As residential solar gears up for another record breaking year, increasing attention has been paid to exactly what types of customers have been fueling this explosive growth. Given the difficulty of pinpointing exactly which houses have gone solar, previous demographic analyses of solar customers have been limited to broad zip code estimates and reliant on often incomplete state incentive databases. It also does not account for the often wide income gaps present in certain zips. The result has been a murky picture of exactly which customers are reaping the benefits of solar energy and which are being left in the dark.
A new study published by PowerScout and GTM Research, however, uses PowerScout’s state of the art image recognition technology to detect solar panels in satellite images of homes. Based on a Convolutional Neural Network (CNN) algorithm, this type of machine learning is behind the latest advances in self-driving cars, Facebook’s auto-tagging algorithms, and dozens of other computer vision applications. We used this algorithm to pinpoint where residential solar is being installed down to the household level. Next, we used robust consumer marketing databases to determine the income bracket for each solar home, and compared the distribution of solar installations across income brackets to the distribution of the general population.
While the findings of the PowerScout and GTM Research study remain largely consistent with previous research, there are unexpected differences that have implications for policymakers, installers, and investors.
Some of the key findings in the study are:
- Over 70% of solar households have annual income between $45,000 and $150,000, a range roughly aligned with “middle-income”. This is compared with 65% of the general population in the same middle income bracket, meaning that in general middle income homes are overrepresented in the solar sample. The most overrepresented demographic is that in the $100,000-150,000 range, which does indicate that solar households tend to have a higher income than the overall population.
- Low income households, while underrepresented as solar homes compared to the general population, have still installed a substantial amount of solar capacity with a GTM estimated 530MW of installations in the four states covered in the study.
- Finally, interesting differences emerged between states, with the solar population in MA and NJ more closely representing the general population than CA and NY, where affluent homes tended to be more overrepresented in the solar population.
Download a full copy of the study to see exactly how income and solar adoption vary across four key solar states and what could be driving the differences here.
Also published on Medium.