Utility Customer Supply of Demand Response Capacity.

• Author: Stewart, James I.

1. INTRODUCTION

   Despite technological efficiency improvements, large public energy-efficiency investments, and the adoption of energy-efficiency building codes and appliance standards, peak demand for utility-supplied electricity in the United States continues to grow. Figure 1 shows a commonly-used indicator of peak demand growth, the ratio of annual peak demand to average demand for utility-supplied electricity. This ratio increased significantly between 2006 and 2016. Income growth, increasing penetration of central air conditioning, and warmer summer temperatures have all contributed to higher peak demand. (1) Peak demand is costly to serve because power producers must build and maintain peak generating capacity, but this capacity remains underutilized for most hours of the year.

   An alternative to building peak generating units is to encourage utility customers to consume less energy on peak. Demand response is a utility's management of customer demand to address the high cost of supplying electricity during periods of peak demand. (2) Demand response results from an agreement between the utility and a customer to reduce electricity demand or to pay a higher retail price for electricity when the utility is short of supply. Demand response events, which typically last from one to five hours, can involve a temporary increase in electricity rates; the dispatch of alerts, notifications, or other signals to customers to reduce demand; or the activation of utility controls of customer appliances, equipment, or energy management systems. Economists often distinguish price-based demand response programs, which require participants to pay a higher price for peak electricity, from incentive-based programs, which pay utility customers for reducing their peak consumption relative to a fixed baseline (Bo[beta]mann and Eser, 2015). Examples of price-based utility demand response include critical peak pricing and variable peak pricing, while examples of incentive-based demand response include air conditioner or water heater direct load control, commercial and industrial load curtailment programs, and behavioral demand response. Since 2010, U.S. utilities have invested $8.6 billion in demand response programs. (3)

   While more utilities are investing in demand response programs, customer participation remains voluntary. A utility's demand response capacity depends on the willingness of its customers to reduce or shift their electricity demand during hours when overall demand for utility-supplied electricity is greatest. Utilities can increase the supply of demand response capacity by offering more generous incentives to customers.

   This paper estimates the long-run price elasticity of supply of demand response capacity to U.S. electric utilities from residential, commercial, and industrial customers. I analyze panel data on utility incentive-based demand response programs between 2010 and 2016 from the Energy Information Administration (EIA) Form 861, which collects information about utility peak demand response capacity and utility payments to customers for supplying such capacity. The supply of demand response capacity represents an equilibrium between the buyer (the utility in this case) and sellers (individual utility customers or demand response aggregators). However, since, in general, utilities with high demand for peak capacity will be willing to pay customers more for demand response, the equilibrium price and the supply of peak capacity are endogenous. This means that ordinary least squares (OLS) estimation of the supply relationship may result in biased elasticity estimates. Accordingly, this study estimates the supply elasticities by instrumental variables two-stage least squares (IV-2SLS), using variation in utility generation capacity costs as an instrumental variable for the utility's capacity payments to customers.

   The results indicate that the supply of demand response capacity by U.S. utility customers was price inelastic. The IV-2SLS estimate of the elasticity of supply of demand response capacity between 2010 and 2016 equaled 0.51 for all utility customers, 0.51 for residential customers, 0.58 for commercial customers, and 0.41 for industrial customers. The first-stage estimates support the IV-2SLS model assumptions, showing that capacity payments to utility customers were strongly correlated with utility costs for acquiring peak generation capacity. Moreover, OLS estimation produces smaller estimated supply elasticities, consistent with the hypothesized endogeneity between utility demand response capacity and customer incentive payments. The IV-2SLS elasticity estimates are robust to alternative model specifications and analysis sample selection criteria.

   A main contribution of this research is to analyze the supply of demand response capacity through incentive-based programs, which constitute the greatest source of capacity for electric utilities. Most economic research has focused on price-based demand response, through which utility customers reduce their peak electricity demand in response to price signals (Wolak, 2011; Faruqui, Sergici, and Warner, 2017). Researchers have paid less attention to the supply of peak capacity from incentive-based demand response programs, despite the greater prevalence of and larger utility investment in such programs. (4) By analyzing incentive-based demand response capacity, this analysis provides new evidence about the willingness of utility customers to supply demand response capacity.

   A second contribution of this research is to estimate long-run demand response supply elasticities that measure changes along the extensive and intensive margins of supply. When a utility increases (decreases) the price it pays for demand response capacity, customers can respond in two ways. The first is to enter (exit) the capacity market. The second is for existing suppliers to increase (decrease) the capacity supplied. In general, previous studies of individual demand response programs have only estimated short-run supply elasticities, how much program participants reduced electricity demand in response to an increase in peak prices during one year. (5) These studies do not observe customer entry and exit in response to changes in capacity payments. This is a limitation of even highly credible randomized experiment evaluations of demand response programs (Fowlie et al., 2017; Gillan, 2017; Ito, Ida, and Tanaka, 2018). Furthermore, while studies of electricity markets have demonstrated that demand response can benefit consumers by lowering wholesale and retail electricity prices without increasing carbon emissions (Smith and Brown, 2015; Eryilmaz, Smith, and Homans, 2017; Dahlke and Prorok, 2019), the analysis in most such studies has been limited to a single year. As a result, it is not possible to infer long-run changes in the supply of demand response capacity. However, with data on demand response capacity and payments for a large number of utilities over multiple years, this study can estimate long-run responses.

   A third contribution of this study is to demonstrate that utility customers are heterogeneous in their supply of demand response capacity. Electricity demand studies document heterogeneous customer responses to retail electricity prices (Reiss and White, 2005; Ito, 2014; Ito, 2015; Alberini, Gans, and Towe, 2016; Allcott and Kessler, 2019). This study adds to that literature by demonstrating that the long-run supply of demand response capacity is also heterogeneous. Using data from the U.S. Census Bureau and variation between utilities in customer demographics and housing characteristics, this study estimates the impacts of customer characteristics (such as home heating fuel type, customer education, and customer income) on the supply of demand response capacity. The results show that using electricity for home space heating, having a college-level education, and residing in an urban area were associated with higher supply elasticities, revealing that the long-run supply of demand response capacity varies between utility customers when they can enter or exit the market.

   These results should be of interest to utilities and policymakers alike. Utility demand for peak capacity is likely to grow, and demand response will often be the lowest-cost option for meeting this demand (FERC, 2009). However, utilities will only be able to acquire more demand response capacity to the extent that utility customers are willing to supply it. This study provides utility resource planners and program administrators with new insights about the long-run customer supply of demand response capacity. Utility planners and program administrators who ignore supply from incentive-based programs or extensive margin changes may underestimate the availability of demand response capacity. Furthermore, as the results of this study suggest, utility program administrators may be able to improve the cost-effectiveness of their demand response programs by enrolling customers with higher expected supply elasticities. Targeting such customers would reduce the utility's costs and allow for demand response capacity to compete more effectively with generation resources in utility resource planning.

2. MARKETS FOR DEMAND RESPONSE CAPACITY

   Utility customers supply demand response capacity to utilities through utility programs or through third-party aggregators who buy from customers and then resell the capacity to utilities. (6) When a customer participates in a utility demand response program, the customer cedes the right to electricity it would normally consume on peak at the standard rate. A customer's willingness to participate depends on the price the utility offers for capacity, the likelihood that the utility will exercise its option, and the cost of foregoing peak consumption.

   Between 2010 and 2016, an average of 376 U.S. utilities possessed demand response capacity. These constituted only...