Estimating the Value of Destination Disclosure: A Dynamic Structural Model in a Transportation Network

Abstract

Information sharing can induce supply-side selectivity in on-demand platforms such as transportation network companies (TNC) that reduce efficiency and raises equity concerns (e.g., discrimination against demand attributes). We quantitatively compare the efficiency and equity performance of TNCs under various information-sharing models associated with destination-related attributes. Specifically, we first develop and estimate a dynamic structural model of TNC drivers' valuations under a \emph{disclosure model} in which various destination-related attributes that induce drivers' dynamic consideration are revealed to drivers before their acceptance. We then leverage the model estimates to run counterfactual experiments on alternative information-sharing models, including a non-disclosure model in which those attributes are concealed before acceptance, and partial-disclosure models in which those attributes are partially revealed. From a per-period (ride) perspective, the model estimates illustrate drivers' selectivity in terms of ride attributes (e.g., trip distance), and its heterogeneity across drivers. With dynamics modeled, the simulation shows a decreasing average acceptance rate, a more dominantly increasing average per-ride utility, an increasing average cumulative utility and decreasing variance in utility of accepted rides as the information models disclose more destination-related attributes. The results substantiate that there is a clear trade-off between driver's acceptance along with the equity performance versus the utility (profit) maximization. The quantified marginal effects further uncover drivers' inter-ride, forward-looking behavior and suggest that the two partial-disclosure models are cost-effective solutions for improving equity.