Industrial energy intensities in the UK: is there a deterministic or stochastic difference among sectors?

Abstract

Energy intensities of industrial subsectors differ widely due to differences in the final product and ultimately in the production process. The aim of this article is to assess whether these differences are stochastic or deterministic. The analysis is implemented for a number of British industrial subsectors over the 1970–2004 and 1978–2004 time periods. It turns out that the results of the tests are very influenced by whether one allows for the presence of structural breaks. Only when modelling structural breaks, one can conclude that the evidence in favour of the long-term differences being deterministic outbalances the evidence pointing to their nature being stochastic. This supports the adoption of policy instruments which are applied across productive sectors in a way which is not affected by the short-run evolution of the sectors.

Acknowledgements

The authors would like to thank Ron Smith of Birkbeck College, University of London for the encouragement and support received during the preparation of this article and Kirsten Wiebe of the Institute of Economic Structures Research, Osnabrueck for feedback on a previous draft of this article. Funding from the Anglo-German Foundation under the programme Resource Productivity, Environmental Tax Reform and Sustainable Growth in Europe is also gratefully acknowledged.

Notes

¹ Energy and GVA data are available separately for the sectors wood and wood products, and paper, pulp and print. The decision to consider these two sectors together is related to the quality of the data on energy consumption in the wood and wood and products. Energy consumption decreases from 127 Thousand Tonnes of Oil Equivalent (TTOE) in 1991 to 21 TTOE in 1992 while GVA decreased from 2783 to 2751 million pounds. The breaks in the series disappear when the energy consumption from wood and wood products is added to paper, pulp and print due to the considerable higher consumption of the latter.

² In fact, as shown in Fig. 1(b), the metal sector has by far the highest energy intensity, i.e. twice as big as the second highest sector. When the metal sector is aggregated to other three sectors in OTH₁, see Fig. 1(a), the energy intensity is similar to that of food and tobacco.

³ Those occurring in 1970–72 and 1984 in the textile sector and in 1989 and 2003 in the mining sector are all caused by sudden changes in energy consumption unmatched by changes of the GVA. A sudden decrease in the energy consumption is also the cause of the marked decrease in the construction sector in the last two years of the sample.

⁴ In the case of the , i.e. MZ_α , MZ_t , MSB and MP _T in the table, GLS detrended data are used to construct the series on which the test is run and the autoregressive spectral density estimator , as suggested in ERS (1996) and NP (2001). The Modified Akaike Information Criterion (MAIC) has been used to determine the lag in the estimation of and in the unit root regressions. When using the MAIC, the maximum lag was set at l12. The PP test was run for a bandwidth going from l0 to l12. The deterministic components in the ADF and DF^GLS tests were chosen according to the value of the Schwarz and Akaike Information Criterion (SC and AIC, respectively). The specification selected for the ADF test is also used for the PP, DF_MAX and the DF^RMA tests while that selected for the DF^GLS test is also used for the tests.