Wage–Productivity Relationship in Indian Manufacturing Industries: Evidences from State-level Panel Data

Abstract

Indian manufacturing industries have experienced the major effects of economic reforms. Since the effect of any policy is ultimately transferred to workers, this article is an attempt to determine the extent to which labour productivity and wages are inter-related in manufacturing industries. The study uses state-level panel data of manufacturing industries and empirically tests the relevance of marginal productivity theory of wages and efficiency wage theory. After confirming the stationarity of the series, various empirical tests such as cointegration, vector error correction mechanism and Granger causality are applied to check the long-run equilibrium relationship between wage growth and productivity growth.

The study finds a divergence between wages and productivity in India as well as in the states. The empirical analysis confirms the existence of long-run relationship between the two variables and finds efficiency wage theory to be more appropriate as its long-term disequilibrium correcting process is quicker as compared to the marginal productivity theory. The study suggests for having a skill intensity matching with capital intensity. Appropriate level of skill and training among the workers will, on the one hand, increase their bargaining strength for more compensation, and on the other hand, will encourage them to produce more.

JEL: J31, J24, O30, F66, L66

Keywords

Real Wages Labour Productivity Capital Intensity Contractualisation Manufacturing Industries India

1. Introduction

Standard micro-economic theory postulates that in the short run, when capital stock is assumed to be constant, output is sensitive to the labour employed. Wage rate is, therefore, determined by the marginal productivity of available labour¹ or simply labour productivity.² The theory posits a clear relationship between wage growth and productivity growth in the short run with wages adjusting to changes in productivity, productivity developments being exogenous (Meager & Speckesser, 2011, p. 9). Higher labour productivity is therefore reflected in higher wages. The theory assumes static conditions in which economic growth or change do not influence the population, capital and underlying techniques of production. In the long run, the firms can alter employment levels as well as the capital stock. The increase or decrease in the price of labour in relation to capital, therefore, allows firms to substitute capital for labour or labour for capital. Thus, whereas in the short run, wage increases have only a scale effect, in the long run, they result in both scale and substitution effects (Meager & Speckesser, 2011, p. 8). Since the substitution between the factors follow their marginal productivities under conditions of perfect competition, this indicates that in the long run also, wages depend upon productivity of labour.

In a nutshell, the stylised fact is that micro-economic theory considers that in a competitive economy, there is a close association between wage rates and labour productivity, both in the short run and long run, and that wages follow productivity developments. The theory does not allow for any influence of wage-setting on productivity. A similar rule is followed in wage-setting through collective bargaining, that is, when labour productivity increases, trade unions demand higher wages. Thus, the marginal productivity theory and bargaining theory assert that the causality runs from productivity to real wages, that is, productivity increases precede real wages (Yusof, 2008, p. 392). From the macro-economic perspective, however, changes in productivity are associated with movements in real wages. This is explained by the efficiency wage theory which states that higher real wages increase the opportunity cost of job loss, which can stimulate greater work effort to avoid redundancy. Thus, firms can retain more skilled workers as newly hired workers may not be as productive as experienced workers. This theory suggests that the causality runs from real wages to productivity. Thus, real wage increases (decreases) precede productivity increases (decreases). This is applicable if economic growth is associated with the generation of more employment opportunities.

Wages and productivity are important economic indicators. Productivity measures the output produced by workers in various sectors of the economy. Wage is the reward (remuneration) to workers for producing the output, and is part of the total cost of producing the output. Increases in productivity determine economic growth, and increases in wages improve the standard of living and reduce poverty. Based on this, a rise in labour productivity is a basic source of improvement in remuneration and thus living standards (Tsoku & Matarise, 2014). An appropriate balance between both variables is important on equity and efficiency grounds, as the distribution of income between the factors depends upon it. When wages rise in line with a productivity increase, they are both sustainable and create incentives or a stimulus for further economic growth by increasing household purchasing power (International Labour Organization [ILO], 2015). Real wages falling behind productivity growth depresses demand prospects, which also determines investment. Depressed wages do not provide an incentive for investment in technology and thus can hamper future productivity growth. For transition economies, a low-wage trap can be a barrier for the long-term catching up process. However, an increase in wages without a corresponding increase in productivity could result in an increase in the cost of production, which could aggravate inflationary pressures as well as erode a country’s international competitiveness and its attractiveness as a profitable centre for foreign investment (Goh, 2009). If wages are growing faster than labour productivity, an increase in labour productivity stimulates economic growth.

In a liberalised policy framework, when economies are open to trade and investment, capital and technological inflows are high, particularly in emerging market and developing economies. Consequently, wage-setting in manufacturing is not confined to labour productivity, as labour-displacing capital-intensive technologies are also important determinants.³ In India, major economic transformations have taken place during the post-reform period (since 1991). The liberal inflow of capital and technology as well as increased competition has made Indian manufacturing firms more capital intensive and technology oriented. As a consequence, productivity in all sectors, particularly manufacturing and services, has increased substantially. Here, India’s remarkably high economic growth, averaging around 7–8 per cent annually over a decade, is worth mentioning. However, the increase in the share of contract workers among total workers in the past decades is a negative outcome for labour welfare. Since India is a labour-surplus economy and wages and productivity are the key indicators for economic growth and well-being, understanding the dynamism between wages and labour productivity in the Indian context is important.

In India, manufacturing is the highest employment-providing sector and thus has a major role in wage-setting. The economic reforms of 1991 were heavily focused on manufacturing industries, and the subsequent easy access to capital and technology from abroad has altered the production and employment scenarios domestically. Against this background, this article determines the extent to which wage growth and productivity growth in manufacturing industries are inter-related. Empirical analysis of state-level panel data for manufacturing industries gives an insight into the short- and long-term association between the growth of wages and productivity and also sheds light on the direction of causality.

The article is organised as follows: the second section briefly reviews some recent empirical studies on the wage–productivity relationship in manufacturing industries. The third section discusses the methodology and explains the empirical approach adopted in the study. The fourth section examines the trends in wages and productivity in Indian manufacturing. The fifth section determines the differences in manufacturing wages and labour productivity across Indian states. The sixth section presents the empirical findings of the stationarity tests, cointegration test, vector error correction model (VECM) test and Granger casuality test, and finally the ninth section concludes.

2. Review of Related Studies

The wage–productivity relationship has received considerable space in the theoretical and empirical literature. Several studies have tried to explain the dynamic linkage between the two variables in the respective countries they have studied. There is a consensus regarding the positive relationship between productivity and real wages, though a number of studies have claimed the existence of this relationship in the long run (Kumar, Webber, & Perry [2012], Strauss & Wohar [2004], Wakeford [2004], among others). However, regarding the direction of the causal links, opinions differ. Yusof (2008) considers real wage to be the main variable that adjusts to maintain cointegration, while Nwaokoro (2006), Khoon and Nyen (2010), Nikulin (2015) and others find that productivity is important in explaining real wage rates. Some of the studies such as Kumar et al. (2012) and Millea (2002) reports empirical evidence about the bi-directional relationship between wages and productivity, considering the nature of the wage-setting process in different countries.

Regardless of the long-term association, the studies have not found a one-to-one relationship between the two variables. As per Strauss and Wohar (2004), increases in productivity are associated with a less than a unit increase in real wages in the USA. Barkery (2007) in a study on Washington, DC, for the period between 2001 and 2006 noted an increase in productivity by 17.9 per cent, although real wages barely moved up over the period. Sharpe, Harrison, and Arsenault (2008) found a meagre increase in the median real earnings of Canadians between 1980 and 2005, and a very high increase in labour productivity (upto 37.4 per cent). Wakeford (2004) in a study on South Africa found an elasticity coefficient of 0.58, indicating that productivity has grown more rapidly than wages. However, for the short run, real wages negatively influence productivity. Goh (2009) in a study on Malaysia found a negative relationship between wages and labour productivity in the short run. Goh as well as Ho and Yaap (2001) found the elasticity coefficient between wages and productivity to be greater than unity for Malaysia. In the Indian context, Goldar and Banga (2005) analysed time-series cross-section data for three-digit industries and found a positive relationship between labour productivity and the wage rate, but the marginal effect of labour productivity on the wage rate as well as on elasticity was found to be low. The study concluded that only a small part of the gain in labour productivity gets translated into a wage increase. Rath (2006) also asserted that growth rates of productivity, wages and employment move in the same direction in the long run, but in the short run, higher labour productivity has a negative impact on employment growth and changes in manufacturing wages. Sabharwal (2007) explained that in India, the relationship between wages and labour productivity during the post-reform era has weakened, which indicates that the proportionate gains from the productivity increase were not distributed among the workers.

The studies indicate that the gap between real wages and productivity varies according to the type of sector and type of labour. According to Mishel and Shiezholz (2011), labour compensation growth was particularly low in the private sector, while the growth of average wage was particularly weak for college-educated public workers. Lopez-Villavicencio and Silva (2010) analysed a macro-economic panel of OPEC countries and found that wage increases have exceeded productivity growth for permanent workers, while the opposite is true for temporary workers, in line with their lower bargaining power. Harrison (2009) in a study on the US advocated that the depreciation of fixed assets has increased as a result of the adoption of new technologies, which has tended to push the labour share downwards. Dunne, Foster, Haltiwanger, and Troske (2004) in a study on wage productivity dispersion in US manufacturing found that a significant fraction of the rising dispersion in wages and productivity is accounted for by changes in the distribution of computer investment across plants. According to Fleck, Glaser, and Spragne (2011), the wage–productivity gap increases, in part, from the measures of labour productivity not having been adjusted for compositional changes in the workforce, and from the choice of different price indices to adjust to inflation. Goldar and Banga (2005) opined that labour market conditions matter a lot in wage-setting: the stronger the trade unions, the higher the wages earned by industrial workers. Greater labour market flexibility tends to push wages down. Rath (2006) and Millea (2002) also considered the importance of workers’ unions on the wage–productivity setting for industrialised countries, while a good industrial climate raises industrial wages. Yildirm (2015) found that the absence of a link running from productivity to real wages in the Turkish manufacturing industry is due to lower bargaining power and structural problems, including high unemployment, a huge tax burden on wages and the large share of the informal sector. Fedderke and Mariotti (2002) conclude that this phenomenon results from changes in employment and the accumulation of skill intensity of production. Budd, Chi, Wang, and Xie (2014) and Besley and Burgess (2002) confirm that union density does not affect the average wage level, but is positively associated with aggregate productivity and output. The divergence between wages and productivity has also been discussed in terms of the impact of heavy and autonomous capitalisation on rising productivity (Nwaokoro, 2006), differences in the wage-setting processes in different countries (Millea, 2002), which is closely connected both with the labour market and the consumer goods market (Nikulin, 2015), market frictions, such as search costs (Zoega & Booth, 2005), economic growth, education and skill-level differences (Semeels, 2005) and other social norms (Frazis & Loewenstein, 2006), lags in adjustment and imperfect competition in the product and labour market (Sharpe et al., 2008), degree of unionisation and firm’s size (Organization for Economic Cooperation and Development [OECD], 2001) and labour market policies.

From the literature review, it is evident that the majority of the studies on the wage-productivity relationship are based on developed countries. However, labour market conditions and production scenarios in developing countries are different from those in developed countries. Accordingly, the inflow of capital and technologies on account of liberalisation combined with national labour-related policies should have different implications for the association between wages and productivity.

3. Methodology

3.1. Model and Data

Following the available literature on the subject, an empirical model of the following specification is estimated:

L (W G R) = α + β_{1} L (L P R) + β_{2} L (K L) + β_{3} L (C N R) + e_{t}

(1)

The variables are in the natural logarithmic form prefixed with ‘L’. Therefore, the estimates of β_s determine the elasticity of the dependent variable with respect to the independent variables. WGR here is the wage rate that is defined as real emoluments per employee. Emoluments refer to payments made to employees plus the imputed value of other benefits in kind. Employees here refer to the total number of people engaged in the manufacturing firm. LPR is labour productivity measured as the ratio of output to labour input. Manufacturing output is calculated as the real gross value added (GVA) by the manufacturing industry and labour input is the total number of people engaged in the manufacturing industry.

Since higher labour productivity is reflected in terms of higher wages, the parameter of labour productivity is expected to have a positive sign. KL is the capital–labour ratio that is defined as the real fixed capital per employee. In Annual Survey of Industries (ASI), fixed capital includes all types of assets, namely, land, building, plant and machinery, transport equipment, etc. as well as rented-in assets without adjusting for depreciation. The capital–labour ratio is a proxy for technology, and an increase in it, on the one hand, increases firm’s productivity, and on the other, decreases the demand for labour that adversely affects real wages.⁴ Improved technology requires more skilled workers, and thus labour productivity tends to rise, as a result of which the wage rate also goes up. On this basis, the relationship between real wages and capital intensity could be positive or negative, that is, it is open-ended. CNR is the contractualisation, measured as percentage of contract workers to total workers. An increase in contractualisation⁵ indicates the weakening of trade union strength and thus their bargaining power; it also implies an increase in productivity on one hand, and a decrease of wage share in GVA on the other. Accordingly, the effect of contractualisation on wage rates could be negative. In fact, both capital-intensity and contractualisation capture the effects of economic reforms and labour-related policies. Finally, e_t is the disturbance term that may be serially correlated.

The a priori values of β_s as stated in Equation (1) are as follows: β₁0β₂0β₃0

This model is based on the marginal productivity theory. In order to confirm the efficiency wage theory, labour productivity and wage rate are interchanged as dependent and explanatory variables, respectively. The rest of the explanatory variables are same.

The study uses annual time-series data covering the period from 2000–2001 to 2015–2016. Indian states differ in terms of social, cultural, economic, geographic and institutional characteristics. These factors simultaneously drive firm output and the other explanatory variables. The impact of state-level disparities on growth has always been an important part of social science research (Bhide, Chadha, & Kalirajan, 2005; Brar et al., 2014; Das & Barua, 1996; Kalirajan, Bhide, & Singh, 2009; Krishna, 2004; Kurian, 2000; OECD, 2017 among others). However, state-level characteristics have hardly been observed in studies on the wage–productivity relationship, though there are ample studies on the issue undertaking firm-specific effects. Through a set of state-level panel data for the manufacturing industries, these unobserved state-specific heterogeneities are controlled.

The main data source for this study is the ASI.⁶ These data are collected and compiled by Central Statistical Organization (CSO). The data included in the study are the total number of people engaged, emoluments, GVA and fixed capital. The monetary values are at current prices, so they are made comparable by deflating them with the wholesale price index of manufactured products.⁷ Data on the wholesale price index of manufactured products are collected from RBIs’ Handbook of Statistics on the Indian Economy. The study uses the statistical package EVIEWS for its empirical analysis.

3.2. Econometric Procedure

The following steps are employed for estimations from the panel of state-wise time-series data on manufacturing:

3.2.1 Unit Root Test

The characteristic of the time series is identified through unit root tests. If the series possess a unit root, then it is a non-stationary time series. A non-stationary series can be made stationary by taking first difference. If Y_t (t = 1, 2, ……) is non-stationary but ΔY_t is stationary, then Y_t is called a ‘difference stationary process’ or ‘integrated of first order’ or I(1). For the present study, the Augmented Dickey Fuller (ADF) (Dickey & Fuller, 1979, 1981) and Phillips Perron (PP) (Phillips & Perron, 1988) statistics are employed to examine the unit roots. These tests involves the fitting of the following regression equations:

The ADF test

Δ Y_{t} = γ Y_{t - 1} + \sum_{i = 1}^{P} δ_{i} Δ Y_{t - i} + μ_{t} ​

(2)

The PP test

Δ Y_{t} = α + γ Y_{t - 1} + e_{t}

(3)

where Y_t stands for each of the variables presented in Equation (1), Δ is the first difference operator and p is the lag order of autoregressive process. The residuals µ_t and e_t are assumed to be normally distributed. The null hypothesis of both the tests is the non-stationarity of the series. If the null hypothesis is not rejected, then the difference operator is applied to the series. In testing for the equations, both the inclusion of the intercept and inclusion of the intercept and trend terms are used.

The PP test has an advantage over the ADF test in that it is robust to general forms of heteroscedasticity in the error term.⁸ Also, the user is not required to specify the lag length for the test regression as in the ADF.⁹ Since the PP test is based on an asymptotic theory, it works well only in large samples; therefore, making use of more than one test and finding the matching results is often suggested.

3.2.2 Cointegration Test

When the series is stationary in first difference (i.e., I[1]), the cointegration test is conducted to examine the long-run relationship between the variables. The underlying logic is that if a set of variables are cointegrated, they will never move too far and will be attracted to their long-run equilibrium relationship (Engle & Granger, 1987). With the help of the Monte Carlo experiment, Xiao and Phillips (2002) and Westerlund (2005) found that the estimated parameters tend to be stable over time whenever the variables are found to be cointegrated.

In the present study, the existence of a cointegrating relationship between the selected non-stationary variables is tested by using the Pedroni residual-based panel cointegration test. Basically, it employs four panel statistics and three group-panel statistics to test the null hypothesis of no conintegration against the alternative hypothesis of cointegration. In the case of panel statistics, the first-order autoregressive term is assumed to be the same across all the cross-sections, while in the case of the group-panel statistics, the parameter is allowed to vary over the cross-sections. If the null hypothesis is rejected in the panel case, then the variables are cointegrated for all the cases. On the other hand, if the null hypothesis is rejected in the group-panel case, then cointegration among the relevant variables exists in at least one case. The statistics are distributed, in the limit, as standard normal variables with a left-hand rejection region, with the exception of the variance ratio statistic (the Pedroni cointegration test method is explained in Appendix).

The presence of cointegration among the series confirms a long-run equilibrium relationship between the variables and is an indication that the variables move collectively over time. So the short-run shocks from the long-term changes will be corrected.

3.2.3 The Vector Error Correction Model (VECM)

If the variables are cointegrated, there exists a long-term equilibrium relationship between them. The existence of cointegration implies that there is a cause-and-effect relationship among the variables at least in one direction (Granger, 1988). However, given that the causality cannot be estimated with certainty and remain an empirical difficulty, Granger causality is estimated in the framework of the error correction mechanism (ECM). The residuals from the cointegration equilibrium regression can be used to estimate the short-term and long-term effects of one time series on the other. In case of no cointegration, ECM is no longer required and we can proceed to the Granger causality tests to establish causal links between the variables. The term error correction as developed by Engle and Granger relates to the fact that last period’s deviation from a long-run equilibrium, the error, influences its short-run dynamics. Thus, the error correction model directly estimates the speed at which a dependent variable returns to equilibrium after a change in the other variables.

The present study employs the VECM, which is just a special case of the VAR for the variables that are stationary in their differences (i.e., I[1]). The specification used for the VECM is as follows:

Δ Y_{t} = α_{0} + \sum_{i =0}^{n} β_{1 i} Δ Y_{t - i} + \sum_{i =0}^{n} δ_{m i} Δ X_{t - i} + γ E C T_{t - 1} + u_{t} ​

(4)

where Y_t is the dependent variable, X_t is the vector of observations of included explanatory variables in Equation (1), Δ is the first difference operator, m is the number of regressors and u_t is the error term. ECT is the error correction term captured from the cointegration regression. It represents the adjustment of the variables towards a long-run equilibrium value. A negative and significant coefficient of ECT (i.e., e_t–1) indicates that any short-term fluctuations between the independent variables and the dependent variables will give rise to a stable long-run equilibrium between the variables. γ is the speed of the adjustment parameter, as it measures the speed of adjustment of the dependent variable to the equilibrium level. The size and statistical significance of the coefficient of the ECT measures the tendency of each variable to return to equilibrium. A significant coefficient implies that past equilibrium errors play a role in determining the current outcomes captured in the long-run impact. The larger the γ, the greater is the response of the variable to the previous period’s deviation from long-run equilibrium.

The VECM model is applied with one cointegration equation. Under the EVIEWS environment, OLS estimates were obtained for a set of equations ordered by each variable. Short-run effects were captured through individual coefficients of the differentiated terms. A Wald test for the joint significance of the coefficients of the lagged variables was conducted to test the existence of a short-run relationship. The Chi square test is used for testing the existence of a short-run relationship between the variables.

3.2.4 Granger Causality Test

Finally, in order to establish the direction of causality between the variables, Granger casuality tests are performed. The basic idea is that if the past values of X are significant predictors of the current values of Y, even when past values of Y have been included in the model, then X exerts a causal influence on Y. The panel Granger casuality test has the following regression framework:

Y_{i t} = α_{i} + \sum_{k =1}^{K} γ^{(k)} Y_{i . t - k} + \sum_{k =1}^{k} β^{(k)} X_{i t - k} {+ε}_{i t} ​

(5)

where α_i captures the individual specific effect across i and the coefficients γ^(k) and β^(k) are implicitly assumed to be constant for all i. The lag order K is assumed to be identical for all individuals and the panel is balanced. Since the variables that enter into the system need to be covariance-stationary, the test is applied on the first-differenced values of the series.

As in Granger (1969), the procedure to determine the existence of casuality is to test for significant effects of past values of X on the present value of Y. The null hypothesis is therefore defined as: H₀: γ_il = ……… = γ_ik = 0 i = 1, ……, N

which corresponds to the absence of causality for all individuals in the panel.

The relationship may have uni-directional causality (γ$ 0, β = 0 or β $ 0, γ = 0), bi-directional causality (γ$ 0, β $ 0) or no causality (γ= 0, β = 0). To decide this, an F-test is carried out to examine the null hypothesis of non-causality.

4. Wages and Productivity in Indian Manufacturing Industries

The real wage is the mechanism through which the benefits of productivity growth are transferred to workers (Schwellnus, Kappeler, & Pionnier, 2017; Sharpe et al., 2008). This implies that movements in real wages are associated with changes in productivity. However, for Indian manufacturing industries, one can observe a divergence between wage rates and labour productivity: wages have been lagging behind labour productivity (Figure 1). This indicates that gains in labour productivity are not translated into higher wages, resulting in an increasing share of profits and decreasing share of wages in total output. The increasing gap between real wages and labour productivity is a cause of concern, as it may worsen the production and investment scenario by adversely affecting aggregate demand on the one hand, and on the other, it may decrease workers’ willingness to contribute to production activity. Higher wage rates stimulate labour productivity via an efficiency wage arrangement (Basu & Felkey, 2009).

Figure 1

Source: ASI, various issues.

After smoothening the effect of fluctuations by computing a three-yearly moving average of growth in real wages and labour productivity (Figure 2), it is clear that the growth trend in productivity is decreasing while there is a slight increase in the growth trend of real wages. A decrease in the gap between the growth trends of real wages and labour productivity is evident. Does this show the possibility of coinciding growth trends in future? This is discussed later in this article.

Figure 2

Source: Author’s calculation based on the ASI (various issues).

Further, the growth trends in the wage rate are compared with the growth trends of the policy variables, particularly related to technological inflows and labour market flexibility. Technological inflow is reflected in terms of capital intensity, while labour market flexibility outcomes are in the form of contractualisation of workers, which is reflected in terms of a decrease in the share of wages in GVA in the manufacturing sector. Table 1 shows the growth trends of selected variables, indicating a wide gap between the growth rate of real wages and that of labour productivity.

Table 1

Growth Trends in Selected Variables (2000–2001 to 2015–2016)

Variables	Growth Trends (Equation: log Y_t = a + bt)
Variables	2000–2001 to 2007–2008 (Period 1)	2007–2008 to 2015–2016 (Period 2)	Overall
Real wage rate	0.006** (R² = 0.69)	0.008** (R² = 0.85)	0.009** (R² = 0.94)
Labour productivity	0.038** (R² = 0.98)	0.020** (R² = 0.93)	0.028** (R² = 0.96)
Capital intensity	0.010** (R² = 0.72)	0.036** (R² = 0.93)	0.031** (R² = 0.92)
Wage share in GVA	–0.032** (R² = 0.97)	–0.014** (R² = 0.89)	–0.019** (R² = 0.92)

Source: Author’s calculations based on the ASI.

Note: ** indicates significance at the 1% level.

During the period 2000–2001 to 2015–2016, labour productivity had an average growth rate of 2.8 per cent while the real wage rate grew at an average rate of 0.9 per cent. The growth in labour productivity is associated with higher capital intensity¹⁰ as observed from the similar average growth rates (3%) of both variables. After economic liberalisation, the ease of availability of capital and technologies from abroad has resulted in a substitution of capital for labour in Indian industry. The flexibility of the labour market has resulted in contractualisation of workers and thus the share of wages in the GVA has declined at an average rate of 2 per cent. For further understanding, the selected time period is divided into two periods: Period 1 (2000–2001 to 2007–2008) and Period 2 (2007–2008 to 2015–2016). A comparison of the growth trends in the two periods indicates that the average growth in wage rates has slightly increased by 0.2 per cent in the later period, while the growth rate of labour productivity has decreased by nearly 2 per cent.

There is an increase in the average growth rate of capital intensity from 1 per cent in Period 1 to 3.6 per cent in Period 2. The growth rate in the share of wages in the GVA is negative in both periods, with a slight improvement observed in the later period. Despite an increase in the average growth rate of capital intensity, labour productivity growth has decreased, while growth in wage rates has slightly improved. This is due to the fact that (a) capital-intensive technology requires skilled workers who demand a higher compensation, and (b) the contractualisation of workers¹¹ makes them footloose, frequently moving from one job to another; this decreases their efficiency and hence the overall output of the manufacturing firm. An increase in wages and job security binds the worker to the production activity and strengthens the employer-employee relationship (Jain, 2014).

5. Wages and Labour Productivity in Indian States

The gap between the growth in wage rates and labour productivity in manufacturing varies among the Indian states (Table 2), but the growth rate in real wages lags behind the growth in labour productivity in all the states. The annual growth rate of labour productivity is highest in Himachal Pradesh followed by Maharashtra, Jharkhand, Jammu and Kashmir, Orissa and Uttarakhand. It is low in Assam, Kerala, Punjab, West Bengal, Bihar, Tamil Nadu and Haryana. The annual growth rate in real wages is high in Himachal Pradesh, Maharashtra, Andhra Pradesh, Haryana, Karnataka, Madhya Pradesh and Orissa. It is low in Bihar, Uttarakhand, Assam, Jammu and Kashmir, Kerala, Punjab and West Bengal.

It is observed that states which have a high growth rate of labour productivity also have a relatively large spread between the growth rates of labour productivity and real wages, and those states which have a low growth rate of labour productivity have a relatively smaller spread between the growth rates of labour productivity and real wages. The correlation coefficient between the two variables is found to be positive and significant at the 5 per cent level. However, the regression coefficient is 0.069 which is significantly lower than one, implying that an increase in labour productivity leads to a much less than proportionate increase in real wages.

Table 2
Statewise Labour Productivity and Wages (Organised Manufacturing Industries)¹²

State Growth in Labour Productivity
(% per annum) Growth in Real Wages
(% per annum) Difference

Andhra Pradesh 2.0 0.5 1.5

Assam 0.8 0.3 0.5

Bihar 1.3 0.1 1.2

Chhattisgarh 2.8 0.4 2.4

Gujarat 2.4 0.4 1.9

Haryana 1.5 0.5 1.0

Himachal Pradesh 4.1 0.6 3.5

Jammu & Kashmir 3.1 0.3 2.8

Jharkhand 3.2 0.6 2.7

Karnataka 2.4 0.5 1.8

Kerala 0.8 0.3 0.5

Madhya Pradesh 1.8 0.5 1.4

Maharashtra 3.2 0.6 2.6

Orissa 3.1 0.5 2.6

Punjab 1.1 0.3 0.8

Rajasthan 2.0 0.4 1.6

Tamil Nadu 1.4 0.4 0.9

Uttarakhand 3.1 0.1 3.1

Uttar Pradesh 2.1 0.4 1.7

West Bengal 1.1 0.3 0.8

Total 2.1 0.4 1.6

Correlation coefficient = 0.449*
Regression coefficient = 0.069*

State	Growth in Labour Productivity (% per annum)	Growth in Real Wages (% per annum)	Difference
Andhra Pradesh	2.0	0.5	1.5
Assam	0.8	0.3	0.5
Bihar	1.3	0.1	1.2
Chhattisgarh	2.8	0.4	2.4
Gujarat	2.4	0.4	1.9
Haryana	1.5	0.5	1.0
Himachal Pradesh	4.1	0.6	3.5
Jammu & Kashmir	3.1	0.3	2.8
Jharkhand	3.2	0.6	2.7
Karnataka	2.4	0.5	1.8
Kerala	0.8	0.3	0.5
Madhya Pradesh	1.8	0.5	1.4
Maharashtra	3.2	0.6	2.6
Orissa	3.1	0.5	2.6
Punjab	1.1	0.3	0.8
Rajasthan	2.0	0.4	1.6
Tamil Nadu	1.4	0.4	0.9
Uttarakhand	3.1	0.1	3.1
Uttar Pradesh	2.1	0.4	1.7
West Bengal	1.1	0.3	0.8
Total	2.1	0.4	1.6
Correlation coefficient = 0.449* Regression coefficient = 0.069*

Source: Author’s calculation based on ASI.

Note: * indicates significance at 5% level.

6. Empirical Findings

From the above analysis, it is can be concluded that the growth rates in real wages and labour productivity have a very weak bonding: a very small portion of labour productivity translates into wages. Further, the divergence between the two variables is increasing. We conduct empirical analysis in order to find the possibility of a long-term equilibrium between the two variables. The results are discussed as under:

6.1 Unit Root Test Results

Since the validity of the econometric estimates depends upon the stationarity of the series, unit root tests are performed on log values of the variables. The results of the ADF and PP unit root tests are reported in Table 3. Both the tests show that the null hypothesis of all the series is not rejected in their level forms. This show the existence of unit root in all the series, indicating that all the series are non-stationary in their level form. The tests are again applied on the first differenced values of all the series. Now the null hypothesis of non-stationarity is rejected for all the series at 1 per cent level of significance. This suggests that all the variables are integrated of order one, that is, I(1). This confirms that the selected variables are stationary in their first difference. This allows us to proceed for cointegration test to ascertain long-run relationship between the variables.

Table 3
Summary of the Panel Unit Root Test

(a) ADF Test

Variables Level
First Difference
Inference

Individual Intercept Individual Intercept and Trend Individual Intercept Individual Intercept and Trend

WGR 7.57 21.50 149.64 138.66 I(1)

LPR 39.34 57.08 183.75 152.24 I(1)

KL 20.56 48.14 175.61 163.49 I(1)

CNR 32.97 55.41 163.81 125.52 I(1)

(b) PP-Fisher Chi Square

Variables
Level
First Difference
Inference

Intercept
Intercept and Trend
Individual Intercept
Individual Intercept and Trend

WGR 7.33 25.98 169.66 229.38 I(1)

LPR 44.17 47.73 192.87 198.48 I(1)

KL 20.99 59.56 232.60 235.03 I(1)

CNR 46.35 63.94 198.49 226.69 I(1)

(a) ADF Test
WGR	7.57	21.50	149.64**	138.66**	I(1)
LPR	39.34	57.08	183.75**	152.24**	I(1)
KL	20.56	48.14	175.61**	163.49**	I(1)
CNR	32.97	55.41	163.81**	125.52**	I(1)
(b) PP-Fisher Chi Square
Variables	Level	First Difference	Inference
Intercept	Intercept and Trend	Individual Intercept	Individual Intercept and Trend
WGR	7.33	25.98	169.66**	229.38**	I(1)
LPR	44.17	47.73	192.87**	198.48**	I(1)
KL	20.99	59.56	232.60**	235.03**	I(1)
CNR	46.35	63.94	198.49**	226.69**	I(1)

Source: The author.

Note: ** implies significance at the 1 per cent level.

6.2 Cointegration Test Results

The results of the Pedroni residual cointegration test (Table 4) reveal that of the 11 outcomes, 6 statistics reject the hypothesis of no cointegration at the 1 per cent significance level. Since the majority of the tests show strong evidence of cointegration between the selected variables, it is empirically proven that the selected time series share a long-run equilibrium relationship, meaning that they could well move together in the long run.

Table 4

Pedroni Residual Cointegration Test

	Constant		Constant and Trend
Alternative Hypothesis: Common AR Coefs (Within-dimension)
	Statistic	Weighted Statistic	Statistic	Weighted Statistic
Panel v-statistic	–0.610	–1.131	0.081	–2.110
Panel rho-statistic	2.216	1.809	3.475	3.531
Panel PP-statistic	–1.777*	–4.402**	–5.971**	–9.238**
Panel ADF-statistic	–3.416**	–4.745**	–5.050**	–5.798**
Alternative Hypothesis: Individual AR Coefs (Between-dimension)
Group rho-statistic	3.600		5.084
Group PP-statistic	–6.791**		–9.534**
Group ADF-statistic	–5.550**		–6.117**

Source: The author.

Note: * and ** represent significance at the 5% and 1% levels, respectively.

6.3. Long-run and Short-run Causality (The VECM Approach)

The cointegration analysis provides support for analysing the long- and short-run causality between the dependent and independent variables using the VECM. Following the literature, the relationship between the real wage rate and labour productivity is examined both ways, namely, the causality running from labour productivity to the wage rate and vice versa. Accordingly, two models are tested. The first model explains wage rate with respect to labour productivity, capital intensity and contractualisation, and the second model explains labour productivity with respect to the wage rate, capital intensity and contractualisation.

The long-run causality between the time series variables is tested from the sign and magnitude of ECT_t–1 (Table 5). A significantly negative sign for ECT implies that all the disequilibrium in the short periods has been corrected to achieve long-run equilibrium. From the results, it is evident that both models have a tendency towards a long-run equilibrium, but the speed of adjustment to the long-term equilibrium is high in the second model as compared to the first model. In the first model, the system corrects its disequilibrium at the rate of 3.7 per cent per annum, indicating that it will take 27 years for the system to reach the equilibrium level. In this model, long-run causality runs from labour productivity and capital intensity to the wage rate. Labour productivity has a strongly converging effect, while capital intensity has a diverging effect on the wage rate. Since both variables act in opposite directions, the overall effect becomes relatively slow, thus, restoring the equilibrium by adjusting labour productivity and capital intensity takes a longer time. In contrast, in the second model, the system corrects its short-run disequilibrium at the speed of 20 per cent per annum which means that it takes five years for the system to restore equilibrium. In this model, the long-run causality runs from wage rate and capital intensity to labour productivity, and both variables have a converging effect. Thus, correcting the equilibrium by adjusting wage rates and capital intensity is comparatively fast.

Table 5
VECM Estimates

Dependent Variables

Model 1
ΔWGR (Marginal Productivity Theory) Model 2
ΔLPR (Efficiency Wage Theory)

Long-run Causality

Cointegrating equation: ECT_{t – 1} = 1.00 WGR_{t – 1} – 1.68 LPR_{t – 1}** + 0.64 KL_{t – 1}* – 0.06 CNR_{t – 1} + 0.62 Cointegrating equation: ECT_t–1 = 1.00 LPR_{t – 1} – 0.60 WGR_{t – 1} – 0.38 KL_{t – 1} + 0.04 CNR_{t – 1} – 0.37

ECT_{t – 1} = –0.037* ECT_{t – 1} = –0.221

Short-run Causality

ΔWGR_{t – 1} –0.28 LPR_t–1 –0.04

ΔWGR_{t – 2} 0.06 LPR_t–2 –0.06

ΔLPR_{t – 1} –0.07 WGR_t–1 –0.07

ΔLPR_{t – 2} –0.12 WGR_t–2 0.19

ΔKL_{t – 1} 0.07 KL_t–1 –0.02

ΔKL_{t – 2} 0.16 KL_t–2 0.03

ΔCNR_{t – 1} –0.12 CNR_t–1 –0.01

ΔCNR_{t – 2} –0.11 CNR_t–2 –0.05

C 0.03 C 0.036

Joint test of significance (χ2 test)

LPR, H₀: C(4) = C(5) = 0 " H₀ rejected
KL, H₀: C(6) = C(7) = 0 " H₀ rejected
CNR, H₀: C(8) = C(9) = 0 " H₀ accepted WGR, H₀: C(4) = C(5) = 0 " H₀ accepted
KL, H₀: C(6) = C(7) = 0 " H₀ accepted
CNR, H₀: C(8) = C(9) = 0 " H₀ accepted

Dependent Variables
Long-run Causality
Cointegrating equation: ECT_{t – 1} = 1.00 WGR_{t – 1} – 1.68 LPR_{t – 1}** + 0.64 KL_{t – 1}* – 0.06 CNR_{t – 1} + 0.62	Cointegrating equation: ECT_t–1 = 1.00 LPR_{t – 1} – 0.60 WGR_{t – 1} – 0.38 KL_{t – 1} + 0.04 CNR_{t – 1} – 0.37
ECT_{t – 1} = –0.037*	ECT_{t – 1} = –0.221**
Short-run Causality
ΔWGR_{t – 1}	–0.28**	LPR_t–1	–0.04
ΔWGR_{t – 2}	0.06	LPR_t–2	–0.06
ΔLPR_{t – 1}	–0.07	WGR_t–1	–0.07
ΔLPR_{t – 2}	–0.12**	WGR_t–2	0.19
ΔKL_{t – 1}	0.07	KL_t–1	–0.02
ΔKL_{t – 2}	0.16**	KL_t–2	0.03
ΔCNR_{t – 1}	–0.12	CNR_t–1	–0.01
ΔCNR_{t – 2}	–0.11	CNR_t–2	–0.05
C	0.03	C	0.036
Joint test of significance (χ2 test)
LPR, H₀: C(4) = C(5) = 0 " H₀ rejected KL, H₀: C(6) = C(7) = 0 " H₀ rejected CNR, H₀: C(8) = C(9) = 0 " H₀ accepted	WGR, H₀: C(4) = C(5) = 0 " H₀ accepted KL, H₀: C(6) = C(7) = 0 " H₀ accepted CNR, H₀: C(8) = C(9) = 0 " H₀ accepted

Source: The author.

Notes: Δ denotes first difference; ** and * indicate 1% and 5% levels of significance, respectively.

Further, the short-run effects are gauged from individual estimates of the cointegrating equation. The short-run effects indicate the rate at which the previous period disequilibrium in the system is corrected. The results of the joint test of significance of the lags are given in Table 5. In the first model, it is observed that the short-run causality runs from labour productivity and capital intensity to the wage rate that is significant at the 1 per cent level. Capital intensity, however, has a diverging effect. In the second model, there is no evidence of a short-run causality. The overall analysis of the VECM model shows the significance of the efficiency wage theory in correcting the wage–productivity relationship. This suggests matching wages with skill requirements in order to enhance labour productivity. Contractualisation of workers does not have any significance in helping attain the wage–productivity equilibrium, either in the short run or in the long run.

6.4. Results of the Panel Granger Causality Test

The direction of causality between the variables is checked and the predictive ability of the variables is determined. The results of panel Granger causality test are reported in Table 6. Since the test is applied on the first difference of the logged variables, the estimates indicate the causality between the growth rates of the variables. The empirical results state that the causality between the growth in wage rates and labour productivity emerges with a time lag. The growth in labour productivity Granger causes growth in the wage rate at lag 8, while the growth in wage rates Granger causes growth in labour productivity at lags 5 and 6. The variables have bi-directional causality at lags 9 and 10. The Granger causality between the growth in wage rates and growth in capital intensity, and also between growth in wage rates and increase in contractualisation, exists up till lag 5. Upto this period, the results show both uni-directional and bi-directional causality between the variables. The bi-directional causality between the wage rate and capital intensity indicates that an increase in capital intensity creates a demand for skilled workers, resulting in an increase in the wage rates of skilled workers. A higher wage rate also motivates workers to acquire more skills, and the resulting increase in productivity encourages firms to employ more capital-intensive technologies. The bi-directional causality between the growth in wage rates and growth in the hiring of contractual workers corroborates the micro-economic theory of distribution that at higher wage rates, employment is lower (of regular workers). To remain competitive, firms try to increase their profits by cutting down on regular employment. An increase in capital-intensity Granger causes a rise in labour productivity only at lags 9 and 10, indicating the time required to acquire skills in accordance with the available technologies.

Table 6
Results of Panel Granger Causality Test

Null Hypothesis Lags F-Statistic Decision

Δ(LPR) does not Granger cause Δ(WGR) 2 2.70 Do not reject

Δ(WGR) does not Granger cause Δ(LPR) 0.05 Do not reject

Δ(KL) does not Granger cause Δ(WGR) 2 6.52** Reject

Δ(WGR) does not Granger cause Δ(KL) 3.09* Reject

Δ(CNR) does not Granger cause Δ(WGR) 2 3.15* Reject

Δ(WGR) does not Granger cause Δ(CNR) 3.88* Reject

Δ(KL) does not Granger cause Δ(LPR) 2 0.29 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 0.71 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 3 2.62* Reject

Δ(WGR) does not Granger cause Δ(LPR) 1.05 Do not reject

Δ(KL) does not Granger cause Δ(WGR) 3 4.30** Reject

Δ(WGR) does not Granger cause Δ(KL) 2.63* Reject

Δ(CNR) does not Granger cause Δ(WGR) 3 1.88 Do not reject

Δ(WGR) does not Granger cause Δ(CNR) 4.67** Reject

Δ(KL) does not Granger cause Δ(LPR) 3 0.29 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 0.43 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 4 0.83 Do not reject

Δ(WGR) does not Granger cause Δ(LPR) 1.35 Do not reject

Δ(KL) does not Granger cause Δ(WGR) 4 3.02* Reject

Δ(WGR) does not Granger cause Δ(KL) 1.67 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 4 2.97* Reject

Δ(WGR) does not Granger cause Δ(CNR) 2.49* Reject

Δ(KL) does not Granger cause Δ(LPR) 4 2.13* Reject

Δ(LPR) does not Granger cause Δ(KL) 0.85 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 5 0.44 Do not reject

Δ(WGR) does not Granger cause Δ(LPR) 5 2.27* Reject

Δ(KL) does not Granger cause Δ(WGR) 5 2.18 Do not reject

Δ(WGR) does not Granger cause Δ(KL) 5 1.59 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 5 2.52* Reject

Δ(WGR) does not Granger cause Δ(CNR) 5 2.27* Reject

Δ(KL) does not Granger cause Δ(LPR) 5 1.12 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 5 0.82 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 6 1.40 Do not reject

Δ(WGR) does not Granger cause Δ(LPR) 6 2.32* Reject

Δ(KL) does not Granger cause Δ(WGR) 6 1.99 Reject

Δ(WGR) does not Granger cause Δ(KL) 6 1.44 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 6 1.49 Do not reject

Δ(WGR) does not Granger cause Δ(CNR) 6 1.57 Do not reject

Δ(KL) does not Granger cause Δ(LPR) 6 0.60 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 6 1.01 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 7 1.59 Do not reject

Δ(WGR) does not Granger cause Δ(LPR) 7 1.40 Do not reject

Δ(KL) does not Granger cause Δ(WGR) 7 1.08 Do not reject

Δ(WGR) does not Granger cause Δ(KL) 7 0.86 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 7 0.88 Do not reject

Δ(WGR) does not Granger cause Δ(CNR) 7 1.31 Do not reject

Δ(KL) does not Granger cause Δ(LPR) 7 0.42 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 7 0.91 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 8 2.48* Reject

Δ(WGR) does not Granger cause Δ(LPR) 8 1.93 Do not reject

Δ(KL) does not Granger cause Δ(WGR) 8 1.67 Do not reject

Δ(WGR) does not Granger cause Δ(KL) 8 0.63 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 8 0.49 Do not reject

Δ(WGR) does not Granger cause Δ(CNR) 8 1.09 Do not reject

Δ(KL) does not Granger cause Δ(LPR) 8 1.17 Do not reject

Δ(LPR) does not Granger cause Δ(KL) 8 1.29 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 9 4.60** Reject

Δ(WGR) does not Granger cause Δ(LPR) 9 2.24* Reject

Δ(KL) does not Granger cause Δ(WGR) 9 0.85 Do not reject

Δ(WGR) does not Granger cause Δ(KL) 9 0.58 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 9 1.67 Do not reject

Δ(WGR) does not Granger cause Δ(CNR) 9 1.06 Do not reject

Δ(KL) does not Granger cause Δ(LPR) 9 2.56* Reject

Δ(LPR) does not Granger cause Δ(KL) 9 0.51 Do not reject

Δ(LPR) does not Granger cause Δ(WGR) 10 8.04** Reject

Δ(WGR) does not Granger cause Δ(LPR) 10 2.10* Reject

Δ(KL) does not Granger cause Δ(WGR) 10 2.29* Reject

Δ(WGR) does not Granger cause Δ(KL) 10 3.37 Do not reject

Δ(CNR) does not Granger cause Δ(WGR) 10 NA NA

Δ(WGR) does not Granger cause Δ(CNR) 10 NA NA

Δ(KL) does not Granger cause Δ(LPR) 10 2.33* Reject

Δ(LPR) does not Granger cause Δ(KL) 10 0.34 Do not reject

Null Hypothesis	Lags	F-Statistic	Decision
Δ(LPR) does not Granger cause Δ(WGR)	2	2.70	Do not reject
Δ(WGR) does not Granger cause Δ(LPR)		0.05	Do not reject
Δ(KL) does not Granger cause Δ(WGR)	2	6.52**	Reject
Δ(WGR) does not Granger cause Δ(KL)		3.09*	Reject
Δ(CNR) does not Granger cause Δ(WGR)	2	3.15*	Reject
Δ(WGR) does not Granger cause Δ(CNR)		3.88*	Reject
Δ(KL) does not Granger cause Δ(LPR)	2	0.29	Do not reject
Δ(LPR) does not Granger cause Δ(KL)		0.71	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	3	2.62*	Reject
Δ(WGR) does not Granger cause Δ(LPR)		1.05	Do not reject
Δ(KL) does not Granger cause Δ(WGR)	3	4.30**	Reject
Δ(WGR) does not Granger cause Δ(KL)		2.63*	Reject
Δ(CNR) does not Granger cause Δ(WGR)	3	1.88	Do not reject
Δ(WGR) does not Granger cause Δ(CNR)		4.67**	Reject
Δ(KL) does not Granger cause Δ(LPR)	3	0.29	Do not reject
Δ(LPR) does not Granger cause Δ(KL)		0.43	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	4	0.83	Do not reject
Δ(WGR) does not Granger cause Δ(LPR)		1.35	Do not reject
Δ(KL) does not Granger cause Δ(WGR)	4	3.02*	Reject
Δ(WGR) does not Granger cause Δ(KL)		1.67	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	4	2.97*	Reject
Δ(WGR) does not Granger cause Δ(CNR)		2.49*	Reject
Δ(KL) does not Granger cause Δ(LPR)	4	2.13*	Reject
Δ(LPR) does not Granger cause Δ(KL)		0.85	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	5	0.44	Do not reject
Δ(WGR) does not Granger cause Δ(LPR)	5	2.27*	Reject
Δ(KL) does not Granger cause Δ(WGR)	5	2.18	Do not reject
Δ(WGR) does not Granger cause Δ(KL)	5	1.59	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	5	2.52*	Reject
Δ(WGR) does not Granger cause Δ(CNR)	5	2.27*	Reject
Δ(KL) does not Granger cause Δ(LPR)	5	1.12	Do not reject
Δ(LPR) does not Granger cause Δ(KL)	5	0.82	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	6	1.40	Do not reject
Δ(WGR) does not Granger cause Δ(LPR)	6	2.32*	Reject
Δ(KL) does not Granger cause Δ(WGR)	6	1.99	Reject
Δ(WGR) does not Granger cause Δ(KL)	6	1.44	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	6	1.49	Do not reject
Δ(WGR) does not Granger cause Δ(CNR)	6	1.57	Do not reject
Δ(KL) does not Granger cause Δ(LPR)	6	0.60	Do not reject
Δ(LPR) does not Granger cause Δ(KL)	6	1.01	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	7	1.59	Do not reject
Δ(WGR) does not Granger cause Δ(LPR)	7	1.40	Do not reject
Δ(KL) does not Granger cause Δ(WGR)	7	1.08	Do not reject
Δ(WGR) does not Granger cause Δ(KL)	7	0.86	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	7	0.88	Do not reject
Δ(WGR) does not Granger cause Δ(CNR)	7	1.31	Do not reject
Δ(KL) does not Granger cause Δ(LPR)	7	0.42	Do not reject
Δ(LPR) does not Granger cause Δ(KL)	7	0.91	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	8	2.48*	Reject
Δ(WGR) does not Granger cause Δ(LPR)	8	1.93	Do not reject
Δ(KL) does not Granger cause Δ(WGR)	8	1.67	Do not reject
Δ(WGR) does not Granger cause Δ(KL)	8	0.63	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	8	0.49	Do not reject
Δ(WGR) does not Granger cause Δ(CNR)	8	1.09	Do not reject
Δ(KL) does not Granger cause Δ(LPR)	8	1.17	Do not reject
Δ(LPR) does not Granger cause Δ(KL)	8	1.29	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	9	4.60**	Reject
Δ(WGR) does not Granger cause Δ(LPR)	9	2.24*	Reject
Δ(KL) does not Granger cause Δ(WGR)	9	0.85	Do not reject
Δ(WGR) does not Granger cause Δ(KL)	9	0.58	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	9	1.67	Do not reject
Δ(WGR) does not Granger cause Δ(CNR)	9	1.06	Do not reject
Δ(KL) does not Granger cause Δ(LPR)	9	2.56*	Reject
Δ(LPR) does not Granger cause Δ(KL)	9	0.51	Do not reject
Δ(LPR) does not Granger cause Δ(WGR)	10	8.04**	Reject
Δ(WGR) does not Granger cause Δ(LPR)	10	2.10*	Reject
Δ(KL) does not Granger cause Δ(WGR)	10	2.29*	Reject
Δ(WGR) does not Granger cause Δ(KL)	10	3.37	Do not reject
Δ(CNR) does not Granger cause Δ(WGR)	10	NA	NA
Δ(WGR) does not Granger cause Δ(CNR)	10	NA	NA
Δ(KL) does not Granger cause Δ(LPR)	10	2.33*	Reject
Δ(LPR) does not Granger cause Δ(KL)	10	0.34	Do not reject

Source: Author’s calculation.

Note: * and ** indicate significance at the 5% and 1% levels, respectively.

7. Concluding Remarks

As in other economies, in India there is divergence between real wages and labour productivity in manufacturing industries. In the country as a whole, as well as in all the states, wage growth lags behind productivity growth. It is observed that productivity growth is associated with a change in capital intensity. The low share of wages in GVA combined with an increase in capital intensity implies that instead of using the increased productivity to benefit workers, firms have been increasing their profit share and using it to increase their technological orientation. This decreases the status of workers, which is indicated by the increasing share of contract workers among total workers.

The empirical results of our study find the existence of a long-run cointegration relationship between wage growth and productivity growth. The speed of adjustment towards equilibrium in the long run is found to be high through efficiency wages than labour productivity. Wage growth and labour productivity growth have a bi-directional causality in the long run, while capital intensity and contractualisation have a short-run causality. The results of the study corroborates the efficiency wage theory as both wage rate and capital intensity have a converging effect; in the marginal productivity theory, due to the diverging effect of capital intensity, the overall impact becomes slow. As a result, correcting the equilibrium by following the efficiency wage theory would be faster than through the marginal productivity theory.

Since capital-intensive technologies require skilled workers, it is essential to match wages with skill requirements to maintain the demand for goods and services in the market, and also to encourage workers to raise their skill sets, which would be reflected in higher labour productivity and higher manufacturing output. There is no evidence that the contractualisation of workers helps attain a long-term wage–productivity equilibrium. Since the labour market has become highly flexible, it has become difficult to retain the regular status of workers and for them to unionise. In such a situation, workers themselves are expected to increase their bargaining strength through skills upgradation.

The study indicates the need for a two-pronged approach to correct the wage–productivity relationship: from the workers’ side, skill development is essential and from the firms’, efficient wages are required (i.e., wages that match skills), otherwise contractualisation would distort the situation resulting in economic loss as well as social unrest.

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

Pedroni Residual-based Panel Cointegration

Pedroni (1999) derives seven panel cointegration test statistics. Of these seven statistics, four are based on within-dimension and three are based on between-dimension. For the within-dimension statistics, the null hypothesis of no cointegration for the panel cointegration test is: H₀ : γ_i = 1 for all i = 1,…, N H₀ : γ_i = γ < 1 for all i = 1,…, N

For the between-dimension statistics, the null hypothesis of no cointegration for the panel cointegration test is: H₀ : γ_i = 1 for all i = 1,…, N H₀ : γ_i < 1 for all i = 1,…, N

The starting point of the residual-based panel cointegration test statistics of Pedroni (1999) is the computation of residuals of the hypothesized cointegrating regression. In the most general case, this may take the form: y_i,t = α_i + δ_it + β_1iχ_1i,t + β_2iχ_2i,t + . . . + β_Miχ_Mi,t + e_i,t t = 1,.. T; i = 1,…N y_i,t = α_i + δ_tt + β_1ix_1i,t + β_2ix_2i,t + . . . + β_Mix_Mi,t + e_i,t t = 1,.. T; i = 1,…N (1)

where T refers to the number of observation over time, N refers to the number of the individual members in the panel, and M refers to the number of independent variables. Here x and y are assumed to be integrated of order one. The slope coefficients β_1i, β_2i,…, β_Mi and specific intercept α_i vary across individual member of the panel.

To estimate the residuals from Equation (1), the seven Pedroni’s statistics are:

Panel υ-statistics:

T^{2} N^{\frac{3}{2}} Z_{\hat{ν} N, T} \equiv T^{2} N^{\frac{3}{2}} {(\sum_{i =1}^{N} \sum_{t =1}^{T} {\hat{L}}_{11 i}^{- 2} {\hat{e}}_{i . t - 1}^{2})}^{- 1} ​

Panel ρ –statistics:

\begin{matrix} T \sqrt{N} Z_{{\hat{ρ}}_{N . T^{- 1}}} \equiv T \sqrt{N} {({\sum^{​}}_{N}^{i =1} {\sum^{​}}_{T}^{t =1} {\hat{L}}_{11 i}^{- 2} {\hat{e}}_{i, t - 1}^{2})}^{- 1} \\ {\sum^{​}}_{N}^{i =1} {\sum^{​}}_{T}^{t =1} {\hat{L}}_{11 i}^{- 2} ({\hat{e}}_{i, t - 1} Δ {\hat{e}}_{i, t} - {\hat{λ}}_{i}) \end{matrix}

Panel t-statistics:

Z_{t_{N, T}} \equiv {({\hat{σ}}_{N, T}^{2} \sum_{i = 1}^{N} \sum_{t = 1}^{T} {\hat{L}}_{11 i}^{- 2} {\hat{e}}_{i, t - 1}^{2})}^{- \frac{1}{2}} \sum_{i = 1}^{N} \sum_{t = 1}^{T} {\hat{L}}_{11 i}^{- 2} ({\hat{e}}_{i, t - 1} Δ {\hat{e}}_{i, t} - {\hat{λ}}_{i})

(Non parametric)

Panel t-Statistics:

Z^{*} ​_{t_{N, T}} \equiv {({\tilde{S}}_{N, T}^{*2} \sum_{i =1}^{N} \sum_{t =1}^{T} {\hat{L}}_{11 i}^{- 2} {\hat{e}}_{i, t - 1}^{*2})}^{- \frac{1}{2}} \sum_{i =1}^{N} \sum_{t =1}^{T} {\hat{L}}_{11 i}^{- 2} {\hat{e}}_{i, t - 1}^{*} Δ {\hat{e}}_{i, t}^{*}

(Parametric)

Group ρ-statistics:

\begin{matrix} T N^{- \frac{1}{2}} {\tilde{Z}}_{{\hat{ρ}}_{N, T^{- 1}}} \equiv T N^{- \frac{1}{2}} \sum_{i = 1}^{N} {(\sum_{t = 1}^{T} {\hat{e}}_{i, t - 1}^{2})}^{- 1} \\ \sum_{t = 1}^{T} ({\hat{e}}_{i, t - 1} Δ {\hat{e}}_{i, t} - {\hat{λ}}_{i}) \end{matrix}

Group t-statistics:

N^{- \frac{1}{2}} {\tilde{Z}}_{t_{N, T}} \equiv N^{- \frac{1}{2}} \sum_{i =1}^{N} {({\hat{σ}}_{i}^{2} \sum_{t =1}^{T} {\hat{e}}_{i, t - 1}^{2})}^{- \frac{1}{2}} \sum_{t =1}^{T} ({\hat{e}}_{i, t - 1} Δ {\hat{e}}_{i, t} - {\hat{λ}}_{i})

(Non parametric)

Group t-statistics:

N^{- \frac{1}{2}} {\tilde{Z}}_{t_{N, T}}^{*} \equiv N^{- \frac{1}{2}} \sum_{i =1}^{N} {({\hat{S}}_{i}^{*2} \sum_{t =1}^{T} {\hat{e}}_{i, t - 1}^{*2})}^{- 1} \sum_{t =1}^{T} {\hat{e}}_{i, t - 1}^{*} Δ {\hat{e}}_{i, t}^{*}

(Parametric)

\begin{matrix} {\hat{λ}}_{i} = \frac{1}{T} \sum_{S = 1}^{k i} (1 - \frac{S}{k_{i} + 1}) \sum_{t = s + 1}^{T} {\hat{μ}}_{i, t} {\hat{μ}}_{i, t - s}, {\hat{S}}_{i}^{2} \equiv \frac{1}{T} \sum_{t = 1}^{T} {\hat{μ}}_{i, t}^{2}, {\tilde{σ}}_{i}^{2} \\ = {\hat{S}}_{i}^{2} + 2 {\hat{λ}}_{i}, {\hat{σ}}_{N, T}^{2} \equiv \frac{1}{N} \sum_{i = 1}^{N} {\hat{L}}_{11 i}^{- 2} {\hat{σ}}_{i}^{2} \end{matrix}

\begin{matrix} {\hat{S}}_{i}^{*2} \equiv \frac{1}{t} \sum_{t =1}^{T} {\hat{μ}}_{i, t}^{*2}, {\tilde{S}}_{N, T}^{*2} \equiv \frac{1}{N} \sum_{t =1}^{N} {\hat{S}}_{i}^{*2}, {\hat{L}}_{11 i}^{- 2} = \frac{1}{T} \sum_{T}^{1} {\hat{η}}_{i, t}^{2} \\ + \frac{2}{T} \sum_{s =1}^{k i} (1 - \frac{S}{k_{i} +1}) \sum_{t = s +1}^{T} {\hat{η}}_{i, t} {\hat{η}}_{i, t - s} \end{matrix}

and where the residuals ${\hat{u}}_{i, t}, {\hat{μ}}_{i, t}^{*}, {\hat{η}}_{i, t} $ are obtained from the following regressions:

\begin{matrix} {\hat{e}}_{i, t} = {\hat{γ}}_{i} {\hat{e}}_{i, t - 1} + {\hat{u}}_{i, t}, {\hat{e}}_{i, t} = {\hat{γ}}_{i} {\hat{e}}_{i, t - 1} + \sum_{k =1}^{K_{i}} {\hat{ρ}}_{i, k} Δ {\hat{e}}_{i, t - k} + {\hat{μ}}_{i, t}^{*}, \\ Δ y_{i, t} = \sum_{m =1}^{M} {\hat{b}}_{m i} Δ x_{m i, t} + {\hat{η}}_{i, t} \end{matrix}

Note: All statistics are from Pedroni (1997).

After the calculation of the panel cointegration test statistics, the appropriate mean and variance adjustment terms are applied, so that the test statistics are asymptotically normally distributed.

\frac{χ_{N, T} - μ \sqrt{N}}{\sqrt{υ}} \underset{\to}{d} N (0,1)

where χ _N,T is the standardized form of the test statistics with respect to N and T. Here µ and v are Monte Carlo-generated adjustment terms.

Notes

References

Andrews

D. W. K.

(1991). Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica, 59(3), 817–858.

Backman

, & Gainsurgh

M. R.

(1949). Productivity: Productivity and living standards. ILR Review, 2(2), 163–194.

Barkery

(2007). Behind the gap between productivity and wages, growth (Issue Brief 2007-05, pp. 494–516). Washington, DC: Center for Economic and Policy Research.

Basu

, & Felkey

A. J.

(2009). A theory of efficiency wage with multiple unemployment equilibria: How a higher minimum wage law can curb unemployment. Oxford Economic Papers, 16(3), 494–516.

Besley

, & Burgess

(2002). Can labour regulation hinder economic performance? Evidence from India. Quarterly Journal of Economics, 119(1), 91–134.

Bhandari

A. K.

, & Heshmati

(2006). Wage inequality and job insecurity among permanent and contract workers in India: Evidence from organized manufacturing industries (IZA Discussion Paper No. 2097). Retrieved from https://ssrn.com/abstract=900362.

Bhide

, Chadha

, & Kalirajan

(2005). Growth interdependece among Indian states: An exploration. Asia-Pacific Development Journal, 12(2), 59.

Brar

, Gupta

, Madgavkar

, Maitra

B. C.

, Rohra

, & Sundar

(2014). India’s economic geography in 2025: States, clusters and cities-identifying the high-potential markets of tomorrow. McKinsey & Company, Delhi.

Budd

J. W.

, Chi

, Wang

, & Xie

(2014). What do unions in China do? Provincial-level evidence on wages, employment, productivity and economic output. Journal of Labour Research, 35(2), 185–204.

10.

Clark

J. B.

(1899). The distribution of wealth: A theory of wages, interest and profits. New York, NY: Macmillan.

11.

Das

S. K.

, & Barua

(1996). Regional inequalities, economic growth and liberalization: A study of the Indian economy. The Journal of Development Studies, 32(3), 364–390.

12.

Dickens

W. T.

, & Katz

L. F.

(1986). Inter-industry wage differences and industry characteristics (Working Paper No. 2014). Cambridge, MA: National Bureau of Economic Research.

13.

Dickey

D. A.

, & Fuller

W. A.

(1979). Distributions of the estimators for autoregressive time series with a unit root. Journal of American Statistical Association, 74(366a), 427–481.

14.

Dickey

D. A.

, & Fuller

W. A.

(1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49(4), 1057–1072.

15.

Dunne

, Foster

, Haltiwanger

, & Troske

K. R.

(2004). Wage and productivity dispersion in United States manufacturing: The role of computer investment. Journal of Labor Economics, 22(2), 397–429.

16.

Engle

and Granger

(1987). Co-integration and error correction: Representation, estimation and testing. Econometrica, 55(2), 251–276.

17.

Fedderke

J. W.

, & Mariotti

(2002). Changing labour market conditions in South Africa: A sectoral analysis for the period 1970–1997. South African Journal of Economics, 70(5), 830–864.

18.

Fleck

, Glaser

, & Sprague

(2011). The compensation-productivity gap: A visual essay. Monthly Labor Review, Retrieved from http://www.bls.gov/opub/mlr/2011/01/art3full.pdf

19.

Frazis

, & Loewenstein

M. A.

(2006). Wage compression and the division of returns to productivity growth: Evidence from EOPP (BLS Working Paper No. 398). Washington, DC: BLS, Department of Labour Statistics.

20.

Goh

S. K.

(2009). Is productivity linked to wages? An empirical investigation in Malaysia (MPRA Paper No. 18095). Penang: Centre of Policy Research & International Studies, University Sains Malaysia (USM).

21.

Goldar

, & Banga

(2005). Wage-productivity relationship organised manufacturing in India: A state-level analysis. The Indian Journal of Labour Economics, 48(2), 259–272.

22.

Granger

C. W. J.

(1969). Investigation of causal relations by econometric model and cross spectral methods. Econometrica, 37(3), 424–438.

23.

Granger

C. W. J.

(1988). Some recent developments in a concept of causality. Journal of Econometrica, 39(1–2), 199–211.

24.

Harrison

(2009). Median wages and productivity growth in Canada and the United States (Center for the Study of Living Standards Research Note 2009–2). Retrieved from http://www.csls.ca/notes/note2009-2.pdf

25.

L. P.

& Yap

S. F.

(2001). The link between wages and labour productivity: An analysis of the Malaysian manufacturing Industry. Malaysian Journal of Economic Studies, 38(1–2), 51–57.

26.

International Labour Organization (ILO) (2015). Global wage report: Wages and income inequality 2014/15. Geneva: International Labour Organization.

27.

Jain

(2014). Wages and wage disparity in organized manufacturing industries: An analysis of post reform era. The Indian Journal of Labour Economics, 57(4).

28.

Jean

, & Nicoletti

(2002). Product market regulation and wage premia in Europe and North America: An empirical investigation (Working Paper No. 318). Paris: Economics Department, Organization for Economic Cooperation and Development (OECD).

29.

Kalirajan

, Bhide

, & Singh

(2009). Development performance across Indian states and the role of the governments (ASARD Working Paper 2009). Retrieved from https://crawford.anu.edu.au/acde/asarc/pdf/papers/2009/WP2009_05.pdf

30.

Khoon

S. G.

, & Nyen

W. K.

(2010). Analyzing the productivity-wage-unemployment nexus in Malaysia: Evidence from the macroeconomic perspective. International Research Journal of Finance and Economics, 53(12–10), 145–156.

31.

Krishna

K. L.

(2004). Patterns and determinants of economic growth in Indian states (Working Paper No. 144). New Delhi: ICRIER.

32.

Krugman

P. R.

(2000). Technology, trade and factor prices. Journal of International Economics, 50(1), 51–71.

33.

Kumar

, Webber

D. J.

, & Perry

(2012). Real wages, inflation and labour productivity in Australia. Applied Economics, 22(23), 2945–2954.

34.

Kurian

N. J.

(2000). Widening regional disparities in India—some indicators. Economic and Political Weekly, 35(7), 18.

35.

Levine

D. L.

(1992). Can wage increases pay for themselves? Tests with a production function. Economic Journal, 102(414), 1102–1115.

36.

Lopez-Villavicencio

, & Silva

(2010). Employment protection and the non-linear relationship between the wage-productivity gap and unemployment. Scottish Journal of Political Economy, 58(2), 200–220.

37.

Meager

, & Speckesser

(2011). Wages, productivity and employment: A review of theory and international data (European Employment Observatory Thematic Expert Ad-hoc Paper). Brighton: Institute of Employment Studies.

38.

Millea

(2002). Disentangling the wage-productivity relationship: Evidence from select OECD member countries. International Advances in Economic Research, 8(4), 314–323.

39.

Mishel

, & Shierholz

(2011). The sad but true story of wages in America (Economic Policy Issue Brief No. 297). Retrieved from https://www.epi.org/files/page/-/old/issuebriefs/IssueBrief297.pdf

40.

Murphy

K. M.

, Riddell

W. C.

, & Romer

P. M.

(1998). Wages, skills and technology in the United States and Canada (Working Paper No. 6638). Cambridge, MA: National Bureau of Economic Research.

41.

Newey

W. K.

, & West

K. D.

(1987). A simple positive semi-definite, heteroscedasticity and autocorrelation consistent covariance matrix. Econometrica, 55(3), 703–708.

42.

Nikulin

(2015). Relationship between wages, labour productivity and unemployment rate in new EU member countries. Journal of International Studies, 8(1) 31–40.

43.

Nwaokoro

N. A.

(2006). Real wage rate and productivity relationship in the declining US steel industry. International Business and Economic Research Journal, 5(2), 87.

44.

OECD (2001). Measuring productivity. OECD manual. Retrieved from http://www.oecd.org/sdd/productivity-stats/2352458.pdf

45.

OECD (2017). OECD economic surveys India. Retrieved from www.oecd.org/eco/surveys/economic-survey-india.htm.

46.

Pedroni

(1997). Panel Cointegration; Asymptotic and finite sample properties of pooled time series tests, With an application to the PPP hypothesis: New results. Working Paper, Indiana University.

47.

Pedroni

(1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61(0), 653–670.

48.

Phillips

P. C. B.

, & Perron

(1988). Testing for a unit root in time series regression. Biometrika, 75(2), 335–346.

49.

Rath

B. N.

(2006). Labour productivity determinants in Indian manufacturing: A panel data analysis. Indian Journal of Labour Economics, 49(1), 113–119.

50.

Sabharwal

(2007). Wage productivity relationship: Theoretical issues. New Delhi: Deep and Deep.

51.

Schwellnus

, Kappeler

, & Pionnier

P. A.

(2017). Decoupling of wages from productivity: Macro-level facts (Economics Department Working Papers No. 1373). Paris: OECD.

52.

Semeels

(2005). Do wages reflect productivity? (GPRG-WPS-029). Retrieved from http://www.gprg.org/pubs/workingpapers/pdfs/gprg-wps-029.pdf

53.

Sharpe

, Harrison

, & Arsenault

J. F.

(2008). The relationship between labour productivity and real wage growth in Canada and OECD countries (CSLS Research Report No. 2008–8). Ottawa: Centre for the Study of Living Standards.

54.

Sidhu

(2008). Wage disparity and determinants of wages in Indian industry. The Indian Journal of Labour Economics, 51(2), 249–261.

55.

Strauss

, & Wohar

(2004). The linkage between prices, wages and labour productivity: A panel study of manufacturing industries. Southern Economic Journal, 70(4), 920–941.

56.

Tsoku

J. T.

, & Matarise

(2014). An analysis of the relationship between remuneration (real wage) and labour productivity in South Africa. Journal of Educational and Social Research, 4(6), 59.

57.

Wakeford

(2004). The productivity-wage relationship in South Africa: An empirical investigation. Development South Africa, 21(1), 109–132.

58.

Westerlund

(2005). New simple tests for panel cointegration. Econometric Reviews, 24(3), 297–316.

59.

Xiao

, & Phillips

P. C. B.

(2002). A CUSUM test for cointegration using regression residuals. Journal of Econometrica, 108(1), 43–61.

60.

Yildirm

(2015). Relationships among labour productivity, real wages and inflation in Turkey. Economic Research-Ekonomska Istraživanja, 28(1), 85–103

61.

Yusof

S. A.

(2008). The long-run and dynamic behaviors of wages, productivity and employment in Malaysia. Journal of Economic Studies, 35(3), 249–262.

62.

Zoega

, & Booth

A. L.

(2005). Worker heterogeneity, intra-firm externalities and wage, compression (Birkbek Working Papers in Economics and Finance BWPEF 0515). Retrieved from http://www.bbk.ac.uk/ems/research/wp/PDF/BWPEF%200515.pdf

(a) ADF Test
Variables	Level		First Difference		Inference
Variables	Individual Intercept	Individual Intercept and Trend	Individual Intercept	Individual Intercept and Trend	Inference
WGR	7.57	21.50	149.64**	138.66**	I(1)
LPR	39.34	57.08	183.75**	152.24**	I(1)
KL	20.56	48.14	175.61**	163.49**	I(1)
CNR	32.97	55.41	163.81**	125.52**	I(1)
(b) PP-Fisher Chi Square
Variables	Level		First Difference		Inference
Variables	Intercept	Intercept and Trend	Individual Intercept	Individual Intercept and Trend	Inference
WGR	7.33	25.98	169.66**	229.38**	I(1)
LPR	44.17	47.73	192.87**	198.48**	I(1)
KL	20.99	59.56	232.60**	235.03**	I(1)
CNR	46.35	63.94	198.49**	226.69**	I(1)

Dependent Variables
Model 1 ΔWGR (Marginal Productivity Theory)		Model 2 ΔLPR (Efficiency Wage Theory)
Long-run Causality
Cointegrating equation: ECT_{t – 1} = 1.00 WGR_{t – 1} – 1.68 LPR_{t – 1}** + 0.64 KL_{t – 1}* – 0.06 CNR_{t – 1} + 0.62		Cointegrating equation: ECT_t–1 = 1.00 LPR_{t – 1} – 0.60 WGR_{t – 1} – 0.38 KL_{t – 1} + 0.04 CNR_{t – 1} – 0.37
ECT_{t – 1} = –0.037*		ECT_{t – 1} = –0.221**
Short-run Causality
ΔWGR_{t – 1}	–0.28**	LPR_t–1	–0.04
ΔWGR_{t – 2}	0.06	LPR_t–2	–0.06
ΔLPR_{t – 1}	–0.07	WGR_t–1	–0.07
ΔLPR_{t – 2}	–0.12**	WGR_t–2	0.19
ΔKL_{t – 1}	0.07	KL_t–1	–0.02
ΔKL_{t – 2}	0.16**	KL_t–2	0.03
ΔCNR_{t – 1}	–0.12	CNR_t–1	–0.01
ΔCNR_{t – 2}	–0.11	CNR_t–2	–0.05
C	0.03	C	0.036
Joint test of significance (χ2 test)
LPR, H₀: C(4) = C(5) = 0 " H₀ rejected KL, H₀: C(6) = C(7) = 0 " H₀ rejected CNR, H₀: C(8) = C(9) = 0 " H₀ accepted		WGR, H₀: C(4) = C(5) = 0 " H₀ accepted KL, H₀: C(6) = C(7) = 0 " H₀ accepted CNR, H₀: C(8) = C(9) = 0 " H₀ accepted