From Total Effects to Underlying Mechanisms via Causal Mediation: An Empirical Example from Ride-Hailing Platforms in the United States

Abstract

For over a decade, IS researchers have explored causal relationships between IT artifacts and various outcomes. They focus on estimating average treatment effects, where IT artifacts cause outcomes. Causal mediation analysis helps identify critical channels and estimate causal mediation effects, revealing the mechanisms behind treatment effects. In this paper, we discuss the benefits and challenges of mediation in causal inference, highlighting differences from previous approaches and offering implementation guidance. We apply causal mediation to investigate an open question in IS literature: How does the demand for public transportation mediate the impact of ride-hailing platforms on US traffic congestion?

Keywords: ride-haling platforms, causal mediation analysis, traffic congestion, public transit ridership