Understanding Hourly Electricity Demand: Implications for Load, Welfare and Emissions.

• Author: Karimu, Amin

1. INTRODUCTION

   Electricity markets are undergoing significant changes, driven largely by two developments: sizable increases in renewable electricity generation and technical development in smart grids. These developments, which are likely to accelerate in the near future for many countries including Sweden (Energimarknadsinspektionen, 2010, 2014), represent both opportunities and challenges to the electricity system. In the resulting new electric grid, the roles of supply and demand are likely to be reconfigured, with active involvement of final demand in wholesale and retail markets viewed as important enabling factor for assisting the scale-up of intermittent generation capacity (Wolak, 2017). In light of these developments, there is an increasing interest in understanding the within day residential demand for electricity, particularly since opportunities for influencing residential demand at the hourly level appear higher than before. Yet, not enough is currently known regarding how individual households consume electricity at the hourly level, including aspects such as the factors driving consumption at these time-scales, and how these consumption patterns are likely to influence hourly retail pricing.

   The present study explores these aspects by allowing utility to depend explicitly upon electricity consumption across hours within a day. Further, and uniquely, hourly demand is also allowed to depend upon within-hour end-use specific demand. Consequently, our analysis accounts for the substantial preference (and demographic) heterogeneity documented in the energy literature (Smale et al., 2017; Strengers, 2012; Yohanis et al., 2008). As discussed in prior studies (e.g. Simshauser and Downer, 2016), households' load curves vary far more than the underlying price variation that households face (which is often minimal). This aspect, true across many household demand contexts, suggests the possibility that household-specific characteristics and heterogeneity in appliance-utilization may in fact act as more than the demand-shifters commonly assumed. A related aspect is the possibility that the welfare implications of prices differ across appliances and hours, i.e., that a given price has different welfare implications across different hours. The demand framework used here explicitly accounts for both aspects by constructing household-specific price indices, exploiting the variation in composition of expenditure within hours.

   This framework is used, together with a unique dataset on appliance-level electricity usage for Swedish households facing monthly retail pricing, to answer the following two questions: (i) what is the response of the household load profile to a shift to hourly retail pricing? and (ii) what are the welfare and carbon emission implications of this shift? The contributions of this paper are twofold: First, we provide a consistent econometric framework for understanding how household response to hourly retail pricing is affected by the characteristics of its electricity demand, allowing us to evaluate welfare and carbon emission implications of different price profiles. Second, we outline a framework in which household-specific hourly information, including end-use consumption data, can be used to drive the hourly demand model. We emphasize that in the framework we develop, the effect of hourly pricing on load and emissions are based upon substitution parameters estimated using hourly aggregate and end-use demand, in contrast to much of the simulation-based prior literature that assumed price responsiveness (e.g., Borenstein and Holland, 2003 and Savolainen and Svento, 2012). While our application is to the Swedish context, the framework developed here can be fruitfully applied across a broad range of dynamic pricing contexts with detailed household data but little price variation. These scenarios are increasingly likely in view of smart metering, leading to hourly or even sub-hourly data availability, being more common than dynamic pricing, see e.g. Simshauser and Downer, 2016.

   For a representative sample of Swedish households faced with plausible hourly pricing scenarios, our main findings are that households reduce demand during peak periods and increase demand during the off-peak. These effects however are rather small, at less than three and two percent respectively. Welfare, measured by the cost of living, is slightly reduced, with some heterogeneity across household types. Finally, we find a very small reduction in emissions consequent to the hypothetical hourly-retail pricing scenarios we consider.

   The rest of the paper is structured as follows: the introduction proceeds by briefly motivating, in Section 1.1, the necessity of analysing high-frequency electricity consumption patterns in light of the literature on dynamic pricing. The broad ideas regarding a demand system that will flexibly encapsulate the ideas sketched out here are described in Section 2 while the data used to estimate the demand systems is described in Section 3. The implications for load shifts, welfare and emissions resulting from two hourly pricing scenarios are discussed in Section 4. Section 5 concludes with a discussion and some policy implications.

   1.1 Motivation

   Swedish policy makers have focused on the salience of hourly pricing and other time-varying pricing schemes far more than in other setting: to our knowledge, Sweden is one of few countries wherein households across the country have the right to ask for, and receive, an hourly pricing-based contract from their retailer, without having to pay for the metering and associated costs (source: [section]3 Chap 11 in the the Swedish electricity act). The rationale offered by Swedish policymakers and regulators for hourly pricing (Energimarknadsinspektionen, 2010, 2016) differs to some extent from that discussed in the largely U.S.-based setting of the prior literature, and is related to the challenges posed by the substantial increases in intermittent generation anticipated in the Swedish context (see Vesterberg and Krishnamurthy, 2016). To illustrate, Swedish climate and energy policy stipulates, among other things, 100 percent renewable generation by 2045. Hourly pricing is seen as an important tool for efficiently accommodating these developments by incentivizing households to adjust their usage to the availability of electricity. This thrust on hourly pricing by policy makers is not, however, reflected in consumer eagerness to adopt these types of contracts, with less than one percent of Swedish household actually on these contracts (Energimarknadsinspektionen, 2014). (1)

   The by-now large literature regarding hourly pricing largely reports sizeable welfare gains (see, e.g., Borenstein, 2005, 2013; Savolainen and Svento, 2012), although recent literature (Gam-bardella et al., 2019) questions some of these benefits (see also Leautier, 2014). These welfare gains are often presumed upon an individual households' willingness to shift load across time, from high-price hours to low-price hours. However, empirical evidence for such substitution is rather scare, and the findings to-date are mixed. For example, Filippini (1995) uses data from Switzerland and finds that demand for peak and off-peak electricity is elastic, and that cross-price effects are positive. Faruqui and Sergici (2011) provide econometric results from the US and report negative substitution elasticity. Allcott (2011) provides evidence that hourly pricing leads to peak conservation but no increase in off-peak demand, suggesting that there is little substitution between peak and off-peak. (2) Furthermore, Allcott (2011) also illustrates how welfare gains at the household level may be small. (3)

   An ancillary benefit of RTP may well involve emission reduction. Yet, as pointed out in previous literature, the implications of hourly retail pricing for emissions depends crucially on the production mix: with little emission in the first place, emission reductions will be small. Furthermore, if emissions per kWh are constant across hours, then only load reductions will have any effects on emissions. Holland and Mansur (2008) explore the effect of load shifting on emissions in the U.S., and find a reduction in emissions in regions where peak demand is met largely by oil-fired capacity, but an increase in emissions in regions that have peak demand met largely by hydro-power (although these effects are small). Similarly, Zivin et al. (2014) also find small increases in emissions from hourly pricing and load shifting in the US. In particular, they find that increases in off-peak consumption may counteract reduced peak demand and peak emissions. It is worth noting that these studies investigate the effects of load shifting only, holding total load constant. Allcott (2011), allowing for effects from both load shifting and energy conservation, finds emission reductions of four percent from the introduction of hourly pricing in California using hourly household consumption data. To our knowledge, there are no previous studies quantifying the emission effects from the introduction of hourly pricing in the Nordic context using household data.

   Two aspects of the prior literature are worth noting. First, despite many studies suggesting at least some response to dynamic pricing, the results presented are not easily generalized to other contexts and settings. For example, weather and climate, as well as institutional settings and pricing structures, may differ substantially between countries. These differences may result in different usage patterns and restrictions on the substitutability of electricity demand. Furthermore, many of these prior studies use household-level data that is either aggregated over time (e.g. weekly/monthly) for many households or disaggregated over time but for household samples from small-scale experiments or pilot programs, where the participating group of households may...