To assess the impact of EVCS installations on local businesses, we implement an identification strategy that leverages spatiotemporal variations in customer counts and spending at points of interest (POIs) surrounding these stations, enabling the establishment of causal relationships. Our analysis incorporates three categories of POIs: (1) accommodation and food services; (2) retail trade; (3) arts, entertainment, and recreation. To account for variations in the timing of EVCS installations during our study periods, we have employed a DID methodology, controlling for POI-specific and county-by-month fixed effects. Our analytical process is distinctively structured for two study periods, namely the year 2019 and the period spanning January 2021 to June 2023. Within this framework, our treatment group encompasses POI locations featuring newly introduced EVCS within 500 m during the study period. In contrast, the control groups are selected from POIs devoid of proximate EVCS that were commissioned during the study period.
One potential limitation to the validity of this approach is that EVCS providers might strategically place EVCS in locations where the benefits of EVCS deployment are maximized, potentially impacting our results. To address this endogeneity concern, we employ propensity score matching, pairing treatment POIs with control POIs of the same category and similar characteristics (see spatial distributions of treated and control POIs in Downtown San Francisco and Downtown Los Angeles in Fig. 1c–f). Our formal analysis then focuses exclusively on the matched POIs, resulting in a smaller but more reliable sample for analysis. The summary statistics for the variables about treatment, control, and unmatched POIs are presented in Supplementary Tables S1 and S2.
Table 1 shows the estimated impacts of installing a single EV charging port on the percentage increase in customer counts and spending at surrounding POIs. The analysis examines effects across all regions, with a particular focus on underprivileged regions. These regions are defined as disadvantaged and/or low-income communities according to designations by both California and Justice4014. Given the emphasis on equity in both federal and California government support for EVCS deployment5,15, strengthened by the federal administration’s Justice40 initiative16, it is crucial to investigate the localized effects of EVCS installations on businesses in underprivileged communities.
Our findings consistently reveal significantly positive effects of newly installed EVCS on customer count and spending at surrounding POIs across all scenarios. Specifically, the introduction of an additional charging port resulted in a 0.21% increase in customer count and a 0.25% increase in spending in 2019. In the subsequent period spanning 2021 – 2023, although the effects are somewhat diminished, they remain statistically significant, yielding a 0.14% increase in customer count and a 0.16% increase in spending. The moderation in magnitude during this period might be attributed to factors such as constraints on customer buying power influenced by the COVID-19 pandemic, as well as variations in the utilization rate of public EV chargers over time.
It is noteworthy that when considering the average number of ports in a single EVCS, which was 5.4 in 2019 and 5.0 in 2021–2023, our findings indicate a more substantial impact. Specifically, the addition of a single EV charging station leads to a 1.4% increase in spending in 2019 and a 0.8% increase in spending in 2021–2023. While these effects may seem minor, they carry substantial importance given the context of low EV adoption rates (2.61% for Battery Electric Vehicles and 1.15% for Plug-in Hybrid Electric Vehicles in California by the end of 202217) and the typically low utilization of EVCS (often averaging fewer than one session per port per day in the U.S.18,19,20).
Within underprivileged regions, the effects persist and maintain statistical significance. In 2019, an additional charging port led to a 0.17% rise in customer count and a 0.29% increase in spending. Between 2021 and 2023, the effects, although relatively smaller, remain notable, resulting in a 0.08% increase in customer count and a 0.09% increase in spending. We also conducted a sensitivity analysis, defining underprivileged communities based on designations from California and Justice40 separately. The estimated treatment effects remain significant and consistent in magnitude with the main specification (detailed in Supplementary Section 7).
Spatial and temporal variations in treatment effects
To provide a more detailed understanding of our estimated treatment effects, we conduct a comprehensive analysis of how these effects vary across both spatial and temporal dimensions. We first categorize the distances between treatment POIs and their nearby EVCS into five distance bins, ranging from 0 to 500 m with 100 m increments. We investigated how the treatment effect differed across these distinct distance bins (refer to the model details in the Methods section).
Our analysis reveals that the magnitude of treatment effects varies considerably across different distance bins from EVCS, as depicted in Fig. 2a (for 2019) and 2b (for 2021–2023). As may be expected, the closest distance bin (0–100 m) exhibits the most substantial effect. Specifically, the introduction of an additional charging port within this proximity results in a noteworthy increase of ~0.5% (in 2019) and ~0.6% (in 2021–2023) in both customer counts and spending at local businesses. Conversely, as the distance from EVCS increases to the range of 400–500 m, the magnitude of the treatment effect diminishes, settling at ~0.2% for both customer counts and spending. The full results are reported in Supplementary Table S4. These findings align with urban planning literature which has established that convenience of access greatly influences consumer choices13,21,22.
After estimating the marginal effect of adding one charging port, we find that the average marginal effect of adding one EV charging station on POIs within 100 m is quite substantial. This effect amounts to 2.7% in 2019 and 3.2% in 2021–2023 when multiplying the estimated marginal effect of adding one charging port with the average number of charging ports in a station.
To explore the temporal dynamics of the treatment effect, we implemented an event study strategy (detailed in the Methods section). Figure 2c (for 2019) and 2d (for 2021–2023) depict the marginal effects of an additional nearby EV charging port on the percentage change in spending. The results indicate that the marginal effects did not exhibit statistical significance before the installation of EVCS for both study periods, supporting the parallel trend assumption of the DID strategy. However, following the installation of EVCS, the marginal effect became significantly positive, validating the positive impact of EVCS installations on local business spending. It is noteworthy that in 2021–2023, the marginal effect declined in the later stage of the study period. This decline may be attributed to the saturation of the effect, particularly the decrease in utilization rate per EVCS port as the proliferation of EVCS may outgrow their demand. Alternatively, it could be influenced by behavioral factors, such as individuals’ preference for exploring new destinations, diminishing the effectiveness of older EV charging stations in attracting customers to surrounding businesses. Further research is necessary to pinpoint the precise reasons.