Modelling Required Energy Consumption with Equivalence Scales.

AuthorYe, Yuxiang

    Energy consumption is different from many other consumer goods, because it is often considered a basic need (Welsch and Biermann, 2017); its satisfaction is necessary for an acceptable quality of life. Furthermore, modern energy services, such as electricity and liquefied petroleum gas are instrumental for most of human basic needs (Chakravarty and Tavoni, 2013). However, energy consumption contributes to household expenses, and (especially) when households are income-poor, energy consumption may be compromised for other purchases. These concerns underpin Sustainable Development Goals (SDG) 7--which seeks to "Ensure access to affordable, reliable, sustainable and modern energy for all" (United Nations, 2021). In other words, there is a concern that a non-negligible proportion of the world population is not able to purchase enough reliable energy for their needs, and are, therefore, energy poor. However, the determination of energy poverty, like any measure of poverty, requires an estimate of "need". For example, Boardman's (1991) energy poverty ratio requires "theoretical energy expenditure" or "required energy expenditure", which are measures of need. Intuitively, required expenditure focuses on the acquisition of adequate energy services (Liddell et al., 2012). (1)

    Despite its intuitive appeal, or maybe because of it, as well as the fact that it cannot be observed, adequate energy services, and thus, required energy consumption (REC), are open to debate and interpretation. Three approaches are widely applied in the literature to measure required or adequate expenditure. One option models energy demand following engineering methods, which are based on detailed domestic energy usage (in kWh), appliances and/or building characteristics information. For instance, the United Kingdom's (UK) Building Research Establishment Domestic Energy Model (BREDEM) (BRE, 2015) requires extensive engineering calculations localized to the country to account for dwelling characteristics and energy usage; such data is not widely available, if at all, in many countries. Papada and Kaliampakos (2018, 2020) suggest a stochastic model for required energy estimation founded on the UK's BREDEM. However, their research does not account for household heterogeneities that affect need. Another option is to conduct a purposive survey incorporating the relevant aspects of energy usage (Ntaintasis et al., 2019); however, such surveys are expensive to conduct, and, therefore, difficult to replicate widely. A further approach uses actual energy expenditure, instead of required (Heindl, 2015; Legendre and Ricci, 2015; Mohr, 2018). Although actual expenditure is expected to capture localized conditions and differences across households, it is unlikely to correctly capture need, because some households will reduce energy consumption to fulfil other needs. We are not aware of any method for the determination of required energy consumption that: (i) is underpinned by readily available data, (ii) accounts for household heterogeneities, (iii) captures localized conditions and (iv) incorporates relevant aspects of energy usage.

    In this paper, we propose such a method for estimating REC--a necessary parameter for the analysis of energy poverty. Our method is especially useful, when detailed engineering energy modelling and usage data are not available, as is the case in most developing country contexts. Instead, it is based on household expenditure surveys, which are collected nearly everywhere for the purpose of determining consumer price indexes and, thus, inflation. The method is also flexible enough to incorporate additional information about households, including information that correlates with household energy expenditure, as well as additional features that researchers or policymakers deem to be relevant, assuming such information is available. Specifically, we use semiparametric regression--a nonlinear multivariate regression--to incorporate a range of both household characteristics and energy usage factors. Finally, since the data is based on observed behaviour, it is able to capture localized information, such as that related to the weather.

    Since the proposed approach makes use of widely available income and expenditure data and can be estimated from multivariate regression, it offers: (i) a clear and viable option for those interested in designing policies to meet SDG 7, (ii) a method that can be applied by those interested in energy poverty and policy analysis in otherwise limited data settings and (iii) a method that can be applied similarly across a variety of countries for comparability purposes. We apply the approach in a case study of South Africa to gauge its reasonableness finding intuitive, and, therefore, reasonable results. In particular, we estimated REC for low- and middle-income households to be well above actual energy expenditure. This result is intuitively appealing, because South Africa is an unequal country, where poverty is rife (Leibbrandt et al., 2016); thus, we would expect poor households to require more energy than they are currently using.

    The remainder of this paper is organised as follows. Section 2 provides an overview of the existing literature in this field. Section 3 describes the methodology for required energy consumption modelling. We discuss the South African case study in Section 4 and present the results and discussion in Section 5. Section 6 concludes.


    When it comes to required household energy consumption for energy poverty measurement, the literature incorporates a range of ideas. For example, the low-income-high-cost (LIHC) indicator (Hills, 2012) uses an energy consumption threshold determined by required household energy costs. Household required energy is underscored by engineering models of energy use that incorporate building specifications, such as insulation levels, heating systems, geographical location of the dwelling and construction type (BRE, 2015), amongst other things. Taking advantage of detailed dwelling and household information provided by the English Housing Survey (DCLG, 2009), the BREDEM calculates total household energy requirements for space and water heating (to meet defined standards), energy for lights and appliances (including requirements for pumps, fans and electric showers, and energy generated by renewables) and energy for cooking (BEIS and BRE, 2018). Once required energy usage has been determined, it is multiplied by the relevant energy price to derive required energy expenditure. Thus, required energy expenditure is often underpinned by detailed knowledge of the building stock and its energy efficiency (Rademaekers et al., 2016).

    However, BREDEM is sensitive to the values of multiple parameters. Herrero (2017) shows that actual energy expenditure is well below (BREDEM) modelled energy expenditure, even in higher income deciles. Such results imply that wealthy households in the UK are energy poor, which is difficult to reconcile with its developed country status. These results also suggest that overestimation of household energy requirements is possible under the BREDEM model, limiting its value and generalizability in application. Furthermore, accessing such information may be problematic in many circumstances, especially in developing countries.

    Similarly, Papada and Kaliampakos (2018, 2020) develop a model for Greek energy consumption taking into account all domestic energy uses (space heating, space cooling, electricity-cooking-lighting and domestic hot water), although it ignores a number of differences that exist across households, such as housing quality and the number of household members. Although that concern is partially addressed by Ntaintasis et al. (2019), who consider floor area, type of residence, age of residence, energy prices and annual specific electrical and thermal energy consumption required across different types of Greek residential buildings, their research is built upon their own survey, and, therefore, may be difficult to replicate in many developing country contexts.

    The aforementioned REC models are relatively complex and require extensive data, or expensive data to collect. Apart from these difficulties, one must also address household heterogeneity, such that different types of households can be reasonably compared. Incorporating heterogeneity, as we do through the equivalization of household energy consumption, is uncommon. In the case of energy poverty, it is often ignored or assumed to be the same as income equivalence (Herrero, 2017; Hills, 2011). For example, Legendre and Ricci (2015) apply the OECD-modified income equivalence scales to adjust household income for the energy poverty ratio calculation, but there is no equivalization of energy expenditure. (2)

    As acknowledged by Hills (2011), the OECD-modified scales do not reflect how energy requirements vary between households. Instead, Hills (2012) proposes an alternative underpinned by three years of the English Housing Survey. (3) Hills' equivalization factors are calculated from ratios between median required household energy expenditure within different household groups and median required energy expenditure in two-adult households. Heindl (2015) estimates German energy poverty rates using this median-to-median ratio scale; however, the scales are used to equivalize actual energy expenditure rather than required, which, even in a highly developed country like Germany are not observed. Even though actual expenditure is not expected to capture need, these ratio-to-ratio scales, which represent an adjustment factor, offer one convenient way to deal with heterogeneity.

    Conceptually, our approach is most similar to Hills's (2012) equivalization factors, although we offer further generalizations and improvements. In most developing countries and even in many developed ones, clear estimates of energy need are not available; thus...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT