Sentiment Analysis of Solar Energy in U.S. Cities: A 10-Year Analysis Using Transformer-Based Deep Learning

Abstract This study examines U.S. public sentiment toward solar energy from 2013 to 2022, analyzing 8 million social media posts using RoBERTa, a transformer-based deep learning algorithm. Sentiment has generally been positive but has declined since 2016 due to increasing negativity. Significant regional and state differences exist, with these disparities deepening over the decade, suggesting growing polarization in views toward solar energy. Negative sentiment is more pronounced in Republican-leaning and Southern U.S. regions. A two-way fixed-effects panel analysis at the municipality level shows positive sentiment in areas with high solar radiation and a large remote-working population, while negative sentiment is more common in regions with high wind speeds and significant multifamily housing. These findings underscore the complex interactions among environmental, socioeconomic, political, and technological factors, highlighting the need for tailored strategies to address regional and demographic disparities in solar energy sentiment. The computational approach developed and tested in this study, which leverages natural language processing and geospatial analysis of social media data, offers broader applicability and provides a scalable framework for analyzing sentiment across various regions and topics beyond the specific context of solar energy in the United States.