Estimating Sparse Spatial Demand to Manage Crowdsourced Supply in the Sharing Economy

This paper develops a structural approach to guide decisions regarding the acquisition, retention, and development of individual providers by a sharing economy platform that crowdsources supply, which we call provider relationship management. Taking the context of a French car-sharing platform for which we have historical data, we lay out a random coefficient logit model of spatial demand combined with a Bertrand model of price competition between providers. Sparsity brings challenges in demand estimation; we resolve them through an approximation that brings new insights on a recent model with Poisson consumer arrivals. We then perform counterfactuals to evaluate the incremental value brought by existing potential providers to the platform. The results show that ignoring externalities between providers leads to large biases: provider incremental values are overestimated by 40% on average, and customer scorings are substantially impacted, resulting in suboptimal reward allocation. We also evaluate the potential impact of an advertising campaign to illustrate how our approach can be used to target acquisitions in specific locations, and we study the impact of activities that may increase the value of existing providers through price and/or location changes.