Testing the CO 2 Emissions Convergence: Evidence from Asian Countries

Abstract

The purpose of this article is to examine the convergence in CO₂emissions across 18 Asian countries over the period of 1972–2010, based on data obtained from World Development Indicators. The study used β and σ convergence for parametric tests and kernel density estimates for non-parametric tests of convergence hypothesis and found convergence in CO₂ emission in all the tests. More recent trends have been considered for the tests rather than the long-term trends in CO₂ convergence. During the initial period (i.e., 1972–1982 and 1982–1992), the values, although converging, are not significantly depending upon the previous period. In contrast, the results for recent decades show convergence in CO₂ emissions significantly. Sigma (σ) convergence estimates also indicates that variation declined more in recent decades. Thus, the empirical findings of the study fulfil the major tests of convergence and provide evidence that many Asian countries’ relative per capita CO₂ emissions are converging over time. The study suggests that these results should be looked at in multilateral negotiation agreements.

Keywords

convergence parametric and non-parametric kernel density

Introduction

The Kyoto protocol was negotiated in 1997 as a first step in addressing the problem of global climate change. It set targets for emission of greenhouse gases by industrialized countries for the period of 2008–2012 (Frankel, 2004). However, it did not provide targets for the developing countries, whereas these countries (e.g., China) are the major emitters. The main objective of Kyoto was to reduce CO₂ emissions which could not be achieved unless and until implemented in developing Asian countries. Thus, it becomes very important for policy-makers to understand the distribution of pollutants across countries and their growth pattern. Although the geographic distribution of greenhouse gas emissions does not affect the climate impact of the emissions, it does affect the political economy of negotiating multilateral agreements (Aldy, 2006). CO₂ is the most contributing pollutant in greenhouse gases and contributes more than 70 per cent of the atmospheric concentrations. CO₂also has the longest life-cycle.

The relationship between economic growth and pollution (i.e., the Environmental Kuznets Curve [EKC] hypothesis) has been tested in various studies, but there is a lack of literature in finding the convergence in pollution. The present study is aimed at testing the convergence in per capita CO₂emissions in Asian countries for the period of 1972–2010. It revisits the distributional analysis of carbon emissions convergence in a novel fashion.

Asian countries play a major role in the overall emission of CO₂ in the world. Among the top 10 emitters in the world, four are from Asia (China, India, Japan and South Korea) and the share contributed by them has been increasing over time. Thus, the study of Asian countries in this respect becomes very essential, but unfortunately the major studies have focused their attention to the developed or the industrialized world. Keeping this objective in view, the present work goes on to test whether per capita CO₂ emissions in Asian countries is converging or diverging over the time period of 1972–2010. This article first presents a review of past studies and the methodological issues adopted in the study, and then goes on to discuss the results and policy implications.

Background and Literature Review

Convergence Hypothesis

The concept of convergence was originally developed to show that the inequality in economic growth between countries and regions should reduce over time. It empirically validates the claims of neoclassical growth models about economic growth in cross-section economies (Barro, 1991; Baumol, 1986; Sala-i-Martin, 1996). In general, convergence in economic growth is verified when there is a negative relationship between the growth rate of income and the initial level of income, in which poor countries tend to grow faster than the wealthy ones. However, the concept is not restricted to economic growth literature only, but has frequently been applied to various fields of study. The two well-known convergence hypotheses used in this study are β-convergence and σ-convergence, for which both standard parametric tests and non-parametric tests were conducted. The main limitations in using β-convergence are the assumption of linearity in growth regression and the impossibility to detect convergence clubs, etc. (Goli, Perianayagam, & Reddy, 2013; Johnson, 2000; Kumar & Russell, 2002; Quah, 1993). The concept of σ-convergence became popular through the work of Quah in the 1990s. Sigma-convergence determines whether or not the distribution of income across economies is becoming more equitable over time. The application of non-parametric methods provide an alternative to standard parametric methods, as non-parametric methods do not assume that data follow normal distribution and are also useful in capturing short-term divergent paths that cannot be identified by Barro-regression (Raileanu Szeles, 2011).

There is ample literature showing the relationship between economic growth and CO₂ emission, but only few studies have tried to identify the convergence in CO₂emission across countries (Aldy, 2006; Brock & Taylor, 2004; Criado & Grether, 2011; Van, 2005). Van (2005) has tested convergence in large samples of developed and developing countries and found convergence in per capita CO₂. Aldy (2005 & 2006) found some historical divergence and forecasted continued divergence over several decades. Strazicich and List (2003) were the first to test CO₂ convergence for 21 OECD countries.¹ They found existence of absolute β-convergence and the value was significant. Brock and Taylor (2004) tested absolute as well as conditional β-convergence using their Green–Solow model for OECD countries for the period of 1960–1998. The test also found significant negative value of β.

Overall, the literature found in the area is mostly based on OECD countries and at the world level. While β-convergence has been found in the data on OECD countries, the world-level data has shown σ-convergence.

From the literature, we found less evidence about developing countries, especially for Asian countries. Hence, this study first explores the current literature by testing CO₂ convergence in Asian countries. Second, in the available literature, there has been little use of non-parametric methods for testing CO₂convergence. Thus, this study also inputs some methodological issues.

Mixed results are found in the literature for convergence and divergence. Recently, attention has been given to the examination of cross-country convergence in CO₂ emissions. The existence or non-existence of cross-country convergence in CO₂emissions is of considerable interest for policy-makers, as the assumption of convergence is inherent in the majority of the projection models used to prepare climate change policy proposals. Panopoulou and Pantelidis (2007) tested club convergence for 128 countries. They found convergence in the early years of their sample. Their study shows that high-income countries seem to converge and there is also slow convergence among the middle-income countries. On the other hand, the results show that low-income countries diverge. However, the study used a sample of data from the European Economic and Monetary Union (EMU) and OECD countries. A study by Westerlund and Basher (2008) shows convergence for 28 developing and developed countries by implementing panel unit root test.

Methodology

Data

The data used in this study consists of the annual CO₂ emissions measured in metric tons per capita, which have been taken from the World Bank database of World Development Indicators. To normalize the data, the study used natural log of per capita CO₂ emission. The dataset covers 18 Asian countries that have been selected on the basis of data availability for the years 1972–2010. To test the convergence hypothesis that countries with initial low per capita emission tend to pollute more with time, absolute β-convergence and σ-convergence have been used.

Methods

This study uses not only parametric methods but also non-parametric methods to test the convergence hypothesis. Non-parametric methods are used to overcome the linearity and normality assumption of the growth regression. It also tests the more recent trends rather than the long-term trends in CO₂ convergence.

Parametric Models of Convergence

Absolute β-convergence

Absolute β-convergence is used where the gap between the rich and the poor countries shrinks especially due to greater progress in the laggard countries, a concept that was introduced by Barro and Sala-i-Martin (1992). In the present work, the absolute β-convergence is tested by using the following equation:

\ln [\frac{Y ​_{i}^{}_{, ​ t}^{} + k}{Y ​_{i, t}}] = α + β . \ln (Y ​_{i, t}) + ε_{i, t}

where $\ln [\frac{Y _{i, t + k}}{Y _{i, t}}]$ is the mean annualized growth rate of the variable Y in country i in the period (t, t+k), Y_i,t is the value in the initial time t and ε_i,t is the corresponding residual.

Here, Y represents per capita CO₂ emission and ln represents its natural log. β-convergence is found if countries with initial emission beneath the average emission display faster growth than countries with initial emission higher than the average emission. For β-convergence to exist, the growth rate of Y_i,t must be negative (positive) if the initial value of Y_i,t is positive (negative). In other words, α and β must produce dissimilar signs (Solarin, 2014).

Douris (2008) suggested that progress that took place in recent periods within the larger period is more important for a policy perspective. For this objective, the study also estimated piecewise β-convergence by disaggregating the longer period.

Sigma (σ) convergence

Another important indicator of convergence is sigma (σ) convergence. It is usually measured either by standard deviation or by coefficient of variation in two different periods of time. Presence of β-convergence does not give the guarantee for σ-convergence. So, σ-convergence is usually used as a complement to the β-convergence. One can use standard deviation or coefficient of variation for testing σ-convergence. If it tends to decline over time, it is evident that there is convergence across the countries because the variability is decreasing.

The σ-convergence shows whether a variable tends to be similar across the countries or not. If the coefficient of variation tends to decline over time, σ-convergence exists. The second stage of the study tested absolute σ-convergence as a method of convergence analysis. This is done by using coefficient of variation, which is defined as:

C V = \frac{σ}{μ}

where σ is standard deviation and μ is mean.

Non-parametric Convergence Models

Parametric convergence models, although are useful to examine the convergence pattern across countries, have been criticized for using unreal assumptions, such as normal distribution and linearity. Therefore, this study used non-parametric tests of convergence as they do not assume that the data will follow a normal distribution and, hence, are helpful to capture short-time divergent paths that occur along the convergence process (Goli et al., 2013). Among the non-parametric estimates, histogram density estimates and kernel density estimates are widely used methods. The smoothness of kernel density estimates are better interpreted compared to the discreteness of histogram, as the former converge faster with the true underlying density for continuous random variables (Silverman, 1978). The study used Gaussian kernel which minimizes the mean integrated square error (MISE) in deriving h. A general form of kernel densities is estimated by using the following equation:

f (x) = \frac{1}{h n} \sum_{i = 1}^{n} k [​]

where, f(x) is the density estimation of the variable x, n is the number of observations, h is the bandwidth (smoothing parameter) and k(.) is the smooth and symmetric kernel function integrated to unity.

The main objective of this study is to test the convergence in CO₂ emissions across Asian countries. The focus is to test the spatial distribution of CO₂ emission across the countries. The study has used data at four time points. It has not attempted to forecast because time series data has not been used. Also, the study has used a comparatively small sample (18 Asian countries), and so forecasting emissions may not be accurate.

Results

Parametric Tests

Absolute β-convergence

Table 1 presents the results of absolute β-convergence models for CO₂ emissions per capita. The results show that in a span of 38 years, the per capita CO₂emissions show absolute β-convergence (β = –0.9619) across the 18 Asian countries and the value is statistically significant at 5 per cent level of significance. Piecewise regression estimates show a mixed pattern. During the initial period (i.e., 1972–1982 and 1982–1992), the values, although converging, are not significantly depending upon the previous period. In contrast, the results for recent decades show convergence in CO₂ emissions significantly.

Table 1.

Absolute β-convergence Model Estimates for Per Capita CO₂ Emissions of 18 Asian Countries for 1972–2010

Period	β-coefficient	p-value	Adjusted R²	n	df
1972–2010	–0.9619**	0.018	0.2595	18	17
1972–1982	–0.5618	0.349	–0.0040	18	17
1982–1992	–0.2749	0.536	–0.0366	18	17
1992–2002	–0.4933***	0.006	0.3445	18	17
2002–2010	–0.8214**	0.012	0.2874	18	17
Constant^a	6.2454***	0.001

Source: Authors.

Notes: n = sample size, df = degree of freedom; Level of significance: * = p < 0.1 ** = p < 0.05 *** = p < 0.01.

Sigma (σ) convergence

Sigma (σ) convergence is estimated by using coefficient of variation and is presented in Figure 1. The estimation of trends in coefficient of variation for per capita CO₂ emissions is found to be decreasing over the period of 1972–2010. A more careful examination of the trends reveals that the convergence has accelerated during the recent periods of 1992–2002 and 2002–10.

Non-parametric Tests

Kernel density estimates

Figure 2 represents the results of Gaussian kernel density estimates of the log of per capita CO₂ emissions. From the figure, it is clear that the mass of highly peaked and right-skewed carbon density in 1972 tends to migrate toward larger CO₂ emissions per capita. In 2010, the kernel density curve becomes smoother compared to earlier periods. The existence of small variations in earlier periods is clear from the kinks in the figure.

Figure 1.

Source: Authors.

Figure 2.

Source: Authors.

Findings

The study shows evidence for convergence in CO₂ emissions across Asian countries. Earlier studies showed different results for different groups of countries, especially for developed countries. However, in case of Asian countries, there exists a perfect mix of developed as well as developing countries, which are also big emitters of CO₂ on a global scale. Our findings show that present emission has been largely affected by the recent past, which means that convergence has been happening at an increasing rate. It is a clear indication that developed countries are controlling their emissions but countries with initially low emissions are emitting more and more. Such an analysis has not been done in previous studies. These results are also verified in this study by both parametric and non-parametric tests of convergence.

Conclusion and Policy Inputs

Previous studies have shown mixed results about the convergence in CO₂ emissions. The purpose of this study was to test the convergence is CO₂ emissions in Asian countries, which are some of the big emitters of the world today. For the stated purpose, it used both parametric and non-parametric convergence tests. The findings are important in all the tests applied. Convergence hypothesis is being proved not only across the time period of 1972–2010 but also for recent decades. The result of coefficient of variation is consistent with the other method and has provided evidence for emissions convergence across the Asian countries. The empirical analysis shows that per capita emission is an important factor in the projection of emission. Hence, the evidence presented through this study proposes a policy towards reducing CO₂emissions in emerging countries of Asia. Asian countries must also start looking for ways to reduce their emissions to check the deteriorating condition of climate in the continent. They should adopt per capita emissions allocation schemes with the need for substantial resource transfers. In conclusion, the present work is an attempt to stimulate empirical research on emission distribution in Asia. It is envisioned that the work will be expanded by using other forms of emissions and pollutants and also by including longer time periods.

Footnotes

Note

References

Aldy

J. E.

(2005). Per capita carbon dioxide emissions: Convergence or divergence? (Resources for the Future Discussion Paper No. 05-53). Retrieved September 28, 2016, from http://www.rff.org/files/sharepoint/WorkImages/Download/RFF-DP-05-53.pdf

Aldy

J. E.

(2006). Divergence in state-level per capita carbon dioxide emissions (Resources for the Future Discussion Paper No. 06-07). Retrieved September 28, 2016, from http://www.rff.org/files/sharepoint/WorkImages/Download/RFF-DP-06-07.pdf

Barro

R. J.

(1991). Economic growth in a cross-section of countries. The Quarterly Journal of Economics, 106(2), 407–443.

Barro

R. J.

, & Sala-i-Martin

(1992). Covergence. Journal of Political Economy, 100(2), 223–251.

Baumol

W. J.

(1986). Productivity growth, convergence and welfare: What the long-run data show. American Economic Review, 76(5), 1072–1085.

Brock

W. A.

, & Taylor

M. S.

(2004). The Green Solow model (NBER Working Paper No. 10557). Cambridge, MA: National Bureau of Economic Research (NBER).

Criado

C. O.

, & Grether

J. M.

(2011). Convergence in per capita CO₂ emissions: A robust distributional approach. Resource and Energy Economics, 33(3), 637–665.

Douris

S. F.

(2008). Global demographic convergence? A reconsideration of changing inter-country inequality in fertility. Population and Development Review, 34(3), 519–539.

Frankel

J. A.

(2004). The impact of a revised Kyoto protocol on developing countries. Retrieved September 28, 2016, from https://www.hks.harvard.edu/fs/jfrankel/ImpactOfRevisedKyotoProtocol.pdf

10.

Goli

, Perianayagam

, & Reddy

(2013). Socio-economic progress across the major Indian states: Converging or diverging (MPRA Paper No. 59668). Munich, Germany: Munich University Library.

11.

Johnson

P. A.

(2000). A non-parametric analysis of income convergence across the US states. Economic Letters, 69(2), 219–223.

12.

Kumar

, & Russel

R. R.

(2002). Technological change, technological catch-up, and capital deepening: Relative contributions to growth and convergence. American Economic Review, 92(3), 527–548.

13.

Panopouloe

, & Pantelidis

(2007). Club convergence in carbon dioxide emissions (IIIS Discussion paper no. 235). Institute for International Integration Studies, Trinity College Dublin, The University of Dublin.

14.

Quah

D. T.

(1993). Empirical cross-section dynamics in economic growth. European Economic Review, 37(2–3), 426–434.

15.

Raileanu Szeles

(2011). Exploring the economic convergence in the European Union’s new member states by using non-parametric models. Romanian Journal of Economic Forecasting, 14(1), 20–40.

16.

Sala-i-Martin

(1996). Regional cohesion: Evidence and theories of regional growth and convergence. European Economic Review, 40(6), 1325–1352.

17.

Silverman

B. W.

(1978). Choosing the window width when estimating a density. Biometrika, 65(1), 1–11.

18.

Solarin

S. K.

(2014). Convergence of CO₂ emission levels: Evidence from African countries. Journal of Economic Research, 19(May), 65–92.

19.

Strazicich

M. C.

, & List

J. A.

(2003). Are CO₂ emissions levels converging among industrial countries? Environmental and Resource Economics, 24(3), 263–271.

20.

Van

P. N.

(2005). Distribution dynamics of CO₂ emissions. Environmental and Resource Economics, 32(4), 495–508.

21.

Westerlund

, & Basher

S. A.

(2008). Testing for convergence in carbon dioxide emissions using a century of panel data. Environmental and Resource Economics, 40(1), 109–120. DOI: 10.1007/s10640-007-9143-2