1. Buy, Lease, or PPA? Drivers of the Mode of Consumer Energy Technology Adoption Varun Rai 1, D. Cale Reeves 1, Robert Margolis 2 1 The University of Texas at Austin 2 National Renewable Energy Laboratory 33 rd USAEE/IAEE North American Conference, Pittsburgh 27 October 2015

2. Acknowledgements This work was funded by the National Renewable Energy Laboratory (NREL) (Subcontract # XGG-3-23326-01). 2

3. 3 Information in Big Ticket Adoption Decisions Focus on BOTH Social and Economic Factors Information and the costs of acquiring it are central to understanding individual decision-making (Nelson, 1970). Behavioral factors (such as risk aversion, anchoring, decision heuristics, etc.) and social factors (such as norms, trust-based information networks, etc.) are of special interest because of the way they expose different individuals to different sets of information (Dietz, 2010; Kemp & Volpi, 2008; Stern, 1992), thereby giving rise to the bounded rationality of individual decision- makers (Wilson & Dowlatabadi, 2007). An important question (in disaggregated modeling) is: how and from whom do individual decision-makers seek the necessary information and form expectations about technological trajectories to aid their decision-making? We address this question in the context of residential solar PV adoption process.

4. 4 Household-level Heterogeneity Much of the current work in solar PV aggregates household-level heterogeneity in behavioral drivers and underlying information processes. Understanding household-level heterogeneity is important for: Estimating demand (Gillingham et al., 2014), Building detailed spatio-temporal models of adoption (Rai & Robinson 2015; Robinson & Rai, 2015; Zhang et al. 2014 ), and Identifying and addressing information gaps and other barriers faced by different customer segments (Rai & Beck, 2015). A couple of modeling examples...

5. 5 Source: Souyris, Rai, Duan, and Balakrishnan, In preparation, 2015. Key Inputs for Disaggregated Demand Models

6. Integrated Decision-Making Framework Based on Deep Data and Predictive Modeling 6 Household-level Data  Adopter and non-adopter  Surveys  Appraisal district rolls  Solar program data  Installer surveys Multi-method  Econometric analyses  Financial modeling  GIS integration  Agent-based modeling (ABM)  Dynamic discrete choice modeling Source: Rai & Robinson, Environmental Modelling & Software, 2015.

7. 7 Hardware costs have been declining rapidly, thereby spurring demand.

8. 8 Information in the Adoption of Residential Solar PV: Available Everywhere, but What to Trust? However, system cost is not the only barrier faced by solar adopters; potential adopters also face various informational barriers, leading to high indirect costs during the information search process, which can be quite intensive for a capital-intensive durable like a solar PV system. Umm?!

9. 9 Peer Effects in the Adoption of Residential Solar PV Bollinger & Gillingham (2012) [zip-code level data, CA; CSI data until 2011]: Peer effects increase the probability of adoption. Rai & Robinson (2013) [household-level data, TX; data until 2012]: Peer effects accelerate adoption by about 2 to 5 months (~ 30% - 60%). Other papers: Grazziano & Gillingham (2014); Noll, Dawes, & Rai (2014);... 12 Length of Decision Period (Months) 8.9 (Mean)6 (Median) $ Not Very Important Peer Effects Neighborhood Contact Lease 4.72.7 2.31.4 Adj. R 2 = 0.24P < 0.0001All IV p < 0.01 Source: Rai & Robinson (2013), Environmental Research Letters, 8(1), 014044.

10. 10 Drivers of the Growth of Third-Party Ownership? Source: GTM Research, 2015. Residential Third-Party Ownership (TPO) Penetration and Installations by Ownership Type Is TPO cannibalizing the conventional “bought” model? Are there any key differences in the preferences/attributes of buyers vs. leasees?

11. 11 Mode of Adoption: Buy vs. Lease What Barriers Does the Leasing Model Address? Using aggregate data, Drury et al. (2012) posit three characteristics that explain the increasing trend toward third-party ownership (TPO), which includes leasing and power purchase agreements (PPA), of residential solar PV in California: it removes some of the complexities and uncertainties associated with adopting a new technology; it reduces the upfront financial burden – both overnight costs and costs associated with securing financing; (found demographic differences too) it re-frames the financial benefits of adoption as a simple-to-perceive monthly savings. Rai & Sigrin (2013) find that...

12. Long-Term Focused Short-Term Focused Leasing is Preferred By Customers Who Have a Tighter Cash Flow Situation. Buyers Use a MEAN Discount Rate 8 – 14% Lower Than Leasers In the early phases in a market the TPO model is able to penetrate the “information ready but cash poor” market segments. Source: Rai, V., Sigrin, B. (2013). Diffusion of environmentally-friendly energy technologies: Buy vs. lease differences in residential PV markets. Environmental Research Letters. 8(1), 014022 (1-8).

13. 13 Research Questions Where do potential adopters get their information from during the research (i.e., information gathering) period, and especially what is the role of peer effects and installers (i.e., the supply side) in this process? Previous analyses use early CSI data before 2012 No household-level analysis exists for California What drives the decision to buy versus lease solar? No household-level analysis exists for California To our knowledge, no disaggregated study to date looks at the effects of O&M and financial factors on the mode of adoption

14. 14 Data We use a dataset composed of a survey of households in northern California that chose to adopt solar PV systems. Between April and June of 2014 the survey was sent to 2131 customers, of which a total of 380 completed responses were received (18% response rate). The survey collected information from respondents across eight broad categories: system details, purchase/leasing/power purchase agreement (PPA) details, decision-making process, financial aspects, sources of information, post- installation evaluation compared to prior expectations, environmental attitude, and demographics. These data are matched by customer to the solar program dataset, which contains system-level details of the individual PV systems such as nameplate capacity, date of interconnection, total system cost, and rebate received.

15. 15 Neighborhood-level peer effects play a significant role in providing valuable information to potential adopters of solar Of the 170 respondents that had pre-existing PV systems in their neighborhoods, 49 (29%) agreed or strongly agreed that, “Solar systems in the neighborhood motivated me to seriously consider installing one.” 22 of these 170 respondents (13%) further recognized the importance of neighbors as information pathways in their decision by agreeing or strongly agreeing that without PV systems in their neighborhood they would not have chosen to install. Thus peer effects impacted the decision-making of about 30% of the respondents (exposed to peer effects) and fully determined the behavior of 13% (who would not have adopted solar otherwise). Note: Peer influence on behavior is typically underreported (Nolan et al., 2008).

16. 16 Peer effects important, but may not be a necessary condition for adoption However, the majority (87%) of these 170 respondents disagree that they would not have installed solar PV systems had there been no previously installed PV systems in their neighborhood. Adopters who received door-to-door marketing, value information from PV-owning neighbors less. (Fisher’s exact test for independence) Direct marketing appears to mitigate the impact of peer effects. Overall, this suggests that peer effects translate into a higher motivation and confidence for non-adopters, but that peer effects may not be a necessary condition for adoption. Or, does the marginal utility of direct marketing increases in the presence of peer effects? (Open question)

17. 17 Determinants of Adoption: Buy or Lease? The relationship of technological uncertainty and financial returns with the decision to pursue third-party ownership (given that the respondent had third-party ownership options available) is explored in a logit model. Survey used to operationalize aspects of technological uncertainty (“concerns about O&M”) and financial returns (“importance of financial returns”), along with several other control variables.

18. 18 Determinants of Adoption: Buy or Lease? Overall, when controlling for neighbor contact, decision period, decision- making factors, spark events, and relevant demographics, importance of financial returns and concerns about O&M are significant determinants of mode of adoption. Ceteris paribus, respondents that regard O&M concerns as very important or extremely important, have about 40-50% higher odds of pursuing TPO. Ceteris paribus, respondents that value financial returns as very important or extremely important are about 25% less likely to pursue TPO. These findings describe a subset of adopters that are financially savvy and approach the decision to install solar PV largely as an investment vehicle, and find buying as the right mode to do so. This group appears to be evaluating the benefit-cost of solar over a longer time horizon than the 15-20 typical in TPO contracts. Thus this group appears to be using a lower (implied) discount rate.

19. Determinants of Spatio-Temporal Patterns in Adoption 19 Financial aspects of the solar-adoption decision performs well in predicting the rate and scale of adoption Accounting for agent-level attitude and social interactions are critical for predicting spatial and demographic patterns of adoption with high accuracy Fit to minimize cumulative Source: Rai & Robinson, Applied Energy, 2015.

20. Workforce Training and Reinforcement: 90% engagement / 90% accuracy Data Proof of knowledge and skill retention Reinforce critical information without taking time away from work Multiple interactions daily Residential EE & Solar Adoption Behavior: An Online Gamification Study

21. 21 Main Findings Disaggregate, household-level data show that peer effects are important in the adoption of solar PV in CA. However, Peer effects do not appear to be a necessary condition for adoption, Direct marketing appears to substitute for peer effects. Substitution or complementarity of peer effects and direct marketing needs further investigation. Different business models address important barriers associated with PV adoption differently and hence unlock different segments of potential adopters.

22. References Rai, V. and Beck, A. L. (2015). Public perceptions and information gaps in solar energy in Texas. Environmental Research Letters. 10, 074011 (1-9). Robinson, S. A. and Rai, V.. (2015). Determinants of spatio-temporal patterns of energy technology adoption: An agent-based modeling approach. Applied Energy. 151, 273-284. Rai, V. and Robinson, S. A. (2015). Agent-based modeling of energy technology adoption: Empirical integration of social, behavioral, economic, and environmental factors. Environmental Modelling & Software. 70, 163-177. Noll, D., Dawes, C. and Rai, V. “Solar community organizations and active peer effects in the adoption of residential PV,” Energy Policy, 67:330–343, 2014. Rai, V. and Robinson, S. A. Effective Information Channels for Reducing Costs of Environmentally-Friendly Technologies: Evidence from Residential PV Markets, Environmental Research Letters, 8(1), 014044(1-8), 2013. Blackburn, G., Magee, C., and Rai, V. Solar Valuation and the Modern Utility's Expansion into Distributed Generation, The Electricity Journal, 26(11), 18-32, 2014. Rai, V. and Sigrin, B. Diffusion of Environmentally-friendly Energy Technologies: Buy vs. Lease Differences in Residential PV Markets, Environmental Research Letters, 8(1), 014022 (1-8), 2013.

23. 23 Solar “Spark” Events: Beginning of the Decision Process Direct marketing by a solar company was the most popular spark event, followed closely by planning for retirement and a recent increase in electricity rates. These three spark events each operated on roughly 30% of the respondents. Comparatively, both of the two neighbor-related spark events “seeing a neighbor install solar panels” and “conversation with a neighbor who had solar panels” combined only sparked less than 15% of the respondents. However, looked at alone this metric underestimates the prevalence of neighbor- related spark events. One really needs to look at this metric conditional upon a respondent having access to a neighbor with solar PV already installed. Of the subset of respondents that report having at least one solar PV system in their neighborhood (170), 27 reported their interest was sparked by seeing a neighbor install a system (16%), 26 report that their interest started after a conversation with a neighbor (15%), and 12 (7%) report that their interest was sparked by both seeing a neighbor install a system and having a conversation with a neighbor. Thus, conditional upon access to neighborhood peer effects, the two neighbor-related spark events combined are nearly as common as direct marketing, retirement planning, and increasing electricity rates.

24. 24 Solar “Spark” Events: Beginning of the Decision Process

25. Strong peer effects observable as PV installation density increases Under certain conditions, peer effects also accelerate adoption decisions. Policy outcomes could vary significantly depending on these effects No PV Dense CommunitiesWith PV Dense Communities Word-of-Mouth/Peer Effects Source: Rai & Robinson (2013), “Effective information channels for reducing costs of environmentally-friendly technologies: evidence from residential PV markets,” Environmental Research Letters, 8(1), 014044 (2013).

26. SCO-driven active peer-effect process. The dotted arrows from the top layer represent the observation that SCOs can form organically, even without the input of resources from other entities. Solar Community Organizations (SCOs) Source: Noll, Daniel, Dawes, Colleen, Rai, Varun. (2014). Solar community organizations and active peer effects in the adoption of residential PV. Energy Policy. 67, 330-343.

27. Scenario Very Conservative ConservativeBaselineOptimistic Very Optimistic Electricity Cost Growth 0.0%/yr2.6%/yr 3.3%/yr5.0%/yr System Life 20 yrs 25 yrs System Loss Rate 0.75%/yr0.5%/yr 0.25%/yr Maintenance Costs 0.5% /yr0.25%0.25%/yr0.15%/yr0%/yr Inverter Replacement Cost $0.95/W $0.7/W None Electricity Plan After PV Adoption Keeps same REP and plan post-installation; no outflows Adopts solar plan if offered by current REP Adopts solar plan if offered by current REP; min. 7.5¢/kWh outflow Adopts plan with max. value among current market solar plans or BAU plan Same as Scenario 4 SCENARIO APPROACH TO PARAMETER UNCERTAINTY Scenarios Approach to Parameter Uncertainty

28. Buyers: Mean: $6.2/W SD: $1.4/W Leasees: Mean: $8.3/W SD: $0.53/W Buyers: Mean: $2.6/W SD: $0.95/W Leasees: Mean: $0.70/W SD: $0.13/W Installed Cost vs. Cost of Ownership Source: Rai & Sigrin, “Diffusion of environmentally-friendly energy technologies: buy versus lease differences in residential PV markets,” Environmental Research Letters, 8(1), 014022 (2013).

29. MEASUREMENT OF IMPLIED NET PRESENT VALUE (NPV) This consumer is indifferent to paying $3000 more for the system– the implied NPV!

30. Too Optimistic Too Pessimistic BUYERS ARE OPTIMISTIC IN PAYBACK PERIOD ESTIMATE BASELINE SCENARIO Buyers Are Optimistic In Payback Period Estimate

31. HOWEVER BUYING MAKES SENSE UNDER OPTIMISTIC ASSUMPTIONS VERY OPTIMISTIC SCENARIO However, Buying Makes More Sense Under Optimistic Assumptions

32. Perception and Attitude of Solar Non-Adopters Huge gap between attitude (5.31/7) and perceived affordability (3.15/7) of solar – Perception of ability to afford solar low (3.15 on a 7-pt Likert scale) – Solar perceived as expensive due to incomplete information about performance, leasing, and incentives Only 16% reported awareness of any incentives – Addressing info gap could open up large potential demand Customer awareness of the cost of solar has not caught up with – Available incentives and rebates, declining prices, and lease options that are quickly increasing the affordability of solar energy Descriptive norms significant for solar models – Respondents feel more knowledgeable and confident with energy conservation than with solar energy; thus, looking to others for information and/or confirmation has greater benefit – Consistent with recent literature: Bollinger & Gillingham (2012); Rai & Robinson (2013) 32The University of Texas at Austin

33. Experiment Overview Use initial survey to capture existing attitudes and intentions regarding energy conservation and solar, as well as other controls Create two randomized cohorts: – Control – Treatment (Gamified information) Employ trivia-style mobile gaming platform to succinctly deliver key information to the Game cohort Use final survey to capture changes in attitudes and intentions regarding energy conservation and solar, and perceived effectiveness of gamified platform 33 Q1 2015 Progress ReportDOE SunShot SEEDS: UT Austin

34. Game Platform: Ringorang® A clue gives a little hint for players new to energy topics A question conveys actionable or educational information An insight provides more context or information about the topic < 1min A “learn more” link to a web site for additional research or information on incentives A sliding scale for points based on how quickly you answer A leaderboard to compete with other players 34 Q1 2015 Progress ReportDOE SunShot SEEDS: UT Austin

35. Energy Games: Impact The interactive nature of a trivia game tests respondents perceived knowledge – More “aha” moments (vs. say a newsletter) with the gamified version  Higher perceived affordability Awareness of incentives significantly increased, which indicates that incentive programs may not be well publicized for passive audiences Likelihood of calling to request a solar quote increased following the game. This is one of the key factors to influence as it is a critical and necessary hurdle in the solar adoption process 35