Research Articles

Environmental Total Factor Productivity in China's Industrial Sector: Trend of Interregional Convergence

Pengfei Sheng, Jun Yang and Ding Lu

Abstract: The Convergence Hypothesis suggests the possible long-run trend for economies to converge in standards of living and growth rate of per capita income. The Hypothesis of Environmental Catch-up points to the tendency for late-developing economies to improve environmental performance as they catch up with those more developed ones. To test these hypotheses, this study compiles a Malmquist-Luenberger index for environmental total factor productivity with the industrial data of Chinese provinces from1998 to 2012. The test results are positive for the convergence of labor productivity but reveal a diverging trend for environmental total factor productivity through the period under study. The divergence is found to be resulted from deterioration of environmental performance among the provincial economies with relatively low labor productivity levels. The findings imply that most parts of China are still experiencing the rising part of the Environmental Kuznets Curve.

 

Keywords: Convergence, Environmental Catch-up, Environmental Total Factor Productivity, Environmental Kuznets Curve

JEL Classification: O13 O44 O47 Q51 Q56 E16

1 Introduction

Economics of development is about convergence: the underdeveloped poor economies find ways to catch up with the developed rich countries and converge to the latter's level of development. In neoclassical economics, the convergence hypothesis refers to the tendency for economies of similar conditions to converge to a long term steady state of growth in per capita income. An extension of the theory concerns the economists because of the possible convergence of productivity growth across economies. In recent literature, the production of both desirable outputs of goods and services and their by-products of undesirable outputs (such as pollutants) have entered into the calculation of environmental total factor productivity. That opens a new research field for the study of economic convergence in the perspective of development effectiveness and efficiency.

Inspired by this discourse, we apply the Malmquist-Luenberger productivity index (Chung et al., 1997) with a modified direct distance function to compile an environmental total factor productivity index for industries among China's provincial economies from 1998 to 2012. We then test the hypotheses of convergence of labor productivity and convergence of environmental total factor productivity during the period under study. The results were consistent with the convergence of labor productivity but reveal a diverging trend of environmental total factor productivity, which is found to be resulted from deterioration of environmental performance among the provincial economies with relatively low labor productivity levels. The findings have nevertheless provided empirical evidence consistent with the Hypothesis of Environmental Catch-up. In particular, the results show that most Chinese provincial economies are still experiencing the rising part of the Environmental Kuznets Curve.

The rest of the paper is organized as follows: the next section is a review on the convergence literature and its relevance to the issue of environmental catch-up and the concept of environmental total factor productivity. The third section explains the conceptual framework of this study. Data and analytical methodology are presented in the fourth section and the results are shown in the fifth section. The last section discusses the findings and summarizes the contributions of the paper.

2 Literature Review

2.1 Convergence Hypothesis and Environmental Catch-Up

Solow (1956, 1957) uses the accumulation of factor inputs (capital) and the growth of total factor productivity to account for differences in per capita output growth. Following this thinking, economics of development has long hypothesized that per capita income and standards of living would converge across nations. That tendency implies opportunities for poor developing economies to catch up with the rich developed ones in the level of development. The hypothetical convergence may be driven by two forces (Todaro and Smith, 2014). One is that technological progress in poor developing economies may occur faster, thanks to technology transfer from rich developed countries. The technology transfer allows the late developers to leapfrog over early stages of technological development and immediately adopt the state-of-the-art technologies. The other force is faster pace of capital accumulation in poor economies, where diminishing marginal returns to physical or human capital would make investment in these places more lucrative and attractive.

Empirical tests for convergence hypothesis can therefore be conducted at two levels. One is at the level of per capita income or per worker output growth. The other is at the level of its two hypothetical driving forces, the total factor productivity (TFP) growth and capital accumulation rates. Empirical tests at the first level generally suggest that absolute convergenceof per capita income only exists for developed countries (e.g. Baumol, 1986; DeLong, 1988; Barro, 1991; Barro and Sala-i-Martin, 1992, 1995, Dowrick and Nguyen, 1989, etc.). At the second level, Miller and Upadhyay (2002) discover that human capital has a significant effect on output and that evidence of absolute convergence of total factor productivity is stronger and more prevailing than that of real output per worker. A number of studies (Bils and Klenow, 2000; Klenow and Rodríguez-Clare, 1997; Hall and Jones, 1999; Parente and Prescott, 2000) have also provided evidence that most of the international differences in output per worker are driven by total factor productivity. However, with a new approach to estimate human capital stock, Manuelli and Seshadri (2014), suggest that human capital accumulation is more important than TFP in accounting for per capita income differences across countries.

The discourse on development convergence and catch-up process takes on a new dimension when environmental costs of economic growth are taken into account. As stated by Färe et al. (2007), there is no fire without smoke, which highlights the typical joint production of the desirable goods and the undesirable by-products in the course of economic development. According to the Environmental Catch-up Hypothesis (Brock and Taylor, 2004), economic development might experience an inverted U-shaped relationship between the level of per capita income and the level of environmental pollution, or the so-called Environmental Kuznets Curve (EKC). As an economy develops from very low income status to high income levels, the environmental quality tends to deteriorate at first before it gradually improves overtime. This transition is driven by the society's perceived opportunity costs of environmental degradation, which is low at early stage of industrialization when poverty alleviation and profit making were the priorities but will later increase substantially when standard of living rises beyond certain threshold level. The theory of Environmental catch-up hypothesizes that if both rich and poor economies start with pristine environments, the qualities of their environments at first diverge and then converge, due to the evolution of the opportunity cost of pollution abatement at different levels of development.

The Environmental Catch-up Hypothesis sheds new lights into the convergence debate and has inspired empirical studies to test international convergence of the pollutant emissions. These include Ezcurra (2007), Criadoand Grether (2011), and Herrerias (2013), who test cross-country convergence in carbon dioxide emissions, and Zhu et al. (2014), who investigate international differences in rates of reducing carbon intensity of economic growth.

Since modern economic growth has the by-products of undesirable damage to environment, the neoclassical definition of total factor productivity is insufficient to gauge growth efficiency. To resolve this issue, Chung et al. (1997) introduce the Malmquist-Luenberger productivity index as a productivity performance measure that credits the reduction of undesirable outputs such as pollution as well as the production of desirable goods. With the available technique to measure environmental total factor productivity (ETFP), new insights can be produced in empirical research on economic convergence. Jeon and Sickles (2004), for instance, use the Malmquist-Luenberger productivity index to account for carbon dioxide (CO2) as a bad byproduct of economic growth in their study of TFP growth across OECD (1980-90) and Asian countries (1985-95). They find that OECD accomplished growth in a lesser carbon-emitting way so as to achieve significant productivity growth while the Asian countries (except Japan) showed significantly negative productivity growth.

The growth accounting that takes into account the by-products of environmental negativities is still in its infant stage of practice. More empirical studies of ETFP are needed to enhance our understanding of the trends of economic convergence in the course of development.

 

2.2 Empirical Studies on China's Internal Convergence

The recent fast economic development in China has provided a rich soil for empirical research on economic convergence: it is a rare natural experiment to observe the occurrence of decades-long dynamic economic growth in the world's most populous country of vast geographic varieties within a consistent socialpolitical structure.

Most studies on per capita income or output convergence have used provincial per capita income data of various time horizons and focused on the period after market-oriented reform started in 1978. Significant β-convergence can only be found with pre-mid-1990s data, such as in Choi and Li (2000), who use shrinkage method for panel data models for period of 1978 to 1994. However, most studies reject the hypothesis of national convergence in either β-or σ-version.Some studies, including Zhang et al. (2001), Weeks and Yao (2003), Maasoumi and Wang (2008), and Lau (2010), find existence of regional convergence clubs by geographic subgroups. In a similar vein, Fredrik et al. (2013) indicate that provincial growth rates would diverge over the short-term but converge to two growth clubs in the long-run. In contrast, Pedroni and Yao (2006), who apply the non-stationary panel techniques, reject the view that provincial divergence can be attributed to the presence of regional convergence clubs. Westerlund et al. (2010) also conclude that the increased regional divergence is due to both region-specific disparities and to disparities between clubs of regions.

There is also some research on the convergence of pollutant emissions in China. For example, Huang and Meng (2013) observe a significant convergence in per capita carbon dioxide emissions in urban China from 1985 to 2008, of which the rate increases when its spatio-temporal dependency is considered. The study by Dong et al. (2013) confirms the existence of a convergence trend of carbon dioxide emissions in the whole country and in the three major economic regions. The convergence test by Herrerias and Liu (2013) reveals the convergence of electricity intensity in the majority of Chinese provinces with some regional exceptions. Wang and Zhang (2014) also confirm that per capita carbon dioxide emissions converge in all industrial sectors across provinces in China and rates of convergence are affected by factors like per capita income, industrialization process and population density. A recent study is Zhang et al. (2014), which constructs a Malmquist index to evaluate regional productivity performance for the removal of two pollutants, Chemical Oxygen Demand (COD) and sulfur-dioxide (SO2).Applying the measurement to provincial industrial data from 1992 to 2009, the authors find an inverted U-shaped relationship between per capita gross regional product and the pollutant-removal productivity growth.

In the literature of the Chinese economy, the issues of per capita output (income) convergence and convergence of per capita pollutant emissions are dealt with separately in most studies. In recent years, a few studies have incorporated both desirable output (income) and undesirable byproducts (i.e. pollutant emissions) in an integrated environmental total factor productivity measurement as inspired by the Malmquist-Luenberger index approach, pioneered by Chung et al. (1997). Using this index to account for sulfurdioxide emission as an undesirable output, Zhang et al. (2011) estimate the environmental total factor productivity in Chinese provincial regions and show its growth to have been much slower than the total factor productivity measured by traditional Malmquist index. Their results suggest that ignoring undesirable outputs has led to overestimation of China's productivity growth. That finding is confirmed by Chen and Golley (2014), who use similar approach to a dataset of the country's 38 industrial sectors to account for carbon dioxide emissions as an undesirable by-product. In a similar vein, Zhang and Choi (2013) use the method of metafrontier non-radial Malmquistindex to measure the totalfactor carbon dioxide emission performance for China's fuel power plants while Zhang and Zhou (2015) apply the same approach to study China's transportation industry. All these recent studies have offered valuable information about productivity performance in the Chinese economy by taking into account the undesirable by-products of pollution. However, they are not designed to test economic convergence in the perspective of environmental total factor productivity.

This paper aims to fill this gap by testing the convergence of environmental total factor productivity in industries across Chinese provincial economies. Applying the Malmquist-Luenberger productivity index introduced by Chung et al. (1997), we conduct tests of inter-regional convergence in a way similar to Jeon and Sickles (2004). In light of these tests, we may find out whether an environmental catch-up process has occurred in China's industrial sector during the period under study.

3 Conceptual Framework

3.1 Environmental Total Factor Productivity

Economic development raises standards of living and expands the range of economic and social choices through increased production of goods and services.

Efficiency of doing so is measured by productivity, i.e. the output of goods and services that can be produced by the given amount of production factors, such as capital and labor. The classical measure of production efficiency is the total factor productivity, which is the output increase that is not accounted for by the input of production factors.

Apart from desirable outputs of goods and services, however, production process also generates undesirable outputs such as pollutant emissions that degrade our living environment. Two broad approaches have emerged in the literature to account for the effects of pollutant emissions on productivity growth. One is to treat pollutant emission as a contributing input in the production process and measure the output elasticity of pollution (e.g. Kalaitzidakis et al., 2007; Empora and Mamuneas, 2011). The other is a joint production model that measures environmental total factor productivity, which evaluates how efficiently the production process generates the desirable output while contains or reduces the undesirable output (e.g. Jeon and Sickles, 2004; Zhang et al., 2014). The latter approach is initiated by Chung et al. (1997), whose directional distance function (DDF) is the framework of our analysis in this study. We assume that every decision-making unit (DMU) produces the desirable output Y (such as value added of industry) as well as the undesirable output B (such as the emission of industrial sulfur dioxide) with the inputs of labor (L) and capital (K) with a joint production function:

The above function (1) satisfies the following features:

1) The undesirable output is weakly disposable, which is expressed as: if (0＜κ＜1)and(Y, B) ∈P(L, K),then(κY, κB)∈P(L, K),implying that a reduction of B is feasible only if Y is simultaneously reduced.

2) The desirable output meets the property of free disposability: if Y'<Y, then (Y', B)∈P (L, K), which means that Y can be reduced without a reduction in B.

3) The desirable output and the undesirable output are “Null-Joint” such that if B=0thenY=0.

Following Färe et al. (1994) and Chung et al. (1997), a directional distance function (DDF) can be defined as:

in which Do denotes the proportion by which the desirable output can be increased and the undesirable output can be decreased.

SY and SB are the slacks by which the desirable output and the undesirable output can be decreased. Note here that Do is different from its definition in Chung et al. (1997), whose DDF tends to decrease the undesirable output and increase the desirable output by the same proportion, leaving slacks with the estimation of the standard data envelopment model (Aparicio et al., 2013). Using the slack-based model (SBM) proposed by Tone (2001), we build a modified DDF of Do to describe the proportion by which the desirable output can be increased and the proportion by which the undesirable output can be decreased. Then, the DMU's performance can be described follows Chung et al. (1997) as:

where ETE(Environment Technology Efficiency)denotes the performance of the ith DMU at period t,which is measured by relative distance between the actual point of production and the production frontier.ETE ranges from 0 to 1,with 1 representing the optimal performance.

Based on the DDF of Do,the growth of environmental total factor productivity is defined by the Malmquist index (Caves et al., 1982; Chung et al., 1997), which measures the change of ETE. Supposing that there are t=1, …, T periods, we can express the change of environmental total factor productivity as follows:

in which ML denotes the environmental total factor productivity,which describes the increase or decrease of ETE from period t to period t+1.(Kt,Lt;Yt,Bt) represents the efficiency value with inputs and outputs in period t and the production set of period t+1. Other values of DDF indicate the relative results.

 

3.2 The Convergence Model with Environmental Total Factor Productivity

On the basis of the neoclassical growth theory (Solow, 1956), Barro and Sala-i-Martin (1992) built the standard β convergence model, which is defined as:

In equation (5), g denotes a change in output per capita from period t to period t+T, y is the initial output per capita in period t, and α is determined by the steady-state growth rate and the rate of technological progress, which is assumed to be equal across different DMUs. The coefficient β<0 implies a negative correlation between growth rate and initial per capita output.

The neoclassical growth theory attributes per capita output growth to the accumulation of factor inputs and the growth of total factor productivity. The β convergence is perceived mainly as a result of diminishing marginal returns to capital. In recent literature, total factor productivity growth is no longer exogenous. Thanks to technological catch-up and learning-by-doing, total factor productivity may grow faster in lower income economies and thus display a trend of convergence in the following way as suggested by Miller and Upadhyay (2002):

where gp is the change of total factor productivity from period t to period t+T, TFP is the total-factor productivity in the referenced period of t.When β,the coefficient of ln(TFP), is negative, it indicates that the DMUs that have the lower total factor productivity have experienced higher growth rates of total factor productivity and therefore convergence might have occurred.

Taking the form of Cobb-Douglas production function, total-factor productivity can be defined as:As stated above, total factor productivity as the classical measurement for productive efficiency only represents the production of desirable outputs, but does not take into account the by-product of undesirable environmental degradation. In light of the Environmental Catch-up Hypothesis, it is intellectually interesting to apply the β-convergence approach to test whether there is evidence of convergence in environmental total factor productivity as defined in equations (3) and (4):Following the definition of total factor productivity and the connotation of environmental total factor productivity (Farrel, 1957; Caves et al., 1982; Chung et al.,1997),the ETE can be presented as the ratio of the output set to the input set:

where a and b are the output elasticity coefficients for capital (K) and labor (L) respectively. Then, the convergence for total-factor productivity can be rewritten as:

By definition, since Y is a desirable output while B an undesirable one, it must be that n1>0, n2<0. Since K and L are productive inputs, it also follows that m1>0 and m2>0.

Plugging equation (10) into equation (9), the modified β convergence model becomes:In equation (11), if λ1<0, λ2>0, λ3>0 and λ4>0, then it implies that all of the changes in output, undesirable output, capital input and labor input support that productivity convergence has occurred across different DMUs.

4 Data and Methods

4.1 Background of Industrial Development

As a manufacturing-based economy, nearly 40% of China's gross domestic product (GDP) comes from industrial added values. Meanwhile, the industrial sector has also become the largest source of many pollutants.

As shown in Figure 1, from 1998 to 2012, the industrial added value per worker increased from 24,685 RMB to 99,157 RMB with an annual growth rate of 10.4%. Remarkably, the provinces with lower initial industrial added per worker value in 1998 experienced higher growth rates than those with higher initial levels, suggesting a likely scenario of β convergence of per worker industrial output.

4.2 Data

In this study we use annual added value of industry (Y) as a proxy for the desirable output and annual industrial emission of sulfur dioxide (S) as a proxy for undesirable output. Industrial added value is measured in 100 million RMB Yuan at the constant prices of 1998. Labor input (L) is the average number of employees at the beginning and the end of the year. Capital stock (K) is the initial value of fixed assets minus accumulated depreciation, also at the constant price of 1998. The sample consists of thirty Chinese mainland provinces (excluding Tibet, Hong Kong, Macau and Taiwan) and covers the period from 1998 to 2012. All of the data are collected from China Statistical Yearbook, China Industry Economy Statistical Yearbook and Statistical Yearbook of the Chinese Investment in Fixed Assets.Table 1 reports the descriptive statistics of all variables.

4.3 Analytical Methodology

The SBM model proposed by Tone (2001) is applied to solve the output-oriented DEA in equations (2) and (3). Since the production frontier attained in this method may be affected by the uncertainties in the sampling variation and the estimated results are only the upper bounds of the real value, it is necessary to assess the sensitivity of environmental total factor productivity by bootstrapping the index. We follow Simar and Wilson (1998) to bootstrap the values and apply the method of Kneip et al. (2008) to calculate the confidence intervals. The number of replications is set to 2000.

To estimate coefficients in model (11), the Wooldridge (2002) approach is first used to test for serial correlation and determine whether a dynamic panel model should be used. Then the method of system generalized moment (GMM) (Arellano and Bover, 1995; Blundell and Bond, 1998) is applied for the estimation so as to eliminate the impact of endogeneity on the results. As stated by Arellano and Bond (1991), since the standard errors are biased in the process of the two-step GMM, the results of the one-step GMM are recommended. However, the research of Windmeijer (2005) indicates that the results of a two-step GMM are preferred if the standard errors are robust. Therefore, we use the two-step GMM to adopt the best model and the robust method proposed by Windmeijer (2005) to conduct the statistical inference.

5 Results

5.1 Environmental Total Factor Productivity Growth

Figure 2 reports the results of China's regional environmental total factor productivity growth from 1999 to 2012.3The fluctuations of environmental total factor productivity growth appear to go in tandem across regions in most years. The eastern China had higher growth than western and middle China from 1999 to 2006, but lost its lead after 2007 when both western China and middle China caught up.

Figure 3 reports the provinces with a lower initial per worker industrial added value that tend to experience slower growth in environmental total factor productivity.

Meanwhile, there is no apparent correlation between annual environmental total factor productivity growth and the initial levels of industrial sulfur dioxide per worker. These pictures indicate that the convergence in environmental total factor productivity may not have happened during the period under study.

 

5.2 Serial Correlation in Panel-data Models

Prior to the convergence test, the Wald test presented by Wooldridge (2002) is carried out to verify the serial correlation in the panel-data models. Table 2 reports the results of the tests for equations (5) and (11). All the results reject the null hypothesis that there is no first order autocorrelation in the panel data.

The possible existence of a serial correlation in the idiosyncratic error term may cause the standard errors to be biased and the efficiency of the estimators to be lower under the standard Pooled OLS (OLS), fixed effect OLS (FE), and random effect OLS (RE). Therefore, the system GMM (One-step and Two-step) is the preferred method to include the dynamic variation in the estimation of convergence.

 

5.3 Tests for β-convergence of Labor Productivity

The Sargan test accepts the null hypothesis of over-identifying restrictions and the AR (2) test accepts that there is no second-order autocorrelation in the first-differenced errors. Therefore the use of two-step system GMM should be efficient.

In the framework of equation (5), the estimated value for β coefficient of the lagged value added per worker is significantly negative in either one-step system GMM or two-step system GMM results. Results in Table 3 confirm the existence of robust convergence across provincial economies.

5.4 Tests for β-convergence of Environmental Total Factor Productivity

Tests for β-convergence of environmental total factor productivity are conducted using the mean or median values of bootstrap results of equation (9) as the dependent variable for estimating model (11). The results are presented in Table 4 and Table 5 respectively. Both the statistics of Sargan and AR (2) show that the estimates by two-step GMM are preferred and valid. In both Table 4 and Table 5, the GMM estimates result in the same signs and statistical significance for the estimated coefficients, confirming the robustness of the estimated results of model (11).

With regards to the lagged added value of industry, the estimated coefficient, λ1in model (11), is significantly negative, supporting the convergence hypothesis. The estimated coefficient of the lagged capital stock, λ3in (11), is significantly positive, also confirming the hypothesis. The estimated coefficient of lagged emission of SO2, λ2in (11), is positive but insignificant while the estimated coefficient of lagged labor, λ4in (11), is insignificant with the “wrong” (negative) sign. The two results thus do not support the convergence hypothesis. Given these results, we conclude that only lagged output and lagged capital stock have moved in the direction of the environmental total factor productivity convergence while the lagged pollutant emission and labor input did not. The hypothesis of convergence of environmental total factor productivity thus cannot be supported by the data.

5.5 Robust Checks by σ-Convergence

The σ-convergence refers to overtime declining of the cross-economy dispersion of the development indicator. The existence of β-convergence tends to reduce dispersion but the effect can be offset by random shocks that increase dispersion (Barro and Sala-i-Martin, 1992). In other words, existence of σ-convergence is a sufficient evidence for β-convergence but not vice-versa. It is therefore meaningful to use σ-convergence as a robust check to test Type II errors (“false negative”) in the results of tests for β-convergence.

The σ-convergence can be measured by standard deviation of the indicator (Miller and Upadhyay, 2002). To compare σ-convergence trends of labor productivity and environmental total factor productivity in a period of rapid economic growth, it is appropriate to use the coefficient of variation (CV), i.e. the standard deviation normalized by the sample mean, in this study.Figure 4 presents the coefficients of variation for value-added per worker and Environment Technology Efficiency [ETE, based on equation (3)]. Both labor productivity and ETE trend upward in most years, rejecting the hypothesis of σ-convergence. In particular, through the 14 years under study, CV of labor productivity rose from 39.8% to 48.8% of the mean value. It, however, experienced a significant decline from 55.6% to 48.8% in the last 3 years in the study period. In comparison, the CV of ETE increased from 19.3% to 28.9% of the mean value and through the whole period there was no significant decline in the CV value.

Therefore the σ-convergence results cannot deny the possible Type I errors (“false positive”) of the positive test results for β-convergence in the case of labor productivity. The σ-convergence results nevertheless confirm the negative test results for β-convergence in the case of environmental total factor productivity.

6 Concluding Remarks

Using industrial data of Chinese provincial economies from 1998 to 2012, we have tested the convergence hypotheses of both labor productivity and environmental total factor productivity. Our results show a significant β-convergence of labor productivity, but suggest that the convergence of environmental total factor productivity has not happened. Specifically, the results reveal significant converging effects of lagged output and lagged capital stock on environmental total factor productivity. However, changes in pollutant emissions and labor input do not appear to have facilitated convergence of environmental total factor productivity.

For the insignificant labor effects on environmental total factor productivity convergence, a plausible explanation is that, in most years of the study period (1998-2012), industrial expansion had not exhausted the rural surplus labor and the economy was yet to reach the so-called Lewis turning point (Lewis, 1954). During the first decade of this century, China's urbanization rate rose from 36% of the population to 50%, with annual increase of urban population in the rage of 18 million to 25 million. Most of the increased urban population consists of rural migrant workers. With such a rapid expansion of labor supply, the industrial production is far less constrained by labor than by capital. The abundance of labor supply implies slacks of labor constraints in the data-envelope-analysis (DEA) that defines the environmental total factor productivity in this study. Therefore, lagged labor input displays insignificant impact on the convergence of environmental total factor productivity.

The insignificant effects of sulfa dioxide emission on environmental total factor productivity suggest that the changing trend of this undesirable by-product of industrial growth has not helped convergence of environmental total factor productivity across provincial economies, or in other words, the poor regions tend to perform worse in pollutant emissions or pollution abatement. This finding is nevertheless consistent with the Environmental Catch-up Hypothesis:As per capita income rises, the perceived opportunity costs of environmentally degrading by-products become increasingly unbearable and will prompt the society to seek greener and cleaner modes of production.

During the study period, the most developed provincial economies might already have reached the peak of the Environmental Kuznets Curve and started to contain the rate of pollutant emissions or make efforts to clean up the pollution while many of the poorer provincial economies were still in the rising part of the Environmental Kuznets Curve, freely puffing out more pollutants as they pumped up the industrial output. In fact, among Chinese provincial economies, a significant Environmental Kuznets Curve has been identified in several studies such as Brajer et al. (2011), Song et al. (2013), and Zhang et al. (2014).

This evolving Environmental Kuznets Curve can also be observed in the results of this study. As shown in Figure 5, there have been a positive correlation between environment technology efficiency (ETE) and value-added per worker across provincial economies. This correlation grew stronger from 1998 to 2012, evident in the rise of R-squared value from 0.389 to 0.682. The stronger correlation indicates a growing gap in environmental total productivity performance between the provincial economies with high labor productivity levels and those with mediocre or low labor productivity levels. It is noteworthy that the growing gap in performance is not due to the improvement in environmental total productivity performance of the provincial economies with the highest labor productivity but due to the deterioration of performance among most provinces, especially those with relatively low labor productivity levels (see the encircled data dots in Figure 5).

The findings of this study have thus provided new evidence to verify the Hypothesis of Environmental Catch-up and Environmental Kuznets Curve in the context of the Chinese economy. Our finding that most Chinese provincial economies are still in the rising part of Environmental Kuznets Curve poses a severe challenge to the country's environmental sustainability. Given the already grim situation of environmental degradation, public policies should be geared to incentivize industries to contain the growth of pollutant emissions, especially in less developed regions. Regional performance in economic development should be evaluated not by the traditional yardstick of GDP or productivity growth but by a comprehensive environmental total factor productivity index, which takes into account not just the desirable outputs of goods and services but also the undesirable byproducts like pollutions.

This study contributes to the literature by testing the convergence of environmental total factor productivity in the Chinese economy. The novelty of our methodology allows us not only to gauge the convergence trend of environmental total factor productivity but also identify which of its components has or has not facilitated the convergence. While this study only includes one undesirable by-product, i.e. sulfur dioxide, the model can be applied to other undesirable “bads” of economic growth and industrial development. The calculated environmental total factor productivity may serve as a useful tool for public policy making and policy evaluation in industrial and regional development.

References

Aparicio, J., Jesus T. Pastor and Jose L. Zofio (2013). “On the Inconsistency of the Malmquist-Luenberger Index,” European Journal of Operational Research 229, 738-742.

Arellano, M., O. Bover (1995). “Another Look at the Instrumental Variable Estimation of Error-Components Models,” Journal of Econometrics 68,29-51.

Arellano, M., S. Bond (1991). “Some Tests of Specification for Panel Data:Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies 58,277-297.

Barro, Robert J. (1991). “Economic Growth in a Cross Section of Countries,” Quarterly Journal of Economics 106(5),407-443.

Barro,R.J.,X.Sala-i-Martin(1995).Economic Growth,New York:McGraw Hill.

Barro, R.J., X. Sala-i-Martin (1992).“Convergence,” Journal of Political Economy 100(4),223-251.

Baumol, William J. (1986). “Productivity Growth, Convergence, and Welfare:What the Long-run Data Show,” American Economic Review 76 (11), 1072-1085.

Bils, Mark, Peter J. Klenow (2000). “Does Schooling Cause Growth?”American Economic Review 90(5),1160-1183.

Blundell, R., S. Bond (1998). “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,” Journal of Econometrics 87,115-143.

Brajer V., R.W. Mead and F. Xiao (2011). “Searching for an Environmental Kuznets Curve in China's air Pollution,” China Economic Review 22,383-397.

Brock, William A., M. S. Taylor (2004). “Economic Growth and the Environment: A Review of Theory and Empiric,” NBER Working Paper 10854.

Caves, D.W., L.R. Christensen and E. Diewert (1982). “The Econometric Theory of Index Numbers and the Measurement of Input, Output, and Productivity,” Econometrica 501,1393-1414.

Chen S., J. Golley (2014). “‘Green' Productivity Growth in China's Industrial Economy,” Energy Economics 44,89-98.

Choi, H., Hongyi Li (2000). “Economic Development and Growth Convergence in China,” The Journal of International Trade & Economic Development:An International and Comparative Review 1(9),37-54.

Chung, Y.H., R. Färe and S. Grosskopf (1997). “Productivity and Undesirable Outputs:A Directional Distance Function Approach,” Journal of Environmental Management 51(3),229-240.

Cobb, C.W., P.H. Douglas (1928).“A Theory of Production,” American Economic Review 8,139-165.

Criado, C. Ordás, J.M. Grether (2011). “Convergence in Per Capita CO2 Emissions:A Robust Distributional Approach,” Resource and Energy Economics 33 (3), 637-665.

Dong, F., X. Li, R. Long and X. Liu (2013). “Regional Carbon Emission Performance in China According to A Stochastic Frontier Model,” Renewable and Sustainable Energy Reviews 28(11),525-530.

Dowrick, Steve, Duc Tho Nguyen (1989). “OECD Comparative Economic Growth,1950-85:Catch-Up and Convergence,” American Economic Review 79 (11), 1010-1030.

Empora, Neophyta, Theofanis P. Mamuneas (2011). “The Effect of Emissions on U.S.State Total Factor Productivity Growth,” Review of Economic Analysis 3, 149-172.

Ezcurra, R. (2007). “Is There Cross-Country Convergence in Carbon Dioxide Emissions?”Energy Policy 35(2),1363-1372.

Färe, R., S. Grosskopf and C.A. Pasurka Jr. (2007). “Environmental Production Functions and Environmental Directional Distance Functions,” Energy 32,1055-1066.

Farrell,M.J.(1957).“The Measurement of Productivity Efficiency,” Journal of Royal Statistical Society A20,281-283.

Andersson, Fredrik, N.G., David L. Edgerton and S. Opper (2013). “A Matter of Time: Revisiting Growth Convergence in China,” World Development 45, 239-251.

Grossman, G.M., A.B. Krueger (1995). “Economic Growth and the Environment,” Quarterly Journal of Economics 110,353-377.

Hall, Robert E., Charles I. Jones (1999). “Why do Some Countries Produce so much more Output per Worker than Others?”The Quarterly Journal of Economics 114(1),83-116.

Herrerias, M.J. (2013). “The Environmental Convergence Hypothesis: Carbon Dioxide Emissions According to The Source of Energy,” Energy Policy 61(10), 1140-1150.

Herrerias, M.J., G. Liu (2013). “Electricity Intensity across Chinese Provinces:New Evidence on Convergence and Threshold Effects,” Energy Economics 36 (3), 268-276.

Huang B., and Meng L. (2013). Convergence of Per Capita Carbon Dioxide Emissions in Urban China:A Spatio-Temporal Perspective.Applied Geography 40, 21-29.

Jeon, Byung M., R. C. Sickles (2004). “The Role of Environmental Factors in Growth Accounting,” Journal of Applied Econometrics 19(5),567-591.

Kalaitzidakis, P., P. M. Theofanis and T. Stengos (2007). “The Contribution of Greenhouse Pollution to Productivity Growth,” University of Guelph, Department of Economics, Working Papers No. 0802.

Klenow, Peter J., Andrés Rodríguez-Clare (1997). “The Neoclassical Revival in Growth Economics: Has it Gone too Far?” in Ben S. Bernanke and Julio J. Rotemberg (eds.) National Bureau of Economic Research Macroeconomics Annual 1997,Vol.12,Cambridge,MA:The MIT Press,73-114.

Kneip, A., L. Simar and P.W. Wilson (2008). “Asymptotics and Consistent Bootstraps for DEA Estimators in Nonparametric Frontier Models,” Econometric Theory 24,1663-1697.

Lau, Chi Keung M. (2010). “New Evidence about Regional Income Divergence in China,” China Economic Review 2(21),293-309.

Lewis, W. A. (1954). “Economic Development with Unlimited Supplies of Labor,” The Manchester School 28(2),139-191.

Maasoumi, E., L. Wang (2008). “Economic Reform, Growth and Convergence in China,” Econometrics Journal 1(11),128-154.

Manuelli, R., A. Seshadri (2014). “Human Capital and the Wealth of Nations,” American Economic Review 104(9),2736-2762.

Miller, S., M. P. Upadhyay (2002). “Total Factor Productivity and the Convergence Hypothesis,” Journal of Macroeconomics 24,267-286.

Parente, Stephen L., Edward C. Prescott (2000). Barriers to Riches, Cambridge, MA: The MIT Press.

Pedroni,P.,J.Y.Yao(2006).“Regional Income Divergence in China,” Journal of Asian Economics 2(17),294-315.

Simar, L., P.W. Wilson (1998). “Sensitivity Analysis of Efficiency Scores:How to Bootstrap in Nonparametric Frontier Models,” Management Science 44, 49-61.

Solow, Robert M. (1956). “A Contribution to the Theory of Economic Growth,” Quarterly Journal of Economics 70(1),65-94.

Solow, Robert M. (1957). “Technical Change and the Aggregate Production Function,” The Review of Economics and Statistics 39(3),312-320.

Song, M. L., W. Zhang and S.S. Wang (2013). “Inflection Point of Environmental Kuznets Curve in Mainland China,” Energy Policy 57,14-20.

Todaro,Michael P.,S.C.Smith(2014).Economic Development,New Jersey:Pearson Education.

Tone, K.A. (2001). “Slacks-Based Measure of Efficiency in Data Envelopment Analysis,” European Journal of Operational Research 130,498-509.

Wang, J., K.Zhang (2014). “Convergence of Carbon Dioxide Emissions in Different Sectors in China,” Energy 65,605-611.

Weeks, M., J.Y. Yao (2003). “Provincial Conditional Income Convergence in China,1953-1997:A Panel Data Approach,” Econometric Reviews 1(22),59-77.

Westerlund, J., David L. Edgertong and S. Opper (2010). “Why is Chinese Provincial Output Diverging?”Journal of Asian Economics 21,333-344.

Windmeijer, F. (2005). “A Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM Estimators,” Journal of Econometrics 126,25-51.

Wooldridge,J.M.(2002).Econometric Analysis of Cross Section and Panel Data,Cambridge,MA:The MIT Press.

Zhang C., H. Liu, H. Bressers and K. Buchannan (2011). “Productivity Growth and Environmental Regulations – Accounting for Undesirable Output:Analysis of China's Thirty Provincial Regions Using the Malmquist-Luenberger Index,” Ecological Economics 70,2369-2379.

Zhang, N., Y. Choi (2013). “Total-Factor Carbon Emission Performance of Fossil Fuel Power Plants in China: A Metafrontier Non-Radial Malmquist Index Analysis,” Energy Economics 40,549-559.

Zhang, N., P. Zhou and C. Kung (2015). “Total-Factor Carbon Emission Performance of the Chinese Transportation Industry: A Bootstrapped Non-Radial Malmquist Index Analysis,” Renewable and Sustainable Energy Reviews 41, 584-593.

Zhang, Z., A. Liu and S. Yao (2001). “Convergence of China's Regional Incomes,” China Economic Review 12,243-258.

Zhang, Keyun, Y. Yi and W. Zhang (2013). “Environmental Total Factor Productivity and Regional Disparity in China,” Letters in Spatial and Resource Sciences 7(1),9-21.

Zhu, Z. S., H. Liao and H. S. Cao (2014). “The Differences of Carbon Intensity Reduction Rate across 89 Countries in Recent Three Decades,” Applied Energy 113(1),808-815.