Sharing the Sharing Economy: Policy Design for Social Good

The sharing economy has transformed society and enabled unprecedented growth by private sector technology giants such as Google, Amazon, and Lyft. However, harnessing its power in the context of public and civic sector operations often requires new approaches due to different objectives and constraints. Unfortunately, despite the outsized importance of governmental and nonprofit organizations in promoting social good, they frequently lack the resources to innovatively take advantage of online platforms. This dissertation sheds light on how such organizations can implement policies which improve their efficiency and growth, with a particular focus on the utilization of crowdsourcing and platform-based markets. For most public and civic sector organizations, success crucially depends on the efficient use of every available resource, be it volunteer labor or scarce social goods. This dissertation begins by studying the optimal design of volunteer notification systems, which need to balance the benefits of reaching more volunteers with the costs of notification fatigue. It then turns to the question of allocating social goods in an efficient yet fair manner when recipients\u27 needs realize sequentially. For both problems, online policies are presented that achieve constant-factor guarantees relative to benchmarks with knowledge about the future. These policies also demonstrate strong numerical performance, which complements their theoretical developments. In support of the aforementioned improvements in efficiency (which rely on centralized decision-making and are most effective when implemented at scale), the second half of this dissertation considers two decentralized approaches for generating growth: (i) leveraging word-of-mouth effects in social networks, and (ii) using a commitment lever known as adoption to improve the engagement and retention of volunteers. These approaches are particularly useful in public and civic sector contexts, where offering monetary incentives is often impractical. Studying the former approach relies on a tractable characterization of diffusion in random networks, while studying the latter requires the development of a model for repeated matching in a two-sided marketplace. This work is motivated in part by a collaboration with Food Rescue U.S. (FRUS), a leading online food recovery platform that uses volunteers to transport donated food. By combining data-driven insights from FRUS with rigorous modeling frameworks, this dissertation provides actionable recommendations which can improve the design of the FRUS platform while also providing guidance for a wide range of similar platforms