Managing Multihoming Workers in the Gig Economy

Abstract

Gig economy platforms compete to source labor from common pools of workers, who multihome by dynamically allocating their services in real-time across multiple platforms. The question of how such workers choose between competing platforms has grown in salience. However, the unavailability of comprehensive data has limited our understanding of workers' dynamic multihoming decisions and their impact on the labor supply and operations of platforms.
We address this gap by integrating a major ride-hailing platform's proprietary data regarding individual drivers' detailed trips with public data on the drivers' outside options. Using empirical methods that overcome the remaining data limitations using simulation and machine learning, we structurally estimate the perceived costs that motivate drivers' forward-looking decisions and behavior. Our analysis reveals that workers are short-sighted and value sequences of consistent rewards (e.g., jobs and pay) over highly variable ones. Based on these findings, we explore the impact of compensation and incentives on the platform's labor supply and operations. Our counterfactual analyses indicate that consistent pay dominates variable pay in retaining multihoming workers. For gig platforms seeking to maintain a stable platform workforce, they can further control their labor supply by rewarding uninterrupted work or delaying quits. For policymakers, our research gives insights regarding the design of gig economy incentive schemes and regulations, including specifically New York City's 2018 Driver Income Rules, and their impact on multihoming behavior.

Keywords: gig economy, multihoming, structural estimation, applied generative adversarial networks, incentives, empirical operations, behavioral operations