ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Economic Reform, Output and Employment Growth in Manufacturing Testing Kaldor's Hypotheses

The basic object of this paper is to explore what role manufacturing output growth has had on overall economic growth and on employment growth in manufacturing industries in India in the pre- and post-deregulation phases of the country. In the 1960s Kaldor put forward certain hypotheses linking growth of output, employment and productivity in the manufacturing sector. Testing those hypotheses with Indian data may illuminate the nature of the growth process in Indian manufacturing. In particular, they may shed new light on differences in regional patterns of growth in India over the period 1970-71 to 2002-03. The paper focuses on two states, namely, West Bengal and Gujarat, experiencing different types of growth.

Economic Reform, Output and Employment Growth in Manufacturing: Testing Kaldor’s Hypotheses

The basic object of this paper is to explore what role manufacturing output growth has had on overall economic growth and on employment growth in manufacturing industries in India in the pre- and post-deregulation phases of the country. In the 1960s Kaldor put forward certain hypotheses linking growth of output, employment and productivity in the manufacturing sector. Testing those hypotheses with Indian data may illuminate the nature of the growth process in Indian manufacturing. In particular, they may shed new light on differences in regional patterns of growth in India over the period 1970-71 to 2002-03. The paper focuses on two states, namely, West Bengal and Gujarat, experiencing different types of growth.

PANCHANAN DAS

I Introduction

I
n the 1960s, Kaldor put forward certain hypotheses linking growth of output, employment and productivity in the manufacturing sector. Testing those hypotheses with Indian data may illuminate the nature of the growth process in Indian manufacturing. In particular, they may shed new light on differences in regional patterns of growth in India over the period from 1970-71 to 2002-03. The period has been subdivided into the period of the licence permit raj that became attenuated in 1985 and the period of deregulation thereafter. Economic reform of the 1990s made a dramatic change in the political economy of India. Until the mid-1980s, the state controlled large-scale industrial production through the complex system of licensing to start and expand industrial enterprises, and also protect them against foreign competition. In 1991, the government of India introduced a new industrial policy in the form of deregulation and also announced the reform of stateowned industrial production by allowing disinvestment even in profit-making public enterprises. Further, trade has been liberalised, tariff rates have been brought down considerably and quantitative restrictions on imports have been by and large removed in the new trade regime. For manufacturing products, the average rate of tariff was lowered from about 120 per cent in 1989-90 to about 33 per cent in 1997-98. Non-tariff barriers in manufacturing industries were reduced from 87 per cent at the end of 1980s to 46 per cent in 1995-96, and further to 28 per cent in 1999-2000. Currently, almost all commodities are free from quantitative restrictions on imports.

There has been recent concern about deceleration in the growth of industrial output along with high unemployment in the Indian economy, particularly since the second half of the 1990s. Between 1995-96 and 2000-01, about 1.1 million workers, or 15 per cent of workers in the organised manufacturing sector lost their jobs [Nagaraj 2004] and such losses have been spread across major states and industry groups. One of the most disconcerting developments of the 1990s in India, and indeed the world over, has been that economic growth has been accompanied by a much lower rate of growth in employment, and by zero or negative growth of what the International Labour Organisation has termed “decent work” [Bagchi 2004b].

This paper examines what role manufacturing output growth has had on overall growth and also on employment growth in manufacturing industries in India. As India is a country of a large number of heterogeneous states in socio-economic and political character, state level studies have immense importance. In India, as in other countries, different regions have been growing at uneven rates. The regional disparities in growth have been highly associated with unequal incidence of industrial development [Kaldor 1970]. In terms of new industrial investment, the western and southern states have gained, and the eastern states are in decline. The western region states have continued to dominate the eastern region states in terms of the shares of value added and employment in the factory sector of the country. In this paper we have concentrated on West Bengal and Gujarat, the leading industrial states in the eastern and western parts of the country, because they have experienced, in some way, different types of structural change [Das 2006a]. They also differ extremely in political-economic character. The government policies followed by the distinct political parties in power in the respective states have been dissimilar. The Left Front government in West Bengal had in fact opposed neoliberal reforms of the kind introduced in the 1990s and sympathised with pro-labour measures. The government of Gujarat, on the other hand, had responded eagerly in implementing neoliberal reforms.

Kaldor himself tested the hypotheses using data for 12 Organisation for Economic Cooperation and Development countries over the period 1953-54 to 1963-64. Since this pioneering work, there has been a jumble of studies testing the hypotheses across countries, across regions and also across industries [Thirlwall 1983 for notable studies]. There have, however, been hardly any studies of testing Kaldor-type hypotheses for a developing country like India. This study is an attempt to test the hypotheses by applying the vector auto regressive (VAR) model which will hopefully throw a new light on the Indian growth process.

The paper is organised in the following way. Section II is a brief discussion on Kaldor’s hypotheses on growth. Econometric models and database used in testing the hypotheses are discussed in Section III. The bird’s eye view of growth pattern across states in the country and the relative position of West Bengal and Gujarat are displayed in Section IV. In Section V empirical results relating to the validity of two major hypotheses on growth proposed by Kaldor are analysed. Section VI concludes.

II Kaldor’s Hypotheses on Growth

Kaldor (1966) articulated a series of hypotheses in his theoretical frame of growth analysis mainly for elucidating the causes of growth differential between advanced capitalist countries at roughly similar stages of development. The first hypothesis states that faster the rate of growth of manufacturing output, the faster will be the rate of growth of GDP, not simply in a definitional sense but in a fundamental causal sense. This is sometimes referred to as, “manufacturing as the engine of economic growth” hypothesis. Secondly, the growth of labour productivity in the manufacturing sector is positively related to output growth because of static and dynamic increasing returns to scale. It has its predecessor in the dynamic relationship between the productivity growth and the output growth as investigated by P J Verdoorn and popularly known as Verdoorn’s Law. Kaldor observed a highly significant relationship suggesting that output growth played a major role in determining productivity growth and also employment growth in the manufacturing sector. The higher rate of growth of manufacturing output leads to higher rates of productivity growth, but not to a faster rate of growth of manufacturing employment.

The main driving force behind the relationship between output growth and productivity growth is the dynamic increasing returns to scale associated with the invention and innovation in manufacturing industries. Endogenous growth theory has also acknowledged the importance of increasing returns in generating economic growth, but in fundamentally a different way. In Kaldor’s theory, demand constrains the growth of output, while in endogenous growth theory the supply of human and other capital constrains the growth of output in the economy. Kaldor’s (1962) growth model had recognised the role played by endogenously determined technical change and technological learning, and emphasised the importance of the expanding market to explain the presence of increasing returns. But in most of the endogenous growth models, some variables such as R&D and improved human capital help to overcome the supply constraints and sustain growth in the long run only by way of externalities. Kaldor’s empirical analysis of economic growth is generally seen as being macroeconomic, driven by economies of scale that are generated endogenously through technological innovation and innovation embodied in new investment.

There are some other interrelated propositions in Kaldor’s growth theory. The faster the rate of output growth in manufacturing industries, the faster will be the rate of labour transference from outside the manufacturing sector, particularly from land-based activities. The result of increasing returns in manufacturing on the one hand and induced productivity growth in non-manufacturing by the process of labour transfer on the other is expected to raise productivity growth in the economy as a whole. In this way the growth of GDP is positively related to the growth of manufacturing output. As the scope for labour transfer from non-manufacturing dries up, the role of the manufacturing sector as engine of growth will likely to be reduced. In this sense a country with little or no surplus of labour in non-manufacturing activities experienced a deceleration of growth. What is surprising in the Indian economy is that the output growth in manufacturing activities decelerated in the mid-1990s in spite of no labour shortages in non-manufacturing. Moreover, Kaldor observed that the growth of manufacturing output is fundamentally constrained by demand and particularly by the rate of growth of exports, not by labour supply. Within this framework, manufacturing output growth determines employment growth, not the other way round. Through the link between output growth and productivity growth, the lower costs of production in fast growing countries make it difficult for developing countries to establish export activities with favourable growth characteristics, except through exceptional industrial enterprise.

III Methodology

The data: We have used data from National Accounts Statistics (NAS), for estimating growth rates of domestic products at the national level from 1970-71 to 2000-01.

In estimating output growth at the state level, the data base on states’ domestic products (SDP) provided by the Central Statistical Organisation and published by the EPW Research Foundation, 2003, have been utilised. There are four series of data on SDP for the base periods 1960-61, 1970-71, 1980-81 and the latest 1993-94. The new series (1993-94) of SDP data are based on the system of national account (SNA) suggested

Table 1: Growth Rates of Domestic and Manufacturing Output at Constant (1993-94) Price across Major States of India: 1970-71 to 2000-01

States Manufacturing
1970-2000 1970-1985 1986-2000
NSDP Output NSDP Output NSDP Output
Andhra Pradesh 5.04 7.81 3.97 6.23 5.85 8.76
Bihar 3.17 4.23 3.61 7.80 2.02 -1.51
Gujarat 5.34 7.26 4.93 6.46 6.51 7.81
Himachal Pradesh 4.33 9.78 2.78 4.79 5.63 13.09
Haryana 5.40 7.71 5.05 7.83 5.21 5.62
Karnataka 5.06 6.70 3.82 6.53 6.57 6.35
Kerala 3.57 4.20 1.68 2.97 5.96 6.22
Maharashtra 5.61 5.93 4.43 4.98 6.95 6.61
Madhya Pradesh 4.01 5.97 2.89 4.23 4.95 8.10
Orissa 2.88 0.90 2.42 1.60 2.63 -5.11
Punjab 4.84 7.90 4.99 8.63 4.50 6.42
Rajasthan 4.97 4.85 3.16 2.93 6.34 6.75
Tamil Nadu 4.43 4.25 2.57 5.34 6.37 4.09
Uttar Pradesh 4.12 6.90 3.77 7.14 3.80 4.48
West Bengal 4.52 3.01 3.22 1.72 6.08 4.80
All India 4.82 5.70 3.67 4.90 5.62 5.60

Note: Growth rates are estimated by using semi-logarithmic trend equation by applying pooled regression fixed effect model and dummy variables are introduced to allow the growth rates to differ among states and between the two sub-periods.

Source: NAS and State Domestic Products of India, Central Statistical Organisation, Government of India.

Economic and Political Weekly September 29, 2007 by the United Nations in 1993. In this series the base year for prices is changed to 1993-94, and the product composition is revised as well in a number of sectors such as real estate and finance. Since the different series are based on different base years, we have constructed a consistent chain linked time series of SDP by extending the 1993-94 series backwards, i e, by converting the old series to the series based on the new base year (1993-94) – following the splicing method recommended by the CSO.

The prime data source of registered industrial sector in India is Annual Survey of Industries (ASI) published by the CSO. The ASI consists primarily of all factories which are required to be registered under sections 2m(i) and 2m(ii) of the Factories Act 1948. Information on different characteristics in the registered sector are processed by the CSO from the survey conducted every year (excepting in 1972) by the National Sample Survey Organisation (NSSO) and are available up to 2002-03. We have filled the data gap for 1972-73 by using simple interpolation. The EPWRF has published an electronic database collecting from the ASI up to the period 1997-98. The data for the period after 1997-98 have been collected from Annual Series, Volume 1, published by the ASI for every year on the basis of National Industrial Code (NIC) 1998.1

The primary unit of enumeration in the ASI is a factory in the case of manufacturing industries and the ASI records rele vant figures on the basis of reporting units. The data base suffers from incomplete coverage of factories and underreporting of workers in the factories covered. Nagaraj (1999) has noticed a decline in the proportion of factories in the ASI to the Economic Census from 48 per cent in 1980 to 43 per cent in 1990. As the number of non-reporting units varies randomly from year to year across states, we normalise the value of gross output and number of workers by the number of reporting factories. The wholesale price index for industrial products provided by the CSO is used in calculating real values of output from the nominal values. In this paper, real gross output and number of workers per factory unit reported in the ASI are treated as measures of output and employment respectively. Econometric model: Most of the macroeconomic time series follow either trend stationary process or difference stationary process or both. To capture trend behaviour of output and employment we have used semi-logarithmic trend model. For state level analysis, generalised least square (GLS) method in the frame of fixed effect pooled regression is applied.2 The popular model in estimating trend is

= αi + βit+uit (1)

YitYit be the natural logarithm of real value of a macroeconomic variable in cross section unit i at time period t.

The relationship between growth rates of the relevant variables is first examined by following the methodology used by Kaldor. Most of the available studies and even Kaldor himself tested the validity of the hypotheses by applying simple regression analyses using cross section data across the countries. If the regression coefficients are found to be statistically significant and positive, it is then concluded that the hypotheses are valid. But, as the relationships between the growth rates are dynamic in nature, this kind of methodology may not be right and sufficient to test the hypotheses. This is because simple regression equations used in the previous studies can only show the presence of a statistical correlation between the variables, but have no bearing on the causal relationship between them. Thus we think that the validity of the Kaldor hypotheses requires not only the existence of the significant correlation between the variables but the causality running from one to another as well.

In this paper we have applied cointegration theory developed in Engle and Granger (1987). Two variables are cointegrated if each is non-stationary but a linear combination of the two is stationary. For the validity of the hypotheses in a causal sense the variables should be cointegrated of order one, or else they will be drifting further apart over time, in which case the regression relationship between them may not be meaningful and indeed becomes spurious.3 The concept of cointegration is normally related to the idea of common stochastic trends.4 Ignoring the presence of deterministic components, however, leads to some misleading inferences. In order to concentrate on the issues of cointegration, we consider the presence of stochastic trend in the time series of growth rates.

The two-step procedure developed by Engle and Granger first involves estimating the long run relationship using the cointegrating regression; in the second step, a general dynamic model is estimated usually expressed in an error correction form, which incorporates the estimated disequilibrium errors

Table 2: Output and Employment Growth in Registered Industries in Major States: 1970-71 to 2002-03

States Output Growth Employment Growth
1970- 1970- 1986- 1970- 1970- 1986-
2002 1985 2002 2002 1985 2002
Andhra Pradesh 4.16 -0.25** 7.28 -0.74* -3.07 1.91
Assam 5.30 3.37 4.91 0.65 0.45** 0.91*
Bihar 3.40 3.17 -1.04** -1.69 -2.35 -6.12
Gujarat 5.71 4.07 6.95 -1.38 -1.65* -1.63
Haryana 4.83 1.63 5.80 -0.81 -3.03 -1.29
Himachal Pradesh 6.83 7.89 3.96 -0.28** 2.77* -6.59
Jammu and Kashmir 6.57 7.67 2.30 -0.97* 2.77 -0.51**
Karnataka 5.96 3.90 6.71 0.01** -0.14** 0.33**
Kerala 3.84 4.63 3.18 -1.71 -2.30 -1.69
Madhya Pradesh 6.35 4.32 5.51 -0.19** 0.33** -2.04
Maharashtra 4.45 3.30 4.24 -1.84 -2.37 -1.99
Orissa 5.62 4.39 2.69 0.11** 0.59** -2.63
Punjab 5.96 6.81 3.55 1.28 2.86 -1.76
Rajasthan 4.08 3.06 2.72 -2.07 -0.73** -3.99
Tamil Nadu 3.02 2.09 3.12 -1.69 -2.86 -1.14
Uttar Pradesh 4.79 2.66 4.17 -2.36 -0.64 -3.22
West Bengal 2.93 3.30 1.80 -1.78 -0.58** -2.96
All India 4.50 2.93 4.75 -1.31 -1.41 -1.55

Notes: Methodology of estimating growth rates is the same as for Table 1.

* Significant at 5 per cent level, ** Insignificant, rest are significant at 1 per cent level.

Source: Annual Survey of Industries Time Series Data, Central Statistical Organisation, Government of India.

Table 3: Regression Results for Testing First Hypothesis: All India

Regression equation: gyt = a + b gmt + ut (Method: Least Squares)

a b t-Statistic Prob

Period: 1970- 2002 0.029 0.315 2.842 0.0079 r2 =0.206, D-W=2.580 Period: 1970- 1985 0.019 0.356 1.625 0.1263 r2 =0.158, D-W=2.919 Period: 1986-2002 0.040 0.248 2.786 0.0138 r2 =0.341, D-W=2.265

Note: Last two columns give t statistics and probability values for the regression coefficient only, not for the intercept term; D-W indicates Darbin-Watson statistics.

Source: As for Table 1.

from the first step. The most widely used vector error cor rection model is represented as

Δgyt = aut-1 + b1Δg yt-1 + b2 Δgxt-1 + εt (2) where gx and gy represent growth rates, measured by the log differences between the current and the previous period, of X and Y respectively. ut-1 = (gy -βgx -c)t-1, is estimated residual from the cointegrating regression representing the past period’s disequilibrium. The direction of causality between the growth rates of Y and X can be detected by estimating this type of model. If the coefficient b2 in equation (2) is statistically significant then the null hypothesis that gx does not Granger cause gy is rejected.

IV Exploring Regional Growth in India

Growth of domestic output and manufacturing output across Indian states: The growth behaviour of the manufacturing sector as of other sectors of an economy largely depends upon government policies. During the regulated-policy regime, the government directly and indirectly influenced the pace of industriali sation of the country. The fate of large-scale industry in any constituent state was primarily determined by the industrial licences approved for the state by the central government, the allocation of central public investment in industry and infrastructure, and the allocation of credit by banks and term lending institutions under the control of the central government. But in the post-liberalisation phase, every state government has been enjoying some freedom in attracting industrial investment by offering various packages of incentives. Thus comparison of growth performance across major states over different policy regimes assumes significance.

Table 1 gives estimates of trend growth rates of net domestic products (NDP) and of its part originating from manufacturing at constant (1993-94) factor prices across major states and at all-India level over 1970-71 to 2000-01 as well as sub-periods covering the phase of state regulation and deregulation. The rate of growth of NDP of the country was below 5 per cent and the rate was less than the manufacturing growth rate during the total period of estimation. The overall economy as well as its constituent states grew at a faster rate after the mid-1980s than in the earlier period. Different states, however, performed at uneven rates both in terms of overall growth and manufacturing growth. West Bengal grew much more slowly than Gujarat and many other states of the country. But the rate of improvement of growth performance of the former after the end of state regulation has been relatively faster than the latter and many other industrial states.

As we go through the growth performance across Indian states, no strong correlation between the rates of growth of overall output and the rate of growth of manufacturing output has been found in any period specified above. Maharashtra registered the highest growth of state’s income, but it lagged behind many other states in terms of growth rate of income originating from manufacturing. On the other hand, in Himachal Pradesh, while income from manufacturing grew at the highest rate, the overall economy had grown at relatively lower rate.

Output and employment growth in registered manufacturing industry: There has been recent concern about deceleration in the growth of industrial output along with high unemployment in the Indian economy, particularly since the second half of the 1990s. Table 2 presents trend growth rates of output and employment per factory reported in the ASI over different periods. The growth rates of manufacturing output shown in Table 2 are different from those shown in Table 1 owing to the differences in the data source as well as in methodologies. While manufacturing output used in Table 1 is the total output originated from the registered and unregistered sectors recorded in SDP series, output figures used in Table 2 is the output per factory registered in the ASI.

Interestingly enough, the states with little share of manufacturing activities experienced higher growth of output per unit factory. Himachal Pradesh, Jammu and Kashmir, for example, showed more than 6.5 per cent output growth during 19702002 contributing less than 1 per cent of country’s output in the registered sector. On the other hand, Maharashtra and Gujarat together contribute more than 30 per cent of the country’s output in registered industries, while output in this sector of the states grew at 4.4 and 5.7 per cent respectively during the same period. In terms of trend growth rate of real output in registered manufacturing, West Bengal is found to be lagging behind not only Maharashtra and Gujarat but many other states as well.

Although at the national level output growth has gone up after the mid-1980s compared to the earlier period, the growth performance has not been uniform across the major states.

Table 4: Regression Results for Testing First Hypothesis: West Bengal and Gujarat

Regression equation: gyt = a + b gxt + ut Method: GLS (Cross Section Weights)

States a b t-Statistic Prob

1970-2000 West Bengal 0.031 0.429 3.174 0.0024 Gujarat 0.022 0.407 2.649 0.0104

r2 0.342 D-W stat 2.773

1970-1985 West Bengal 0.025 0.417 1.786 0.0848 Gujarat 0.049 -0.092 -0.281 0.7805

r2 0.134 D-W stat 2.960

1986-2000 West Bengal 0.047 0.228 1.878 0.0715 Gujarat 0.014 0.594 4.332 0.0002

r2 0.801 D-W stat 2.320

Source: As for Table 1.

Table 5: Regression Results for Testing Second Hypothesis: All India

Period 1970-71 to 2002-03: ge = Period 1970-71 to 85-86:ge = Period 1986-87 to 2002-03:ge = Figures in parentheses give t value 0.036 + 0.608 gy, ( -4.105) ( 5.898) -0.030 + 0.647 gy, (-1.969) (3.988) 0.039 + 0.555 gy, (-4.071) (4.322) r2 = 0.536, D-W Stat.= 1.598 r2 = 0.550, D-W Stat.=1.447 r2 = 0.554, D-W Stat.=1.632
Source: As for Table 2.
Economic and Political Weekly September 29, 2007 3981

While in some states, growth rate of output per factory went up, in many other states the rate had fallen during the period 1986-2002 compared to the rates observed in 1970-1985. Gujarat improved its growth performance at a higher rate than Maharashtra in the western part of the country after the mid-1980s. In the southern region, Andhra Pradesh had managed to revive its growth performance in the reform period from the phase of negative and even insignificant growth in the regulated regime. West Bengal along with many other states, on the other hand, showed a marked deceleration in the growth of manufacturing production in the registered sector after the ending of licence permit raj. Contrasting to the result displayed in Table 2, the relatively better performance of West Bengal manufacturing in the reform period shown in Table 1 has been mainly due to the growth acceleration of the unorganised sector after the mid-1980s.

What is more concerning in our exercise is that the actual fall of employment per factory throughout the period and in many cases job losses are more severe in the period of deregulation (Table 2). The phenomenon of jobless growth at least in the 1980s at the national level is well-established. But the estimates in Table 2 undoubtedly disclose job destroying growth in the registered industries across the major states since the 1970s and the rate of dislocation of workers in this sector has increased significantly in the post-deregulation period. Punjab is the only state displaying positive employment growth not only in the licensing period but in the period of deregulation as well although the growth rate had fallen drastically. Growth rates of employment shown in Table 2 are somehow different from that documented in Nagaraj (2004) and Goldar (2000) because of methodological differences. In them total number of workers recorded in the ASI are considered in estimating employment growth in registered manufacturing. But we have already mentioned that the total figures of principal characteristics shown in the ASI may not be comparable truly across different states or across industry groups within a particular state.

V Testing Kaldor’s Hypotheses

Kaldor’s first hypothesis on growth puts forward the causal relationship between industrial growth and overall economic performance:

g yt = f (gmt ), f'`> 0 (3) where gyt is the growth rate of domestic output and gmt is that of manufacturing output in real terms.5.

This hypothesis captures a trait of a developing economy characterised by immaturity in terms of lower productivity in land-based activities. Kaldor hypothesised and also found empirically that, in an intermediate stage of economic development, a fast rate of economic growth is associated with a fast rate of output growth of the manufacturing sector.

According to the second hypothesis higher output will induce higher labour productivity in manufacturing industries. There are two different ways of looking at the second hypothesis by using simple regression analysis. One is to regress productivity growth (gpt) on output growth (gyt) and the other is to regress employment growth (get) on output growth (gyt). The linear specification of the first:

gpt = a +bgyt (4)

In the second case we have: get = gyt – gpt = -a + (1 – b)gyt (5)

Coefficient b in equation (4) or (5) is the Verdoorn coefficient and its value nearer to unity signifies the existence of substantial increasing returns. One extreme value of the

Table 6: Regression Results for Testing Second Hypothesis: West Bengal and Gujarat

Regression equation: g = a1 + b2 g Method: GLS (Cross Section Weights)

em

States a1 b2 t-Statistic Prob

Sample period: 1971-2002

West Bengal -0.039 0.853 5.619 0.0000 Gujarat -0.056 0.751 7.998 0.0000 r2 0.429 D-W stat 2.392 Sample period: 1971-1985 West Bengal -3.695 0.425 4.363 0.0000 Gujarat -2.724 0.279 8.124 0.0000 r2 0.552 D-W stat 2.115 Sample period: 1986-2002 West Bengal -0.046 0.729 2.701 0.0074 Gujarat -0.049 0.477 3.460 0.0006 r2 0.391 D-W stat 2.527

Source: As for Table 2.

Table 7: Error Correction Model for Testing the First Hypothesis

All India = 0.126 ut-1 - 0.595Δgyt-1+ 0.009Δgmt-1

Δgyt

(0.892) (-3.212) (0.052)

ut-1=gyt-1 - 1.680gmt-1+ 0.041 (-1.953) (-0.914)

West Bengal = - 0.575ut-1 - 0.428Δgyt-1 - 0.012Δgmt-1

Δgyt

(-1.801) (-2.078) (-0.062)

ut-1= gyt-1 -0.859 gmt-1 -0.018 (-7.604) (-4.175)

Gujarat Δgyt = - 0.861 ut-1 - 0.369Δgyt - 0.385Δgmt (-2.576) (-1.804) (-1.961)

ut-1= gyt-1 - 0.921gmt-1 + 0.013 (-3.687) (0.634)

Notes: Figures in parentheses give the t values; gyt and gmt indicate overall output growth and manufacturing output growth respectively in time t; Δ indicates first difference.

Source: As for Table 1.

Table 8: Error Correction Model for Testing Second Hypothesis

All India Δglt = -1.418 ut-1 + 0.002 Δglt-1 +0.100Δgmt-1 (-6.010) (0.013) (0.727) ut-1 = glt-1 + 0.108 gmt-1 + 0.012

(0.588) (1.297) West Bengal Δglt = -0.330 ut-1 -0.763 Δglt-1+ 0.744 Δgmt-1

(-1.827) (-2.721) (2.991) ut-1 = glt-1 + 1.470 gmt-1 -0.023

(1.470) (-0.729) Gujarat Δglt = -1.595 ut-1 + 0.210 Δglt-1 + 0.049Δgmt-1

(-5.872) (0.877) (0.287) ut-1= glt-1 + 0.094 gmt-1 + 0.010

(0.314) (0.561)

Notes: Figures in parentheses give the t values; glt and gmt indicate employment growth and manufacturing output growth respectively in time t; Δindicates first difference.

Source: As for Table 2.

coefficient (b = 1) indicates no variation of employment growth and the other extreme value (b = 0) gives no response of productivity growth due to the change in output growth. We have used equation (5) for getting statistical relationship between employment growth and output growth in the registered manufacturing sector. Linear regression results: We have first estimated a statistical relationship between the overall output growth (gyt) and manufacturing output growth (gmt) by following the methodology used by Kaldor and utilising time series of NAS data and also panel data of SDP series for West Bengal and Gujarat over the period 1970-71 to 2000-01. The relationship between employment growth (get) and output growth (gyt) in the manufacturing sector is estimated by using ASI time series data from 1970-71 to 2002-03 in a panel frame for the states.

The pattern of growth rates of domestic product and manufacturing output across the major states shown in Table 1 suggests no strong correlation between them. Similar findings have been revealed in our linear regression results. The correlation coefficients between the growth rates of overall output and manufacturing output are reasonably small. Table 3 presents OLS estimates of regression coefficients of gyt on gmt for the total as well as different sub-periods. The estimated value of the regression coefficient is less than unity implying that the greater the excess of the rate of growth of manufacturing output over the rate of growth of the economy the faster the overall growth rate. But, since it is sufficiently smaller than unity, the overall growth rate would be negligible even if the rate of manufacturing output growth exceeds the rate of NDP growth.6

Table 4 gives GLS estimates of regression coefficients of growth rates of domestic output on manufacturing output for West Bengal and Gujarat in a frame of fixed effect panel regression. It is clear that NSDP growth of either of the state has responded more or less at the same rate (0.4) with 1 per cent rise in manufacturing output growth over the period 1970-71 to 2000-01. The estimated value of r2 suggests that just below 35 per cent in the growth rate of NSDP is asso ciated with the growth of manufacturing output during the same period. But a contrast is observed between the states in the effect of manufacturing growth on the overall growth both in the period of licensing as well as in the deregulation phase. In Gujarat, manufacturing growth had no significant effect on growth of state’s total income despite of its higher rate of growth of manufacturing output during 1970-71 to 1985-86. In West Bengal, on the other hand, the regression coefficient is found to be 0.4 during the same period. However, the relationship is very strong and statistically significant after mid-1980s. Overall, the growth of the Gujarat economy has responded at a considerably higher rate by a 1 per cent increase in of manufacturing output in this period.

These simple regression results, however, are open to the charge of spuriousness. For the hypothesis to be valid, even in the sense of statistical relationship, there must also be a positive association between the growth rate of domestic products and the excess of rates of manufacturing output growth over non-manufacturing output growth [Kaldor 1966]. But the regression results estimated from both the national level data and the state level panel data do not confirm it (Table A1). The relation between the rates of growth of non-manufacturing output and manufacturing output is also statistically insignificant (Table A2). Thus we do not reject the spuriousness of the regression result in finding the role of manufacturing sector as a driving force behind the overall growth at the national as well as state levels. The statistical relationship between overall output growth and manufacturing output growth is simply a definitional one implying manufacturing output is a part of total output.

The estimated coefficients of the regression equation of employment growth on output growth in the registered manufacturing sector are shown in Tables 5 and 6 for all India and for the states respectively. The regression coefficient (1-b) in equation (5) measures employment elasticity of output growth. The relationship between employment growth and output growth in registered manufacturing industries is found to be strong and statistically significant for the total as well as for the sub-periods. The effect of output growth on employment growth in this sector has been significantly higher in West Bengal than in Gujarat. While at the national level the growth effect on employment has fallen after the mid-1980s compared to the earlier period, such effect has increased in both of the states and at a significantly higher rate in West Bengal during the same period. As the growth rate of gross output per factory in West Bengal declined sharply during 1985 to 2002, employment growth turned to be more negative in this period (Table 2). Vector error correction model: One must beware of attributing causal significance to the statistical relationship given in equations (3) and (5) for testing the validity of Kaldor’s hypotheses. The causal relationship is largely accounted for by the productivity growth, which is basically technological one. If productivity, both in terms of level and in terms of the rate of growth, in manufacturing industries is higher than in other activities as observed in many countries, the expansion of the manufacturing sector pulls up the total output growth rate. The existence of increasing returns in the manufacturing activities, the well known contention in the classical literature, causes productivity to increase in response to the rise in output in such activities. In the classical view, productivity growth ultimately is demand constrained, while the neo-classicals ignored this. Manufacturing production has an influence on the overall rate

Table A1: Regression of Overall Output Growth on Excess of Manufacturing over Non-manufacturing Output Growth

All India: gyt = 0.046 - 0.163(gmt – gnmt), r2 = 0.063, D = 2.013 (-1.450) West Bengal: gyt = 0.038 -0.265(gmt – gnmt), r2 = 0.280, D = 2.511 (-2.105) Gujarat: gyt = 0.056 -0.336(gmt – gnmt), r2 = 0.280, D = 2.511 (-2.827)

Notes: gyt, gmt and gnmt indicate growth of total output, manufacturing output and non-manufacturing output respectively. Figures in parentheses give t values.

Source: As for Table 1.

Table A2: Regression of Non-manufacturing Output Growth on Manufacturing Output Growth

All India: gnmt = 0.034 + 0.199 gmt, r2 = 0.072 , D-W = 2.566

(1.561) West Bengal: gnmt = 0.039 +0.284 gmt, r2 =0.214 , D-W = 2.918

(1.674) Gujarat: gnmt =0.028 + 0.248gmt, r2 =0.214 , D-W = 2.258

(1.299) Figures in parentheses give t values.

Source: As for Table 1.

Economic and Political Weekly September 29, 2007 of economic growth partly on account of its influence on the rate of growth of productivity in this sector itself and partly because of its indirect effect to raise the rate of productivity growth in other sectors.

As the first differences of the logarithmic time series, representing growth rates, of output and employment do not contain unit roots [Das 2006b], the regression results shown above make sense statistically, but they do not necessarily make causal sense. As, the logarithmic series are characterised by integrating of order one, I(1), and their first differences are I(0), performing cointegration tests for them is theoretically possible. For any two variables x and y, the long-run relationship between their first differences (ΔLog(x) and ΔLog(y)) can be detected by the cointegration method developed by Johansen and Juselius (1990).

Table 7 presents the estimated results of vector error correction model stated in equation (2) for testing Kaldor’s first hypothesis. The cointegrating coefficients in Table 7 suggest that there has been no significant long run relationship between overall output growth and manufacturing output growth at the national, but a significant relationship between them has been persisted both in West Bengal and Gujarat. The error correction parameter is negative as expected for both the states, but is positive for the all India level. Thus while the statistical relationship between the growth rates of domestic output and manufacturing output is meaningful dynamically for the states, the relationship is found to be spurious in a dynamic sense for the country as a whole over the period 1970-71 to 2000-01.

Our primary concern has been on the causality from manufacturing growth (gm) to overall growth (gy) as hypothesised by Kaldor. Table 7 gives an idea on the causal relationship. As the coefficient of Δgmt-1 is statistically insignificant in all cases, one can accept no causality from manufacturing growth to overall growth in India. Thus the Kaldor hypothesis on manufacturing output growth as a driving force behind overall output growth of an economy has been rejected in causal sense although they have a significant statistical relationship.

The estimated results of vector error correction model for testing Kaldor’s second hypothesis are shown in Table 8. The long run dynamic relationship between employment growth and output growth in registered manufacturing is not statistically significant. Thus although the statistical relationship between employment growth and output growth shown in Tables 5 and 6 is statistically significant, they become spurious in a dynamic sense. We can have a sense of causality from output growth to employment growth by judging the coefficient of Δgmt-1 in Table 8. West Bengal has registered a significant causal relationship from output growth to employment growth in registered manufacturing industries, while manufacturing in Gujarat as in the country does not show any significant causality from output growth to employment growth. Thus the null hypothesis of no-causality from manufacturing output growth to employment growth has not been rejected for Gujarat and the country as a whole, but is rejected for West Bengal.

VI Conclusion

The rate of manufacturing growth as of overall economic growth has increased after the mid-1980s compared to the earlier period. The recent growth in India has been driven to some extent by the explosion of information technology (IT) related services. But IT cannot be a long run source of growth because it currently accounts for a very small share of GDP and employment and even an explosive growth will not change its relative insignificance. Industries such as telecom and entertainment have also registered impressive growth. The automobile industry has improved its growth capability after the abolition of licensing. But the portrait of these two industries alone cannot give us a rounded picture of the Indian manufacturing or services and higher growth in the post-reform period conceals much turbulence. Economic reforms including the withdrawal of subsidies and reservation for small enterprises have opened up new opportunities for the growth of giant enterprises, particularly transnational corporations, and have created increasing vulnerability of small enterprises.

This paper finds an important contradiction in the pace of industrialisation in India over 1970-71 to 2000-01. Manufacturing industries in India did not play any significant role as the engine of growth over the past three decades. The services sector has grown at higher rate than the manufacturing sector in the phase of neoliberal reform. But, as is well known, the high growth of services may not be sustainable without significant growth of the commodity sectors. Job destroying industrial growth is well documented in this study. The incidence of labour displacement in the process of industrial growth is stronger in the industrially advanced states.

A large number of theoretical and empirical research supports Kaldor’s hypotheses on growth. In this paper, the hypotheses are tested at the national level as well as across two major states displaying contrasting growth pattern. While simple regression results give a strong statistical relationship between employment and output growth, there has been no causality between them at the national level and also in industrially advanced states like Gujarat. They are cointegrated in many cases implying long run equilibrium, but there has been a short run disequilibrium relationship failing to follow causal relationship in the Granger sense.

EPW

Email: dasp1962@yahoo.co.in

Notes

[The author is gratefully indebted to Amiya Kumar Bagchi for helpful comments and discussions.]

1 Since 1998-99 there have been some changes in the coverage of

industrial units under the purview of ASI: new series of classification

(NIC 1998) had been introduced; all electricity undertakings other

than captive units and all departmental undertakings such as railway

workshop have been kept outside the frame of the ASI. 2 We have used panel regression model applying generalised least

squares (GLS), partly because such model provides rich environment

for estimation techniques and better estimated values [Maddala

1993 and Baltagi 1995]. Fixed effect is chosen on the basis of standard

diagnostic test.

3 Yule (1926) first showed that spurious correlation could persist in non-stationary time series even if the sample size is very large. For extensive Monte Carlo simulations on spurious regression see Granger and Newbold (1974).

4 The presence of stochastic trend in the logarithmic series of output and employment is confirmed by the augmented Dickey-Fuller unit root test.

5 We define growth rate of a variable y in a time t as the log difference of its value in the previous year from the current year value: g=Δyt =logyt–logyt-1.

6 If we set gyt = in the first estimated equation in Table 3, gyt

gmt

would be 0.042.

References

Bagchi, A K (2004a): ‘Nanny States for Capital and Social Darwinism for the Workers’, The Indian Journal of Labour Economics, Vol 47, No 1, pp 69-79.

  • (2004b): ‘Demographic Regimes and Social Reproduction of Labour’, Economic and Political Weekly, Vol 39, No 30, pp 3367-77.
  • (1998): ‘Studies on the Economy of West Bengal since Independence’, Economic and Political Weekly, Vol 33, November 21, pp 2973-78. Baltagi, B (1995): Econometric Analysis of Panel Data, John Wiley and Sons, New York.
  • Das, P (2006a): ‘Deregulations, Output Growth and Structural Change: A Comparative Study of West Bengal and Gujarat’, Social Change and Development, Vol 4, pp 83-102.
  • (2006b): ‘Growth and Structural Change in West Bengal: 1960-61 to 2000-01’, paper presented at the Symposium on Development Economics in Mumbai University on February 21-22.
  • Engle, R F and C W Granger (1987): ‘Cointegration and Error Correction: Representation, Estimation and Testing’, Econometrica, Vol 55, No 2, pp 251-76.

    EPWRF, Annual Survey of Industries: 1973-74 to 1997-98, A database on the Industrial Sector in India, A Sameeksha Trust Publication, Mumbai.

  • ‘Domestic Products of States of India: 1960-61 to 2000-01’. Goldar, B (2000): ‘Employment Growth in Organised Manufacturing in India’, Economic and Political Weekly, Vol 35, No 14, pp 1191-95. Granger, C W J and P Newbold (1974): ‘Spurious Regressions in Econometrics’, Journal of Econometrics, Vol 2, pp 111-20.
  • Government of India: Annual Survey of Industries, various issues, Central Statistical Organisation.
  • National Accounts Statistics, 1950-51 to 2002-03, Central Statistical Organisation.
  • Johansen, S and K Juselius (1990): ‘Maximum Likelihood Estimation and Inference on Cointegration with Applications to the Demand for Money’, Oxford Bulletin of Economics and Statistics, Vol 52, pp 169-210.

    Kaldor, Nicholas (1966): Causes of the Slow Rate of Growth in the UK, Cambridge University Press, London reprinted in Targetti and Thirlwall (1989).

    – (1970): ‘The Case for Regional Policies’, Scottish Journal of Political Economy, Vol XVII, No 3, November.

    Kaldor, N and J A Mirrlees (1962): ‘A New Model of Economic Growth’, Review of Economic Studies, Vol 29, No 3, reprinted in N Kaldor, 1978, Further Essays on Economic Theory, Vol 5, Duckworth and Co Ltd, London.

    Maddala, G (1993): The Econometrics of Panel Data, Vols I and II, Vt E E Elgar, Brookfield. Nagaraj, R (1999): ‘How Good Are India’s Industrial Statistics’?, Economic and Political Weekly, Vol 34, No 6, pp 350-55.

  • (2000): ‘Organised Manufacturing Employment’, Economic and Political Weekly, September 16, pp 3445-48.
  • (2004): ‘Fall in Organised Manufacturing Employment: A Brief Note’, Economic and Political Weekly, Vol 39, No 30, pp 3387-90. Nouroz, Hasheem (2001): Protection in Indian Manufacturing: An Empirical Study, MacMillan India, Delhi. Thirlwall, A P (1983): ‘Symposium on Kaldor’s Growth Laws’, Journal of Post Keynesian Economics, Vol 5, pp 345-58.
  • Yule, G U (1926): ‘Why Do We Sometimes Get Nonsense Correlations between Time Series? A Study in Sampling and the Nature of Time Series’, Journal of Royal Statistical Society, Vol 89, pp 1-64.

    Economic and Political Weekly September 29, 2007

    To read the full text Login

    Get instant access

    New 3 Month Subscription
    to Digital Archives at

    ₹826for India

    $50for overseas users

    Comments

    (-) Hide

    EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

    Back to Top