Time-of-Use Electricity Pricing and Residential Low-carbon Energy Technology Adoption.

• Date: 01 May 2020
• Author: Liang, Jing

1. INTRODUCTION

   Energy efficiency and solar energy are two measures promoted by policy makers to reduce residential fossil fuel energy consumption and the associated greenhouse gas emissions. Not surprisingly, various policies and financial incentives (e.g., tax credits, direct rebates, etc.) exist to encourage the adoption of these technologies. For example, the cost of typical financial incentives (including direct rebates and tax credits) for the adoption of a solar panel system is on the magnitude of $5,493-$9,156 (Solar Energy Industries Association, 2014; Hughes and Podolefsky, 2015; Gillingham and Tsvetanov, 2019). However, despite these costly policy instruments, the penetration of energy efficiency and solar energy is still relatively low. Many organizational, behavioral, and market factors have been analyzed in the existing literature to explain the low adoption level. Yet, the impact of one particular factor (electricity rate structure) on energy efficiency investment and solar panel adoption is often overlooked in empirical studies (Novan and Smith, 2018). In this paper, we show empirically that consumers facing Time-of-use pricing (TOU) are positively correlated with the adoption of solar energy, compared to consumers on non-dynamic pricing plans. Our results have important implications for policy makers to promote the adoption of solar panels and TOU pricing.

   TOU, one of the most widely adopted dynamic pricing programs, charges different electricity prices depending on the time of the day, i.e. higher prices during peak hours (e.g. late afternoon in summer months) and lower prices during non-peak hours. TOU plan provides benefits to the utilities because it helps decrease peak load, which has a higher marginal cost of electricity supply compared to that of the base-load. In addition, reducing peak load helps utilities maintain the grid stability through the reduced likelihood of blackouts during peak hours. TOU can also potentially help the consumers save on energy bills if they switch part of their usage from peak to off-peak hours. This study focuses on another potential positive welfare impact of TOU--its correlation with low-carbon technology adoption, i.e. energy efficiency and solar panel installation.

   Figure 1 shows how the price and hour-of-day relationship of typical TOU price plans in Arizona (our study area) corresponds to the timing of electricity savings from solar panels and energy efficiency. The hourly savings from energy efficiency is obtained by recovering the data from Boomhower and Davis (2019), and hourly solar panel electricity generation is obtained by converting hourly solar data from the typical meteorological year (TMY2) dataset using the PVWATTS model (Ong et al., 2010). The figure shows that a significant portion of energy savings happen during peak hours when electricity prices are high. Naturally, this correlation might incentivize consumers to adopt energy efficiency and solar panels if they are on TOU pricing. However, there is little empirical analysis quantifying the correlation between TOU and the adoption of these technologies. This study provides the first empirical evidence of such correlation and fills the gap in existing literature along three dimensions.

   First, many studies have shown that the penetration of energy efficiency and solar panels falls short of optimal levels, which is widely referred to as "energy efficiency gap" (Jaffe and Stavins, 1994). Energy efficiency gap is attributed to various organizational, behavioral and market factors (Hirst and Brown, 1990; Weber, 1997; Gillingham et al., 2009; Gillingham and Palmer, 2014; Qiu et al., 2014; Qiu et al., 2017a), such as inefficient pricing of electricity (Gillingham et al., 2009), lack of information (Ramos et al., 2015), and the principal-agent problems (Davis, 2011; Gillingham et al., 2012). Meanwhile, the low adoption of solar energy is also attributed to a range of technical, financial and institutional barriers (Margolis and Zuboy, 2006; Timilsina et al., 2012; Zhang et al., 2012), including high initial cost, technology risk and complexity (Drury et al., 2012), information barriers during information-search process (Rai et al., 2016) and a lack of incentives. However, rate design is often a factor missed in existing empirical studies (Novan and Smith, 2018). This study contributes to this strand of studies by exploring empirically whether rate design is correlated with solar panel and energy efficiency adoption.

   Second, there have been many studies focusing on the impacts of TOU rates on energy consumption behaviors and the resulted change in social welfare. Some studies find that consumers shift peak load consumption to off-peak hours (Faruqui and Sergici, 2010; Qiu, et al., 2017a) while others do not find such load shifting behavior (Torriti, 2012; Faruqui et al., 2014). The load shifting behavior could be a result of technology adoption (e.g., demand-side management technology and renewable energy technology), and/or purely shifting energy consuming activities such as watching TV or washing clothes from peak to off-peak hours. This study contributes further to studying the impact of TOU on energy consumption behaviors by examining whether TOU is correlated with energy technology adoption. The adoption of energy efficiency and solar panels can serve as one underlying explanation for the observed load shifting behaviors in existing studies.

   Third, despite simulation or systems type of modeling on the impact of rate design on solar panel adoption, there is a lack of empirical evidence for such impacts. Existing simulation studies show that solar adoption should be sensitive to the rate structure (Darghouth et al., 2011; Ong et al., 2012; McLaren et al., 2015; Darghouth et al., 2016). Two seminal empirical studies support that a relationship exists between rate design and adoption of energy efficiency or solar PV Borenstein (2007 & 2017) show that tariff design provides indirect economic incentives for solar adoption. Specifically, Borenstein (2017) illustrates that the incentive from a tiered tariff is as much as the 30% federal tax credit in California. The calculation also indicates that the lifetime savings could be $7000 more under a tired tariff (increasing block rate) than a flat rate structure. Our empirical results of the correlation between TOU and solar adoption can help verify the simulation studies and further assist policymakers in choosing the appropriate rate designs that better reflect the social cost of providing electricity and potentially encourage the adoption of energy efficiency or solar panels (Ong et al., 2010).

   We compare adoption decisions in energy efficient appliances and solar panels between consumers on non-dynamic rates (marginal electricity prices are constant throughout the day) and those on TOU rates. We use household-level data in Phoenix, Arizona from an appliance saturation survey of 16,035 customers conducted by a major electric utility in 2014 for empirical verification. Probit model and statistical matching methods are employed, and robustness checks are conducted using multinomial logit model, bi-variate probit model, and machine learning matching method.

   We do not claim that our current finding of the correlation between TOU and technology adoption is causal, although we take steps to try to eliminate confounding factors and endogeneity issues for causal identification. There are two potential threats to causal identification: reverse causality and selection bias. Reverse causality could happen if households first adopt solar panels and then switch to TOU pricing. In our customer level dataset, for all solar customers, only 7 solar customers (less than 1.4% solar customers) switched to TOU after they adopted solar panels. We dropped these 7 solar customers in order to help avoid reverse causality. Also on average solar customers adopted solar panels several years after they enrolled in TOU pricing. In terms of selection bias, since TOU is not mandatory, it is possible that some consumers are more likely to enroll in TOU compared to others while these households are also more likely to adopt energy efficiency and solar panels. If these households have specific characteristics that are not observable to us such as environmental awareness and knowledge on energy usage, a potential self-selection bias exists. We apply a matching approach and include a rich set of covariates to help deal with such selection bias. For a customer that is on TOU pricing, we find a control customer that is similar in terms of home and socio-economic characteristics and that is not on TOU pricing. In addition, we use the adoption of programmable thermostat as a proxy for environmental awareness.

   Our empirical evidence suggests that TOU consumers are associated with a 27% higher likelihood to install solar panels, but not more likely to adopt energy efficient AC. Despite our efforts in overcoming the threats to causal identification, due to limitations on non-experimental cross-sectional data, there could still be remaining issues such as other omitted variables that could affect both TOU enrollment and technology adoption. However, even if our empirical finding of the correlation between TOU and solar adoption is not fully causal, quantifying such correlation is still valuable to policy makers. As discussed earlier, both TOU and solar adoption themselves could improve social welfare. TOU is found to enhance social welfare through aligning marginal electricity prices with marginal costs of electricity supply (Qiu et al., 2018; Train and Mehrez,1994). A positive correlation between these two adoptions after controlling for other types of confounding factors implies that if policy makers could encourage these two adoptions together either through informational/educational programs or financial incentives, then consumers could have a higher likelihood of enrolling...