Incentive-driven transition to high ride-sharing adoption

Sustainable urban mobility is crucial for a viable urban life. Ride-sharing (ride-pooling) offers a promising alternative to single-occupancy vehicles, currently dominating urban transport. Digital app-based ride-hailing services facilitate ride-sharing by easily matching trips with similar origins and destinations. Combining trips increases vehicle utilization, reduces the number of vehicles needed, and mitigates congestion and environmental impacts. Previous research focused on algorithms for large-scale ride-sharing and potential efficiency gains. These analyses suggest that large-scale ride-sharing is most efficient in densely populated areas where matching rides without significant detours is easier, improving economic and environmental efficiency and service quality. However, the conditions under which people are willing to adopt ride-sharing remain unclear. This article investigates the incentive structure governing ride-hailing users' decisions to share rides. Using a game-theoretic model of a one-to-many demand scenario, the authors illustrate how user interactions lead to two distinct ride-sharing adoption regimes: a low-sharing regime where adoption decreases with increasing demand, and a high-sharing regime where adoption is independent of demand. Analysis of ride-sharing decisions from approximately 250 million ride requests in New York City and 110 million in Chicago supports the model's predictions, suggesting a small increase in financial incentives could disproportionately increase ride-sharing adoption for certain user groups.

Existing literature explored the development of algorithms for efficient large-scale ride-sharing and the potential efficiency gains from aggregating rides, particularly in dense urban areas. Studies highlighted the increased efficiency and reduced environmental impact associated with higher ride-sharing utilization in high-demand zones. However, a gap remains in understanding user willingness to adopt ride-sharing, and how to encourage increased adoption among ride-hailing users. Prior research touched upon the incentives and disincentives influencing ride-sharing choices, such as financial discounts, expected detours, and the inconvenience of sharing, but a comprehensive model integrating these factors and predicting adoption patterns remained elusive.

The study employs a game-theoretic model to analyze the incentive structure behind ride-sharing decisions. The model considers three key incentives: financial discounts (proportional to the single ride distance), expected detours to pick up or drop off other passengers, and the inconvenience of sharing a ride with strangers. The overall utility of a shared ride is calculated as the sum of the utility of a single ride plus the difference in utility due to these three factors (financial discounts, detours, and inconvenience). The decision to share is determined by the expected utility difference between a shared and a single ride. The model uses a stylized city network with a central origin and multiple destinations to simulate ride-sharing decisions. The network topology combines star and ring topologies with uniformly chosen destinations. Ride requests are matched along adjacent branches to reduce the total distance driven and return to the origin. A one-to-many demand constellation is analyzed, where users have the option to choose a single or shared ride. Users observe the utility difference over multiple rides and adjust their sharing decision to maximize expected utility. This leads to equilibrium probabilities reflecting optimal responses and maximizing the utility for users to each destination. The model simulates this process using discrete-time replicator dynamics to model the users’ learning of their optimal equilibrium strategies, which converges to an equilibrium adoption probability for ride-sharing. The model analyzes ride-sharing decisions in both homogeneous (all users have the same preferences) and heterogeneous preference scenarios. For the empirical analysis, the study utilizes ride-request data from New York City (over 250 million requests) and Chicago (over 110 million requests) in 2019. The data includes information about origin and destination, time, and whether the request was for a shared or single ride. The authors compute the average request rate and the fraction of shared rides for different areas in both cities. The model's predictions are then compared with the observed ride-sharing adoption patterns in these cities.

The model reveals two distinct ride-sharing adoption regimes. In the low-sharing regime, ride-sharing adoption decreases as the total number of requests increases, even though the potential for sharing increases. This is because the expected detour and inconvenience increase with the number of users, outweighing the financial incentives. In the high-sharing regime, ride-sharing adoption is independent of demand, implying that the financial incentives fully compensate for the expected inconvenience. The transition between the two regimes is discontinuous at high demand, meaning there's an abrupt shift from low to high adoption. The empirical analysis of ride-sharing data from New York City and Chicago confirms the coexistence of both regimes in these cities. In New York City, locations with low request rates show a relatively high proportion of shared rides increasing linearly with the number of requests. High-request rate locations exhibit varying levels of ride-sharing adoption. In Chicago, high-demand zones show a constant number of shared rides, consistent with the low-sharing regime. The observed patterns in both cities are consistent with the model's predictions considering heterogeneous user preferences and mixed sharing states. The model suggests that the ratio of financial discounts to the inconvenience tolerance acts as a control parameter defining the high-sharing or low-sharing regime. A phase diagram illustrates this relationship, showing that when financial discounts fully compensate for expected inconvenience, all users choose shared rides. Otherwise, the system transitions to the low-sharing regime where the number of shared rides saturates. The transition between regimes is discontinuous in the high-demand limit for homogeneous user preferences. However, heterogeneous user preferences lead to a mixture of high and low-sharing behaviors at different locations, obscuring the abrupt transition.

The findings address the research question by demonstrating the existence and interplay of two distinct ride-sharing adoption regimes driven by the balance between financial incentives and disincentives (detours and inconvenience). The significance lies in providing a mechanistic explanation for the often-observed low ride-sharing adoption rates even in high-demand situations. The results complement previous findings on the increased potential shareability of rides at high demand by emphasizing the crucial role of individual incentives. The model's ability to accurately predict the observed patterns in New York City and Chicago strengthens its validity and offers a robust framework for understanding ride-sharing dynamics. This model can guide the design of effective strategies to increase ride-sharing adoption. The model is robust to changes in parameters and model details such as matching strategies, network topology and demand distribution, making the results more reliable.

This study presents a game-theoretic model explaining the complex relationship between incentives and ride-sharing adoption, revealing two distinct regimes and a discontinuous transition at high demand. The model's predictions align with empirical data from New York City and Chicago, highlighting the importance of balancing financial incentives with user preferences. Future research could investigate the impact of additional factors like sustainability attitudes and risk aversion. The sharp transition to high ride-sharing adoption suggests that even moderate improvements in financial incentives or service quality could significantly increase adoption.

The model utilizes simplified assumptions about user preferences and city topology. The linear utility function might not capture all aspects of user behavior, and the stylized city network may not fully represent the complexity of real urban environments. The empirical data may not capture all relevant factors influencing ride-sharing decisions, such as socio-economic factors and the impact of multiple ride-sharing service providers. Additional research could explore the effect of non-local matching and more detailed modeling of user heterogeneity.