On Rider Strategic Behavior in Ride-Sharing Platforms

TL;DR

The paper investigates rider-side strategic behavior in surge-prone ride-sharing markets, introducing a two-region model (surge vs non-surge) where riders may relocate across the boundary according to a function of demand disparity. It derives the dynamical updates for unmet demands $D_s(t)$ and $D_{ns}(t)$, proving that surge demand dissipates in finite time while non-surge demand follows a single-peaked trajectory, distinguishing localized and spill-over surges and establishing convergence bounds. The authors extend the analysis with stochastic demands and an agent-based, game-theoretic simulation that incorporates drivers, riders, and platform pricing, showing that rider relocation can smooth demand and reduce the price gap across zones; the model calibrates with an optimal relocation parameter and aligns with theory under realistic conditions. Practical implications include informing platform incentive design (e.g., walk-and-save schemes) and pricing policies to mitigate surges and improve equity across regions.

Abstract

Over the past decade, ride-sharing services have become increasingly important, with U.S. market leaders such as Uber and Lyft expanding to over 900 cities worldwide and facilitating billions of rides annually. This rise reflects their ability to meet users' convenience, efficiency, and affordability needs. However, in busy areas and surge zones, the benefits of these platforms can diminish, prompting riders to relocate to cheaper, more convenient locations or seek alternative transportation. While much research has focused on the strategic behavior of drivers, the strategic actions of riders, especially when it comes to riders walking outside of surge zones, remain under-explored. This paper examines the impact of rider-side strategic behavior on surge dynamics. We investigate how riders' actions influence market dynamics, including supply, demand, and pricing. We show significant impacts, such as spillover effects where demand increases in areas adjacent to surge zones and prices surge in nearby areas. Our theoretical insights and experimental results highlight that rider strategic behavior helps redistribute demand, reduce surge prices, and clear demand in a more balanced way across zones.

Figures (12)

• Figure 1: The above map schematic represents a geography serviced by a ride-sharing platform. The entire region is divided into hexagonal blocks and the price in each block depends on the market conditions "locally". Ride-sharing platforms like Uber employ surge pricing algorithms at a "hyperlocal" level uber_pricing, and Uber in particular uses such hexagonal regions for pricing uber_hex. The red block represents a region with abnormally high demand, leading to a surge in prices. We call this zone the "surge zone". In all other surrounding blocks (marked in green), demand reflects usual levels; we call them the "non-surge zone".
• Figure 2: Spill-over surge: Demand curves over time for $D_0 = 1000$, $d_0 = 200$, $\lambda = 30$ and $\mu = 50$. We set $k = 0.005$ and simulate over $T = 65$ time steps. $D_s(t)$ converges in $\tau_s = 33$ time steps while $D_{ns}(t)$ converges in $\tau_n = 30$ time steps. Observe that they lie within our theoretically computed bounds in Theorems \ref{['thm:Ds_conv']} and \ref{['thm:Dns_conv']} (indicated by dotted lines).
• Figure 3: Localized surge: Demand curves over time for $D_0 = 1000$, $d_0 = 200$, $\lambda = 30$ and $\mu = 50$. We set $k = 0.001$ and simulate over $T = 65$ time steps. $D_s(t)$ converges in $\tau_s = 41$ time steps while $D_{ns}(t)$ converges in $\tau_n = 20$ time steps (within theoretically computed bounds). Note the the distinctive difference in shape of $D_{ns}(t)$ from Figure \ref{['fig:spill_surge']}.
• Figure 4: Surge Inversion: Demand curves over time for $D_0 = 1000$, $d_0 = 300$$\lambda = 30$, $\mu = 50$. We set $k = 0.05$. Observe that for this set of parameters, $D_{ns}(t)$ exceeds $D_s(t)$ at $t = 1$.
• Figure 5: Illustration of surge dynamics in the deterministic setting with variation in parameters $\mu$ and $\lambda$. For all sets of results here, the following parameter values are fixed: $D_0 = 1000$, $d_0 = 200$, and $k = 0.005$.

Theorems & Definitions (25)

• Example 2.1
• Theorem 3.1
• proof
• Theorem 3.2
• proof
• Definition 3.1: Localized Surge
• Definition 3.2: Spill-over Surge
• Claim 3.3
• proof
• Theorem A.1