Industrial energy demand, a forecasting model based on an index decomposition of structural and efficiency effects

• Author: François Lescaroux
• Published date: 01 December 2013
• Date: 01 December 2013
• DOI: http://doi.org/10.1111/opec.12023

Industrial energy demand, a forecasting

model based on an index decomposition of

structural and efficiency effects

François Lescaroux

Senior Market Research Analyst, Qatar Petroleum, Strategic Planning, Qatar Petroleum Headquarters, P.O.

Box 3212, Doha, Qatar. Email: lescaroux@qp.com.qa

Abstract

This paper proposes an approach for explaining and forecasting global industrial energy demand,at

the country level, that could be seen as a trade-off between the twomain options in use: ‘top-down’

and ‘bottom-up’ models. It relies on a two-term decomposition of industrial energy intensity, one

evaluating the contribution of changes in the industrial structure, the other one reflecting the con-

tribution of changes in sectoral efficiency. The former can be projected ‘bottom-up’ using

microeconomic forecasts. The latter is modelled ‘top-down’as a function of real per capita g ross

domestic product and electricity’s market share. Hence, this approach enables both to project future

demand and to disentangle the effects of structural mutations and efficiencygains for explaining past

or future changes. The latter driver might further be decomposed into two factors: efficiencygains

resulting from fuel substitutions and ‘other’sectoral efficiency gains.

1. Introduction

Industrial energy consumption represented 28 per cent of total final demand worldwide in

2008 (IEA, 2010a). From an accounting point of view, it consists of mining and quarrying

of non-hydrocarbon materials (NACE112–14),constr uction (NACE 45) and manufactur-

ing activities (NACE15–37) excluding ‘refining’ (NACE 23).

It is a highly heterogeneous sector of energy demand as the industrial structure is very

different from one country to another; the sub-sectors’ market shares evolvewith time and

when the economydevelops, and they depend as well on local features. Further, while non-

electricity resources are mainly used to produce heat and steam and can generally be

substituted one for the other, some sub-sectors have very specific requirements, like

‘chemicals’, ‘steel’ or ‘paper’.

Industrial energy demand results from the interplay of three main drivers. The most

important one for developing countries is the growth of industrial output.The second one,

that is most relevant for middle-income countries, is the changing shape of the industrial

structure: $1 value-added in the ‘electronic equipment’ or ‘construction’ sectors has not

477

© 2013 Organization of the Petroleum Exporting Countries. Published by John Wiley & Sons Ltd, 9600 Garsington

Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

the same energy requirements as $1 value-addedin the ‘steel’ sector. Eventually, efficiency

gains at the sectoral level is the principal factor for developedcountries; this can fur ther be

decomposed into two terms: efficiency gains resulting from the optimisation of the energy

mix and the use of higher-quality resources (mainly electricity) on one hand,all other effi-

ciency gains on the other hand.

Two alternative approaches commonly compete in industrial energy demand model-

ing: one consists of handling it ‘top down’ and trying to relate aggregate consumption to

macro variables likeg ross domestic product (GDP) or industrial production (Medlock and

Soligo, 2001; Adeyemi and Hunt, 2007; Lescaroux, 2011), the other is to look at each sub-

sector individually and explain total demand from the bottom (Schurr et al., 1960; MGI,

2007; Agnolucci, 2009; EIA, 2010; IEA, 2010b).

The second option looks preferable on paper but it is hampered empirically by several

limits. Firstly, building a global—or large-scale at least—model of this kind implies esti-

mating a lot of equations. Secondly, the required detailed data is available only on short

time samples for most countries. Thirdly, the quality of the data is rather poor2so that

sectoral energy intensity and market share series mayfollow chaotic trajectories and use to

show wide differences in their levels from one country to another. Fourthly and mostly,

simulating such a model implies making a lot of assumptions regarding potential effi-

ciency gains in each industrial subsegment; this requires sectoral case studies (MGI, 2007)

or similar heavily resource- and time-consuming analyses.

While aggregate values are more acute than sectoral ones, the first option faces major

difficulties as well.The industrial structure of an economy changes as it develops: both the

share of manufacturing activities in total value-added and the share of heavy industries in

manufacturing tend to decrease. As a consequence, industrial energy demand is a non-

linear function of GDP or industrial production. Such non-linear relationship is all the

more difficult to estimate that time samples are short even at the aggregate level and there

is no universal pattern (one country might be structurally more industrialised or more

heavily industrialised than another for geographic, historic or political reasons—access to

raw materials, voluntary or imposed trade limitations, earlydevelopment of automobile or

chemical industries for example) so cross-sections or panel regressions do not help so

much. In addition, it is not possible to disentangle structural and efficiency effects when

looking only at aggregate values. So top-down models are a convenient way to simulate a

business-as-usual (BAU) case but forecasters find themselves alone when they have to

build an alternative scenario like a ‘deglobalisation’case or anything out of the trend.

Our purpose is to build a tractable forecasting model of global industrial energy

demand, at the country or regional level, that enables one to easily simulate alternative,

out-of-the-trend scenarios (unlike top-downmodels that are best suited for BAU forecasts)

without requiring a lot of detailed assumptions (unlike bottom-up models that need to be

fed with lots of sectoral hypotheses).

François Lescaroux478

OPEC Energy Review December 2013 © 2013 Organization of the Petroleum Exporting Countries