Nature Energy: Unequal Solar Photovoltaic Performance Among US Social Groups

Original Information

 

Original Title: Unequal solar photovoltaic performance by race and income partly reflects financing models and installer choices

 

Published Journal: Nature Energy

 

Published Date: March 2025

 

Authors: Mircea Gherghina, Fedor A. Dokshin & Benjamin Leffel

 

Original Link: https://www.nature.com/articles/s41560-025-01743-7

Abstract

 

Residential solar photovoltaic (PV) is critical for rapid decarbonization strategies. To paint a picture of equitable energy transition, researchers have measured inequalities in residential PV adoption and identified factors driving group differences. However, little is known about people’s experiences after solar installation. Electricity generation differences between PV systems can be substantial, potentially leading to inequitable outcomes even when adoption disparities are narrowed. This study uses data measuring actual monthly electricity generation from over 26,000 PV systems installed in Connecticut to determine significant differences in community income and race in system output. Results show that financing model choices (purchase vs. lease) partly explain the observed disparities. In addition, installers vary substantially in system generation, highlighting the role of firm behavior in driving inequitable outcomes. These findings underscore the importance of measuring both the quantity and quality of renewable energy projects to ensure an equitable energy transition.

 

1. Research Background

 

Residential solar photovoltaic (PV) electricity generation is a crucial component of energy transition. However, in the United States, the deployment of residential PV systems exhibits significant inequalities, showing installation differences across multiple socio-demographic dimensions such as wealth, race, and urban-rural areas. These differences are not trivial; they have real-world impacts because solar PV systems can bring multiple economic, environmental, and health benefits to installers and their communities. Although the deployment of distributed PV systems is accelerating, existing research rarely focuses on usage after system installation and lacks in-depth exploration of whether these social disparities extend to the long-term performance of the systems. Current research on installation differences often ignores key factors that may affect system performance, such as installation location, environmental factors, panel orientation, system component configuration, equipment condition and maintenance, and design and construction quality.

 

This study focuses on the interaction of two key factors to explore their potentially differential impacts on solar PV deployment: financing models and installer behavior. First, the choice of financing method may directly affect the performance of the PV system. Homeowners can choose to purchase the system directly or use one of two common third-party ownership (TPO) models—leasing or power purchase agreements (PPAs). Because TPO models help homeowners avoid the high upfront costs of solar systems, this approach is particularly popular among low-income groups and minority communities, significantly expanding the accessibility of solar energy. Second, installers and TPOs, as key intermediaries, profoundly influence users’ access to solar PV and the overall user experience. Although installers play a central role in shaping the solar experience for residential users, current research on PV deployment disparities still pays relatively little attention to installer behavior.

 

2. Methods

 

This paper uses ordinary least squares regression to predict monthly electricity generation as a function of key independent variables and controls. Standard errors are clustered at the project level to account for the dependence between observations from the same project. The complete model specification is as follows:

 

 

To test for performance variation by installer organization, a linear model is specified that introduces the installer as a categorical variable:

 

 

Finally, considering the heterogeneous effects of financing models, equation (1) is modified to include interaction terms between financing model and income decile and financing model and majority status, as shown below:

 

 

3. Research Results

 

3.1. Solar PV Performance Differences

 

First, the distribution of all systems (Figure 1a) and the ethnic and racial composition (Figure 1b) are shown by low-income community status. The vast majority of low-income communities adopt solar PV through TPO options, with only 14.5% of installations purchased, while leased or PPA-purchased systems account for 41.5% and 44%, respectively. This contrasts sharply with non-low-income communities, where purchasing is more common than leasing (33.1% vs. 27.4%), although PPAs remain popular (accounting for 39.5% of all installations). Comparing financing models across race-majority categories, the highest rate of purchased systems is among white-majority census tracts (32.3%). In Black-majority, Hispanic-majority, and non-majority census tracts, purchased systems account for only 8.8%, 9.0%, and 13.1% of installed solar PV systems, respectively. Purchasing is more common in Asian-majority census tracts, but there is only one such area in Connecticut, so caution is needed in generalizing from this result. Therefore, except for Asian-majority communities, the vast majority of minority and non-majority census tracts install solar PV through leasing and PPAs. PV systems installed in different communities also vary in size; on average, larger systems are installed in high-income, predominantly white communities.

Figure 1 Distribution of financing model choices by low-income status and area ethnic-racial majority

 

3.2. Differences by Demographics and Financing Characteristics

 

Before discussing organizational behavior, we consider that the impact of financing models may not be consistent across different communities. To test this hypothesis, we introduce interaction terms between financing models and two socio-demographic variables into the model. Figure 2 shows the predicted results of this model. The results show that under the leasing model, the difference in estimated system electricity generation between different income groups is not significant. However, on the racial-ethnic dimension, there is a significant difference in electricity generation from leased systems: leased systems in white-majority areas produce more than those in minority-majority or non-white-majority census tracts.

 

In contrast, in systems using direct purchase, electricity generation shows significant differences on both socio-demographic dimensions. Systems in non-low-income communities generate significantly more electricity than those in low-income communities; similarly, systems in white-majority communities also generate more electricity than those in Black-majority (P<0.05) and non-majority communities (P<0.001). It is particularly noteworthy that this heterogeneity on the racial dimension is particularly pronounced in the "purchase" financing model. Purchasing is more popular in white-majority communities, while leasing is the most common financing method in Black-majority communities (see Figure 1b). Therefore, when comparing the most commonly used financing models in various communities, it can be found that, controlling for other identical system characteristics, PV systems installed in white-majority areas through purchasing generate more electricity per month than those installed in Black-majority areas through leasing

 

Figure 2 Monthly electricity generation predicted by financing model and area low-income status and area ethnic-racial majority status

 

3.3. Organizational and Generational Differences

 

Figure 3b shows the predicted electricity generation in June for typical photovoltaic systems installed by 23 installers (specific characteristics of each system are noted in the figure title). Each of these installers has installed at least 200 residential photovoltaic systems, accounting for over 90% of the installations in the sample. The figure shows that there are significant differences in average monthly electricity generation among different installers, with an estimated gap of over 100 kilowatt-hours between the best and worst performing installers. Furthermore, Figure 4 further shows the installation proportion of these companies in low-income communities (see Figure 3a) and their distribution in different race-ethnicity dominant areas (see Figure 3c). The dashed lines in Figures 3a and 3c represent the average installation proportions of systems in low-income communities (24.2%) and non-white communities (18.1%) in the entire sample, respectively. These results reveal that different installers have clear preferences regarding the income level and ethnic composition of the communities they serve.

 

The data shows that the top-performing companies tend to be more concentrated in high-income and predominantly white communities; while the lowest-performing companies tend to operate in low-income and racialized communities. Specifically, the average installation proportion of the top five installers in low-income communities is only 9%, while for the bottom five installers, this proportion is as high as 39.1%. A similar trend is observed in the race-ethnicity dimension: only 4.6% of the systems installed by the top five companies are in majority-minority or non-majority census tracts, while the bottom five companies have an average installation proportion of 24.3% in these communities, significantly higher than the Connecticut average. We also found that company size (measured by the number of installations) appears to be negatively correlated with system performance. Of the six companies with over 1,000 installations, five are among the bottom nine in terms of electricity generation; and all three companies with over 3,000 installations are among the bottom seven in terms of electricity generation.

 

Figure 3: Monthly electricity generation of typical solar photovoltaic systems installed by different companies and the socio-demographic characteristics of the census tracts where they installed photovoltaic systems, including the proportion of installations by low-income status and the proportion of installations by majority-minority race-ethnicity status in the area.

 

In Figure 4, we highlight the spatial distribution of the service areas of the installation companies associated with the highest (Figure 4a) and lowest (Figure 4b) system electricity generation. Adjacent Figures 4c and 4d show the distribution of low-income census tracts and predominantly non-white areas, respectively. These images further reveal the association between system power generation performance and geographic characteristics, showing that the top and bottom-performing companies exhibit significantly different geographic patterns in their monthly installation locations. Specifically, the highest-performing companies tend to deploy their systems in affluent, predominantly white communities outside of cities (see Figure 4a), concentrating in a few large suburban areas with larger populations. These companies are typically smaller (see Figure 3c) and primarily serve local markets. The significant gaps in urban areas such as Bridgeport and Hartford in Figure 4a further confirm this trend: these densely populated urban communities have almost no installations of these high-performance systems, with some areas completely uncovered and only a very small number of installations distributed among them. In sharp contrast, Figure 4b shows that the five lowest-performing installers primarily serve economically disadvantaged and highly racialized urban communities, such as Hartford, Bridgeport, and New Haven (located northeast of Bridgeport). Notably, these urban areas are not only the primary service areas of these low-performing installers, but also include two of the three largest installers in the sample.

 

Figure 4: Distribution of installations per thousand households by high-performing and underperforming companies by census tract.

 

4 Discussion and Conclusion

 

The rapid and equitable deployment of residential solar photovoltaics is a key component of the US decarbonization strategy. While previous research has measured disparities in photovoltaic adoption rates across different groups, there has been little attention paid to inequalities in the performance of installed systems. This study confirms that there are substantial and meaningful differences in the electricity generation performance of photovoltaic systems, and that these differences manifest along social dimensions of race and income. Our analysis finds that some of the performance differences can be attributed to the choice of financing model. Compared to directly purchasing systems, leased systems tend to perform worse, which may be a significant reason why they generate less revenue for homeowners. At the same time, installation companies, as key intermediaries in the residential photovoltaic market, have a significant impact on system performance. This study reveals significant differences in system performance among different installers.

 

The results show that simply counting the number of photovoltaic system deployments is insufficient to fully understand the inequalities in the process of promoting energy technologies. Post-installation performance evaluation of the systems is equally essential. In addition, negative user experiences with solar energy may inhibit future adoption, weakening the peer effect that has been shown to play an important role in driving photovoltaic diffusion, and may further exacerbate inter-group differences in the deployment process. Current policies place a high priority on the speed of growth in renewable energy capacity; however, achieving an equitable energy transition requires not only "quantity" but also "quality." Solar leasing, as an increasingly popular financing method, has broadened access for vulnerable groups to some extent, but if its performance is significantly lower than that of purchased systems, it may lead to further inequalities in the results of the energy transition. This problem is particularly prominent among large installers, who often serve low-income and minority-majority communities more frequently.

 

This study takes a key first step in measuring differences in the performance of residential solar photovoltaic systems and hopes to pave the way for future research focusing on the installation process and post-installation user experience. Future research should further explore whether poor system performance in vulnerable communities stems from predatory business practices, organizational failures, or other structural factors within the industry. At the same time, it should also explore how system performance differences affect user returns on investment, and integrate power generation data with financing structures, electricity price information, and third-party ownership (TPO) contract terms to comprehensively assess differences in photovoltaic system financial returns and their underlying drivers.

From the Economics of Climate Change