|
Article Excerpt
This paper analyzes the residential natural gas demand in 12 European countries using a dynamic demand model, which allows for country-specific short- and long-run elasticity estimates. The own-price and income elasticities tend to be very inelastic in the short run, albeit with much greater responsiveness in the longer run. Our results support very limited technological substitution possibilities between different energy carriers in the short run. Furthermore, the results suggested structural differences of residential natural gas demand across European countries and provide support for employing a heterogeneous estimator such as the shrinkage estimator.
1. INTRODUCTION
This paper analyzes the residential natural gas demand in 12 European countries using a dynamic model, which allows for country-specific short- and long-run elasticity estimates. Many European countries experienced considerable changes in the natural gas market during the last decades, due to deregulation and considerable increases in supply. Figure 1 presents the development of total residential energy demand (split into types of energy carriers) in the European Union over the period from 1960 to 2002. From 1978 to 2002 the relative changes in residential consumption of natural gas, electricity, petroleum products and coal were 133.5%, 74.3%, -35.2% and -85.2%, respectively. In other words, the residential natural gas consumption has more than doubled over the period, while some other energy carriers have experienced a decline. (1)
[FIGURE 1 OMITTED]
European countries are likely to have large structural differences in residential natural gas consumption. As Figures 2 and 3 show, there are considerable differences in per capita consumption of natural gas and also in the share of natural gas in total residential energy consumption. Natural gas share of total energy consumption in 1978 and 2002 had a standard deviation of 24.8 and 21.6, and varied from 1.8% in Denmark to 78.7% in the Netherlands in 1978, and from 0.9% in Finland to 79.8% in the Netherlands in 2002. The share of natural gas in energy consumption increased from 21.3% in 1978 to 40.6% in 2002.
Energy demand studies based on annual data face several challenges. The potential structural differences between countries in energy demand indicate that using homogeneous type estimators can be inappropriate, but at the same time, the small number of time observations within each country makes the use of individual regressions problematic. Recently, Maddala et al. (1997), Baltagi and Griffin (1997) and Baltagi et al. (2000), have demonstrated that individual cross-section regression models often provide implausible estimates. For example, we observe positive own-price elasticities, as well as larger variations between countries than is plausible from an energy economic perspective. Although homogeneous model parameters provide a higher degree of freedom, they lead to a loss of information by imposing homogeneity across countries and fail to take into account the potential structural differences between the countries. In this paper we employ several estimators which allow for different degrees of heterogeneity in parameters. In addition to OLS and standard fixed and random effects panel data estimators we employ the more novel shrinkage estimator. It represents a compromise between separate OLS estimation on each country, that allows full heterogeneity in all parameters, and panel data estimators that allow only for heterogeneity in the intercept.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
The previous econometric research of European residential natural gas demand uses older data (Pindyck, 1979; Griffin, 1979; Estrada and Fugleberg, 1989; Berkhout et al., 2004; Haas and Schipper, 1998). Energy elasticities have been estimated by various methods and model approximations, and have tended to differ substantially. (2) Less is known about the structure of the residential natural gas demand in European countries when we employ data from more recent years.
What appears from the reviewed literature on energy demand is that (1) there is no 'superior' estimator--it depends on the nature of the data set at hand and model specification, and (2) there is a lack of econometric estimates on European natural gas demand, which is unfortunate given the importance of understanding the structure of this large and growing market.
The paper is organized as follows. Section 2 discusses previous research on residential energy demand, and issues in econometric demand estimation on panel data discussed in the literature. In section 3 we present our demand model specifications, estimators, and data. Section 4 provides the empirical results from the estimated econometric models. Finally, section 5 provides summary and conclusions.
2. ISSUES AND EMPIRICAL RESULTS IN ENERGY EMAND STUDIES
Studies on the estimation of residential energy demand elasticities vary in many aspects, and have used different model specifications and estimation methods. Surveys by Atkinson and Manning (1995) and Madlener (1996) reports a wide range of elasticities. However, less is known about the structure of the residential natural gas demand in European countries after changes over the last years. Furthermore, some of the previous research have imposed strong restrictions on the elasticities and are likely to have provided biased long-run effects.
Studies in the literature use several econometric techniques for estimating elasticities of residential energy demand from panel data. These estimators vary in their degree of parameter heterogeneity, with pooled estimators at the one extreme and individual country estimators at the other. There is a debate in econometric energy demand studies whether to use homogeneous or heterogeneous model parameters over the cross-section (Maddala, 1991; Maddala et al., 1997; Pesaran and Smith, 1995; Baltagi and Griffin, 1997; Baltagi et al., 2000). Intermediate estimators in terms of heterogeneity include standard panel data estimators, i.e. fixed and random effects estimators, and the more novel iterative empirical Bayes estimator advocated in Maddala (1991), hereafter denoted as the shrinkage estimator. The shrinkage estimator is a statistical trade-off between the heterogeneous and homogeneous estimators, dealing with the limited number of time observations, and enabling us to pursue an estimation strategy that discriminates between countries with structural differences in natural gas demand. The estimator 'shrinks' country-specific parameters of the individual countries toward a common probability distribution, leaving the individual country estimates heterogeneous after shrinkage. It represents a compromise between separate OLS estimation on each country, that allows full heterogeneity in all parameters, and panel data estimators that allow only for heterogeneity in the intercept. The shrinkage estimator has become popular in heterogeneous estimation on panel data models since it appears to provide more plausible elasticity estimates (Maddala et al., 1997; Baltagi and Griffin, 1997; Baltagi et al., 2003).
Furthermore, when there is potential parameter heterogeneity between the countries, the fixed effects estimator is likely to impose strong restrictions on the slope parameters. In the case of a dynamic panel data model and coefficients differing between cross-sections, Pesaran and Smith (1995) argue that "pooling and aggregating give inconsistent and potentially highly misleading estimates of the coefficients, though the cross-section can provide consistent estimates of the long-run effects". The larger the degree of parameter heterogeneity, the greater the bias of the long-run effect provided by the homogeneous estimators. When the number of time observations is small, the bias of the pooled estimator is likely to be a serious problem (Pesaran and Smith, 1995). Hence, the long-run elasticities provided by the fixed effects estimator are likely to be biased if there are structural differences between cross-sections.
Previous econometric studies of European residential natural gas demand using aggregated country data have been done by Balestra and Nerlove (1966), Pindyck (1979), Griffin (1979) and Estrada and Fugleberg (1989). Estrada and Fugleberg (1989) analyzed the own-price and cross-price elasticities of residential natural gas demand in France and West Germany over the period 1960-83, and argued that the absolute values of the elasticities are generally higher for West Germany than for France. They reported own-price elasticities in the range from -0.61 to -0.76 for France and from -0.75 to -0.82 for West Germany. Moreover, Pindyck (1979) reported a long-run own-price elasticity of -1.7 for residential natural gas demand using pooled annual country data for a group...
|
