Rebound Effects for Household Energy Services in the UK.

• Author: Chitnis, Mona

1. INTRODUCTION

   Major investments in energy efficiency are central to tackling climate change and to driving green growth. But to assess the effects of these investments on energy demand, it is important to understand the nature and magnitude of any associated 'rebound effects'.

   The term 'rebound effects' refers to a variety of economic responses to improved energy efficiency whose net result is to increase energy consumption and greenhouse gas (GHG) emissions relative to a counterfactual baseline in which those responses do not occur. For example, more energy efficient lighting reduces the marginal cost of lighting which encourages consumers to use more lighting (e.g. to illuminate more areas to higher levels for longer periods), thereby offsetting some of the potential energy and emission savings. The magnitude of these effects has been a source of controversy for years, but an increasing volume of research has reduced some of the key uncertainties (Dimitropoulos et al., 2016; Sorrell, 2007; Turner, 2013). However, for energy efficiency improvements by consumers, the evidence base has three important limitations (Chitnis and Sorrell, 2015).

   First, most studies focus upon rebound effects for car travel, since reliable data on the marginal cost and quantity demanded of other household energy services is much harder to obtain. Studies of rebound effects for lighting, for example, remain relatively rare (Saunders and Tsao, 2012; Tsao et al., 2010).

   Second, most studies focus solely upon direct rebound effects and neglect the associated indirect rebound effects. For example, energy-efficient lighting may encourage consumers to use more lighting (a direct rebound effect), but any remaining savings on electricity bills will be respent on other goods and services. Since the provision of those goods and services necessarily involves energy use and emissions, either directly or indirectly along their global supply chains, this re-spending is associated with additional emissions that further offset the environmental benefits of the energy efficiency improvement (an indirect rebound effect).

   Third, most of the studies that estimate indirect rebound effects focus on the income effects of energy efficiency improvements and neglect the associated substitution effects--or in other words, they rely upon expenditure elasticities rather than cross-price elasticities (Alfredsson, 2004; Bjelle et al., 2018; Murray, 2013; Thomas and Azevedo, 2013). As a result, their estimates of rebound effects are incomplete and likely to be biased (Chitnis and Sorrell, 2015). Since goods may be either substitutes or complements to the energy service, the associated indirect rebound effects may either offset or amplify the original emission savings.

   This paper seeks to overcome these limitations. We estimate the combined direct and indirect rebound effects from improvements in the energy efficiency of six different energy services in UK households, namely: i) space and water heating; ii) lighting; iii) cooking; iv) refrigeration and clothes washing; v) entertainment and computing; and vi) private vehicle travel (i.e. cars, motorbikes and vans). This approach is made possible by a unique database on the consumption and price of those services in the UK over the last half-century (Fouquet, 2008; Fouquet and Pearson, 2006). Our analysis involves estimating a two-stage demand system for household expenditure that includes these energy services as categories of expenditure. Our results suggest that rebound effects have eroded more than half of the potential emission savings from historical improvements in energy efficiency. We estimate that direct rebound effects have eroded as much as 90% of the potential emission savings, but we also find that indirect rebound effects are negative--that is, they contribute additional emission savings. For example, we find that improvements in the energy efficiency of lighting are associated with reductions in the consumption of heating, which contributes additional emission savings. As a result, the total (direct + indirect) rebound effect is less than the direct effect. However, we emphasise that our estimates are subject to considerable uncertainty, owing in part to the difficulty of including energy services within a household demand model.

   The paper is structured as follows. Section 2 describes the background to our approach, and summarises how we use estimates of the own- and cross-price elasticities and emission intensities of different goods and services to derive estimates of the combined direct plus indirect rebound effect. Section 2 describes the economic model used to estimate the elasticities while Section 3 summarises the econometric techniques employed. Section 4 summarises our data sources, and presents our estimates of the price and consumption of the six energy services over the period 1964-2015. Section 5 presents our results including our estimates of the own-price and cross-price elasticities for each energy service, together with the corresponding direct, indirect and combined rebound effects. Section 6 concludes by highlighting the limitations of our approach and providing some suggestions for future research.

2. ESTIMATING THE COMBINED REBOUND EFFECT

   Estimates of the direct rebound effect from an efficiency improvement can be obtained from estimates of the own-price elasticity of demand for the relevant energy service (e.g. the own-price elasticity of lighting) (Sorrell and Dimitropoulos, 2007). Estimates of the indirect rebound effect from such improvements can be obtained by combining estimates of the elasticity of demand for other goods and services with respect to the price of the energy service (e.g. the elasticity of demand for heating with respect to the price of lighting), with estimates of the energy/emission intensity of those goods and services (Chitnis and Sorrell, 2015).

   Both own and cross-price elasticity estimates can be obtained from a household demand model which specifies the expenditure on different categories of goods and services as a function of total expenditure, the price of each category and other variables. To estimate rebound effects, one of the expenditure categories must be the relevant energy service. For example, to estimate the direct and indirect rebound effects associated with energy efficient lighting, the model should include lighting as one of the categories of household expenditure, alongside other categories such as food and heating. This requires data on the price and expenditure share of each category of good and service, including lighting itself (Chitnis and Sorrell, 2015; Sorrell, 2010). The average price of lighting will, in turn, depend upon both the price of electricity and the energy efficiency of the installed stock of light bulbs. Improvements in lighting efficiency will make lighting cheaper, thereby encouraging increased consumption of lighting along with increased (reduced) consumption of goods and services that are gross complements (gross substitutes) to lighting (see Annex A of Chitnis and Sorrell, 2015).

   To convert these elasticity estimates into estimates of rebound effects, it is further necessary to estimate the energy use or emissions associated with household expenditure on each category of good and service. For energy services such as lighting, these primarily derive from the direct energy use and emissions associated with consumption of the relevant energy commodities--such as gas and electricity. (1) For other goods and services such as food and furniture, these derive from the embodied energy use and emissions associated with manufacturing, processing, shipping and retailing those goods and services. Embodied energy use and emissions can be estimated with the help of environmentally-extended input output models (Kitzes, 2013).

   As far as we know, no study has used this approach to estimate the combined (i.e. direct plus indirect) rebound effects for household energy services--owing primarily to a lack of data on the consumption and cost of those services and their share of total household expenditure. However, several studies have estimated combined rebound effects for the energy commodities used to provide those energy services. For example, Brannlund et al. (2007) estimate the combined rebound effect associated with efficiency improvements in household gas use by combining estimates of the own and cross-price elasticities of household demand for natural gas with estimates of the energy or emission intensity of different categories of household expenditure. But this approach has two drawbacks. First, using energy commodity price elasticities as a proxy for energy service price elasticities will lead to biased estimates of the both direct and indirect rebound effect (Chitnis and Sorrell, 2015; Sorrell, 2010). Second, additional bias will be introduced if the relevant energy commodity provides more than one energy service (e.g. electricity provides both lighting and entertainment), and/or the same energy service is provided by more than one energy commodity (e.g. heating is provided by both gas and oil) (Chan and Gillingham, 2015; Hunt and Ryan, 2014).

   Chitnis and Sorrell (2015) use the own and cross-price elasticities of energy commodities to estimate combined rebound effects for UK households over the period 1964-2013. This leads to estimates of 41% for efficiency improvements affecting gas consumption, 48% for improvements affecting electricity consumption and 78% for improvements affecting vehicle fuel consumption. In what follows, we seek to improve upon Chitnis and Sorrell (2015) in two ways. First, we estimate elasticities with respect to the price of energy services rather than the price of energy commodities, thereby allowing individual energy services to be isolated and removing one source of bias. Second, we distinguish between six categories of energy service, namely: i) space and water heating; ii)...