Modeling residential adoption of solar photovoltaic systems

The world is on an urgent transition to renewable energies. Photovoltaic (PV) solar energy is the most viable green energy source to be produced at the domestic level, allowing every individual to contribute. Understanding the factors that influence the adoption of domestic solar energy, how it changes throughout the country and how spatial dependent factors contribute to the promotion of this technology is of the utmost importance to stimulate adoption. As to this day, to the best of my knowledge, these are not yet known. This study aims to contribute to channeling efforts to where adoption is more likely, ultimately accelerating Portugal’s energy transition. Hence, the goal of this study is to build a spatial model that estimates for each spatial unit in Portugal the probability of individuals adopting domestic PV systems. The study uses data related to past solar PV installations as well as socioeconomic and demographic data from public sources. An exploratory spatial analysis including the study of spatial correlation across municipalities confirmed the importance of spatial considerations. Three dependent variables were considered sequentially: installations (binary), number of panels installed (discrete), and installed power (continuous). To model the latter, it being the main focus of the study, eight models were compared: linear regression (OLS), spatial lag (SAR), spatial error (SEM), Kelejian-Prucha (GSM), spatial lag of the explanatory variables (SLX), spatial Durbin (SDM), spatial Durbin error (SDEM), and Manski models. It was concluded that socioeconomic factors do spill over to neighbor locations and in that way influence solar PV adoption, but also that unobserved characteristics result in similar decisions in nearby municipalities. The SDEM was found to be best to fit the data and a final map representing the likelihood of adoption across the different municipalities in Portugal was produced according to its estimations.