What leads individuals to adopt more sustainable patterns of consumption? An understanding of the factors that shape environmental consumption can inform policy interventions designed to promote more desirable environmental behaviors, such as increased support for environmental protection or increased investments in energy-efficient technologies. In this paper, we propose an agent-based model of the adoption of a particular, high-cost environmental technology—residential solar photovoltaic (PV) systems—and we use this model to explain why solar installers (firms that sell or lease solar systems) are incentivized to pursue marketing strategies that, paradoxically, slow the overall adoption of solar PV. This occurs because adoption behaviors are strongly determined by social influence and social learning, where individuals tend to adopt the beliefs and behaviors of others they are connected to in a social network. At the same time, however, social networks often exhibit network segregation, where connections tend to exist between individuals who share the same or similar underlying attributes that also explain solar adoption—such as education, income, nationality, or gender. Network segregation patterns provide a way for installers to quickly reach out to individuals with a high propensity to invest in solar, often through referral programs or geographically focused marketing efforts. Over time, however, this causes certain market segments to be systematically overlooked by solar installers.

These dynamics are at odds with the goals of prominent federal programs (such as the U.S. Department of Energy’s SunShot initiative) that seek practical strategies to speed overall rates of solar PV adoption. We show how this agent-based model can be used to suggest levers for governmental intervention. Depending on levels of network segregation observed in the system, it is possible to predict when governmental interventions will be most useful for promoting a sustained increase in solar PV adoption. These dynamics—including predictions about the effectiveness of government programs meant to speed overall solar adoption trends—are illustrated in empirical application of our model to approximately 50,000 solar installations in southern California.

Overall, this research helps to build emerging theories of social learning and network segregation—understanding these processes is critically important as we seek practical strategies that may be used to enhance pro-environmental behaviors and move towards a sustainability transition. This research illustrates how agent-based modeling approaches may be used to test fundamental theories of environmental behavior that are often not feasible to test using traditional methods of statistical analysis. Moreover, we illustrate how agent-based modeling frameworks may be used to evaluate and experiment with policy interventions in a virtual space, which enhances the scientific basis of environmental policymaking.