Energy transitions from individuals or aggregates? How consumer data sources shape agent-based simulations in the United States

As electricity systems evolve, accurately modeling consumer behavior is crucial for policy design and system planning. This study examines how different approaches to initializing consumer agents in electricity market simulations impact sustainability outcomes. We compare three strategies: (1) aggregate public data distributions, (2) aggregate survey data distributions, and (3) individual-level survey data from 839 respondents. Using New Jersey's electricity market as a case study, we simulate household decisions on solar investments, clean-energy participation, and consumption over 40 years (2010–2050) with an agent-based model, running 500 Monte Carlo simulations per approach, validated against 2010–2020 historical data. Results reveal important trade-offs between modeling approaches. Aggregate public data models most accurately track historical consumption and energy burden, while survey-based models, particularly individual-level, predict higher renewable adoption and program participation rates. The individual survey methodology captures greater behavioral heterogeneity and socioeconomic disparities, revealing potential energy justice concerns that remain hidden in aggregate models. Despite these differences, all approaches maintain comparable accuracy in predicting system-level metrics like total electricity consumption. These findings demonstrate that modeling outcomes are very sensitive to initialization highlighting the importance of aligning model design with the intended research question and available data.