Project key contributions include the following:
(i) Behavior – we conducted a series of studies to unravel the behavioural preferences of both demand-side and supply-side of urban mobility platforms. Om the demand side, our data collection efforts involved a battery of surveys and choice experiments regarding attitudes and choice preferences of travellers towards mobility-on-demand under different settings and in comparison to mobility alternatives. In particular, we focus on the willingness-to-share and perceptions of service reliability as well as the impact of experience and attitudes towards flexibility, sharing, COVID and technology, and perform a market segmentation by estimating latent class models. On the supply side, we carried out a focus group analysis followed by a series of choice experiments designed to extract information on preferences in relation to ride acceptance and relocation decisions for a unique sample of ride-hailing drivers.
(ii) Network and operations – we developed a new algorithm for matching travel requests into pool rides, known as ExMAS (EXact Matching of Attractive Shared rides). The model enables network-wide strategic evaluations of alternative system and environment settings. At the strategic level, the model has been used to assess the shareability potential of alternative spatial demand distributions. At the tactical level, it has been used to model the potential spreading of a virus in pooling networks and assess how matching management strategies can be deployed to mitigate virus spreading. In addition, we developed a game theoretical approach for analysing the impact of passenger delays on co-riders, trading capacities between operators in the event of disruptions and on different means of splitting shared-ride costs.
In a large-scale empirical study we have analysed 3.5 million Uber trip records from three US cities (New York, Washington and Houston) and three in Europe (Amsterdam, Warsaw and Stockholm). For each Uber ride, we have identified the most attractive public transport alternative. We found out that between 1 out of 5 (in the case of Amsterdam) and 2 out of 5 (Houston) of Uber trips in these cities have no viable public transport alternative. We also looked into how Uber and public transport competitiveness impacts the demand for each using detailed smart card data from Washington DC.
(iii) Mass effects – the development of a novel simulation tool for modelling the dynamics of two-sided mobility platforms, MaaSSiM, has enabled the analysis of a wide range of scenarios and the potential for mobility on-demand. We capture labour supply decisions in the model: long-term registration decisions and daily participation decisions. Anticipated earnings and labour opportunity costs are key variables in both decisions. In the registration decision, also capital registration costs are considered. Potential drivers are dependent on information about earnings from drivers with experience. Therefore, we represent the diffusion of information through a pool of potential drivers with an epidemiological model. On the consumer side, we capture the diffusion of information preceding ride-sourcing adoption with a similar model. In our agent-based model, informed travellers can opt for the ride-sourcing service as an alternative to traditional modes as the private car, bike and public transit. We performed scenario-based analyses of variables potentially affecting the co-evolution of supply and demand in the ride-sourcing market ranging from pricing schemes to income inequality.
Our work involved also a series of meetings, workshops and seminars with leading research groups in the field (e.g. NYU, EPFL, UC Davis, TU Dresden), public authorities (Municipality of Amsterdam, Amsterdam Transport Authority), public transport service providers (e.g. WMATA, GVB), mobility on-demand providers (e.g. Uber, ViaVan, felyx) and events and communication with policy makers and the wider public (e.g. MaaS@AMS event, press release).
