|
Article Excerpt
Models used to test whether an environmental Kuznets curve (EKC) can be used to describe the relationship between GDP and energy use and/or carbon emissions may be biased by the omission of energy prices. Here we include real energy prices and fuel shares in models that describe energy use and carbon emissions. We test if these models show a turning point in OECD countries. Results indicate that including energy prices eliminates statistical support for a turning point and suggest that the relationship between income and both energy use and carbon emissions is represented most accurately by diminishing returns. These results imply that economic growth per se will not reduce energy use or emissions that cause global climate change.
1. INTRODUCTION
The atmospheric concentration of carbon dioxide has increased by more than 30 percent since pre-industrial times and still is increasing at an average rate of 0.4 percent per year, mainly due to the combustion of fossil fuels and deforestation (IPCC, 2001). Quantifying the relationship between income and both energy use and carbon emissions is important to understanding the role of economic growth in climate change. This role often is analyzed using an environmental Kuznets curve (EKC), which hypothesizes that the relationship between income and both the use of natural resources and the emission of wastes has an inverted U-shape. Reasons for this inverted U-shaped relation include income driven changes in: (1) the composition of production and consumption; (2) the preference for environmental quality; (3) institutions that are needed to internalize externalities; and/or (4) increasing returns to scale associated with pollution abatement (for a review, see Andreoni and Levenson, 2001).
Empirical tests of the EKC hypothesis use a quadratic specification to estimate the relationship between income and the use of natural resources and the emission of wastes. To date, many empirical analyses generate results that are consistent with the hypothesis of an EKC (e.g. Holtz-Eakin and Selden, 1995; Schmalensee et al., 1998). This consistency is undermined by recent analyses that use more sophisticated econometric techniques (e.g. Richmond and Kaufmann, 2006; Perman and Stern, 2003) and/or expand the variables in the model (e.g. Kaufmann et al. 1998).
Including variables other than income is critical because previous efforts to estimate an EKC for the relationship between income and energy use and/or carbon emissions usually do not include energy prices (Holtz-Eakin and Selden, 1995; Schmalensee et al., 1998; Richmond and Kaufmann, 2006). Higher energy prices reduce energy use and carbon emissions because both firms and households can substitute capital or labor. Energy prices also drive interfuel substitution. As the price of oil increases relative to the price of coal, firms and household substitute coal for oil. This substitution will increase both heat measures of energy use and carbon emissions due to differences in their marginal product and carbon content (Kaufmann, 1994).
To evaluate the effects of price-induced substitution, we include real energy prices in econometric models and use their out-of-sample forecasts to test three alternative hypotheses about the relationship between income and energy use/carbon emissions. Including energy prices eliminates statistical support for a turning point in the relationship between income and energy use and/or carbon emissions. Instead, income gains increase energy use and carbon emissions, albeit at an ever decreasing rate. These results imply that economic growth per se may not reduce energy use or ease global climate change.
These results and the methodology used to obtain them are described in five sections. Section 2 reviews the development of an environmental Kuznets curve and its use to analyze energy use and carbon emissions. Section 3 describes the data and the statistical methodology that is used to estimate the relationship among income, fuel shares, energy prices, and either per-capita energy use or per-capita carbon emissions. Section 4 describes the statistical results. Section 5 discusses how the inclusion of energy prices affects conclusions about whether there is a turning point in the relationship between income and energy use or carbon emissions. Section 6 describes the implications of these results for policies aimed at slowing climate change.
2. AN ENVIRONMENTAL KUZNET CURVE FOR ENERGY USE AND CARBON EMISSIONS
The original Kuznets curve describes the relationship between income and its distribution (Kuznets, 1955). Kuznets postulates that initial gains in per capita income increase income inequality. Beyond some turning point, additional income gains reduce income inequality. This bell-shaped curve is known as the Kuznets Curve and is the basis for Simon Kuznet's Nobel prize in 1971.
The notion of an environmental Kuznets curve was popularized by the World Bank Development Report (Shafik and Bandyopadhyay, 1992). Their study estimates EKCs for ten indicators of environmental degradation, including per capita carbon emissions. Their sample includes observations for up to 149 countries between 1960 and 1990. They use a quadratic specification to estimate the relationship between per capita GDP (US$ 1985 PPP) and carbon emissions per capita. They find an EKC for carbon emissions with a turning point at $4,000 (1985 $US) per capita. The relatively low turning points suggests that climate change can be slowed by policies that promote economic development.
Shafik (1994) tests the degree to which this conclusion is robust by fitting three basic models: log linear, quadratic, and cubic. She finds that the relationship between income and carbon emissions is best represented by a log linear model, and that the turning point represented by the quadratic model occurs at an income level that is well beyond the largest observation in the sample.
Holtz-Eakin and Selden (1995) estimate the relationship between income and carbon emissions using a quadratic model and one specified in natural logs from an unbalanced panel that includes observations from 130 countries for the years 1951-1986. Regression results for the standard quadratic specification indicate a turning point at $35,000 (1985 $US) that is well beyond the maximum value for per capita GDP in their sample. For the quadratic model that uses natural logarithm, the turning point is occurs at a per capita income of $US 8 million.
Tucker (1995) specifies changes in per capita carbon emissions as a function of changes in per capita income and changes per capita income squared and estimates this equation from a balanced panel that includes observations for 137 countries for the years 1971-1990. Based on regression results, Tucker (1995) argues that tuning points exist in the relationship between carbon emissions and that their level decreases over time. Cole et al., (1997) estimate the relationship between income and carbon emissions (and other environmental indicators) from a panel that includes OECD nations. The model includes country-region specific dummy variables that represent the effects of natural endowments on environmental quality, a linear time trend that represents the effects of technological change, and a trade intensity variable that represents trade openness. The turning point in the relationship between income and carbon emissions occurs between $22,500 and $34,700 (1985 $US).
Roberts and Grimes (1997) are the first to test for a turning point in the relationship between income and carbon emissions in time series for individual OECD and non-OECD countries. They specify a log-quadratic model that includes population density. Their results suggest a turning point in the relationship between income and carbon emissions in high income countries. They argue that this, combined with efficiency improvements in wealthy countries, are responsible for the turning point in the relationship between income and carbon emissions estimated by previous studies of cross-sectional data.
This interpretation is supported by Dijkraaf and Vollebergh (1998), who suggest that testing for an EKC for carbon emissions from pooled data is misleading. They estimate this relationship from OECD-wide data from 1960-1990 using a logarithmic model with a third degree polynomial, which simulates changes in the relationship between environmental pressure and economic growth. Results indicate that the relationship between income and carbon emissions varies among nations.
Results generated by Bruyn et al., (1998) are consistent with the notion that an EKC estimated from pooled data need not hold for specific individual countries. They estimate a reduced form model that specifies per capita emissions as a function of per capita income for three types of emissions (carbon dioxide, oxides of nitrogen, and sulfur dioxide) for various time intervals between 1960 and 1993 for Netherlands, Western Germany, the UK, and the USA. Regression results indicate that carbon emissions are positively related to economic growth. They also argue that emission reductions are generated by structural and technological changes.
Unruh and...
|
