Access to reliable and affordable transportation shapes opportunities for employment, education, healthcare, and food. Yet for many urban residents, particularly those without access to a private vehicle, mobility remains a challenge. Public transit networks can fall short in terms of coverage, frequency, and flexibility. Carsharing has long been positioned as one strategy to address these gaps by offering short-term vehicle access without the costs and responsibilities of ownership. Since its early implementation in Zurich in 1948, carsharing has expanded globally and evolved into multiple service models, including roundtrip systems, where vehicles must be returned to their point of origin (e.g., Zipcar), and one-way systems, which allow for pickup and drop-off at different locations. Within the one-way model, some services are station-based, requiring users to pick up and return vehicles to designated stations (e.g., BlueLA), while others operate as free-floating systems that allow vehicles to be parked anywhere within a defined service area (e.g., Evie Carshare). These carsharing systems can offer flexibility and convenience, particularly in dense urban areas.
However, the benefits of carsharing have not been equitably distributed. Research has shown that users tend to be younger, higher-income, and multimodal travelers, while cost, carsharing access, and digital barriers have historically limited adoption among lower-income households. In response to these disparities and energy and environmental concerns, electric vehicle (EV) carsharing has re-emerged in recent years. EV carsharing combines the flexibility of shared mobility with the climate benefits of electrification. While EVs have been included in carsharing programs since the 1990s, advances in battery technology, expanded charging infrastructure, and targeted policy incentives have accelerated their integration. EV carsharing promises to reduce urban air pollution and greenhouse gas (GHG) emissions, but it also introduces new challenges, including infrastructure requirements, operational constraints, and questions around long-term service viability.
My dissertation focuses on one-way EV carsharing as a potential strategy for addressing both environmental and social equity goals. I investigate whether EV carsharing reduces energy consumption and associated GHG emissions, how different user groups engage with these services over time, and whether carsharing improves access to essential destinations such as grocery stores. These questions are especially relevant in underserved communities, where mobility barriers and transportation-related pollution are more pronounced.
Research GapsMy research identifies the following gaps in the EV carsharing literature:
1. Environmental Impact Measurement: Previous studies have extensively documented the environmental benefits of roundtrip and one-way carsharing; however, most of this work focuses on conventionally powered vehicles in higher-income, urban neighborhoods. The electrification of carsharing fleets has been recognized as a promising strategy for reducing emissions, yet studies evaluating EV carsharing’s energy and environmental impacts in underserved communities remain limited. Additionally, prior assessments estimate emissions based on VMT reductions and fuel economy assumptions, rather than directly measuring electricity consumption and the emissions associated with EV charging.
2. User Retention and Behavior Over Time: While previous studies have examined the sociodemographic profiles and tracked usage patterns and impacts such as VMT reduction and changes in vehicle ownership over time, there has been less focus on longer-term engagement and retention.
3. Grocery Access: Although access to a private vehicle is one determinant of food access, few studies examine the relationship between carsharing and grocery store access. Research on food accessibility has typically relied on static proximity metrics, such as straight-line distances or fixed travel times, which are unlikely to capture the accessibility experienced under real-world conditions from both a spatial and temporal perspective.
To address these gaps, my research is guided by the following five hypotheses:1. One-way EV carsharing reduces VMT and GHG emissions, with electrification further amplifying these effects.
2. One-way EV carsharing user activity patterns, including trip frequency, duration, and travel distances, differ between general and low-income qualified members, with low-income users demonstrating higher frequency of use, longer trip durations, and greater travel distances.
3. One-way EV carsharing membership type (i.e., general or reduced-cost for low-income qualified users) influences the duration of user active engagement and retention, with low-income users remaining active in the service for longer periods than general members.
4. Spatial-temporal accessibility from one-way EV carsharing stations to different grocery store types varies based on store type, traffic conditions, and trip departure times.
5. One-way EV carsharing users differ in their perceived grocery access and station usage patterns, with low-income users reporting greater improvements in access.
Research Approach and ContributionsTo test these hypotheses, I follow a mixed-methods approach that combines quantitative analysis of trip activity data with survey data and spatial modeling techniques. I focus on BlueLA, a one-way, station-based EV carsharing service launched in 2018 in Los Angeles, California. In a one-way, station-based system, vehicles must be picked up and returned to designated stations, but not necessarily the same one, allowing for one-way travel between fixed locations. BlueLA was designed to expand clean mobility access in underserved communities and operates under a tiered membership structure that includes both general population (Standard) and low-income qualified (Community) users. Community eligibility is determined by the operator (i.e., Blink Mobility) based on household income thresholds (e.g., $41,700 or less gross annual income for a single-person household) or enrollment in public assistance programs (e.g., SNAP, Medicaid, CalWORKS) . BlueLA’s location in Los Angeles—a city shaped by car dependence, sprawling development, and transportation equity challenges—serves as my empirical case. As a graduate student researcher with the Transportation Sustainability Research Center (TSRC) at UC Berkeley, I worked on a California Air Resources Board (CARB)-sponsored evaluation of BlueLA, which provided access to the data employed in this study.
My analysis is based on four main data sources. First, I use a trip activity dataset (n=59,112 trips from 2021–2022) provided by Blink Mobility in collaboration with CARB. This dataset includes detailed information on trip times, distances, station use, and vehicle charge levels across a two-year period. Second, I analyze a user survey that I co-designed with Professor Susan Shaheen and Dr. Elliot Martin. The survey (n=215 responses) captures information on vehicle ownership, travel behavior, BlueLA’s impact on access and mobility, and user experience. Third, I incorporate a service area resident survey (n=1,017 responses) co-developed with Professors Elizabeth Deakin and Susan Shaheen and Dr. Martin to capture insights from non-users. This addition offers a broader view of adoption barriers and the perceived value of EV carsharing among those living within the BlueLA service area. Lastly, I use several external datasets to support spatial and demographic analysis, including grocery store data from the United States Department of Agriculture (USDA) Supplemental Nutrition Assistance Program (SNAP) Retailer Locator (n=5,888 stores), U.S. Census and American Community Survey data, and routing and store attribute data from Google, Yelp, and Mapbox APIs.
My dissertation makes three primary contributions to the study of one-way EV carsharing and its environmental, behavioral, and equity implications: 1) addressing knowledge gaps by testing hypotheses related to EV carsharing’s impact on emissions, user retention, and accessibility, 2) methodological contributions, introducing novel approaches including a finite state machine model for measuring EV charging emissions, clustering and survival analyses for modeling user retention, and a dynamic accessibility model incorporating real-time travel conditions and store characteristics for assessing grocery access, and 3) empirical contributions, providing evidence on how different user groups engage with EV carsharing, the extent of its environmental benefits, and its potential to improve grocery store access for underserved households.
Addressing Gaps in Knowledge of CarsharingThis dissertation addresses the three research gaps including the: 1) limited understanding of EV carsharing’s environmental impacts in underserved communities and in contrast to the general population of users; 2) lack of research on user retention and engagement patterns across different socioeconomic groups; and 3) underexplored relationship between EV carsharing and access to grocery stores among underserved households. My dissertation contributes to the literature, addressing gaps by testing the five hypotheses noted above, within the BlueLA context.
Methodological ContributionsMy dissertation builds upon established methodologies in carsharing impact assessment, user retention analysis, and accessibility modeling while introducing new approaches to improve measurement precision and applicability. In Chapter 2, I extend the existing framework by Shaheen and Martin (Martin and Shaheen, 2011; Martin and Shaheen, 2016), which combines trip activity data with survey responses to estimate carsharing impacts on vehicle ownership and travel behavior. I contribute by developing a finite state machine model that infers vehicle charging activities from trip data. This approach allows for a direct measurement of electricity consumption, improving upon previous studies that estimated emissions solely based on VMT reductions and fuel economy standards. To calculate emissions, I apply time-variant and efficiency-adjusted emission factors, following the methodology developed by Zhang, Martin, and Shaheen (2025), which accounts for hourly fluctuations in grid emissions intensity and charging efficiency losses. In Chapter 3, I build on previous carsharing research that has largely relied on survey-based demographic analyses by applying clustering analysis to trip activity data, segmenting users based on actual usage patterns rather than just self-reported behavior. Additionally, I employ survival analysis , a technique widely used in engineering (e.g., machinery failure modeling) and consumer analytics (e.g., user retention in social media and ridehailing platforms) to quantify retention patterns and disengagement probabilities in EV carsharing. I also collaborate with Deakin, Shaheen, and Martin to implement a service area resident survey to examine the perceptions of non-users, providing a complementary perspective to the user survey and expanding the understanding of one-way EV carsharing adoption barriers. In Chapter 4, I advance existing accessibility models that rely on static proximity measures such as straight-line distances and cumulative opportunity counts, by integrating spatial-temporal accessibility modeling that accounts for real-world traffic conditions, store operating hours, and price levels. I construct a custom grocery access dataset by augmenting the USDA SNAP Retailer Locator dataset with store attributes from Google and Yelp APIs, allowing for a more realistic assessment of when and where grocery stores are accessible to BlueLA users. Collectively, these methodological contributions establish a replicable framework for evaluating one-way EV carsharing programs across environmental, behavioral, and accessibility dimensions.
Empirical Contributions
Empirically, my dissertation provides data-driven insights into the role of one-way EV carsharing in promoting sustainability, accessibility, and transportation equity. Through the case of BlueLA in Los Angeles, I analyze how a publicly supported one-way, station-based EV carsharing program operates in a complex urban setting characterized by car dependence and transportation inequities. Each chapter focuses on a different dimension: environmental outcomes in Chapter 2, user behavior and retention in Chapter 3, and grocery store access improvements in Chapter 4. These empirical findings offer practical insights for carsharing operators and policymakers seeking to improve EV carsharing services.
The next section presents key findings from each chapter, followed by a discussion of limitations and directions for future research.
Key FindingsSystem Overview
The BlueLA system is composed of 40 stations, each with five dedicated parking spaces equipped with Blink IQ 200 Level 2 alternating current (AC) chargers. The fleet currently includes 100 Chevrolet Bolt EVs, each with a battery capacity of 66 kWh, an estimated range of 260 miles, and a full charging time of approximately 8.5 hours. As of April 2025, Standard members pay $5 per month, and income-qualified Community members pay $1 per month. Standard members are charged $0.25 per minute (i.e., $15 per hour), while Community members pay $0.20 per minute (i.e., $12 per hour). These rates were increased in March 2024; prior to that, Standard and Community members paid $0.20 and $0.15 per minute, respectively, with a 15-minute minimum rental. The minimum rental period is now one hour, with a 12-hour cap. BlueLA also offers two promotional packages: a three-hour package priced at $28 for Standard members and $22 for Community members, and a five-hour package priced at $45 and $36, respectively. After the package period ends, per-minute rates apply.
From an operational perspective, we defined the BlueLA service area as the geographic region within a 30-minute walk from any BlueLA station, based on user-reported walking thresholds from our survey. While the City of Los Angeles currently has a total population of about 3.88 million and 1.46 million households (of which 12% are carless and 18% receive cash public assistance or food stamps/SNAP), the defined BlueLA service area—covering approximately 891,283 residents (23% of the city total) and 377,280 households—exhibits higher proportions of carless (22%) and SNAP-receiving (19%) households. Approximately 81% of residents in this area are age 18 or older and thus eligible to use BlueLA, although users must also hold a valid driver’s license and possess a credit or debit card. Stations are densely clustered in central neighborhoods such as Downtown Los Angeles, Koreatown, and Westlake, which are characterized by multi-story residential buildings and limited parking. In contrast, peripheral areas like Hollywood have sparser station coverage, suggesting that some neighborhoods benefit from more accessible station networks while others rely on fewer stations. Within a five-minute walk of a BlueLA station, 19% of total households and 17% of the population are covered, including 20% of SNAP-recipient households and 26% of carless households. At a 10-minute walk, the shares rise to 49% of households, 46% of the population, 52% of SNAP households, and 62% of carless households.Between 2021 and 2022, the system recorded 59,112 trips taken by 3,074 users. Community members made up 37% of the total user base, but they accounted for 51% of all trips. BlueLA experienced growth across both membership types: in 2021, there were 1,632 users (988 Standard, 644 Community), which increased in 2022 to 2,121 users (1,299 Standard, 822 Community). Trip activity data show that the average trip distance was 35 miles, with a median of 27 miles. Average trip duration was five hours, with a median of 4.8 hours. Despite being a one-way carsharing service, 36% of all trips were roundtrip (i.e., began and ended at the same station).
The user survey (n=215), which we conducted in December 2022, offered further insight into member demographics, motivations, and experiences. We distributed the survey to the entire BlueLA user base at the time (n=4,122), and it was available in both English and Spanish. Participants were incentivized by a raffle for one of forty $50 Amazon gift cards. We employed two types of anonymous user identifiers: one for linking survey responses to trip activity data and one for identifying membership status. We categorized respondents as “active” if they were subscribed at the time of the survey and “inactive” if they had previously subscribed but were no longer active. Of the 215 respondents, 60% were active members and 40% inactive; 56% were Standard members and 44% were Community members. Among all respondents, 71% reported not owning a personal vehicle. Income levels varied significantly by membership type: 74% of Community members reported household incomes under $50,000, while 68% of Standard members reported incomes over $75,000. Regarding educational attainment, 74% of Standard members and 54% of Community members held a bachelor’s degree or higher. This relatively high level of educational attainment is common in surveys and may reflect sampling bias toward more educated individuals. Age distribution among respondents was as follows: 7% under 25, 31% ages 25–34, 28% ages 35–44, 18% ages 45–54, 12% ages 55–64, and 4% age 65 or older. Language use at home was predominantly English (89%), with others reporting Spanish (6%), and the remainder including Arabic, French, Korean, and Tagalog.
Among all survey respondents (n=215), 67% reported using BlueLA for non-shopping errands, 65% for grocery shopping, and 62% for social or recreational activities. Other purposes included healthcare appointments (38%), commuting to or from work (22%), accessing public transit (11%), and attending school (5%). When asked why they joined BlueLA, 63% of respondents reported that they did not have a personal vehicle and joined to gain additional mobility. Other reasons included avoiding replacement of an inoperative personal vehicle (11%), changes in lifestyle requiring a vehicle (7%), the need for an additional vehicle for flexibility (6%), and the convenience of parking BlueLA vehicles (4%). Barriers to continued use were primarily identified by inactive users (n=79). The most common reason for no longer using BlueLA was the difficulty of finding an available BlueLA vehicle nearby (67%). Other challenges included preferring other transportation modes (20%), cost (15%), difficulty finding parking at drop-off (11%), charging station access (9%), loss of access to a personal vehicle (9%), and loss of a valid payment method (4%).
Environmental Impacts of EV Carsharing (Hypothesis One)In Chapter 2, we assessed whether the environmental benefits of traditional carsharing extend to one-way EV-based systems. Using trip activity data (n=59,112 trips from 2021–2022) and a user survey (n=215), we estimated changes in personal vehicle use and associated GHG emissions. I contributed to this work by developing a finite state machine model to infer charging behavior from trip records and applying time-sensitive, efficiency-adjusted emissions factors—based on Zhang, Martin, and Shaheen (2025)—to estimate actual electricity consumption and resulting emissions from EV charging. Our analysis focused on a subset of 81 active BlueLA members whose survey responses were matched with their trip activity data. Based on this sample, we assessed changes in vehicle ownership and driving behavior that respondents explicitly attributed to their BlueLA use. Respondents were asked whether they or their household had shed a vehicle due to increased mobility through BlueLA, whether they would have still shed that vehicle without access to the service, and whether they had postponed the purchase of a vehicle that would otherwise have occurred. We also asked respondents to estimate the change in their annual personal vehicle mileage due to BlueLA access. To account for potential sampling bias, we applied a weighting procedure based on frequency of their BlueLA use. Survey respondents and the broader user population were grouped by usage frequency, and we calculated weights to adjust for differences in representation across these groups. These weights were then used to scale the reported impacts to reflect the full BlueLA membership base (n=3,074 users). After weighting, we estimated that 4% of users shed a personal vehicle due to BlueLA, while 30% postponed a vehicle purchase they otherwise would have made. Additionally, 10% of users reported driving fewer miles annually in their personal vehicles as a result of using BlueLA. These behavior changes formed the basis for estimating system-wide VMT and emission reductions.
The results indicated that participation in BlueLA led to significant environmental benefits. Accounting for vehicle shedding, postponed vehicle purchases, and reduced driving, we estimated a net reduction of 460,845 VMT and 656 metric tons of GHG emissions over the two-year study period. Standard members contributed (n=1,931) 1,001,000 miles of BlueLA travel and Community members (n=1,143) 1,051,000 miles, yet the greater behavioral changes among Community members resulted in larger net GHG reductions (371 metric tons vs. 234 metric tons). On average, each BlueLA vehicle replaced 16 privately owned vehicles. We also compared the VMT and associated GHG emissions produced in the presence and absence of BlueLA. In the presence of BlueLA, we combined VMT and GHG emissions resulting from personal vehicle and BlueLA driving, and we deducted those avoided as a result of vehicle shedding, suppression, and reduced personal vehicle driving behavior. In the absence of BlueLA, we combined VMT and GHG emissions resulting from personal vehicle driving and those previously avoided but now realized from vehicle shedding, suppression, and reduced personal vehicle driving behavior. Comparing system-level scenarios, BlueLA reduced VMT by 34% and GHG emissions by 48% relative to a baseline without the service. Additionally, we compared the emissions produced by the BlueLA fleet to those of a hypothetical gasoline-powered fleet by translating the total VMT driven by BlueLA vehicles into GHG emissions using the EPA’s estimated real-world fuel economy for 2021 of 31.85 miles per gallon (MPG). When compared to a hypothetical internal combustion engine fleet, the BlueLA EV fleet produced 43% lower GHG emissions.
User Behavior, Adoption, and Retention (Hypotheses Two and Three)Chapter 3 examined how activity patterns and retention rates varied across user groups. We applied clustering and survival analyses to the BlueLA trip data and drew on both the user survey (n=215) and the service area resident survey (n=1,017) to understand motivations for use, challenges to continued participation, and differences between Standard and Community members. I contributed by developing the clustering and survival analyses for this study. In this context, retention refers to the length of time a user remained actively engaged with the service, while engagement was measured based on patterns in trip frequency over time. Specifically, we defined disengagement as a statistically significant increase in the inter-trip interval, using a Z-score threshold applied to a rolling window of one week. The analysis revealed distinct behavioral profiles. Community members had longer engagement (median of 449 days vs. 362 days for Standard members), traveled longer distances (27 miles on average vs. 15 miles), and used the service for longer durations (4.8 hours on average vs. 2.4 hours) compared to Standard members. Another segment, composed of both member types, exhibited frequent long-distance use (49 miles and 5.6 hours on average). Retention declined steeply after the first 500 days, but Community members stayed active longer overall. Survey responses showed that lack of vehicle access was the most common reason for joining BlueLA across both groups, while barriers to sustained use included vehicle and charging station availability, affordability, and parking constraints. These findings underscore the importance of accounting for socioeconomic differences in designing and managing EV carsharing systems. Understanding the retention dynamics of different user groups is essential for improving service longevity and equitable access.
Grocery Access and Accessibility Outcomes (Hypotheses Four and Five)In Chapter 4, we assessed whether one-way EV carsharing has the potential to improve access to grocery stores for households without personal vehicles using a spatial-temporal accessibility model, trip activity data, and user survey responses. I contributed by developing the accessibility model, constructing an enriched grocery store dataset using third-party APIs, and integrating these components with trip and survey data to evaluate accessibility outcomes. The findings showed that grocery shopping was a frequent trip purpose: 69% of Community members and 61% of Standard members used BlueLA for this purpose. Reported improvements in grocery access were higher among Community members (84%) than Standard members (71%). While Community members also reported greater perceived benefits in travel distance, speed, and scheduling flexibility, these differences were not statistically significant. The only statistically significant difference was in the ability to make late-night grocery trips, reported by 66% of Community members compared to 37% of Standard members.
Proximity to BlueLA stations was critical for underserved populations. For this analysis, we defined the service area to include all census blocks within a 30-minute walk of any BlueLA station, forming the boundary for our accessibility analysis. Within this area, 26% of carless and SNAP-receiving households were within a five-minute walk and 62% within a 10-minute walk of at least one BlueLA station. Trip activity data revealed that 11 stations accounted for half of all BlueLA trips. While all stations contribute to overall coverage, at least one of these high-volume stations was located within a 30-minute walk for 46% of carless households in our defined service area. More specifically, 11% of carless households were within a five-minute walk and 26% within a 10-minute walk of at least one of these highly active stations. However, differences in station-level activity between Community and Standard members were minor and statistically insignificant. Scenario-based modeling showed that smaller grocery outlets (e.g., convenience stores) were more accessible, while supermarkets and superstores required longer trips and incurred higher costs, particularly as access expanded beyond nearby stores. Under low-traffic conditions, reaching 1–5 of the nearest stores required about 11 minutes of roundtrip travel. Including an average in-store shopping duration of 46 minutes, the cost for Community members—who pay $0.20 per minute with a one-hour minimum—was approximately $12. Accessing 25% of available stores required 20 minutes of roundtrip travel, bringing the total trip time to around 70 minutes and cost to $14. Extending access to 50–100% of stores under higher traffic conditions raised the total trip duration to 79–100 minutes and cost to $16–$20. However, these estimates do not account for the additional time and effort required to walk to and from the nearest station or unload groceries, which may further increase the total trip time and user cost. The timing of grocery trips was also important. Store accessibility peaked between 10 AM and 6 PM, while late-night options were limited, often forcing users to rely on convenience stores with limited product variety. These findings highlight the value of aligning EV carsharing grocery trips with store hours and lower traffic periods to minimize travel time and rental expenses. To further enhance affordability for low-income users, carsharing operators could implement targeted incentives such as discounted short-term rentals or loyalty programs. In addition, grocery retailers could help address late-night accessibility gaps by extending store hours in locations that serve underserved households.
Limitations and Future WorkMy dissertation had several limitations that should be considered when interpreting the findings. One key limitation was the reliance on user survey data with a relatively small sample size, which introduced potential sample and self-reporting biases. While we attempted to mitigate selection bias by including a service area resident survey to capture responses from both users and non-users of BlueLA, the voluntary nature of survey participation meant that respondents may not fully represent the broader user population. Additionally, survey responses were self-reported, which could have led to recall errors or unintentional misrepresentation of travel behaviors. Although we applied weighting techniques to better align responses with BlueLA usage patterns, these adjustments may not have fully accounted for non-response bias or systematic differences between survey participants and non-participants. Future research could improve upon this by expanding survey sample sizes by offering increased incentives to encourage participation, using stratified sampling methods to target population subgroups of interest (e.g., oversampling low-income users in underserved communities), or incorporating alternative data collection approaches such as interviews or longitudinal studies.
Another limitation was the estimation of EV charging behavior and emission impacts. The finite state machine model we developed relied on assumptions about charging times, rates, and maximum charge levels, based on the known specifications of BlueLA’s Chevrolet Bolt EV fleet. These assumptions were necessary due to the lack of direct charging event data, but they introduced uncertainty in the precise measurement of electricity consumption and emission calculations. Additionally, while we incorporated time-variant and efficiency-adjusted emission factors to provide a more accurate representation of environmental impact, the real-world carbon footprint of EV charging can vary based on electricity grid fluctuations, weather conditions, and energy demand cycles. Future research could refine emission assessments by obtaining direct charging data from carsharing operators and integrating higher-resolution energy grid data to improve emission factor precision.
The timeframe of the study (2021 to 2022) also posed challenges in generalizing findings beyond this period. This timeframe coincided with the COVID-19 pandemic and its economic aftermath, which likely influenced travel behavior, car ownership decisions, and mobility preferences. While our analysis found that BlueLA usage remained stable or increased during this period, external factors such as pandemic-related economic conditions, fluctuations in fuel prices, and changes in remote work policies may have impacted both carsharing adoption and retention trends. Additionally, our findings on grocery store access improvements were based on a snapshot of user behavior and store access conditions, which may change over time. Future research could address these concerns by conducting longitudinal studies to track changes in mobility patterns, user retention, and grocery accessibility under varying economic and policy conditions.
The focus on a one-way, station-based EV carsharing model also limited the broader applicability of these results to other carsharing systems. One-way EV carsharing presents unique benefits and challenges compared to roundtrip or free-floating carsharing models, which may exhibit different usage patterns, cost structures, and accessibility outcomes. The findings suggest that Community members relied more heavily on BlueLA, but they also faced retention challenges due to cost and station availability constraints, highlighting the need for a comparative analysis of different carsharing models. Future studies could expand this research by evaluating how roundtrip or free-floating EV carsharing services perform in terms of emission reductions, retention, and grocery accessibility, as well as assessing policy interventions that might enhance carsharing accessibility for low-income users. Additionally, while we observed that low-income users remained active in the service longer than general population users, the reasons for this retention gap remain unclear. Possible contributing factors include the discounted pricing offered to Community members, as well as the greater reported mobility benefits among lower-income users. Future research could explore causal factors that drive longer retention among low-income users, including the role of affordability, unmet mobility needs, or broader lifestyle factors. Future work could also examine the potential of alternative policy interventions—such as providing eligible households with flexible transportation cards usable across public transit, ridehailing, carsharing, or bikesharing platforms—to support mobility and access. A comparative approach contrasting multi-modal integration versus service-specific strategies could reveal whether holistic access or targeted investments are more effective at advancing mobility equity in underserved communities.
The dynamic grocery accessibility analysis provided a more detailed assessment of food access than traditional static distance measures, but certain limitations remained. The accessibility model assumed that larger grocery stores provided greater product variety than smaller outlets, based on prior research, but it did not directly incorporate price or inventory data. The model relied on scenario-based routing and store-level attributes from third-party APIs, which may contain inaccuracies, temporal fluctuations, and incomplete information on store operations. Additionally, travel time estimates were limited to automobile trips and did not account for other transportation modes such as ridehailing or public transit. While the model integrated store operating hours and traffic conditions, it did not fully account for seasonal variations in grocery shopping behavior or day-to-day fluctuations in store availability. Additionally, because we did not have the data, the analysis did not include users’ home locations, limiting the ability to measure door-to-door accessibility. Lastly, while the model quantified access in terms of the number of reachable stores, it did not capture individual decision-making behavior or store choice preferences, where prior research has shown that presenting users with a large number of options can lead to choice overload, particularly in situations involving complex decisions. Future research could refine this approach by integrating residential location data, multi-modal travel scenarios, detailed inventory and pricing data, and behavioral models that account for consumer preferences, perceived effort, and decision complexity. Despite these limitations, my dissertation provides a detailed evaluation of one-way EV carsharing’s environmental, behavioral, and accessibility impacts, offering valuable insights for researchers, policymakers, and carsharing operators. By addressing these challenges in future research, scholars can further refine EV carsharing’s role in sustainable urban mobility, accessibility improvements, and emission reduction strategies.