Abstract
This paper investigates why on-demand carpooling—despite its promise for sustainable mobility—often fails to generate profit. We develop a stochastic demand model that captures match probability, rider beliefs, and network effects. Our results show that profitability emerges only when network effects and market density are sufficiently strong. We find that upfront pricing consistently outperforms contingent pricing when riders hold rational expectations, and this advantage intensifies under higher demand volatility. When rider beliefs are biased, contingent pricing becomes favorable only if riders substantially overestimate match probability. Finally, we evaluate two innovative matching mechanisms and show that pre-matching is generally inferior, while real-time matchability screening can offer short-term gains but loses advantage as beliefs stabilize.