Abstract

                 Using firm-level panel data we investigate whether reform of the trade and industrial policy regimes in India introduced in 1991 resulted in a reduction in market power and/or an acceleration in productivity growth, consequences that have been predicted in theory. Econometric estimation of a suitably transformed production function for every industry group at the two-digit level in India yielded limited evidence of acceleration in productivity growth and no evidence of a reduction in market power. This is interpreted as suggesting that in the case of Indian industry trade liberalisation has not exhibited the potential often attributed to it.

 

JEL Classification: F12.

Keywords: Economic liberalisation in India; Trade policy reform; Market power; Productivity growth.

 

 

 

Liberalisation, market power, and productivity growth in Indian industry

I. INTRODUCTION

This study investigates whether two outcomes commonly associated with economic liberalisation have in fact occurred in Indian manufacturing industry since the launching of reforms in 1991.  The first is a reduction in market power, measured by the price-marginal cost ratio.  The second is whether productivity growth has accelerated.  These two possibilities have been investigated for a very wide range of economies internationally by now, leaving us with greater appreciation of the theoretical aspects of the relationship between trade and industrial evolution and a wide set of results enabling cross-country comparison, both central to development economics. 

As the reforms in India have been the source of very great interest, and much has been written on the topic, we do not spend any time in describing them.  A detailed discussion, including of phases during some four decades since 1947, is provided in Srinivasan (2000).  A less detailed, but arguably more evaluative, discussion of the measures taken since 1991 can be found in Joshi and Little (1997).  We have only the following to state here.  It would be reasonable to say that during the first phase of the economic reforms launched in 1991 a major concern appears to have been the question of how to foster competition in the manufacturing sector.  The principal means chosen were the removal of licensing of capacity and the, at least partial, liberalisation of trade.  We say “partial” to emphasise that the removal of quantitative restrictions to trade in India was to come well over a decade later as part of an international adherence to WTO requirements.  So the liberalisation of trade commencing 1991 took the form of a lowering of tariffs.  However, this by itself was of a quite dramatic order of magnitude. The movement of the tariff by industry group have been recorded by the World Bank (1998) and by Nouroz (2001). Here we provide evidence of the lowering of the effective rate of protection in manufacturing since 1991. Note from Table 1 that within five years this rate had been more than halved across the board. However, be that as it may, the economic reforms in India were a combination of internal and external liberalisation measures aimed primarily at greater competition.  This character of the policy shift is at times overlooked, with the external liberalisation getting emphasised over the internal one.  This could contribute to faulty inference.  It leads us to view the period since 1991 as one of a general economic liberalisation rather than one of merely trade liberalisation.

 

Table 1 Goes Here

In economic theory, one effect of increased competition - whatever be the source - is to lower supernormal profits measured as the excess of price over marginal cost. In fact, the simple intuition that the mark-up is highest in the case of monopoly, zero for the case of perfect competition and somewhere in between for oligopoly can be formalised (Shapiro 1991) in terms of the relation that, given the elasticity of demand, the price-marginal cost ratio is inversely related to the number of firms in the industry.  In the new trade theory this very basic result has been exploited to envisage (Krugman 1992) an additional element over and above the standard gains-from-trade under perfect competition.  So we see that in economic theory there is a definite prediction as to the impact of entry-inducing economic liberalisation on market power.  The relation between liberalisation and productivity growth, however, is not quite so simple.  A nuanced reading of the argument underlying the prescription of free trade would be that it is Pareto-improving in the space of welfare. As, traditionally at least, only movements along the production frontier are considered, technical progress is not even a matter for consideration.  However, the existence of X-inefficiency cannot be ruled out in practice. Therefore, when we address a situation such as the economic reforms in India then we might wish to consider the likely impact on productivity growth of increased competition.  In the literature (Corden 1974) this has been visualised as stemming from trade liberalisation, but the principal result is invariant with respect to the source of the enhanced competition.  Competition, whether induced by trade or industrial policy liberalisation, is expected to affect the incentive to work, thus eliminating X-inefficiency and consequently raising productivity.  This is not implausible.  However, to hold, this argument requires (Rodrik, 1988, 1992) that the managerial-effort supply curve is backward bending in the relevant range and that changes in work incentives go in the same direction in both export-oriented and import-substituting sectors of industry.

Might trade liberalisation have consequences for productivity growth other than through increased competition?  At least in theory, it does. An obvious consequence of a more liberal trade regime is the availability of superior imported technology, often at a lower cost.  Thus Dornbusch (1992) has argued that this has the effect of shifting out the production function, and thus at least a one-shot improvement in productivity.  While this needs to be acknowledged, it pays to record the observation by Tybout (1992) that it is a mistake to think of productivity growth as an orderly shift in the production function of the representative plant. Instead, he asserts, it is gradual processes of technological diffusion or the displacement of inefficient plants with efficient ones that matter.  Summing-up the research efforts up to then, he records that preliminary work suggests no clear link between trade policies and patterns of entry and exit. This had led him to conclude that “…. although many economists believe that there are important links between trade regimes and factor productivity, they have to look elsewhere for formal models that support their priors." (Tybout 1992). We briefly review some such models now.

     Where there is an expansion of output following trade liberalisation – a feature not relevant for import substituting firms of course - productivity growth may result.  However, for productivity growth to follow from economic expansion unexploited economies of scale must exist.  In the presence of decreasing-returns-to-scale-technology productivity growth could actually slow down.  The following appraisal is of interest: “In the presence of imperfect competition and increasing returns to scale, trade liberalisation is compatible both with a magnification of the welfare gains and with welfare losses.  It all depends on how the economy is expected to adjust, which in turn depends on the frustrating ambiguities of oligopoly theory.  At one extreme we could imagine that free entry eliminates all excess profits and that liberalisation rationalizes industry structure by reducing the number of firms and forcing the remaining ones down their average cost curves. ……  .  But at the other extreme, we can imagine a world in which the contracting sectors tend to be those with supernormal profits and unexploited industry-wide scale economies.  The protectionists’ fears may then well be justified.” ( Rodrik  1988: 110-111). Though addressed to the research on the consequences of trade liberalisation per se, these observations are entirely applicable to the consequences of economic liberalisation more generally.  A `rationalisation’ of industry may be expected to follow from entry no matter what the source, i.e., whether foreign or domestic competition.  However, net effect of the rationalisation cannot be predicted in advance.  It appears that the relevance of empirical work to this question cannot be exaggerated.

     The arguments that we have looked at so far link trade and the level of productivity.  However, there have been theoretical efforts to link trade to the rate of technological diffusion, and via innovation, to productivity growth.  Rodrik (1992) argues that a firm’s market share can effect its payoff from innovation.  Now, the impact of trade reform would be ambiguous, for productivity growth is as likely to slow down in import-competing sector as it is likely to accelerate within the export-oriented ones as their respective markets shrink and expand. A general equilibrium representation of the link between trade, innovation and growth appears in the work of Grossman and Helpman (1991).  For Grossman and Helpman private R&D yields newer intermediate goods which enhances final goods productivity. Trade is crucial to the rate of innovation.  Fresh R&D initiatives are based by entrepreneurs upon the variety of substitute products available, and this depends upon the exposure to world markets.  Secondly, market size encourages innovation as larger markets are necessary to support a greater variety of intermediate inputs. These two effects of greater integration with the rest of the world via trade work in opposite directions though, and could lead to zero net outcomes.  The general equilibrium representation of the effect of trade is yet to be considered, however.  According to Grossman and Helpman, R&D may be seen as using labour and capital, which are also used in the production of tradeables.  Now a change in the trade regime that affects the relative price of these two sets of outputs affects the return to new product development and thus the rate of productivity growth.  In our view, the near exclusive focus on the production of new intermediate products as the driver of productivity growth in this version of the Grossman and Helpman view makes it a little less relevant for India where R&D activity by the private corporate sector is not been particularly vigourous historically.  However, another role of trade in influencing national productivity growth has also been emphasised.  In particular for a developing country, “… an open trade policy regime increases the domestic producers’ interaction with foreign produces and buyers.  This interaction stimulates the cross border learning of production methods, product design, organisational methods and market conditions.” (Grossman and Helpman 1991, pp. 166-7)  This possibility cannot be ruled out as less relevant for India.  Nevertheless, even as this view of trade has many followers we bring to notice the paper by Young (1991) where trade in a model of endogenous growth can lead to a permanently lower rate of technical progress for the less developed country. In the model, this occurs as a result of the specialisation that is assumed to accompany the move from autarky to trade. For we would expect the less developed economy to specialise in primary products assumed to be subject to a lower rate of technological progress compared to manufacturing.    

            The pronounced move towards more liberal trade regimes in the developing countries from the nineteen seventies onwards inspired much research on the consequences.  Notable among these are the one on Chile by Tybout, Corbo and de Melo (1991), Turkey by Levinsohn (1993), Cote d’Ivoire by Harrison (1994), and India itself by Krishna and Mitra (1998).  These studies have focussed on the impact of trade reform on one or more variables among productivity growth, market power and scale efficiency.  In this paper we investigate the impact on market power and productivity growth in Indian industry of a significant reversal of the policy regime, which is referred to as ‘liberalisation’ in discussions on India.  We believe that, in the context, India serves as a major test case, for three reasons.  First, economic policy has always been pursued with some vigour in India, making the testing of its consequences - whatever the policy - particularly relevant.  Secondly, when in India trade reform did eventually appear on the scene of a weaker version of the Soviet model of industrialisation covered by a restrictive foreign-trade regime, it came with a bang.  For, though adopted here over a decade after it had gained ascendancy in East Asia including China, the dismantling of quantitative restrictions on industrial imports – other than on consumer goods - was brisk and the scaling down of the tariff barrier impressive.  The third reason for our belief in the relative importance of the Indian case is purely on statistical grounds.  The sheer size of the Indian economy offers the researcher a bonus in terms of the sample size.  For instance, in certain industry groups analysed by us the very number of firms in our sample exceeds the total number of observations for the same or comparable industry groups in the studies cited above.  While individual researchers can take no credit for this, the sheer statistical advantage of a larger sample cannot be exaggerated.

            Four remaining sections of this paper follow.  We commence by setting out a framework of analysis including the derivation of the equation to be estimated.  Next we outline an estimation strategy, present the results and provide an interpretation of the same.  A conclusion is now offered. Finally, as any econometric exercise is only as good as the data base, the construction of the variables and the sources of the data are discussed in some detail in an appendix.

II. THE ANALYTICAL FRAME

A methodology due to Hall (1988) has been applied, with modification, in studies of the consequence of economic reforms for market power and productivity growth in several countries, notably of the Cote d’Ivoire by Harrison (1994).  In our investigation we adopt, after further modification, this very methodology.

Specify the production function for firm i in industry j at time t as:

 

Yijt = AjtfiwitG(Lijt, Kijt, Mijt)                                                                                                          (1)                                                                        

where Y, K, L and M stand for output, capital, labour and materials inputs, respectively, Ajt is an industry-wide index of  Hicks-neutral technical progress, fi is a firm-specific fixed productivity term while wit isa firm-specific transitory productivity shock. (We have followed Grilliches and Mairesse 1998 in the decomposition of the firm-specific productivity into fixed and transitory components. The modification of the Harrison model is aimed at making the theoretical model consistent with the estimation method to be followed.)   

Totally differentiating (1) and dividing throughout by Y we have[2]

(dY/Y)ijt = (dY/dL)(dL/Y)ijt + (dY/dK)(dK/Y)ijt + (dY/dM)(dM/Y)ijt + (dA/A)jt + (dw/w)it.     (2)

From the first-order condition for profit maximisation of a firm in Cournot equilibrium the expression for the marginal product(s) can be written as:

(dY/dL)ijt =  (w/p)jt{1/[1+(sij/ej)]} = (w/p)jtmij                                                                                                    (3a)

(dY/dK)ijt =  (r/p)jt{1/[1+(sij/ej)]} = (r/p)jtmij                                                                                                       (3b) (dY/dM)ijt =  (z/p)jt{1/[1+(sij/ej)]} = (z/p)jtmij                                                                                                                    (3c)

where p is the product price, w, r and z are the price of labour, capital and materials, respectively, sij is the market share of firm i in industry j,  mij is the price-marginal cost ratio and ej  is the price elasticity of demand for product of the jthindustry .   

Anticipating the estimation - which takes the form of estimating an industry-level production function - it is assumed that while the mark-up varies across industries it is common to the firms within an industry.  Now, substituting (3a) to (3c) into (2) and re-arranging terms, we have:

(dY/Y)ijt = mj[(wL/PY)(dL/L)+(rK/PY)(dK/K)+(zM/PY)(dM/M)]ijt + (dA/A)jt  + (dw/w)it.      (4)

Denoting the factor shares (wL/PY), (rK/PY) and (zM/PY) as al, akand am, respectively, the equation (4) may be re-written as:

(dlnY)ijt = mj[al(dlnL) + am(dlnM) + ak(dlnK)]ijt  + (dA/A)jt  + (dw/w)it.                                               (5)

Denoting the sum of factor shares under imperfect competition as b/m, where bis the returns to scale parameter and b= 1 is the constant-returns to scale case, we can(Chambers 1988: 70, Harrison 1994: 56)re-write (5) as a growth-rate version of the production function in intensive form in capital:

dyijt = mj[aldn+amdm]ijt + (bj -1)(dK/K)ijt + (dA/A)jt    +(dw/w)it                    (6)

where the variables y, n and m stand for  ln(Y/K), ln(L/K) and ln(M/K), respectively.

 

We work with equation (6) in the econometric estimation, replacing (dw/w)it by the term eit.  eitincludes (dw/w)it, the firm-specific transitoryproductivity shock, and a further term u accounting for such factors as may havebeen excluded from (6) yielding

dyijt = mj[aldn+amdm]ijt + (bj -1)(dK/K)ijt + (dA/A)jt    +  eit                                            (6a)            Were Equation (6a) to be treated as a regression, estimates of (dA/A)jt and mcan be thought of as the rate of productivity growth and the price-marginal cost ratio in industry j, respectively. Further, an estimate of (bj – 1) that is not significantly different from zero would imply constant-returns-to-scale technology, while a statistically significant positive coefficient would imply increasing returns and a significant negative coefficient decreasing returns to scale.  However, we do not pursue this in our empirical investigation. The above specification can be used to test for a change across policy regimes in market power mand of productivity growth.  This may be implemented by introducing into the regression a dummy variable applied to the variable-input index [aldn+amdm] and the constant term, respectively, in equation (6a). We now have:

dyijt = Boj + B1jD + B2jdxijt + B3jdkijt + B4jdxijt*D + eit                                                                           (7)

where

Bo = dA/A,

B1 = change in B0, B2 = mj

B3 = (bj-1), B4 = change in B2

dx = [aldn + amdm],

dk = dK/K,

and D is a dummy variable accounting for the policy regime during any particular historical phase. Here 1992-93 is treated as the first year of the changed policy regime.  

III. ESTIMATION

III.1: Estimation strategy

The estimation of a production function is perhaps the most widely discussed problem in the literature on applied econometrics (Grilliches and Mairesse 1998).  The main issue is the endogeneity of the regressors which would render the OLS estimates of a production function’s coefficients inconsistent. In our specification a possible endogeneity arises as follows. Note that eitis in two parts, the first one being the firm-specific productivity shock (dw/w)it and the second, u, a stochastic term allowing for measurement errors and any excluded variables.  The first is unobservable to the econometrician, but being observable by the firm could lead to correlation with the inputs that a firm might adjust in light of its productivity rendering OLS estimates inconsistent. u, on the other hand, has no effect on a firm’s input-demand decision. The appropriate econometric response to endogeneity would be to use instruments for the endogenous regressors. Note that it is only the current and future input demand that may respond to the transitory productivity shock wit. This allows for the use of lagged values of the regressors as valid instruments.

     One route to consistent estimates in the presence of endogeneity is to use the Generalised Method of Moments (GMM) procedure. In this study, we use GMM as developed for the panel data context by Arellano and Bond (1991) to estimate the parameters of the model set out above. Other methods of solving the problem of simultaneity of course exist. These are the methods proposed by Olley and Pakes (1996), improved upon by Levinsohn and Petrin (2003), and Blundell and Bond (2000). However, these approaches address, specifically, the estimation of the parameters of a production function. On the other hand, the Hall equation which we work with is a modified production function enabling the simultaneous estimation of both productivity growth and the price-cost ratio, the parameters of interest to us. Note that the simultaneous estimation of  productivity and the price-cost ratio is an advantage as periods of liberalisation may be associated with a change in the latter which affects measured productivity growth. This is demonstrated by Harrison (1994). 

            As our equation (6a) is already in growth rate form and does not contain a fixed effect, no further transformation of the variables is required, and we directly estimate this equation using lagged values of the regressors as instruments. As in Arellano-Bond, values of the regressors from the second lag onwards are used as instruments.

III.2 The Data base

The data for the present study is drawn from the database PROWESS of the Centre for Monitoring Indian Economy (CMIE).  CMIE provide annual data on firms registered with the Bombay Stock Exchange, limited to public limited companies.  Firms have been collected by us under industry groups broadly as under the National Industrial Classification (NIC) at the two-digit level.  It may be noted that we have constituted a group ‘Other Industries’ comprising ‘Leather and Leather products’, `Wood and Wood products’ and ‘Miscellaneous’ in the original source to ensure a reasonable number of observations for the estimation.  Firms for which implausible values (of the variables) were encountered were excluded.  The final data set was an unbalanced panel consisting of 3582 firms for the ten-year period 1988-89 to 1997-98. The last year of the period was determined for us by data availability at the time of the commencement of the study.  On the other hand, the earliest date for which CMIE data are readily available, i.e., in electronic form, is 1988-89. As far as we aware, the time series thus assembled covers a longer period – 1988-89 to 1997-98 - than most comparable studies for India. Firms in the industry groups chosen for the study account for nearly 73 percent of the value of output of the organised manufacturing sector and approximately 70 percent of value added for the same sector in the year 1997-98, the final year for which data are available in our sample.  The distribution of firms across industry groups and the number of observations corresponding to each industry group are provided in Table A1 of the Appendix.  On an average, these firms account for more than 60 percent of the output of the corresponding industrial group reported in the Annual Survey of Industries.

 

III.3Results

Equation (7) was estimated for all industry groups by the software Panel Data Estimation usingDPD for Ox(Doornik et al 2002).  In order to capture the likely consequences of the change in the policy regime initiated in 1991 the dummy variable is set to zero for the years prior to 1992-93 and to one for all subsequent years inclusive of it. This implies that we are allowing for any impact of liberalisation to show up from 1992-93. The choice of the break-date needs to be explained. One reason is purely related to data frequency. The reforms commence in July 1991, or the second quarter of the financial year 1991-92. As the data is annual we allow for a break at the first opportunity. A second reason is more substantive. The liberalisation set rolling in 1991 was combined with a macroeconomic stabilisation programme that included severe import compression to rein-in the balance of payments deficit. That the import compression actually bit is evident from the statistic that non-oil imports regain their pre 1991-92 level only in 1992-93 (see ‘Economic Survey 1993-94’, p. 79, Government of India). For both these reasons it does not appear to us advisable to test for an impact of trade liberalisation commencing 1991-92 as some researchers have done.      

               The estimates are presented in Table 2. First, note that the Sargan-test statistic signals the validity of the over-identifying restrictions, in turn validating the instruments used. Furhter, absence of second-order serial correlation of the errors – indicated by the statisic labelled AR(2) – satisfies the specific requirement for instrument validity when the second lag has been included in the instrument set as has been done here.  Before proceeding to the results of the test of the hypotheses that interest us, viz., the impact of liberalisation, we would like to observe that the point estimates obtained here are reasonable. In particular, note that the estimated price-cost ratio is greater than one, for all cases other than ‘Rubber’, ‘Machinery’ and ‘Metal products’, that is for ten out of the thirteen industries.  This condition, to be expected from economic theory, is often violated in comparable studies where unacceptably low estimated price-cost ratios are presented without comment.           

                                                            [Table 2 goes about here]

                                                            [Table 3 goes about here]

            The sign of the estimated change in productivity growth and in the price-marginal cost ratio for each industry have been gathered together in Table 3, along with an indication of whether the estimate is statistically significant at the five-percent level. We are now in a position to state the principal findings of our study.  With regard to productivity growth, we find that there is a statistically significant rise in this rate for three industry groups, viz., Paper, Basic Metals and Metal Products.  On the other hand, there is a statistically significant decline recorded for two groups, namely, Rubber, Plastic & Petroleum Products and Chemicals. It would, therefore, be correct to say that there is no evidence of a widespread rise across industry groups in the rate of productivity growth since the onset of the reforms for the period that we have studied. With regard to the price-marginal cost ratio, we find no instance of a statistically significant decline in the ratio.  However, the implication of this finding needs to be handled with some care for we may expect to find any decline in market power only in instances where such power actually exists.  Our estimates allow us to test for the existence of market power in the period prior to the policy change.  The results of a test for whether the price-cost ratio is greater than or equal to one are presented in Table 4.  Note that only in the case of two industries, viz., Chemicals and Transport, is there evidence of market power in period prior to 1991. In all other cases the null of a price-cost ratio equal to one cannot be rejected.[3] However, on the basis of the results presented in Table 2, it would be appropriate to conclude that there is no evidence of a decline in market power in Indian industry since the liberalisation of the economy.

[Table 4 Goes About Here]

Before concluding, we evaluate our work in relation to similar studies on Indian manufacturing.  The only comparable study of the impact of liberalisation on market power and productivity growth is that of Krishna and Mitra (1998).  In an early study of the Indian experience, Krishna and Mitra had investigated four industry groups chosen on the basis of the extent of liberalisation defined in terms of tariff reduction.  They report “... strong evidence of an increase in competition (as reflected in price-marginal cost mark-ups) and some evidence of an increase in the growth rate of productivity”  (p. 447).  Krishna and Mitra’s results really are applicable only to a small part, indeed only 2, of the 13 industries that we study, viz., to the groups ‘Machinery’ and ‘Transport’.  While we are able to replicate their finding of an increase in the rate of growth of productivity and a decline in the price-cost ratio in these groups (see Table 3), the estimated changes not statistically significant.  We provide a plausible explanation.  First, we look at differing periods.  Our data starts from 1988-89 while that of Krishna and Mitra start with 1986.  At the other end, our data extend much longer, up to 1997-98 compared to 1993 in the study of Krishna and Mitra. Secondly, all of our input variables are constructed a little differently from that of Krishna and Mitra, notable differences being that we use what are arguably superior deflators for money wages and have attempted to evaluate the capital stock at replacement cost. The precise details may be gathered from the Appendix.  Finally, we have allowed for the endogeneity of regressors, an advisedly superior procedure in the estimation of production functions.  It is unlikely that these aspects would not have influenced the outcome.

            It is of interest to compare the results presented by us here to those of studies of the effect of trade liberalisation elsewhere in the developing world. Generally, there are more studies on productivity growth following liberalisation than there are studies of changing market power, and the comparisons that we will make must reflect this. In a study of Chilean manufacturing over the years 1967 and 1979 Tybout et al (1991: 241) find that there is "... no evidence of overall improvements in productive efficiency for the manufacturing sector", with estimated returns to scale higher in less than 50 percent of the industries studied.  Our estimates for India are exactly in line with this finding.  Tybout et al do not investigate the behaviour of the mark-up across policy regimes just as we do not investigate the change in the returns to scale. On the other hand, in her study of the Cote D'Ivoire Ann Harrison examines the change in both market power and productivity growth across policy regimes.  She reports a statistically significant decline in the mark-up in only one out of nine industries post-trade-reform.  The figure for productivity growth is two out of nine. These comparisons show that our findings for India are broadly in line with those of researchers studying other developing economies. Finally, Paus (2004, p. 427) reports evidence from studies on Latin America as follows: “After 15 years of neoliberal reforms in Latin America productivity growth has been very disappointing.”     Having compared our findings to those of other important studies on the impact of trade on market power and/or scale efficiency internationally, we consider it appropriate to make an observation regarding the database of our study in relation to these.  First, we have used a continuous panel in this study as opposed to comparing parameters of interest at different points in time – prior to and after a regime change - as do Tybout et al.  The sample size is another issue.  Quite often, there are more firms in an industry within our sample than there are number of observations in the Harrison study.  This, of course, has to do with the fact that India has a much larger economy but, as we mention in our introduction, the statistical advantages of sample size cannot be overlooked. 

IV. CONCLUSION

In 1991 a major regime shift was effected in the Indian economy. This sweeping economic liberalisation was led by the removal of industrial licensing and the lowering of import tariffs, both of which had constituted formidable barriers to entry in the past.  Based on the predictions of economic theory and a strongly-held view of international development agencies we have sought evidence of some expected consequences of this liberalisation.  We have provided estimates of the change in productivity growth and in market power since 1991-2 for the entire manufacturing sector disaggregated at the two-digit level of industry.  We find some evidence of an increase in productivity growth, but for less than half the industry groups.  For market power, there is no evidence of decline.  As India is a large country with a substantial private sector, and the reforms initiated in 1991-92 themselves constitute an exogenous policy shock we have in the event something of a controlled experiment.  We believe that the results reported by us here are important, constituting as they do an improvement over what exists in the literature in terms of length of the time series, better variable construction, superior econometric methodology, and greater coverage of the Indian manufacturing sector. 

            Even though the title of our paper refers to liberalisation, as India witnessed both trade and industrial policy results, we believe it appropriate to interpret the results presented here as having a greater bearing on the potency of trade policy in the Indian case.  This may appear as a deviation from our having earlier cautioned against the reading of all outcomes as having been effected by changes in the trade-policy regime. However, it is not so. For while it is correct that in 1991 industrial policy changes had figured prominently in the reforms in India, we would be faulted for ignoring the relative roles of industrial and trade policy in fostering competition and accelerating productivity growth in the short run.  Indeed we believe that a priori trade policy may be expected to act faster. Why do say so?  In our view, the principal feature of the industrial policy reforms in India was the removal of industrial licensing.  At the simplest, this may be read as enabling entry previously restricted.  The pro-competition effects of this may be considered to go in the same direction as the liberalisation of trade with respect to import-competing sectors.  It is our judgement though that, on balance, the competition-enhancing impact of trade policy reform is likely to be more immediately effective than that of the removal of policy-induced barriers to entry for domestic industry.  The removal of such legal barriers to entry does not imply the removal of, conceivably more substantial, economic ones. On the other hand, when capacity exists overseas, imports can cross borders with ease in response to trade liberalisation.  Surely, as far as market power is concerned, the impact of trade liberalisation is likely to be more immediate than that of domestic entry liberalisation, even though a variation across the manufacturing sector may be expected with respect to this feature.  We take the view, therefore, that in the entire set of policy changes that were put into place in India since 1991, at least in the first phase, trade reforms must count for more than industrial policy reform as far as their relative competition-inducing effect is concerned. This we claim on grounds that closure and exit remain a difficult proposition in India’s manufacturing sector, where retrenchment and layoff of workers requires government authorisation.  The removal of legal barriers to entry need not therefore imply that economic barriers to entry have been removed, for exit as an option cannot be assured to the firm. Thus the pro-competition effects of industrial policy changes in India are likely to have been limited, for while they abolished compulsory licensing of capacity they did not alter the feature that an implicit no-exit policy may have continued to restrict entry.  However, while the new industrial policy may have continued to constrain competition arising out of the entry of domestic firms, the trade policy reforms at least may be expected to have introduced competition, prima facie and on the margin, via the threat of imports.  Going by this reasoning, the results reported in this study may be interpreted as suggesting that the beneficial impact of trade on market power and productivity growth may have been exaggerated. This finding for India is broadly in line with what has been reported for several other developing economies where the trade policy regime has been liberalised.

 

V. ENDNOTES





[1]* Corresponding author, Indian Institute of Management, P.O. Kunnamangalam, Kozhikode 673 571, Kerala 673 570, India; e-mail: balan@iimk.ac.in; ** Centre for Development Studies, Thiruvananthapuram, Kerala 675 011, India; *** Indian Institute of Technology Madras, Chennai, India This study originated in a background paper commissioned for the World Bank’s ‘India: Country Economic Memorandum’ of 1996. We thank Zoubida Allaoua, Manuel Arellano, Wendy Carlin, Steve Bond, Deb Kusum Das, Sanghamitra Das, Arunava Sen, Geoff Harcourt, K.L. Krishna, Dilip Mookherjee, Bharat Ramaswami, Abhijit Sen, Kala Krishna, Subrata Sarkar, Rohini Somanathan, and Rajendra Vaidya for discussions, and the Indian Statistical Institute, Centre for Development Studies, and the Indian Institute of Management for support. Errors, if any, would be ours. 

[2]Note that the fixed effect is eliminated by total differentiation.

 

[3]This finding is in line with the observation by Tybout (2000) that evidence of market power in developing country industry is weak.

 

VI. REFERENCES

Arellano M. and Bond, S.(1991), “Some Tests of Specification for Panel Data: Monte Carlo

Evidence and An Application to Employment Equations”, Review of Economic

Studies 58277-97.

Baltagi B. H. (1995), Econometric Analysis of Panel Data, John Wiley & Son: Chichester Blundell, R. and Bond, S. (2000), “GMM Estimation with Persistent Panel Data: An

            Application To Production Functions”, Econometric Reviews, 19 321-340. 

Chambers, R. G. (1988), Applied Production Analysis: A Dual Approach, Cambridge

            University Press: Cambridge.

Coe, D.T., Helpman, E. and Hoffmaister, A.W. (1997), North-South R&D Spillovers, The

            Economic Journal 107134-49.

Das, D.K. (2003), Quantifying Trade Barriers: Has Protection Declined Substantially in Indian

            Manufacturing?,Working Paper No. 105, ICRIER: New Delhi.

Corden, M. (1974) Trade Policy and Economic Welfare, Clarendon Press: Oxford.

Dennison, E. F. (1967), Why Growth Rates Differ: Post-War Experience in Nine Western

            Countries, The Brookings Institution: Washington, D.C.

Doornik J.A., Arellano, M. and Bond, S. (2002) Panel Data Estimation Using DPD for Ox

            http://www.nuff.ox.ac.uk/users/doornik.

Dornbusch, R. (1992), “The case for trade liberalisation in developing countries”, Journal of

            Economic Perspectives 669-85.

Government of India (1994), Economic Survey 1993-94, Ministry of Finance: New Delhi.

Grilliches, Z. and Mairesse, J. (1998) “Production Functions: The Search for Identification”,

in Griliches, Z. (ed.) Practicing Econometrics: Essays in Methods and Applications,

Edward Elgar: Cheltenham, pp. 383-411.

Grossman, G.M., Helpman, E. (1991), Innovation and Growth in the Global Economy,

MIT Press: Cambridge, Mass.

Hall, R.E. (1988), “The Relation Between Price and Marginal Cost in U.S. Industry”,

Journal of Political Economy96 921-47.

Harrison, A.E. (1994), “Productivity, Imperfect Competition and Trade Reform: Theory

            and Evidence”, Journal of International Economics36 53-73.

Joshi, V., and Little, I.M.D. (1997), India’s Economic Reforms 1991-2001,Oxford

            University Press: Delhi.

Krishna, P., and Mitra, D. (1998) “Trade Liberalisation, Market Discipline and Productivity

            Growth: New Evidence from India”, Journal of Development Economics 56 447-62.

Krugman, P. (1992), Rethinking International Trade, MIT Press: Cambridge, Mass.

Levinsohn, J. (1993), “Testing the Imports-as-Market-Discipline Hypothesis”, Journal of

            International Economics35 1-22. 

Levinsohn, J., and A. Petrin  (2003), “Estimating Production Functions Using Inputs to Control

            For Unobservables”, Review of Economic Studies 70 317-41. 

Nouroz, H. (2001), Protection in Indian Manufacturing: An Empirical Study, Macmillan: New

     Delhi.

Olley, S. and Pakes, A. (1996), “The Dynamics of Productivity in the Telecommunications

            Equipment Industry”, Econometrica 65 292-332.

Paus, E.A. (2004), “Productivity  Growth in Latin America: the Limits of  Neoliberal Reforms”,

            World Development, 32 427-445.

Reserve Bank of India (1990), RBI Bulletin, RBI: Mumbai.

Rodrik, D. (1988), “Imperfect Competition, Scale Economies, and Trade Policy in

Developing Countries” in: Baldwin, R. (ed.) Trade Policy Issues and Empirical

            Analysis, University of Chicago Press: Chicago, pp. 109-143.

Rodrik, D. (1992), “Closing the Technology Gap: Does Trade Liberalisation Really Help?”,

            in: Helleiner, G. (ed.) Trade Policy, Industrialisation and Development: New

            Perspectives, Clarendon Press: Oxford, pp. 155-75.

Shapiro, C. (1989), “Theories of Oligopoly Behaviour”, in: Schmalensee, R. and Willig, R.

            (eds.), Handbook of Industrial Organisation, North Holland: Amsterdam,

            pp. 329-414.

Srinivasan, T. N. (2000), “Foreign Trade Policies and India’s Development”, in Kapila, U. (ed)

            Indian Economy since Independence, Academic Foundation Press: New Delhi.

            pp. 667-701.

Srivastava, V. (1996), Liberalization, Productivity and Competition: A Panel Study of Indian

            Manufacturing, OxfordUniversity Press: Delhi.

Tybout, J.R. (1992), “Linking Trade and Productivity: New Research Directions”, The World

            Bank Economic Review  6189-211.

Tybout, J.R. (2000), “Manufacturing Firms In Developing Countries: How Well Do They Do,
               and Why?”,Journal of Economic Literature March, 38 11-44.

Tybout, J.R., de Melo, J. and Corbo, V. (1991), “The Effects of Trade Reforms on Scale and

            Technical Efficiency: New Evidence from Chile”, Journal of International Economics

            31 231-50.

Windmeijer, F. (2002), A Finite Sample Correction for the Variance of Linear Two-Step

            GMM Estimators, Working Paper 00/19, The Institute for Fiscal Studies:  London.

World Bank (1998), India: Macro Economic Update, The World Bank: Washington D.C.

Young, A. (1991), “Learning By Doing and the Dynamic Effects of International Trade”,

            Quarterly Journal of Economics 106 369-405.

                                                                        *****

VII. APPENDIX

VII.1 Descriptive statistics of the variables

(Table A1 goes here)

VII.2 Definition of the variables

As the balance-sheet data provided by the CMIE are in nominal terms, the conversion of these values into a measure of the underlying quantities was the principal data processing involved. This requires deflating the nominal values using appropriate prices. Details are as follows:

Output: The value of output was deflated by the industry-specific wholesale price index. The source of the price series is "Index Numbers of Wholesale Prices in India, base 1981-82=100", Ministry of Industry, Government of India.

Capital: Needed is an estimate of the real value of capital. We follow Srivastava (1996) in attempting a measure of the firm-specific capital stock allowing for vintage.         

            Our data base provides information on gross fixed assets and its components along with depreciation. From this net investment can be obtained as the difference between the current and lagged values of gross assets less depreciation. In principle, this enables one to use the perpetual inventory method to arrive at an estimate of capital stock for each year as follows:

                                              Pt+1 Kt+1 = [Pt+1/Pt].PtKt(1-d) + Pt+1It+1,

where P, K and I refer to the price, physical capital stock and net investment, respectively, while dis the depreciation rate.

However, this procedure can be applied as it is only when the base-year capital stock is P0K0, i.e., in the chosen base year a firm has no inherited capital, as it were. But this is seldom the case, for in any particular year a firm has a mixture of vintages, and, in the context of balance-sheet data, all valued at historic cost.  The problem of arriving at a measure of the real capital stock using the perpetual inventory method is really one of valuing the base-year capital stock.  It is essentially a question of converting balance-sheet data at historic cost into a measure of capital at replacement cost, while at the same time accounting for the vintage mix.  The value of capital at replacement cost for the base year is arrived by revaluing the base year capital as found in the balance sheet. The method adopted for this involves an element of arbitrariness, and one at best arrives at an approximation.  In constructing a ‘revaluation factor’ we depended upon the following three assumptions:

(a) We treat 1997-98 as the base year as the maximum number of observations in our sample corresponded to this year.  We assume that the earliest vintage in the capital mix dates to either the year of incorporation or 1977 if the year of incorporation is earlier than 1977. The year 1977 is adopted on the basis of the ‘Report of a Census of Machine Tools’, Central Machine Tools Institute, Bangalore 1986, which states that the life of machinery in India is, on average, of the duration of twenty years. 

(b) The price of capital changes at a constant rate from 1977 or the year of incorporation upto1997.  The actual value adopted is arrived at from a series of price deflators constructed from the CSO's estimate of gross fixed capital formation published in the National Accounts Statistics.

(c) As with the price of capital, we assume that investment in a firm grows at a constant rate too. The growth of fixed capital formation at 1980-81 prices is applied.  Depending on the year of incorporation, firms will have different annual average growth rates after 1977. The resulting revaluation factor was applied to the capital stock at book value in the chosen base year, converting it into the capital stock at replacement cost. This value was then deflated, to arrive at the real capital stock for that year.  The price deflator used was the price index for Machinery and Machine Tools, as plant and machinery account for 71.5 percent (Reserve Bank of India 1990:  103) of GFA. Investment, measured as (GFAt - GFAt-1) is now added to the estimated real capital stock in the base year to arrive at the figure for the initial year.  The capital stock series is updated using the perpetual inventory method. The estimation procedure has been outlined here onlyvery briefly. Greater detail may be found in Srivastava (1996), as also the algebraic expression of the revaluation factor used.

            It should be noted that we use the gross value of capital in our study.  Gross value is often used by researchers as the estimated net value is found to decline more rapidly than warranted by the facts.  Capital goods are often maintained in a good condition until firms scrap them.  However, use of the gross value involves a stringent assumption that the ability of a capital good to contribute to production remains constant throughout its economic life.  Dennison (1967) thus argues that a correct measure of capital services falls somewhere in between the gross stock and the net stock advocating the use of a weighted average of the two with higher weight for the gross stock as the true value is expected to be closer to it. An attempt at implementing this runs into trouble in the Indian context as a measure of capital consumption is difficult to arrive at. Even when some of the arbitrariness in arriving at a depreciated capital stock may be overlooked, the data requirement exceeds what is available currently. We thus preferred to work with the gross value of capital. 

Labour: The item `Wages and Salaries' is converted into a measure of labour input of firms by administering an estimated Compensation per Worker in the industry in that year.  The resulting measure would be recognised as (Tybout et al 1991: 245) the labour input expressed in `efficiency units'. Compensation per Worker was computed by dividing Total Emoluments by Total Labour Hours as reported in the Annual Survey of Industries (ASI).

Materials: The materials bill is deflated by a materials-input price index. Input-Output coefficients for 1989-90 have been used as weights to combine the wholesale price indices for  the materials individually. The source of the weights is CSO's input-output table for 1989-90 and appropriate price indices were taken from  "Index Numbers of Wholesale Prices in India, base 1981-82=100", Ministry of Industry, Government of India.

 

 

Table 1

               The Effective Rate of Protection (percent) in Indian Manufacturing

 

 

1986-90

1991-95

1996-2000

Intermediate Goods

149.2

87.6

40.1

Capital Goods

 78.5

54.2

33.3

Consumer Goods

111.6

80.6

48.2

Source: Das (2003).


 

Table 2

 

Estimated change in productivity growth and market power

Model: dyit = B0 +B1D+ B2dxit + B3dkit+ B4dxit*D+ uit

 

Industry

B0

B1

B2

B3

B4

Sargan

AR (2)

Food Products

-0.069

(-0.11)

0.070

(0.13)

0.914*

(3.85)

-0.008

(-0.18)

0.313

(1.22)

68.76

(0.93)

0.026

(0.98)

Beverages and Tobacco

-0.045

(-0.87)

0.043

(1.07)

1.379*

(2.47)

0.081

(0.45)

-0.166

(-0.31)

29.92

(0.99)

-0.902

(0.36)

Textiles

-0.017

(-1.39)

0.015

(1.65)

1.673*

(3.19)

0.024

(0.37)

-0.489

(-0.93)

73.23

(0.44)

-1.521

(0.13)

Textile Products

-0.057

(-1.85)

0.031

(1.19)

1.522*

(5.21)

0.078*

(2.20)

-0.255

(-0.80)

45.35

(0.95)

-1.406

(0.16)

Paper

-0.015*

(-2.05)

0.049*

(9.26)

1.482*

(5.67)

-0.008

(-0.13)

-0.241

(-0.93)

66.53

(0.82)

-0.651

(0.52)

Rubber

0.023

(1.81)

-0.035*

(-2.97)

1.069*

(3.15)

0.003

(0.04)

0.179

(0.52)

53.47

(0.95)

0.809

(0.42)

Chemicals

0.033*

(1.96)

-0.028*

(-2.38)

1.349*

(5.69)

-0.072

(-0.767)

-0.015

(-0.06)

56.47

(0.71)

-1.157

(0.25)

Non-metallic mineral products

0.013

(0.29)

-0.005

(-0.07)

1.918*

(2.09)

-0.131

(-0.42)

-0.302

(-0.32)

20.33

(0.91)

-1.702

(0.09)

Basic metals

-0.009

(-0.88)

0.033*

(4.98)

1.332*

(8.64)

-0.085

(-1.54)

-0.054

(-0.39)

44.14

(0.97)

-0.602

(0.55)

Metal Products

-0.037*

(-1.96)

0.031*

(2.02)

0.758*

(2.08)

0.142

(1.84)

0.604

(1.26)

51.65

(0.99)

-0.389

(0.69)

Machinery

0.016

(0.86)

-0.005

(-0.36)

0.708*

(2.22)

0.156

(1.61)

0.567

(1.72)

30.73

(0.99)

-1.178#

(0.24)

Transport

0.023

(0.803)

0.006

(0.228)

1.096*

(2.13)

-0.077

(-0.96)

0.294

(0.54)

52.47

(0.83)

-0.065

(0.95)

Other Industries

-0.028

(-1.02)

0.026

(1.20)

1.806*

(2.42)

0.105

(0.57)

-0.682

(-0.91)

33.19

(0.99)

-0.64

(0.52)

 

 

Note: Notes: t-values computed using robust standard errors in parentheses; standard errors of the two-step GMM estimators are corrected for finite sample bias as in Windmeijer (2000); * indicates significance at 5 per cent level; AR (2) refers to a test for second-order residual correlation - correlation values are given and the p-values are in parentheses; # indicates third-order residual correlation; ‘Sargan’ is a test of over identifying restrictions - chi-square value is given and the p- value is in parentheses.          

 

 

 

 

Table 3

Direction of change in productivity growth and market power

 

Industry

Productivity growth

Price-Cost

Ratio

Food Products

+

-

Beverages &Tobacco

+

+

Textiles

+

-

Textiles Products

+

-

Paper

+ *

-

Chemicals

- *

-

Rubber, Plastic & Petroleum products

 

- *

+

Basic Metals

+ *

+

Non-metallic mineral Products

-

+

Metal Products

+ *

+

Machinery

+

-

Transport

+

-

Other: Leather, Wood

  and Miscellaneous

-

+

                                *  indicates that the estimated change is significant at the 5 per cent level

 

Table 4

Testing for Market Power

 

Industry

B2

t-value

Food Products

0.914

-0.36

Beverages and Tobacco

1.379

0.68

Textiles

1.673

1.29

Textile Products

1.522

1.24

Paper

1.482

1.85*

Rubber

1.069

0.20

Chemicals

1.349

1.47

Non-metallic mineral products

1.918

0.99

Basic metals

1.332

2.15*

Metal Products

0.758

-0.66

Machinery

0.708

-0.92

Transport

1.096

0.19

Other Industries

1.806

1.08

 

                 

                      Note: one-tailed test for B2 ³1 performed; null hypothesis is B2 = 1;

                      * indicates significant at five per cent level.


 

 

 

 

 

Table A1

Sample size

 

Industry

Firms

Observations

Food products

408

1716

Beverages and Tobacco

55

272

Textiles

402

1942

Textile products

160

598

Paper

112

545

Chemicals

609

3145

Rubber, plastic and petroleum products

270

1177

Basic metals

406

2145

Non-Metallic minerals

212

1013

Metal products

134

671

Machinery

561

3024

Transport

170

1065

Miscellaneous

83

298

All Manufacturing

3582

17611