Energy consumption and economic growth in Asian economies: A more comprehensive analysis using panel data

Abstract

This paper applies the most recently developed panel unit root, heterogeneous panel cointegration and panel-based error correction models to re-investigate co-movement and the causal relationship between energy consumption and real GDP within a multivariate framework that includes capital stock and labor input for 16 Asian countries during the 1971–2002 period. It employs the production side model (aggregate production function). The empirical results fully support a positive long-run cointegrated relationship between real GDP and energy consumption when the heterogeneous country effect is taken into account. It is found that although economic growth and energy consumption lack short-run causality, there is long-run unidirectional causality running from energy consumption to economic growth. This means that reducing energy consumption does not adversely affect GDP in the short-run but would in the long-run; thus, these countries should adopt a more vigorous energy policy. Furthermore, we broaden the investigation by dividing the sample countries into two cross-regional groups, namely the APEC and ASEAN groups, and even more important results and implications emerge.

Introduction

How to reduce carbon dioxide (CO₂) emissions yet, all the while, maintain stable economic growth has not just given rise to heated rhetoric but has also been one of the major concerns of energy and environmental protection policy in every corner of the world. The 1997 Kyoto protocol commits industrialized nations that have ratified the treaty to reducing their emissions of greenhouse gases, principally CO₂, by around 5.2% below their 1990 levels over the next decade. Proponents have reached a consensus that this is a necessary step, not least because high-growth, industrialized economies produce nearly 40% of the world's human-generated CO₂, a figure that has accelerated in recent years. But, opponents contest such a policy out of fear that the imposition of a set of such so-called ‘unrealistic’ goals would be detrimental to the global economy, not to mention economic development at home, and subsequently lead to serious unemployment. By extension, many argue that even a shortage of energy would adversely affect income (Masih and Masih, 1998).

Following the work of Stern, 1993, Stern, 2000, Oh and Lee, 2004a, Oh and Lee, 2004b, Ghali and El-Sakka (2004) and Beaudreau (2005) who argue that energy is an essential factor in production, this paper makes it mark in the extant literature by empirically examining long-run co-movement and the causal relationship between energy consumption and real GDP. It does so based on the aggregate production function. And, rather than the bivariate model attempted in numerous earlier studies, the present research employs a multivariate model with energy consumption (EC henceforth), real GDP (GDP henceforth), labor force (LB henceforth) and real capital stock (K henceforth) for 16 Asian economies from 1971 to 2002. This study includes Iran, Jordan and Syria which have not been investigated in most, if not all, recent studies (see Table 1). We also determine the relationship between EC and GDP within a multivariate framework when capital stock and labor input are controlled for; this relationship could feasibly run in either or both directions regardless of whether it is transitory or permanent. For our panel of sample countries, we use a cointegration test which,¹ when compared to the cross-section approach, is more powerful and allows us to increase the degrees of freedom. We then use the fully modified OLS (FMOLS henceforth) technique to estimate the cointegration vector for heterogeneous cointegrated panels. This enables us to correct the standard OLS for bias induced by endogeneity and serial correlation of the regressors. Furthermore, we specify and estimate a dynamic vector error correction model (VECM) that is appropriate for heterogeneous panels and that distinguishes between short-run and long-run causality. Finally, we explore different group issues that are of concern to the Asia Pacific Economic Cooperation (APEC) and the Association of Southeast Asian Nations (ASEAN), and with the results of this study, we are able to examine the deeper characteristics that determine the most efficient policies with respect to energy use.

Most earlier studies have investigated the energy—income/output relationship from one of two perspectives: the demand side (or energy demand function) and the production side (or the aggregate production function). On the demand side, Oh and Lee (2004b) emphasize that this model should be used with three variables, namely energy, GDP and energy price proxied by the consumer price index (CPI). The production side model, however, includes energy, GDP, capital stock and labor in a multivariate production function. Table 1 lists some of the most recent research studies that pertain to energy consumption and income in Asian countries. With the exception of Masih and Masih (1998), Asafu-Adjaye (2000) and Fatai et al. (2004), all of the studies primarily focus on the production side; few papers include labor and capital input in their model. Meanwhile, most scholars have only considered single countries or have only had small samples (see as Stern, 1993, Stern, 2000 and Oh and Lee, 2004a, Oh and Lee, 2004b). Though some have employed the panel data approach, they have mostly ignored the cointegrated relationship among variables (Olatubi and Zhang, 2003). Al-Iriani (2006), however, does apply the panel causality technique to determine the EC–GDP relationship among the Gulf Cooperation Council (GCC) countries. Nevertheless, we note that the error correction term is missing in the panel system.

Owing to the gravity of the two energy crises in 1974 and 1981, the energy consumption—economic growth relationship has been at the core of numerous empirical studies since the late 1970s (see for example, Erol and Yu, 1987, Masih and Masih, 1996, Asafu-Adjaye, 2000, Morimoto and Hope, 2004, Lee, 2006, Lee and Chang, in press-a, Lee and Chang, in press-b). The groundbreaking research by Kraft and Kraft (1978) used data for the 1947–1974 period and found evidence of unidirectional causality running from GDP to energy consumption in the United States. Based on the empirical results from those studies, three interesting, albeit contradictory, theories have been put forth. First, if uni-directional causality runs from income to energy, then the implementation of energy conservation policies would have few, if any, adverse effects on income. Secondly, if uni-directional causality runs from energy consumption to income, then this is indicative of an energy-dependent economy in which energy is an impetus for growth, with the implication being that a reduction in the amount of energy available would likely negatively affect income. Finally, the finding of no causality in either direction has been reported; this, the so-called ‘neutrality hypothesis’, signifies that energy consumption does not affect income whatsoever.

In this regard, there are several reasons for researchers being confounded with these reported causal relationships. For one, earlier studies have used time series data. Apart from this, several have been flawed by the short time spans of the typical datasets.² Others have been estimated using the OLS without any regard for whether the series properties are stationary or not.³ Fortunately, Granger and Newbold (1974) pointed out that not taking non-stationarity into account could result in misleading conclusions about relationships among the variables. Recently, numerous studies have investigated the causal relationship between two series by using a time-series analysis, and in particular, by using the concept of Granger causality. However, despite advances in time series analysis over the last decade, recent tests on the energy consumption–growth relationship have employed bivariate causality procedures.⁴ These tests often fail to detect additional channels (such as capital and labor input) of causality and can also result in conflicting results.⁵ Thus, in the presence of cointegration among the series, what has been long overdue is an alternative, superior econometric method to the vector autoregressive (VAR) method (Asafu-Adjaye, 2000). Here, we use the VECM. Additional information suggests that VAR models may only be able to identify short-run relationships and that they are unable to determine cointegration among variables because a long-run relationship disappears with the first differencing; contrast this with a VECM which is capable of distinguishing between short-run and long-run relationships among variables as well as identify the sources of causation (Oh and Lee, 2004b).

As mentioned above, previous causality test results have mostly been based on individual countries and the use of time series data. However, there are different results for different countries, as well as for different time periods within the same country. Because panel data can provide much more information than either cross-sectional data or time series, and in light of the lack of power of individual unit root tests and traditional cointegration tests, we need to combine information from time series and cross-sectional data once again. Thus, we use the panel unit root tests and heterogeneous panel cointegration tests to prevent further debate and put the issue to rest.⁶ In a departure from previous studies, therefore, this study applies a new heterogeneous panel cointegration technique to re-investigate the relationship between energy consumption and real GDP across 16 Asian economies for the 1971–2002 period. Ideally, the panel methods is better able to catch the cointegrated relationships since a pooled level regression uses cross-section information in the data when estimating cointegrated coefficients.

We then compare the results for an overall more comprehensive, more accurate picture that should, in essence, resolve all the outstanding debates over the relationship between energy consumption and economic growth. This paper is unlike earlier ones in the sense that it compares the relationship between energy consumption and real GDP while controlling for capital and labor input in Asian economic groups rather than individual countries. It is well-known that Asia has an extremely bright future in the world economy, with economies that have the potential to develop quickly as well as those that are already considered developed. According to the population and vital statistics of the United Nations, Asia has the largest population of all major areas and regions in the world. Thus, the relationships among energy consumption, economic growth and environmental protection in such a densely populated area are the focus of this study. This paper targets the Asian region, and at the same time, developed and developing countries are included to gain more insight into relevant energy strategies in the different nations.

The remainder of this paper is organized as follows: In Section 2, we provide a brief discussion of the definitions of the variables and the data description. Section 3 presents the theoretical structures, methods and the empirical results. Finally, Section 4 concludes and suggests an important policy implication.