The Impact of Digital Technologies on OECD Energy Demand
Dr Shivani Taneja, University of Nottingham
Information and Communication Technologies (ICTs) offer many opportunities for energy saving, such as optimising energy use in buildings and industrial processes, but the continuing increases in the number, power and range of applications of ICTs may act to increase energy demand. To date, however, there have been few rigorous, empirical estimates of the net impact of ICTs on energy demand in different sectors. In this paper, we provide such estimates using a panel dataset of 28 sectors in 17 OECD countries over 13 years. By estimating equations for the share of energy in variable costs, we are able to derive estimates of the elasticity of energy use with respect to ICT capital services – both for the whole sample and for individual sectors. Our results suggest that investment in ICT is associated with a modest reduction in energy demand, with the impact being much larger in ‘service’ sectors. These results appear robust to a variety of specifications. The findings are relevant to the role of digitalisation in delivering net-zero emissions, and to the panel session on energy demand.
Employing neoclassical production theory, and building upon upon previous work on ICT and employment, we specify a short run variable cost function for each sector, assuming quasi-fixed capital inputs and using a translog functional form. Using Shepherds Lemma, we derive equations for the share of energy in variable costs as a function of energy prices, labour prices, ICT and non-ICT capital services and value added. We estimate this equation directly for our cross-country and cross-sector panel dataset over the period 1995-2007. Our main data source is the 2009 release of the EU-KLEMS database which provides internationally comparable data on inputs and outputs, distinguishing between capital stocks and capital services and between ICT and non-ICT capital. We supplement this with data from the 2013 release of the World Input Output Database, from the IEA Energy Prices and Taxes Database, and from the OECD. We estimate our cost-share equations with OLS, and explore a number of different specifications (including the use of country, sector and industry dummies) to identify a preferred specification. We then re-estimate this model using quantile regression techniques, which are more robust to outliers.
Our preferred model, which includes country and industry dummies, provides a statistically significant estimate of -0.07 for the elasticity of energy use with respect to ICT capital services for the whole sample. This suggests that investment in ICT was energy saving over this period, but the impact was small. Subdividing the sample, we estimate an elasticity of -0.06 for the manufacturing sectors and -0.33 for ‘services’ – which suggests that ICTs had a much larger impact on energy demand in the latter. We repeat the estimations excluding post-communist countries and/or countries with missing data and find comparable results. Our estimates for individual sectors vary widely, with a particularly large elasticity for the post & telecommunication sector (-0.99) and small and/or positive elasticities for most manufacturing sectors. Quantile regression techniques provide lower elasticity estimates, depending upon the percentile. We provide a number of caveats to these results and indicate priorities for future research.
Our results suggest that the diffusion of ICTs were associated with only modest reductions in energy demand over this period, and that the energy savings were much larger in services than in manufacturing. These results run counter to claims that ICTs are forming the basis of a green industrial revolution, and suggest that the energy savings from ICTs may be offset by counter trends in a number of areas.