The sluggish growth of employment in the organised manufacturing sector in India, has caused much concern. In general, the growth of employment has lagged far behind the growth of output. Between 1950-94, employment in the organised manufacturing sector only grew at an estimated average rate of c3.7% p.a., while real gross output grew by an estimated 8.1% p.a.[1]. There has been a large gap between output growth and employment growth. The present paper tries to evaluate the reasons for this output-employment growth (OEG) gap for the period 1950-94 taken as a whole.
Previous studies have favoured four different reasons for the OEG gap. For instance, Gowda (1984) appears to pick out a low output elasticity of employment that he estimates to be about 0.5. Others ascribe the slow growth of employment to various inhibiting factors, such as the growth of wage rates due to minimum wage legislation, and the rigidities in the labour market induced by job security legislation (Agrawal 1997, Fallon and Lucas 1991). Thus, while stressing the importance of rising wages in this regard, the World Bank points “to a significant trade off between the higher real cost of labour and employment”. Ahluwalia on the other hand identifies policy induced rigidities in the labour market as the principal reason for the slow growth of employment. (Nagaraj 1994).
It has also been suggested in passing that “poor productivity of ... employees leads to very slow growth of employment in the organised sector” (Agrawal 1997, p157). Apparently a decline in productive efficiency is being argued. Aggregate studies do indicate a long run decline in efficiency in Indian industry, reflecting in a negative growth of total factor productivity (Goldar 1983, Ahluwalia 1991). This decline in productive efficiency could have been general or ‘disembodied’ and/or input-specific or ‘embodied’. Any fall in efficiency, whether general or labour-specific, lowers the marginal productivity of labour and ceteris paribus, pulls down its employment. Thus declining productive efficiency could well have kept down the growth of employment in Indian manufacturing in the period of analysis.
Our paper tries to evaluate the different reasons for the OEG gap by estimating the influence of output, product-wages, productive efficiency, and labour market rigidities on employment in organised Indian manufacturing industry in the period: 1950-1994. The approach is aggregative, econometric and seeks generalisations for the entire period.
Source of Data
The study is based on aggregate time series data drawn from CSO abstracts. CMI data till 1958 is followed by ASI Census Sector data after 1959. Since CSO discontinued reporting Census Sector data after 1982, it has been extended to 1994 by splicing with factory sector data after 1982.[2] The CSO abstracts constitute the sole data base for this study.
Theoretical Framework
The Approach
Our object is to estimate the influence of production, product-wages, efficiency changes, and labour market rigidities on employment.
Employment depends upon production plans. Given the level of production, it can be affected by product-wages (which measure the ‘real cost’ of labour to the employer) only if input proportions are variable. To allow for the possibility of variable input proportions, we derive the labour demand function from a CES production function. Although it imposes the restriction of a constant elasticity of substitution (s), the CES production function includes the fixed proportions case and the Cobb-Douglas form as special cases. Hence, the appropriateness of these forms for modelling the techno-economic production relations in organised Indian manufacturing (1950-94) can be tested by using the CES production function.
The efficiency parameters of the CES production function are allowed to change at a constant exponential rate over the period of analysis.[3] These changes have the effect of shifting the production function. Overall or ‘disembodied’ efficiency changes as well as input-specific or ‘embodied’ efficiency changes are allowed for. Overall efficiency changes shift the marginal product curves of all inputs equi-proportionately, whereas input-specific efficiency changes shift the marginal product curves of specific inputs only. This study is concerned only with the net effectof all efficiency changes on employment. It does not try to distinguish between the different types of efficiency changes.
Labour market rigidities may force the industry to realise only a part of their equilibrium (or profit-maximising) employment plans. Industry may adjust employment only partially due to job security legislation[4]. For instance, a proposed cut in employment may be curtailed by labour legislation that requires considerable severance pay and government permission in many cases. On the other hand, expansion in employment may be inhibited by such legislation, since they restrict the scope of future job cuts. Thus, labour market rigidities may force partial adjustments of employment on the industry.
We try to capture partial adjustment of industrial employment through a Nerlovian type of model that makes the actual current employment a weighted function of the ‘equilibrium’ employment and the past employment.[5] The influence of past employment on the actual current employment shows that industry is only partly adjusting its employment to the equilibrium. If past employment exerts no influence on current employment, it means that industry is fully adjusting employment to its profit- maximising or ‘equilibrium’ level.
Equilibrium Demand for Labour
The ‘equilibrium’ influence of output, product-wage and efficiency levels on employment can be deduced by applying the marginal productivity condition to the CES production function. Simple manipulation yields the equation[6]:
lnLt* = B + ewln()t + eqlnQt + r t ... [1]
where Lt* = the ‘equilibrium’ level of employment,
()t = the product-wage i.e., (wage rate)/(product price),
Qt = output level,
ew= the ‘equilibrium’ product-wage elasticity of employment, to be referred as the wage elasticity of employment hereafter. Its absolute value equals the elasticity of substitution (s) of the CES function,
eq= the ‘equilibrium’ output elasticity of employment. Its value equals a compound expression : [] in which nis the returns to scale parameter,
r = the sum of exponential rates of change of overall efficiency (g) and labour efficiency (l).
Eqn.[1] reveals ‘equilibrium’ employment to be a function of the product-wage, output level and time.
Expectations about the parameters
ew— the wage elasticity of employment, is zero on an Lshaped isoquant (fixed proportions case) and negative on other isoquants (variable proportions). Its absolute value equals the elasticity of substitution (s) in the CES function. Hence an absolute value of one for ew, would indicate the appropriateness of a Cobb Douglas form. This suggests the following hypotheses: H0: ew = 0 (rejection of which would cast doubt on the fixed proportion form) and H0: |ew| = 1 (rejection of which would cast doubt on the Cobb Douglas form).
eq— the output elasticity of employment, is expected to be positive, i.e., ‘equilibrium’ employment may be expected to rise as production increases. Gowda (1984) estimates its value to be .5. The technical notes prove that eq= 1, either when the elasticity of substitution = 1 or when there are constant returns to scale, or both. This suggests the hypothesis H0: eq = 1. Rejection of this hypothesis would rule out the Cobb-Douglas form as well as constant returns to scale.
The Keynesian revolution shifted the focus of employment theory from wage-cost considerations to production plans. In the extreme instance, it has been stressed “that it is not the reduction in the real wage that brings about increased employment, but the increased effective demand” (Fletcher 1989, p279). The more moderate General theory however allows for a negative effect of real wage on employment.[7] However, it retains the focus on effective demand. The greater attention paid to real output over real wage would be empirically justified if the output sensitivity of employment were greater than its wage sensitivity. This suggests the plausible hypothesis that the ‘equilibrium’ output elasticity of employment is greater than the absolute value of the ‘equilibrium’ wage elasticity of employment, i.e., eq> ½ew½.
r — the coefficient of time, is the sum of the change in overall efficiency and labour-specific efficiency (see technical notes for derivation). It measures the net change in productive efficiency and its effect on employment. Negative values of TFPG lead us to expect ‘r’ to be negative.
Partial Adjustments due to Labour Market Rigidities
Combining the Nerlovian model with eqn. [1] above, we get (see technical notes):
lnLt = C + lewln()t + leqlnQt + lr t + (1-l) lnLt-1 ... [1a]
where l = partial adjustment factor and (1- l) reflects the influence of job security on employment,
Expectations about 1-l
Eqn.[1a] differs from eqn.[1] in two ways. Firsts, all the coefficients of eqn.[1] are now multiplied by l, which is called the partial adjustment factor. If lis equal to one, there is no partial adjustment and the coefficients in eqn.[1a] are the same as in eqn.[1].
Second, eqn.[1a] has an additional variable: past employment—lnLt-1.
1-l— the coefficient of past employment, shows the presence and extent of partial adjustment. Since partial adjustments represent the effect of labour market rigidities, its value shows the effect of labour market rigidities and job security legislation.
If 1-l, equals zero, labour market rigidities have no influence on employment, there is no partial adjustment and industry fully adjusts employment to the equilibrium level. Further, if 1-l= 0, l= 1, which means that the coefficients of ln()t , lnQt , and t in eqn.[1a] are equal to their equilibrium values: ew,eq, and r. In that case the expectations of the previous section apply to these coefficients. This suggests the hypothesis, H0: (1-l) = 0. Rejection of this hypothesis would show a lack of support for the partial adjustment hypothesis and the labour market rigidities explanation, and enable us to interpret the other coefficients of eqn.[1a] as being equal to their equilibrium values.
Summing up
The subsequent regression analysis will use time series data (1950-94) to throw light on the hypotheses outlined above, i.e.,
H0: (1-l) = 0, H0: ew = 0, H0: |ew| = 1, H0: eq = 1, H0: eq>½ew½, and r < 0.
The principal objective of the analysis, however, will be to explain the output-employment growth gap in the period of analysis.
The Regression Equation
Eqn.[1a] is not yet ready for regression analysis.
First, the directions of influence must be argued. Output and product-wage can be expected to be independent of industrial employment for the following reasons. Industrial output depends on demand side forces, and the availability of inputs such as electricity, etc.. Since employment in the organised manufacturing sector is only a small fraction of the total employment in the country, it can be assumed that it does not affect demand for industrial output or its product price[8]. Likewise, the industrial wage is assumed to be given by complex negotiations in the labour market that ultimately rest on the class balance of forces, although the negotiations may refer to the industrial production and the supply price of labour. On these grounds, it is assumed that industrial output and product-wage are exogenous in eqn.[1a].
Second, eqn.[1a] needs to be transformed to avoid the problems of non-stationarity of the time series variables. It is now understood that regression between non-stationary variables may lead to spurious regressions, or to ‘super consistent’ results that may generally be biased in finite samples (Davidson and Mackinnon 1993, p 721). We adopt the classical approach to this problem by differencing eqn.[1a], which reduces the individual terms to stationary[9] growth rates, yielding the regression form:
gr.Lt = lr + lewgr.()t + leqgr.Qt + (1-l) gr.Lt-1 + ut ... [2][10]
where gr. = exponential growth rate
One significant feature of eqn.[2] when compared with eqn.[1a], is that the coefficients of growth rates are the same as the corresponding coefficients of logarithms of variables. Hence their interpretation remains unchanged.
A second important feature of eqn.[2] is that the coefficient of time — r, in eqn.[1a] has now become the constant term in [2], and ‘t’ has disappeared. Thus, the net effects of efficiency change on employment are now captured by the constant term in [2].
The Regression Estimates
The regression of eqn.[2] yielded estimates that are reported in eqn.[2a] below:
gr.Lt =
- 0.0266
- .2527 gr.()t
+ .8943 gr.Qt
- .0559 gr. Lt-1
... [2a]
(-2.828)
(-2.429)
(11.639)
(-.741)
n = 44,
R2 = .768,
DW = 1.96,
Condition Index
= 2.89
figures in parenthesis are t statistics
The residuals seem to be well behaved and normally distributed.[11]
The coefficient of lagged employment is not significant. It was therefore dropped to improve the precision of the other estimates, and eqn.[2] was re-estimated without lagged employment to yield the 95% confidence intervals that are reported in rel.[2b] below:
-.45 < ew< -.04, .75 < eq< 1.06 and -.046 < r < -.012 ... [2b]
One problem with the data base used for estimation is the switch from census sector data to factory sector data after 1982. In order to see whether this has significantly affected the model, a forecast test is conducted, using the pre-1982 observations as the data set (d), and the post 1982 observations as the forecast set (f) and running two regressions: one on the data set[12] and the other on the data+forecast set.[13] The difference between the two is tested by: F = [(RSSd+f - RSSd)/f ] ¸[RSSd/(n-k)] = 0.7168, which is too small to reject the hypothesis that the model for the data set: 1950-82, continues to hold in the forecast period 1982-94. Evidently, the use of factory sector data after 1982, has not materially affected the results.
Table 1: Estimates of elasticities of employment, rate of labour augmenting technical change and returns to scale from regressions 1950-1980 to 1950-94
Period
Product wage elasticity of employment ( ew) ½ew½= s
Output elasticity of employment ( eq)
Exp. rate of change in efficiciency ( r )
Returns to scale (n )
1950–80
–0.21*
0.95
–0.024
0.94
1950–81
–0.25
0.92
–0.024
0.90
1950–82
–0.27
0.92
–0.024
0.90
1950–83
–0.25
0.91
–0.024
0.89
1950–84
–0.26
0.91
–0.024
0.89
1950–85
–0.26
0.91
–0.026
0.89
1950–86
–0.27
0.91
–0.027
0.90
1950–87
–0.27
0.91
–0.027
0.90
1950–88
–0.26
0.91
–0.028
0.89
1950–89
–0.26
0.91
–0.028
0.89
1950–90
–0.26
0.91
–0.029
0.89
1950–91
–0.27
0.90
–0.028
0.88
1950–92
–0.26
0.90
–0.028
0.89
1950–93
–0.24
0.90
–0.030
0.89
1950–94
–0.25
0.90
–0.029
0.89
Average value
– 0.26
0.91
– 0.027
0.90
Range
– 0.21 < ew< – 0.27
0.90 < eq< 0.95
– 0.024 < r < – 0.030
0.88 < n< 0.94
à significant at 10% level, all other estimates are significant at 5% level
Any discussion of regression estimates is usually hamstrung by the uncomfortable choice between the very precise point estimates in which zero degree of confidence rests, and the highly imprecise 95% confidence intervals that defy firm inferences. In order to steer between these extreme choices, this study runs 15 regressions[14] for time periods ranging from 1950-80 to 1950-94, and uses the average value and range of regression estimates derived therefrom, to draw tentative inferences. The full results are appended in Table 1, while the discussion below utilises the summary information in the last two rows of the table.
Results of the Analysis
1. The coefficient of past employment (1 - l) in eqn.[2a] is not significantly different from zero, implying that statistics do not support the hypothesis of partial adjustment and the ‘labour market rigidities’ explanation (e.g. Ahluwalia 1991) of the slow growth of employment in organised Indian manufacturing. If job security legislation do statistically affect employment, it is not through labour market rigidities.
2. (1-l) = 0 implies that l= 1. This means that the coefficients of gr.()t and gr.Qt in [2a] can be interpreted as the equilibrium wage and output elasticities of employment — ewand eq, and the constant can be interpreted as the net effect of changes in efficiency.
3. Eqn.[2a] shows that the wage elasticity of employment ewis significant and negative, with 95% confidence interval between -0.04 and -0.45 {rel.[2b]}.
This means that as the product-wage increases, employment decreases along an isoquant. Thus, the isoquant is not L shaped. The fixed proportion production function is rejected. Moreover, |ew| is significantly less than one. Since |ew| equals the elasticity of substitution, this implies that the Cobb-Douglas form of production function is also rejected by the data. The results also indicate that the elasticity of substitution for the period as a whole, may be fairly small (< ½).
The range of 15 regressions puts the point estimates of ew between -0.21 and -0.27. This suggests that a 1% increase in the product-wage may reduce employment by 0.21–0.27%, cet.par.. There is indeed “a significant trade-off between the real cost of labour and employment” (cf. Nagaraj 1994).
4. Eqn.[2a] shows that the output elasticity of employment eqis positive and fairly high, with a 95% confidence interval between 0.75 and 1.06 {rel.[2b]}.
The confidence intervals show that eqis not significantly different from one. Hence, the hypothesis of a unitary output elasticity of employment — eq= 1, is not rejected. The range of 15 regressions puts the point estimate of eq between .9 and .95. This means that if output increases by 1%, employment may increase by .9–.95%, cet.par., i.e., which is nearly 1%, and much more than Gowda’s estimate of 0.5.
5. It has been pointed out earlier that eq= 1, either when the elasticity of substitution equals unity or when there are constant returns to scale, or both. The results thus far indicate that eqis possibly = 1, and the elasticity of substitution is not unity. This leaves constant returns to scale with the burden of maintaining the equality. Hence, the hypothesis of constant returns to scale cannot be rejected.
The range of the 15 regressions puts the point estimates of the returns to scale parameter between 0.88 and 0.94, which is less than but fairly close to one.
6. Eqn.[2a] also shows that the output elasticity of employment may be much larger than the absolute value of wage elasticity of employment, and rel.[2b] shows that there is no overlap in the 95% confidence intervals. Thus, employment is far more sensitive to output than to the product-wage. This seems to support the Keynesian pre-supposition that employment is driven more by the demand for output than by wage cost considerations.
7. Eqn.[2a] shows that the exponential rate of change in efficiency — r, is significant and negative, with 95% confidence intervals between -0.046 and -0.012 {rel.[2b]}. This suggests a decline in productivity efficiency in organised manufacturing in the time period of analysis.
The 15 regressions put the point estimates between -0.024 and -0.03. This means that, cet. par., industrial employment declines annually by about 2.4–3% p.a. on account of declining productive efficiency alone.
8. The average values of the point estimates from 15 regressions can be used to decompose the average annual growth rate of employment into the different contributions. For the entire period 1950-94, calculations based on the average point estimates suggest an average growth rate of employment of 3.7% p.a.[15]. This estimate of employment growth decomposes into: +7.3% p.a. due to output growth, -0.7% p.a. due to product-wage growth, and -2.7% p.a. due to declining efficiency.[16]
This means that the period as a whole, an average output growth of 8 % p.a., promised a potentially equal employment growth — 7.3% p.a.. Of this potential employment growth, however, only 50% (3.7% p.a.) actually materialised. The remaining 50% of the potential employment growth failed to materialise, and was lost due to other factors.
About 40% of the lost potential growth in employment was due to declining efficiency (-2.7 ÷ 7.3 = -37%). Thus, declining industrial efficiency was the main reason for the failure of employment growth to keep pace with output growth in organised manufacturing in the period 1950-94.
Only 10% of the lost potential employment growth was due to rising product wages (0.7 ÷ 7.3 = 10%). Thus rising product wages contributed to the gap between output growth and employment growth only in minor way in the period as a whole.
Conclusions
This study finds a near unity output elasticity of employment, a significant trade-off between the real cost of labour and employment, and a sizeable downward pull on employment exerted by declining efficiency in the organised manufacturing sector for the period (1950-94) as a whole .
Output in organised manufacturing grew by c 8% p.a. during the period. However, the growing output failed to pull up employment at the same rate in-spite of a near unity output elasticity of employment. Actual employment grew by only c 3.7% in the period. This paper finds that the principal reason for the output-employment growth gap in the period seems to have been declining industrial efficiency. On an average, the decline in efficiency alone seems to have pulled down employment by 2.4–3% p.a., cet. par.. The growth in product-wages seems to have contributed only in a minor way to the slow growth of employment with an average contribution of -0.7% p.a..
One major finding of the study is the lack of evidence for partial adjustment of employment, that would have been expected from ‘labour market rigidities’. It does not seem likely that labour market rigidities were responsible for the output-employment growth gap in the period. Thus, it is not through labour market rigidities “that overly protective labour laws in the organised sector slow down the growth of employment” (Agrawal 1997). However, it is possible that the job security legislation may have contributed to the slow growth of employment through increasing inefficiency in the production process.
The output elasticity of employment is estimated to be much higher than the product-wage elasticity of employment. This provides support to the Keynesian approach which presupposes that employment is driven more by the demand for output rather than wage-cost considerations.
The study suggests a low value of elasticity of substitution between inputs, with a 95% confidence interval ranging between 0.04 and 0.45. The estimate represents an average for the period, and rejects both the fixed and Cobb-Douglas forms of production functions. The hypothesis of constant returns to scale is not ruled out by the study and point estimates of the returns to scale parameter range between 0.88 and 0.94.
Technical Notes
Factory sector data have been spliced on to the census sector data after 1983 by computing average weights for each variable for 1980-82 in which common observations are available for the two data sets. Wages are taken to mean total emoluments of all employees. Missing observations on other emoluments for some years in the Census sector data have been estimated by regression. The estimates of the year 1972 are calculated as a simple average of the previous and succeeding year’s observation. Product price is taken as the price index of manufactures (1952-53 = 100), which has been extended to 1994 by splicing at various points of time.
The Nerlovian Model
Given the equilibrium demand for labour Lt*, the Nerlovian partial adjustment model suggests that:
Lt - Lt-1 = l(Lt* - Lt-1), which gives: Lt = lLt* + (1 - l) Lt-1 ... [3]
where lis the partial adjustment factor.
The Equilibrium Labour Demand Function
The equilibrium demand for labour Lt*, can be derived from the marginal productivity conditions of the CES production function as follows.
Given, the three factor CES production function with efficiency parameters:
Q = Aegr[aleltL-b+ akektK-b+ amemtM-b]-n/b
where ‘g’ is the exponential rate of change in overall efficiency of the firm, ‘l’ is the exp. rate of labour-specific efficiency change, ‘k’ is the exp. rate of capital-specific efficiency change and ‘m’ is the exp. rate of material-specific efficiency change.
The marginal productivity condition is:
()t = (ale(l+g)tL*t-(b+1))/(nQt-1 -(b/n))
Substitutings= 1/(1+b), r =(l+g), and taking logarithms on both sides, we get:
ln()t = A + r t + [] lnQt - lnLt* which on manipulation yields:
ln Lt* = B + r t + [] lnQt - sln()t ... [4]
Expressing the coefficients of lnQt and ln()t as equilibrium elasticities eq and ewrespectively, we get:
ln Lt* = B + r t + ewln()t + eqlnQt ... [1]
Introducing Equilibrium Labour Demand into the Nerlovian Model
Substituting eqn.[1] in eqn.[3], we get the final mathematical form:
ln Lt = C + lewln()t + leqln Qt + lr t + (1-l) ln Lt-1 ... [1a]
where ew= sand eq= []
Thus, eq= 1, either when , or when , or both.
*The Author is grateful to Dr. Prabha Panth and Prof. S. Kishan Rao, both of Osmania University for their comments and discussion of earlier drafts of this paper.
Notes
1 The average growth rates are estimated for the period 1950-94 by exponential regressions (R2 .952 and .998 respectively), and the estimated exponential growth coefficients ge are converted into the annual compound growth rate by the formula: exp(ge) - 1.
[2] The sensitivity of the results to this treatment is also tested through a forecast test, whose results are reported. Other data adjustments are discussed in technical notes.
[3] We use the term ‘efficiency parameters’ in a broad sense to include both efficiency factors as well as the technological changes that may occasionally underlie them. Thus the changes in parameters encompass both a “broader view of technological progress,” as also “aspects of efficiency with which inputs are used” (Marjit and Singh, 1997). For an explicit statement of the form the production function and the efficiency parameters, see technical notes.
[4] Job security legislation make “for inflexibility in hiring and firing possibilities and rigidities in the labour market” (Ahluwalia 1991).
[5] See technical notes.
[6] See technical notes for derivation.
[7] The key question for Keynes in General Theory was not the negative real wage-employment relation established by the marginal product condition, but whether such a relation could be realised by general money wage cuts across the economy. He argued that a fall in prices would cancel out the effects of money wage cuts, leaving real wages and employment intact - a phenomenon called the “Keynes case” in literature (Reynolds 1987, p133). Later post-Keynesians allow for the possibility of an employment response to money wage cut, i.e., “the classical case”.
[8] It has been suggested that the industrial product price depends upon the price of raw materials, wage rate and productivity (Balakrishnan 1997).
[9] Dickey Fuller tests on the growth rates of employment, product-wage and real gross output based on the form:
Dyt = (a- 1) yt-1 + ut
yield the tstatistics: -8.021, -7.293 and -7.731 respectively (DW: 2, 1.62, 2.1 respectively). These are considerably higher than the critical value at 1%: -2.56 (Davidson and Mackinnon 1993, p 708), prompting us to reject the hypothesis of non-stationarity.
[10] Since the difference between the logarithms of observations of successive years yields the growth rate of the variable between the years.
[11] The JB statistic is 1.617 which indicates a more than 25% probability that the residual term is normally distributed. The auto-correlograms show white noise, with a possibly high autocorrelation at the 7th lag. Hence, the Breusch Godfrey test is run, but it does not reject the hypothesis of no autocorrelation even at a lag of 7 periods. Tests do not support the possibility of heteroskedasticity or ARCH either.
[12] The data set regression 1950-82 gives results that are very similar to the one reported for 1950-94:
gr.Lt =
- 0.02
- .27 gr.()t
+ .9 gr. Qt
- .07 gr. Lt-1
(-1.79)
(-2.23)
(10.88)
(-.85)
R2 = .822
DW = 2.04
[13] The test is sensitive to the assumption of normal distribution of the residuals (Stewart 1991, p82). The residuals from the extended: data+forecast regression have already been shown to be normally distributed with a probability of more than 25%. The JB statistic for the residuals from the regression on the data set is 0.792 which suggests a more than 50% probability of normal distribution.
[14] Eqn. [2] is estimate after excluding lagged employment.
[15] As against this average employment growth from exponential regressions on actual data is estimated to be 3.64% p.a. (R2 = .89). The difference between this and the estimate from calculation using the point estimates is due to rounding off of the point estimates, and the use of the average of 15 regressions.
[16] The estimated annual compound growth rates of output and product-wage during the period are: 8.07% p.a. and 2.75% p.a. respectively (R2 = 0.986 and 0.967 respectively). their contribution to employment growth is calculated by multiplying these average growth rates with the point estimates of their coefficients.
References
Agrawal P. (1997) “Labour Policy: Striking a Balance”, in India Development Report 1997, Ed. Parikh K.S., Oxford University Press, Delhi.
Ahluwalia I.J (1991) Productivity and Growth in Indian Manufacturing, Oxford University Press, Delhi.
Balakrishnan P. (1997) “The Short-run Behaviour of Prices and Quantities in Indian Industry”, in Indian Industry: Policies and Performance, Ed. Mookherjee D., Oxford University Press, Delhi.
Davidson R. & Mackinnon J.G. (1993) Estimation and Inference in Econometrics, Oxford University Press, Oxford.
Fallon P.R. & Lucas R.E. 1991 “The Impact of Changes in Job Security Regulations in India and Zimbabwe”, World Bank Economic Review, 5(3).
Fletcher G.A., (1989) The Keynesian Revolution and its Critics: Issues of Theory and Policy for the Monetary Production Economy, Macmillan, 2nd Edn.
Goldar B.N. (1983) “Productivity Trends in Indian Manufacturing Industry: 1951-1978”, Indian Economic Review, vol. xviii, no. 1.
Gowda M.V.S. (1984) “Organised Sector Employment in India”, IEA Conference Volume, 67th Annual Conference, Agra.
Marjit S. & Singh N. (1997) “Technology and Indian Industry”, in Indian Industry: Policies and Performance, Ed. Mookherjee D., Oxford University Press, Delhi.
Nagaraj R. (1994) “Employment and Wages in Manufacturing Industries: Trends, Hypothesis and Evidence”, Economic and Political Weekly, vol. 29, No. 4.
Reynolds P.J. (1987) Political Economy: A Synthesis of Kaleckian and Post Keynesian Economics, Wheatsheaf Books, New York.
Stewart J (1991) Econometrics, Phillip Allan, London.
