This paper analyzes the changes in the energy consumption of the service sector in France over the period 1995–2006, using the logarithmic mean Divisia index I (LMDI I) decomposition method. The analysis is carried out at various disaggregation levels to highlight the specifics of each sub-sector and end-use according to their respective determinants. The results show that in this period the economic growth of the service sector was the main factor that led to the increase in total energy consumption. Structure, productivity, substitution and intensity effects restricted this growth, but with limited effect. By analyzing each end-use, this paper enables a more precise understanding of the impact of these factors. The activity effect was the main determinant of the increase in energy consumption for all end-uses except for air conditioning, for which the equipment rate effect was the main factor. Structural changes in the service sector primarily impacted energy consumption for space heating and cooking. Improvements in productivity limited the growth of energy consumption for all end-uses except for cooking. Finally, energy efficiency improvements mainly affected space-heating energy use.
While the service sector represents more than 70% of the GDP of OECD countries, it accounts for only 13% of total final energy consumption, behind transport, industry and households (which account for 37%, 25% and 21% of energy consumption, respectively) (IEA, 2008). Furthermore, the service sector is the most heterogeneous sector of the economy, made up of many small energy consumers. Thus, statistical assessment of this sector is complex and varies according to country. These characteristics explain why there has been a lack of investigation in this sector. However, rapid growth in service sector energy consumption makes it necessary to attain a more thorough understanding of this sector, particularly in the context of strengthening energy efficiency policies. The purpose of this paper is to contribute to the improved understanding of the determinants of energy consumption in the service sector in France, through the use of a decomposition method.
Decomposition methods have been widely used in the fields of energy and the environment. They make it possible to evaluate the relative contributions of various factors to changes in energy consumption, energy intensity and CO2 emissions. The increasing use of decomposition methods has been accompanied by methodological advances and the development of new methods.1 The most frequently used methods have been the Laspeyres index and the arithmetic mean Divisia index (Ang and Zhang, 2000). Based on both theoretical and practical criteria, Ang (2004) recommends using the logarithmic mean Divisia index I (LMDI I). First, this method provides a perfect decomposition, i.e. there is no residual term. Certain methods leave a residual term, whose size cannot be foreseen in advance. The residual term may sometimes be negligible but in some cases may be so large that it limits the usefulness of the decomposition analysis. This is the case with the examples given by Albrecht et al. (2002) and Liaskas et al. (2000). Furthermore, the LMDI I method is consistent in aggregation: the value of the index calculated in two steps is equal to the value of the index calculated in a single step with the same data. This property is useful for performing decomposition at various disaggregation levels. Finally, this method has several practical advantages in its implementation when compared with other perfect decomposition methods. In particular, there is a direct relationship between the additive and multiplicative forms,2 which is useful for interpreting the results. Additionally, the decomposition equations have the same mathematical form, irrespective of the number of factors considered. In this paper, we apply the LMDI I method to analyze changes in service sector energy consumption in France over the period of 1995–2006. The decomposition analysis is carried out at three levels of disaggregation to highlight the specifics of each sub-sector and end-use according to their respective determinants.
The following section reviews the literature on energy decomposition analysis in the service sector. 3 Methodology, 4 Data describe the methodology employed and the data used in this paper. Section 5 provides an overview of energy consumption, sub-sectoral composition and end-uses in the French service sector. Section 6 presents and discusses the results. The conclusions are summarized in Section 7.
2. Review of the literature
Golove and Schipper (1997) used a decomposition method based on the Laspeyres index to analyze changes in energy consumption in the United States from 1960 to 1993. The service sector was not disaggregated into sub-sectors and end-uses. Energy consumption was decomposed into an activity effect (represented by service sector value added) and an intensity effect (represented by the aggregate energy consumption per value-added ratio). Energy consumption rapidly increased until 1973, then stabilized until the drop in oil prices in 1986. The authors found that the activity effect was the major factor affecting energy demand and drove energy consumption upward. However, between 1973 and 1986, due to more efficient use of energy, the intensity effect canceled out the activity effect, resulting in stabilization of energy consumption. Following this, energy consumption rebounded due to a reduction in the intensity effect.
The International Energy Agency (IEA, 2004) applied the same method to 11 IEA countries over the period 1973–1998. Between 1973 and 1982, the strong decline in energy intensity was enough to offset the activity effect. Subsequently, the intensity effect decreased and the activity effect became the main factor leading to an increase in energy consumption.
The Asia Pacific Energy Research Centre (APERC, 2001) also used the Laspeyres index methodology, applying it to 13 APEC countries over the period 1980–1998. Three effects were considered: an activity effect (represented by total service sector value added), a productivity effect (represented by the total number of employees per unit of value added) and an intensity effect (represented by total energy consumption per employee). The activity effect was the main factor driving energy demand. The intensity effect also contributed to the growth of energy consumption. Improvements in productivity helped to offset the activity and intensity effects.
Murtishaw and Schipper, 2001a, Murtishaw and Schipper, 2001b decomposed the energy consumption of the service sector of the United States using the adaptive weighting Divisia method. Their two studies deal with the periods 1970–1994 and 1988–1998. Their method included an activity effect (service sector value added), a structure effect (floor area per value added) and an intensity effect (total energy consumption per floor area). The authors found that the activity effect was the dominant driver of energy consumption during the entire period. However, until the early 1990s, the intensity effect and the activity effect offset each other, while the structure effect had very little impact. Subsequently, the intensity effect decreased and the structure effect only partially offset the activity effect.
Krackeler et al. (1998) used the Laspeyres index to decompose carbon dioxide emissions in the service sector of 13 OECD countries over the period 1973–1995. They defined the same effects as Murtishaw and Schipper and completed their decomposition of CO2 emissions with a fuel mix effect and a utility mix effect.
The APERC (2001) conducted a second decomposition analysis taking into account the composition of the service sector. It was applied to a selection of three countries for which data were available. Energy consumption was decomposed into an activity effect (total floor area of the service sector), a structure effect (each sub-sector's share of total floor area), a weather effect (the ratio of energy consumption to weather-adjusted energy consumption for each sub-sector) and an intensity effect (weather-adjusted energy consumption per unit of floor area for each sub-sector). The activity and intensity effects were found to play important roles in changes in energy consumption compared with other effects.
Farla et al. (1998) used the Marshall–Edgeworth decomposition method to analyze changes in energy intensity in the Netherlands over the period 1980–1990, disaggregating the sector into five sub-sectors (including a construction sub-sector, which is not, strictly speaking, part of the service sector). They found that the growth in energy consumption per employee was compensated for by an increase in productivity (the number of employees per unit of value added). The structure effect played a minor role.
Canada's Office of Energy Efficiency (OEE, 2006) had previously used a refined Laspeyres index method in which the residual term is distributed to other effects. This method becomes increasingly complex once the number of factors exceeds three. It was thus abandoned by the OEE in favor of the LMDI I method. The service sector was disaggregated into 10 sub-sectors and 6 end-uses. The OEE separated changes in energy consumption into an activity effect (measured by the total floor area of the service sector), a structure effect (each sub-sector's share of the total floor area) and an intensity effect (energy consumption per floor area for each end-use and sub-sector). A weather effect was taken into account for space heating and air conditioning. Finally, a service level effect measured the impact of the increased penetration of auxiliary equipment. In the absence of complete data on stocks, sales and unit energy consumptions levels related to this equipment, the OEE instead used an estimated index to capture the impact of changes in service level. The activity effect and, to a lesser extent, the service level effect contributed to the increase in energy consumption between 1990 and 2004. Improvements in energy efficiency made it possible to limit this growth until 1999, but since then the intensity effect had decreased.
If you are interested in this research article, you can download the full length article through the download page using the link below, both PDF file and Word file are provided.
No comments:
Post a Comment