A service transition is supposed to lead to the decline of energy intensity (energy/GDP). We argue that this interpretation is overly optimistic because the shift to a service economy is somewhat of an illusion in terms of real production. Several recent studies of structural effects on energy intensity have made the error of using sector shares in current prices, combined with GDP in constant prices, which is inconsistent and ignores the different behaviour of prices across sectors. We use the more correct method of sector shares in constant prices, and make an attempt to single out the effect from the real service transition by using two complementary methods: shift share analyses in current and constant prices, and Logarithmic Mean Divisia Index (LMDI) for 10 developed and 3 emerging economies.
A service transition is rather modest in real terms. The major driver of the decline in energy intensity rests within the manufacturing sector. Meanwhile, the transition to a service sector had a small downward impact on energy intensity in 7 of the developed countries (and no impact in the others). For emerging economies like Brazil, Mexico and India, it is the residential sector that drives energy intensity down because of the declining share of this sector as the formal economy grows, and as a consequence of switching to more efficient fuels.
1. Introduction
Currently, strong concerns about the energy basis for our economic welfare in society exist due to the risks of global warming and potentially increasing costs of energy production. Thus, there is a hope that future energy demands will be less than in the past. One possible solution that could bring about this less energy-intensive future, at least in theory, is the transition to a service economy, because service production is generally less energy-demanding than industrial production in relation to the value that is created.
It is beyond dispute that employment in the service sector has increased drastically over the last several decades and that services make up the lion's share of GDP these days. However, this trend bears little resemblance to what happens in actual production. Kander (2005) raised the concern that the transition to a service economy was merely an illusion when it comes to what matters for energy: the real production structure. Kander's analysis confirmed this for Sweden's production. Data of value added at the four-sector level (industry, agriculture, services and transport) were used to demonstrate that the real share of the service sector did not grow in the long run. Further, the share for transports grew slightly, whereas the manufacturing sector share declined.
This paper expands the analysis to a wider set of countries. The aim of the paper is, first, to explore how the share of real service sector production develops over time and, second, to what degree any decline of energy intensity can be attributed to a (possible) service transition. We define service sector from an energy perspective, that is, one that includes wholesale, retail trade, hotels and restaurants, finance, insurance and real estate, community, personal and government services, post and communications, but excludes transportation. The analysis is mainly based on a set of developed countries, but is subsequently widened by including some of the giant emerging economies: India, Brazil and Mexico. This expansion is motivated because of the widespread concern that developing countries today are taking over the role that England played during the first industrial revolution of being the “factory of the world”. Thus, any transition to a service economy in the developed world may be due to a new division of labour on the global scale, accompanied by developing countries producing energy-intensive exports for the developed world. As some might say, “while we live in the service economy our industrial goods are produced elsewhere” (Hermele, 2002). Because global warming is a truly global phenomenon, as the atmosphere knows no national borders, there are no system gains from a service transition if this logically cannot be generalised over the globe. We will not be able to cover the trade issue in any depth here, but will look into the economic structures of the emerging economies and their energy intensity paths to see if we can find indications that they are being used as the new factories of the world.
In this paper, Section 1 introduces the problem. Section 2 gives an overview of previous research. Section 3 provides a theoretical discussion, which connects the service transition with the decoupling discussion and ends with our hypothesis. Section 4 presents the data. Section 5 describes the analytical method adopted for separating within-sector changes from between-sector changes using the Logarithmic Mean Divisia Index (LMDI). Section 6accounts for the results, and is divided into two parts: one for the developed world (6.1) and one for emerging economies (6.2). Section 7 summarises the main results and broadens the discussion.
2. Previous Research
Kander (2005) used the concept of Baumol's cost disease to explain why both employment and the share of GDP in current prices have grown in Sweden since the 1970s, whereas the share of real service production has not grown. Further, Kander discussed the environmental implications of these findings on energy intensity (energy/GDP). Baumol (1967) used a simple two-sector model of the economy that included the technically progressive sector (industry) and the stagnant sector (services). In the technically progressive sector, labour time is a means to achieve an end, so production can be rationalised by equipping workers with timesaving machines. In the stagnant sector, human time is often an indispensable part of the product itself, and labour productivity cannot rise as fast as in the manufacturing sector. Productivity gains in the progressive sector normally lead to an increase in industrial wages, and consequently, service workers will also demand higher salaries even though their productivity has not risen to the same degree. The result is higher costs for service production relative to manufacturing production. The higher costs in service production will be passed on to consumers, and prices of services will tend to increase compared to manufactured goods. There is a “cost disease” in the service sector, and if people continue to buy services in roughly the same relative amounts as they buy manufactured goods, despite the relatively higher costs, employment will logically have to shift over gradually from manufacturing to services. Thus, the service sector will employ an increasing share of the labour force, and services will become more expensive than industrial products over time. The increased employment, together with the increasing prices, creates an illusion that service production increased its share of GDP in recent decades more than it actually did. This fact becomes obvious from the evolution of the real production structure, which did not exhibit any growth of services, at least not in the Swedish economy (Kander, 2005).
Ever since William Baumol wrote his stimulating 1967 article, the accuracy of his analysis regarding the stagnant nature of service production has been discussed. The doomsday prediction that could be read from it was that overall productivity and growth rates would slow down substantially in the post-industrial societies. It is now widely acknowledged that not all services lag behind manufacturing in labour productivity; it is only those services that have human time as an indispensable ingredient, what we call personal services, that, by necessity, lag. In reality, the service sector consists of a diverse mixture of progressive and stagnant elements, so the generalisation that services are stagnant and manufacturing industries are progressive clearly is too rough (Baumol et al., 1985). Broadberry (2006)points out that certain market services have had high productivity increases, and what determines the level of productivity over the long run is the “industrialisation” of services, which involves both producer services (services provided to business) and the provision of some consumer services in a more mass-market fashion. Especially in the areas of transport and telecommunications, wholesale and retail distribution, and banking and finance, large productivity increases have occurred. van Ark and Piatkowski (2004) compare the importance of ICT capital for productivity in manufacturing industries in the EU15 and former communist countries of East Europe. They find that ICT capital contributes decisively to labour productivity, which brings convergence. Still, even with this industrialisation of services, a large fraction of the sector consists of personal services, such as health care, education, child care, etc., so there is normally some difference between the productivity development of the service sector and that of the manufacturing sector.
All sensible researchers use sector values in constant prices, as opposed to current prices, when they calculate sectoral productivity. This is done to account for price inflation within the sector, which has nothing to do with actual productivity. For some reason, scholars who calculate the impacts of structural shifts on energy intensity do not always use this method correctly and consequently produce skewed results, as seen in Fig. 1, which shows the long-term price development (price deflators) for some main sectors of the Swedish economy, as well as GDP. The price for public services and private services rise more than the price of GDP does, whereas transport and communication is, on average, more technically progressive than GDP and even industry. Price deflators, like the ones in Fig. 1, are used in national accounts to recalculate values in current prices to constant prices to measure the sectoral productivity development. The simple procedure for transforming numbers in current values into real production is to divide the values in current prices by the proper sector price deflator. Naturally, dividing the values in current prices with such different price deflators, as those of Fig. 1, means that sector shares in current prices and in constant prices will differ, as will their growth rates. The more rapid increase of service prices over other prices, as shown in Fig. 1, is an illustration of what Baumol (1967) calls the cost disease of services. Much discussion these days is devoted to the topic of proper price deflators and whether real service production is underrated the way we measure productivity in services, especially in public services, where most often a zero productivity increase is assumed. It is not the intention of this paper to engage in discussions of the quality of price deflators. If services were measured differently, taking better account of their productivity increase, then GDP would have grown faster, and energy intensity would have declined more, some of which would be explained by the service transition. However, that reality is in another world and far from our current situation. In this paper, we confine our argument and investigation to what impact the transition to the service economy has the way it is measured in national accounts today.
Fig. 1
Researchers sometimes overlook these different price developments for sectors. For instance, Hamilton and Turton (2002) find that energy intensity decline has taken place mainly in services and industry in the United States and within services in the European Union. Close scrutiny reveals that the value added by the sectors has been calculated in current prices, which exaggerates the decline in the service sector and underestimates it in industry. The results are as misleading as if energy/GDP was calculated based on current prices, which would show an immense decline over time due to inflation. Schäfer (2005) also uses GDP shares in current prices when he discusses structural change in energy use. He divides the world into 11 regions and finds that for the period from 1971 to 1998, structural changes in the economy are accompanied with energy shifts. When economies move from agricultural to industrial, there is a shift in the final energy use from the former to the latter. More surprisingly, he also finds that the shift to the service sector is accompanied by a shift in final energy use. There are two reasons for this finding in Schäfer's study, and both of them have to do with unconventional methods. One shortcoming is that he uses shares of GDP in current prices to calculate energy intensity, a procedure that tends to overrate the value of the service sector over time, and thus underreports the values for service energy intensity in 1998. The other shortcoming is that the statistics he uses (International Energy Agency) do not enable any separation of energy for commercial transport (which constitutes part of GDP and should be allocated to services) and energy for household cars (this is final energy use, which should not be allocated to the service sector). Instead, he combines all this energy into the service sector. Thereby, levels of energy intensity in services are inflated, perhaps equal to those of the manufacturing industry, which means that if the energy for the household car fleet increases more than service production in general, the overestimation of service sector energy use will grow over time. These flaws in his method distort the results of energy intensity of the service sector in different directions, but mean that the service sector becomes “guilty” of far more energy use than it really is, and therefore there is no environmental relief from the transition to the service economy in his study. Although such a pessimistic conclusion is fairly accurate, as we will see later on in this paper, his results have not been obtained in a convincing manner.
3. Theory and Hypothesis
Up until the 1980s, energy was believed to have an iron-hand relationship with GDP and there was no talk of decoupling, or any growing gap between the energy and GDP curves over time. The couple was theorised to stick together over time. The view changed drastically with an article in Scientific American (Reddy and Goldemberg, 1990) that introduced long-run estimates of energy intensity (energy/GDP) for several countries suggesting a bell-shaped curve, or an inverted u-curve (see Fig. 2). According to this stylised graph, latecomers in the development process benefit from technical transference from their predecessors, so they peak at a lower point. This article served as a major input to what later became known as the environmental Kuznets curve.
Fig. 2
The idea behind the inverted U-curve is clearly related to structural change. During the industrialisation phase, energy intensity would increase, and after a peak, the curve would turn downwards, demonstrating the effect from the transition to a service economy. The idea that industrial societies would eventually see an end was presented in Marxist theory, where it was viewed in negative terms of a collapse. However, Fourastié (1949) introduced the more modern and positive view that declining industrial production would pave the way for a new kind of society based on increased service production.
Panayotou (1993) labelled the inverted U-curve in environmental relations the Environmental Kuznets Curve (EKC), after the famous Kuznets curve for income and equality relations (Kuznets, 1955). Panayotou (1993) explicitly proposed that the transition to a service economy was one of the main reasons for the declining environmental impact of economic growth (see Fig. 3). The idea is quite intuitive; as a country industrialises, it uses more energy and machines, and when it de-industrialises, the process is reversed, resulting in less energy use. Because of its intuitive character, this idea has not met many objections, that is, until Kander (2002, 2005) suggested that it was perhaps based on false beliefs of how a service transition functions.
Fig. 3
This brings us to the issue of what drives a service transition and of what it consists. At the intermediate service production level (services used as inputs of firms), part of the transition is a statistical artefact, as manufacturing firms increasingly outsource some of their service production to consultancy firms (Petit, 1986). Still, with the increasing complexity of goods, innovation and value creation takes place more and more in the after-production stage by companies providing maintenance services to their customers. Thus, there is reason to think that some actual increase in producer service production takes place, which is larger than the growth of manufacturing goods production per se (Perry, 1990, Berggren et al., 2005). When it comes to services for final demand by households, there are two opposing forces. One demand side force is Engel's law, suggesting that when people climb up the income ladder they first meet their basic needs, such as food and shelter, mainly produced in the primary sector (agriculture); second, they meet their less pressing needs, such as refrigerators, televisions and telephones, produced in the secondary sector (manufacturing); lastly, their least pressing needs are met, such as opera visits, psychological therapy, etc. (services).1 Another supply-side force is Baumol's cost disease, which states that certain services tend to price themselves out of the market and thus stimulates the DIY (do-it-yourself) economy, where companies like IKEA have found a niche by concentrating on the technically progressive part of the process: providing flat parcels and leaving the final, time-consuming construction to the costumers to complete. Engel's law is only applicable in one country at a specific point in time and can neither be used to infer expenditure patterns between countries, nor the consumption patterns over time for a country (Ingelstam, 1997). Simply put, when a country gets richer, this does not imply that everyone can afford to hire a cleaning lady or a gardener because the wages of these people follow the general income growth of the country.
Fourastié (1949) expressed these opposing forces as saturation on the demand side and technological progress on the supply side. Krüger (2008) places large emphasis on the theoretical insights provided by Fourastié and then draws the conclusion that “viewed against the empirical pattern of sectoral development, these considerations imply that in the long run the changes of the demand structure dominate the supply-side forces.” It is possible to say that these researchers jumped to their conclusions, based on not paying attention to the fact that sector development looks very different when viewed as shares in current or constant prices. This possibility was addressed by Gerschuny (1978), who questioned the view that modern society is entering a post-industrial phase when it comes to actual production, but from the expenditure side of individuals.
Theoretical discussions give reasons to be cautious in expecting too grand of a service transition in actual production terms. We therefore propose the following double hypotheses, which will be examined in our paper:
- 1)
- A service transition means that services have increased their share of total employment and their share of GDP in current prices (see Fig. 4).
Fig. 4 - 2)
- A service transition does not mean that services have increased their share of GDP in constant prices (or not as much as employment or shares in current prices) (see Fig. 5).
Fig. 5
Our main hypothesis is that the service transition is modest when it comes to actual production in the economy, and that it does not affect the energy intensities of developed countries. This hypothesis is based on the theoretical arguments above and the previous study results for Sweden, which confirmed our hypothesis in an empirical test (Kander, 2005). However, the results need not be as general as we expect; Sweden may be unique with its relatively large public service sector and large fraction of person-to-person services. Therefore, there is reason for investigating the issue across more countries.
4. Data
We use a dataset on energy quantities and economic values for both developed and emerging economies. In the developed group, we include the United States, Japan and 8 European Countries (France, Germany, Italy, the Netherlands, Portugal, Spain, Sweden and the United Kingdom). Emerging economies examined are Brazil, Mexico and India, which were chosen due to the important share that their population and economy represent to their respective regions. The dataset can be divided into three main sections: Long-run macro-economic energy quantities and GDP values, final energy by economic sector and economic structure measured in constant prices.
4.1. Long-run Primary Energy and GDP Data
To get a proper perspective on the purported service transition, it is necessary to study the issue from a long-term, historical perspective. Energy consumption for the 8 European countries during the period of 1860 to 2006 comes from the LEG (Long-term Energy Growth) database (Kander, 2002, Malanima, 2006; Gales et al., 2007, Warde, 2007, Henriques, 2009) and includes, besides commercial modern energy carriers, traditional carriers like firewood and draft animal muscle energy, which have been carefully elaborated. Direct working water and wind are also included in these estimates but make up a negligible part of the total energy in every country. Somewhat rougher estimates of traditional energy carriers have been done for this paper for the other large economies. Japanese energy consumption data was retrieved from Energy data and Modelling Center (EDMC) (2009), complemented with data from Mitchell, 2003, Mosk, 1978 and Historical Statistics of Japan by Ministry of Internal Affairs and Communications (MIC) (2008) for muscle energy. The United States' primary energy data was collected from Schurr and Netschert, 1960, [EIA] Energy Information Administration, 2009 and complemented with data from Mitchell (2003) and U.S. Department of Commerce (1975) for muscular energy. The same aggregates for Brazil, India and Mexico (1971–2006) use IEA energy balances ([IEA], International Energy Agency, 2008a, [IEA], International Energy Agency, 2008b). Primary electricity is expressed in terms of the energy content of the electricity produced in hydro and nuclear plants. GDP values are expressed in 1990 Power Purchasing Parities (PPP) (Maddison, 2008).
4.2. Economic Structure
We present benchmarks on the share of employment and share of the services in constant and current prices for 1950, 1971, 1990 and 2005 to explore our hypothesis of completely different evolutions of the service sector depending on what indicator is used (Table 1).2 For all the developed countries, except the United States of America, we use the newly updated EU KLEMS November 2009 release for 32 industries (EU KLEMS Database, 2009, O'Mahony and Timmer, 2009). In the case of the United States, we had to use the EU KLEMS 2008, SIC version that covers the period from 1970 to 2005 because the EU KLEMS 2009 release — NAISICS version only goes back to 1977 (EU KLEMS Database, 2008, Timmer et al., 2007). The reason why we prefer the EU KLEMS to other databases is because it separates postal services and communications from commercial transport in the service sector; these branches are very different from an environmental perspective. The benchmarks for 1950 are obtained by linking the EU KLEMS database to the 10-sector database of the Groningen Growth and Development Centre (GGDC) (van Ark, 1995). For Portugal in 1970, EUKLEM 2009 values for the employment share are linked with Pinheiro (1997). Data for India, Mexico and Brazil are based on the database in Timmer and de Vries (2007) for Asia and Latin America, which covers the period from 1950 to 2005 and comprises 10 sectors.3
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