You are on page 1of 4

cent during 1991-97.

Goldar explains it

Discussion
Organised Manufacturing reasoning, starting from a wage reduction
leading to an increase in employment, and
hence output growth really holds. This

Employment assumes the labour market to be the


‘active’ market, while the product market
‘follows’ the labour market. If, however,
product market is postulated to be the
R NAGARAJ working day) rose only at about 1½ per active one, then the fall in wages may
cent per year [Nagaraj 1994]. Bhalotra contract the aggregate demand and hence
output growth.2
I n 1993-94, India’s manufacturing sec-
tor employed about 10 per cent of
workforce to produce about 16 per cent
(1998) sought to provide a more rigorous
verification of this proposition.
Before the ink was dry, employment
Analytically, other things being equal,
labour demand is a function not of wage
of measured GDP.1 Registered (organised) growth turned around in the 1990s (Fig- rate (product wage), but wage-rental ratio.
manufacturing – that is, all factories ure 1), providing more fodder for the In other words, a firm’s decision how much
employing 10 or more workers using debate. Recently, Goldar (2000) has sought labour to employ depends not only on the
power, and 20 or workers without power to show that wage moderation explained going wage rate (for the standard working
– made up 2/3rd of measured manufac- the turn around. He attributed it to a faster day), but the wage rate in relation of cost
turing value added (output, hereafter), but employment growth in smaller sized fac- of capital. Even this may be of limited value,
employed only about 1/5 of workers in this tories, in response to the changes in trade when labour productivity is rising. There-
sector. Therefore, the bulk of the workforce and industrial policies initiated since 1991. fore, one should look at the unit wage cost.
in manufacturing is in the unregistered In providing econometric evidence for the Hence, Goldar’s analytical specification is
(unorganised) sector. It also accounts for inverse relationship between employment open to question. Empirically, however,
much of India’s net merchandise exports. and earnings growth, Goldar found no sup- Goldar’s findings may still be valid since
Though quantitatively modest, registered port for the hypothesis of growth in mandays cost of capital (proxied by real interest
manufacturing sector (manufacturing – even for the 1980s. To quote him, rates in the economy) has been higher in
hereafter, for short) attracts academic and Acceleration in employment growth is the 1990s than in the previous decade.
policy attention disproportionate to its size found both at the aggregate level, and for In Goldar’s conception, wages and
– perhaps, for justifiable reasons. In the most industries. This may partly be ex- manday’s are seen as competing explana-
initial years of planning, much of the plained by changes in the size structure tions for labour demand. This is an incor-
development strategy focused on creating in favour of small and medium sized rect specification of the relationship.
a large and efficient capital goods indus- factories. Another important explanation Earnings per worker (wages according to
tries. But in recent years, this sector has for acceleration in employment seems to Goldar) is a product of wage rate (wage
come to symbolise all that is believed to lie in a slow down in the growth in real rate per standard working day) and the
have gone wrong with Indian develop- wages [Goldar 2000:1195]. number of days worked. Thus, a given
ment effort: modest long-term output This note re-examines the analytical and labour demand can be met either by
growth in a comparative economic perspec- empirical bases of Goldar’s conclusions, (i) increasing the number of workers
tive with high capital intensity, and low and suggests an alternative explanation (working the standard working day), or
(or negligible) total factor productivity for the employment expansion in the 1990s. (ii) increasing mandays worked by the ex-
growth. It is widely believed that one of isting workers, or (iii) both.3 In Goldar’s
the principal reasons for the poor perfor- Some Analytical Considerations econometric specification, variation in em-
mance is the policy-induced rigidities in It is a truism to say that in the conven- ployment across industry groups is a
the industrial labour market [for instance, tional competitive equilibrium theory, function of variation in (i) wages
Lucas 1989; Datt Chaudhuri 1995]. employment is inversely related to wages. (ii) mandays and (iii) gross value added
‘Jobless growth’ in manufacturing in However, in theory, if efficiency wage or (GVA) – all in terms of growth rates. Such
the 1980s was considered as the evidence labour quality considerations are intro- a specification is incorrect, because the
for this proposition. During 1980-89, as duced, then the standard results get suit- variable ‘wages’ already includes the effect
the annual earnings per worker (earnings ably modified. Further, in the industrial
Table 1: Growth in Number of
for short) went up by 3.5 per cent, there sector with mark up pricing, with excess Establishments and Employment in
was said to have been a large-scale sub- capacity, and with considerable state inter- Unregistered Manufacturing, 1985 -
stitution of labour for capital that practi- vention in all the markets, one is not sure 1994-95 for all India.
(Percentage change over the previous period)
cally stopped employment growth. This if the inferences of a perfectly competitive
view has been widely contested. It was economy can be directly applied. Therefore, Year Establishments Employment
shown that nearly half the earnings growth in principle, Goldar’s inference that employ- 1978-79
represented an increase in mandays, im- ment revived in the 1990s due to a wage 1984-85 15.1 12.8
plying, during this period workers worked moderation appears somewhat simplistic. 1989-90 3.9 –0.8
1994-95 –9.5 –1.3
longer hours or more number of days, From a different analytical perspective,
while the wage rate (for the standard one can question whether the standard Source: Lalitha (1999).

Economic and Political Weekly September 16, 2000 3445


Figure 1: Employment in Registered Manufacturing Figure 2: Employment and Earnings in Registered
6.9 – Manufacturing
6.85 – 5–
6.8 –
Log of value

Per cent per year


4–
6.75 –
6.7 – 3–
6.65 –
6.6 – 2–
6.55 – 1–
6.5 –
74

76

78

80

82

84

86

88

90

92

94

96

98
0–
1975-80 1981-91 1991-98
Year ending
Years
Employment
Employment Earnings

Figure 3: Growth in Employment and Mandays in Registered Manufacturing 1950, to less than 60 in 1976. This result
3.5 –
was true for most 2-digit industry group
– a glaring fact that many studies on Indian
3 – industry chose to ignore [Sandesera 1979;
Nagaraj 1985]. The trend decline contin-
2.5 – ued in the 1970s and the 1980s as well
Per cent per year

[Nagaraj 1994, Figures 8 and 9]. There-


2 – fore, Goldar’s observation of the changes
in the size distribution of employment is
1.5 –
part of a long-term trend, observed over
half a century. Therefore, it has probably
got little to do with the reforms initiated
1 –
in the 1990s. Thus, the evidence does not
really support his explanation.
0.5 – Finally, and perhaps more significantly,
Goldar’s econometric estimates seem to
0 – have serious problems. Using his data in
1980-91 1992-97
Years Tables 3 and 4 (pp 1193-94), we have re-
Workers Mandays estimated the regression equations, by
introducing the independent variables one
of ‘mandays’. Therefore, by definition, Some Empirical Considerations by one, to understand the incremental
one cannot expect to find an independent contribution of each of them in explaining
effect of mandays reflected on the depen- In Goldar’s view, the faster growth of the variations in the dependent variable
dent variable. Econometrically, there employment in smaller sized factories, and (Tables 2 and 3).6
is bound to be high multi-collinearity loss of employment in larger size classes
between wages and mandays. is in response to industrial and trade policy Table 2 (ii): Inter-correlation Matrix
If Goldar’s argument that much of the reforms initiated in the 1990s. But this
employment has occurred in smaller sized argument ignores the fact that such a change Employ Wages Mandays GVA
factories in response to lower wages, in in the size structure has been taking place Employ 1
a regime of more appropriate industrial over the last five decades. For instance, in Wages –0.197 1
Mandays –0.326 0.622* 1
and trade policy regime, then the same the factory sector, the average factory size
GVA 0.709* 0.426 0.155 1
argument should hold more true for the fell from over 140 workers per factory in
unregistered manufacturing as well. In Table 2 (i): Regression Analysis, 1980–90
principle, these policy changes should –2
increase demand for labour-intensive Sl No Constant Term Wages Mandays GVA R
manufacture that should boost output and 1 2.229 –0.200 0.041
employment growth.4 But the evidence is (0.779)
2 1.903 –0.009 –0.746 –0.009
not consistent with such an explanation.5 (0.028) (1.028)
Successive five-yearly economic census 3 –0.769 –0.507 –0.361 0.526 0.785
data, between 1984-85 and 1994-95, sug- (3.044)* (1.066) (7.258)*
4 – 0.667 –0.618 0.538 0.782
gests a decline in absolute number of (4.709)* (7.480)*
factories (establishments), and employment 5 –1.882 0.394 0.498
during 1984-95 (Table 1). Therefore, (3.895)*
Goldar’s argument that the changes in trade Note: Dependent variable: employment; GVA – gross value added. All variables are in growth rates. It is
and industrial policies have had a a cross-section regression of growth rates of the variables at 2–digit industries as the observation.
favourable impact on smaller sized facto- Number of observations, 17. Data are from Tables 3 and 4 of Goldar’s paper. Figures in brackets are
absolute values of t–statistics.
ries in the factory sector needs to taken * represents statistical significance at least at 5 per cent significance level. This notation holds for the
with great caution. other tables in this study as well.

3446 Economic and Political Weekly September 16, 2000


Figure 4: Fixed Investment Growth in Registered Manufacturing However, as output expanded, the over-
20 –
hand got exhausted, and the employment
growth recovered. Then, the question arises,
Per cent per year

15 –
how was the employment growth sustained
in the 1990s, despite a modestly slower
10 –
output growth compared to the previous
decade? The explanation probably lies not
5 –
in output growth but in investment boom
0 –
that was witnessed in response to the in-
1974-80 1981-86 1987-91 1992-97 dustrial deregulation and trade policy re-
Years form. As Figure 4 shows, average of an-
Gross fixed capital formation nual growth rate of fixed investment (gross
fixed capital formation) went up from 9
We find that simple linear regression is correct, then it could be of some value per cent during 1986-91 to 15 per cent
between employment (dependent variable) in exploring an alternative explanation. during 1992-97.10 Although the interest
and wages (independent variable) – both rates in the economy were high, dampen-
in terms of growth rates for the cross-
An Alternative Explanation ing investment demand, the cost of capital
section of 17 2-digit industry groups – How does one explain the revival of for the large private corporate firms was
has no explanatory power. The explana- employment in the 1990s, if Goldar’s lower during the first half of the 1990s as
tory power does not improve either, when proposition (of wage moderation) is not the stock market boom reduced real cost
‘mandays’ variable is added. This simply valid? Figure 1 plots, on a log scale, of investable funds. This was perhaps aided
confirms our earlier argument that, by employment of workers over the period, by a freer access to international capital
definition, one cannot expect mandays to 1973-74 to 1997-98.8 During 1974-83, markets for large Indian firms. Hence, we
have an independent explanatory power, employment grew steadily at 3.4 per cent would hypothesise this aspect of the prod-
once wages is included. But, interestingly, per year. Then, there was a fall in the uct market is primarily responsible for the
when gross value (GVA) added is in- absolute number of workers for four years, improved employment growth in the 1990s.
cluded, the regression equation becomes between 1982-83 and 1986-87. Employ-
ment growth recovered thereafter – though Conclusions
statistically significant, with both the
wages and GVA turning out to be statis- at a slower rate and with a greater fluctua- To sum up, contrary to the jobless growth
tically significant, with the expected signs. tion – peaking in 1995-96. The trend growth in the 1980s, registered manufacturing
How does it happen? rate in employment during 1987-98 is about employment grew annually at about 3 per
To understand this, we computed inter- 3.2 per cent per year. Evidently, there is cent during 1991-97. Goldar explains it
correlation matrix of all the four variables. no break in the trend in 1991-92. mainly as a response to wage moderation:
Simple (linear) correlation coefficient Since the output growth rate in the 1970s the annual growth rate of earnings per
between employment and wages is insig- was lower, in relation to employment worker declined from 4.8 per cent in the
nificant, though it has the expected nega- growth, there developed an overhang of 1980s, to 2.5 per cent in the 1990s. He
tive sign. However, GVA is correlated, employment in the 1980s (Figure 2).9 As further argued that the decline in wage
statistically significantly, with employ- we hypothesised earlier [Nagaraj 1994], growth partly represents changing size
ment in the 1980s, and with wages in the this is probably what caused the decline in distribution of factories in the 1990s
1990s. This suggests that the associations employment during the mid-1980s, for about
between employment and wages are me- four years. As output demand picked up, Table 3 (ii): Inter-correlation Matrix,
1990-97
diated by GVA. But independent of it, the firms first sought to maximise the use the
variation wages across the industry-groups existing stock of capital and labour re- Employ- Wages Mandays GVA
hardly explains the variation in employ- sources. This got reflected in greater ment
ment, during both the periods. Therefore, mandays than growth in employment. In Employment 1
unless one has a priori explanation as to other words, growth in earnings in the 1980s Wages 0.021 1
Mandays -0.231 -0.303 1
how GVA affects both employment and was more on account of growth in mandays
GVA 0.4695 0.768* -0.396 1
wages across the industry groups, than the growth in employment (Figure 3).
Goldar’s econometric results have little
Table 3: Regression Analysis, 1990-97
economic meaning.7
Here, our earlier analytical discussion
–2
could be of some value. Perhaps, these Sl No Constant Term Wages Mandays GVA R
results suggests that wages and employ- 1 3.097 0.017 0.0
ment are together driven by some aspect (0.081)
of the product market – indicating that the 2 3.454 -0.044 -1.437 -0.079
likely causation runs from the product (0.199) (0.909)
3 -0.002 -0.673 -0.305 0.551 0.389
market to the labour market. This seems (2.713)* (0.246) (3.424)*
consistent with the Keynesian postulate – 4 0.884 -0.673 0.561 0.430
rather than with the neo-classical causa- (2.809)* (3.751)*
5 -1.392 0.239 0.220
tion running from an active labour market,
(2.059)*
to the product market. If this reasoning

Economic and Political Weekly September 16, 2000 3447


towards smaller sized factories, which he, favour of smaller sized factories for the economic reasons in terms of labour quality,
in turn, attributes to the industrial and trade last five decades. efficiency wages, etc, which probably account
for much of the observed wage differential
policy reforms initiated since 1991. More- – Finally, and perhaps most significantly, [Mazumdar 1989].
over, he contested the view that the growth Goldar’s regression results have a prob- 6 Just as Goldar, we have made computations
in mandays played no role in explaining lem of interpretation. Our re-estimation separately for 1980-90, and for 1990-97. The
results for the two decades are broadly similar,
the employment trends – not even in the of Goldar’s model by introducing one though they are generally stronger for the 1980s.
1980s, as some studies have contended. independent variable after another (to 7 In the absence of a clearly articulated economic
This paper re-examined Goldar’s find- understand the incremental contribu- reasoning for including the variables in the
ings, both from analytical and empirical tion of each additional variable intro- regression equation, and the form chosen to
represent them, Goldar’s model remains a mere
points of view. The following are the main duced) suggests that there is, in fact, no statistical exercise.
results of such an attempt: statistically valid relationship between 8 This refers to employment of workers, taken
– The inverse relationship between employ- growth in employment and wages across from the ASI. We have used 1973-74 as the
ment and wages is a truism in a conven- industries (as postulated). The relation- starting point of the series as the official
summary results for the factory sector are
tional equilibrium analysis. However, ship is ‘driven’ by gross value added, available continuously from this year.
when efficiency wage considerations which is statistically significantly asso- 9 One could argue that the overhang is a clear
or quality of labour are introduced, one ciated with both employment in one evidence of the rigidities in the labour market.
cannot be sure if the competitive equi- case and with wages in the other. It, thus, Not necessarily, we would contend, since the
overhand may in fact reflect efficiency wage
librium conditions really hold. suggests that the labour market vari- hypothesis, as industrial skills are in short supply.
– Viewed from a Keynesian perspective, ables seem to be mediated by product 10 In contrast, investment growth in the un-
one could argue that the economic cau- market variables. Therefore, Goldar’s registered manufacturing declined perceptibly
as the real interest rate shot up in the 1990s.
sation could run from the goods market empirical evidence to support the in- Moreover, in the changed policy regime,
to labour market, and not the other way verse relationship between employment reduced credit flow to the small-scale sector
round. Since the evidence from the de- and wages appears seriously flawed. also perhaps contributed to the investment
veloped economies is very mixed on – We hypothesise that the product market decline.
this issue, there is little basis to be sure variable that could explain the employ-
that the transmission mechanism runs ment growth in the 1990s is the growth References
from the active labour market to the in fixed investment. It seems likely that Bhaduri, Amit (1996): ‘Employment, Labour
product market, as postulated in the industrial deregulation led to an invest- Market Flexibility and Economic Liberalisation
neo-classical theory. ment boom, aided by the stock market in India’, The Indian Journal of Labour
Economics, Vol 39, No 1.
– Goldar’s postulated relationship be- boom in the first half of the 1990s that Bhalotra, Sonia R (1998): ‘The Puzzle of Jobless
tween employment and wages is incor- reduced capital cost for large private Growth in Indian Manufacturing’, Oxford
rect, as employment depends not on corporate firms. EPW Bulletin of Economics and Statistics, Vol 60,
No 1.
wages but on wage-rental ratio. How- Datt Chaudhuri, Mrinal (1995): Labour Market
ever, this may not vitiate his results as Notes as Social Institutions in India, Working Paper
the wage-rental ratio could, in fact, have [Following the usual disclaimers, I thank Chandan No 16, Centre for Development Economics,
fallen since the economy’s real interest Mukherjee, K V Ramaswamy and M H Surya- Delhi School of Economics.
narayana for their comments and suggestions on Edwards, Alejandra Cox and Edwards, Sebastian
rate rose rapidly in the 1990s. (1994): ‘Labour Market Distortions and
an earlier draft of this paper; and Ramesh Datt,
– To test if the mandays have an indepen- and Errol D’Souza, for discussions on this study.] Structural Adjustment in Developing
dent influence on employment, Goldar Countries’, Susan Horton et al (eds), Labour
regressed employment on (i) wages and 1 Unless otherwise mentioned, in this study, all Markets in an Era of Adjustment, Volume 1,
variables in value terms are at constant prices. EDI Development Studies, The World Bank,
(ii) mandays, in a cross-section regres- 2 In economic theory there is no consensus on Washington, DC.
sion of 17 industry groups (all in growth which kind of transmission mechanism – neo- Goldar, Biswanath (2000): ‘Employment Growth
rates). This is an incorrect specification classical or Keynesian – is empirically more in Organised Manufacturing in India’,
since wages, by definition, is a product valid. As Amit Bhaduri said, “Economists are Economic and Political Weekly, Vol 35, No 14,
sharply divided regarding the question of the April 1-7.
of wage rate (for the standard working relevant transmission mechanism... It is only Lucas, Robert E B (1988): ‘India’s Industrial
day) and number of mandays. As fair to say that the state of our knowledge in Policy’ in Robert E B Lucas and Gustav
Goldar’s independent variables are this respect is fluid, and the ‘Keynesian Papanek (eds), The Indian Economy: Recent
consensus’ no longer rules supreme; nor do Developments and Future Prospects, Oxford
inter-related, by definition, the effect of monetarist conclusions” [Bhaduri 1996:15]. University Press, Delhi.
mandays will not be captured in such 3 For simplicity, we have ignored intensity (or Mazumdar, Dipak (1989): ‘Microeconomic Issues
a multiple regression equation. More- porosity of the working day), which is amenable of Labour Markets in Developing Countries:
over, econometrically, there is bound to change by piece rate work, by incentive Analysis and Policy Implications’, EDI Seminar
income, or more intensive supervision. Paper No 10, World Bank, Washington, DC.
to be high multi-collinearity. 4 For an exposition of the effects of trade policy Nagaraj, R (1994): ‘Employment and Wages in
– According to Goldar, changes in em- reforms on employment, see Edwards and Manufacturing Industries: Trends, Hypotheses
ployment in favour of smaller sized Edwards (1994). and Evidence’, Economic and Political
factories in the 1990s partly account for 5 One could argue that with a reduction in the Weekly, Vol 29, No 4, January 22.
rigidities in the labour market after the reforms, – (1985): ‘Trends in Factory Size in Indian Industry,
the wage moderation. And he attributes employment would shift from the unorganised 1950 to 1980: Some Tentative Inferences’,
the changes in the size distribution to to the organised. Hence, the evidence in the Economic and Political Weekly, Vol 20, No 8,
the industrial and trade policy reforms 1990s is consistent with the theory. Such Review of Management, February 23.
reasoning may be a bit simplistic because Sandesera, J C (1979): ‘Size of Factory and
initiated since 1991. This view is per- wages differential across the organised and Concentration in the Factory Sector in India,
haps overly simplistic, since size dis- unorganised sectors is not merely due to the 1951 to 1970’, Indian Economic Journal,
tribution has been steadily moving in policy induced rigidities. There are sound Vol 27, No 2.

3448 Economic and Political Weekly September 16, 2000

You might also like