Information Searching in the Residential Solar PV Market.

• Author: Pless, Jacquelyn

1. INTRODUCTION

   The market for solar photovoltaic (PV) systems has experienced tremendous growth over the past decade, with installed capacity in the U.S. expanding from less than 500 MW in 2008 to more than 40 GW in 2016 (SEIA 2017). (1) This growth can be attributed to steep capital cost declines, government financial incentives, reduced technology uncertainty, and market maturation more generally. Some of it also can be attributed to the availability and use of third party ownership (TPO) financing options (Drury et al., 2012; Corfee et al., 2014). (2) The TPO model, which is similar to leasing in other markets (e.g., automobiles), offers customers the option of paying a third-party owner for using a solar system by either signing a lease or a power purchase agreement (PPA) with little or no money down. Increasing in use from just 10% to 20% of solar PV installations in 2009 to roughly 65% in 2013 (GTM Research, 2014), TPO allows customers to overcome some of the key barriers to solar PV adoption by reducing or eliminating upfront costs, technology risk, and the installation process complexity (Margolis and Zuboy, 2006). It also enables the monetization of tax-based incentives, lowering costs to the customer, and provides a source of financing in an early market. On the other hand, host-ownership (HO)--or buying the system outright--requires homeowners to purchase solar panels directly, incur technology risk, and bear installation process complexity, though HO customers also often enjoy larger financial benefits over the system lifetime.

   Innovative approaches to financing technology adoption such as TPO can open up markets to new customer bases and remove barriers to entry (Rai and Sigrin, 2013; Drury et al., 2012; Margolis and Zuboy, 2006). However, while TPO use has increased substantially in some residential solar PV markets, solar PV customer acquisition costs still remain high (in both TPO and HO markets), potentially dampening future technology diffusion. They could be reduced with a better understanding of solar PV adopter preferences. In particular, if consumer preferences for TPO versus HO among solar adopters are correlated with other observed consumption patterns, the amount of time between a customer's initial interest in solar and actual adoption can be improved with customer segmentation and targeted marketing strategies. This can reduce the resources that firms allocate to customer acquisition, potentially helping to accelerate technology diffusion and improving firm competitiveness.

   In this paper, we explore heterogeneity in the preferences of households that adopt solar PV as measured by their information searching behavior. Our objective is not to estimate the causal relationship between information search and the solar PV financing decision, as we do not have an adequate identification strategy or the necessary data for addressing endogeneity concerns such as selection bias, nor can we fully model the decision-making process with our data. Rather, by exploring the correlation between information search and the financing decision of solar PV adopters, we recover information on consumer preferences conditional on adoption, controlling for many other factors that could influence the decision.

   We examine households that installed solar systems in San Diego County between 2010 and the first quarter of 2013, focusing on differences in the research conducted prior to adoption for those that opted for TPO versus HO. We estimate probit and bivariate probit models to examine the relationship between information search and the financing decision of solar adopters. This requires merging multiple proprietary household-level datasets as well as additional public system-level data. We use information from two surveys of San Diego households that were fielded during 2014 that elicited new data exploring factors that determine a household's decision to adopt solar PV, such as their motivations, potential barriers, how they accessed information, and the type of information they sought. Second, we use data on actual TPO contract terms to determine the 'price' that TPO households paid for such systems over their contract lifetimes, as publicly reported TPO pricing data is known to be inconsistent across installers. We also match these datasets to two other public datasets for additional information such as the system size and market concentration.

   Our main results suggest that solar PV households using HO versus TPO seek different types of information throughout their decision-making process. Solar PV households using TPO spend more time researching the required home modifications associated with installing solar while solar PV households using HO spend more time researching expected financial returns. These correlated preferences indicate that information search is heterogeneous for solar PV households that use different financing options for adoption. We also explore how other household and market characteristics are correlated with whether solar PV households use TPO or HO. Interestingly, we find no clear demographic differences between solar PV adopters that use TPO versus HO, although HO customers are more likely to live in slightly larger homes and face slightly higher utility bills.

   This paper makes three main contributions. First, understanding differences in the types of information sought by solar PV households that use alternative financing models can help guide solar companies' marketing strategies, which in turn can reduce customer acquisition costs and accelerate technology diffusion. Our results can be interpreted as a reflection of heterogeneous conditional consumer preferences. On the one hand, TPO customers may place a higher value on reducing home modification hassle related to technology whereas HO customers may be more concerned with long-term investment returns. If these preferences are also correlated with other household consumption patterns, solar companies may be able to identify which households are more likely to respond to marketing materials targeted towards TPO versus HO solar financing options given the consumption of other goods and services. For example, households that value reducing hassle associated with home ownership also may be more likely to hire external assistance to handle other household responsibilities, such as certain home renovations, maintenance, or lawn care. They also may be more likely to lease a vehicle. Identifying these types of households in advance by observing consumption of these other goods and services may help reduce customer acquisition efforts if a firm is aiming to market the TPO option. Similarly, by observing the financing decision of solar adopters, non-solar firms can target marketing of other goods and services to these households more effectively.

   Second, this paper contributes to a growing literature aiming to understand differences in TPO and HO markets in technology adoption. Sigrin et al. (2015) use the same data that we use to observe descriptive differences between TPO and HO customers in the motivations for initially adopting solar PV. Rai and Robinson (2013) compare the length of time of the solar adoption decision-making process between TPO and HO customers. Drury et al. (2012) conduct a correlation analysis and find that solar PV households using TPO in southern California tend to be younger, less affluent, and less educated populations relative to those using HO. On the other hand, Rai and Sigrin (2013) consider the solar PV financing decision in Texas and find that, although TPO seems to have opened up the market to those with a tight cash-flow situation, TPO and HO customers do not differ on socio-demographic variables. Overall, this literature has not yet examined differences in information searching behavior. Doing so can be useful for reducing customer acquisition costs, which remain quite high in the U.S.

   Third, this paper complements the growing body of work exploring the demand for solar PV by focusing on the subsequent decision of how to finance the asset. While we are not able to estimate demand with our data, a better understanding of the financing decision can have implications for demand. Several papers directly examine solar PV demand in residential markets. Bollinger and Gillingham (2012) study peer effects and demonstrate the impact of previous nearby adoptions on PV uptake in California. Graziano and Gillingham (2015) examine the diffusion of residential solar PV systems using installation data from Connecticut to identify spatial patterns of diffusion and clustering of adoptions, finding a strong relationship between adoption and the number of nearby previously installed systems as well as policy variables and the built environment. Richter (2013) asks whether installation rates of solar PV are affected by social spillovers and finds a small but statistically significant neighbor effect in PV system adoption in the United Kingdom. Hughes and Podolefsky (2015) study the impact of subsidies on solar installations, finding that the CSI rebate program has had a large effect on adoptions. Lastly, Gillingham and Tsvetanov (2019) estimate price elasticity of demand for solar PV. None of these studies explore differences between solar PV households that use different technology adoption financing options, however.

   The remainder of this paper is organized as follows. Section 2 describes the factors that influence a household's decision to use TPO or HO when adopting solar PV. Section 3 presents our empirical strategy and Section 4 describes our data and variable construction. Section 5 summarizes our main empirical results, Section 6 demonstrates that our findings are stable across several robustness checks and alternative specifications, and we conclude in Section 7.

2. FINANCING SOLAR PV WITH TPO VERSUS HO

   Abstracting from the details of market structure and demand, the empirical approach we take in...