive environment i.e., postreform period Basic Metals ( 27) was the efficient one as it is emerged from negative phase of efficiency during pre-reform period.

Concluding Remarks
India has experienced policy changes since 1984-85, and these policy changes, in turn, have resulted in considerable changes in the economy.The nature of these reforms and their likely impact on the economy are well documented in the literature.Since the 1991 dramatic economic reform, it appears that the speed of growth engine has slowed down.
In this paper, we evaluate total factor productivity growth and its components of for the Indian manufacturing sector during pre and post 1991 reform periods by using a nonparametric approach of DEA.The empirical estimates suggest that the efficiency has been higher in the pre-reform period as compared to the post-reform period, contradicting the common belief that economic reform can stimulate firms to promote efficiency.The role of labor and capital play a significant role in reform process in improving efficiency.Fikkert and Hasan (1998) also report the importance of scale economies in improving efficiency of Indian manufacturing particularly after the reform period.Bhattacharya et al. (2010) emphasize the role of sluggish labor market in changes productivity after the reform period.
Also cost of capital could make important differences in productivity and efficiency across industries.
We accept that reforms undertaken so far have enhanced the openness of the Indian economy and have facilitated India in integrating with the world market.The challenge for policymakers, therefore, lies in addressing the question as to what reduces efficiency and why the degree of efficiency has eroded despite the introduction of market-oriented reforms.It also must be acknowledged that technical progress does have a role to play and that technological upgrading may be used to improve performance.During the pre-reform period, productivity is driven mainly by efficiency, but in post-reform it is mainly driven by technical progress.This diversity must be borne in mind in policy context.As the economy opens, greater inflow of capital combined with strong labor market reform program will enhance productivity, efficiency and scale economies in accessing world market.
industrial sector to be a DMU.Since the entire sector may not behave like a DMU and since it is an analysis over the aggregate level data, the study is unable to capture firm-specific heterogeneity in their behavior; and (b) The methodology we have followed has certain limitations and is somewhat restrictive in nature.Ray and Desli (1997) and later Simar and Wilson (1998) generalize the decomposition of Malmquist productivity index further and provide more meaningful interpretation of such decomposition and three of its components.
Since any method may have some relative advantages as well as disadvantages over an alternative measure, our results may have some limitations from that point of view also.

Figure
Figure 1: Pre-Reform Period

Table - 1
DEA Results: (Geometric) Average Annual Measures of TFP Change and its Components

Table - 3
Pure Technical Efficiency Change in Indian Manufacturing Industries during Pre (1980-1991) and Post-Reform Period Figures are annual changes for the period ending in the year shown.A value of unity indicates no change.Value less than one indicate declines in efficiency, while values exceeding one indicate improvements in efficiency.