[Paper Review] Forecasting e-scooter competition with direct and access trips by mode and distance in New York City

Mina Lee, Joseph Y.J. Chow+2 more|arXiv (Cornell University)|Aug 21, 2019

Transportation and Mobility Innovations17 citations

TL;DR

This study develops a novel nonlinear, multifactor model to forecast e-scooter trips in New York City by analyzing substitution effects with taxis and public transit access/egress trips. Using log-log regression on demographic data (age, income, etc.) with an R² of 0.663, it predicts 66,000 daily trips and $67M annual revenue, with e-scooters replacing up to 1% of taxi trips and generating $800,000 in revenue from this substitution.

ABSTRACT

Given the lack of demand forecasting models for e-scooter sharing systems, we address this research gap using data from Portland, OR, and New York City. A log-log regression model is estimated for e-scooter trips based on user age, income, labor force participation, and health insurance coverage, with an adjusted R squared value of 0.663. When applied to the Manhattan market, the model predicts 66K daily e-scooter trips, which would translate to 67 million USD in annual revenue (based on average 12-minute trips and historical fare pricing models). We propose a novel nonlinear, multifactor model to break down the number of daily trips by the alternate modes of transportation that they would likely substitute. The final model parameters reveal a relationship with taxi trips as well as access/egress trips with public transit in Manhattan. Our model estimates that e-scooters would replace at most 1% of taxi trips; the model can explain $800,000 of the annual revenue from this competition. The distance structure of revenue from access/egress trips is found to differ significantly from that of substituted taxi trips.

Motivation & Objective

• To address the research gap in e-scooter demand forecasting by developing a predictive model using real-world data.
• To analyze how e-scooters substitute for taxi trips and public transit access/egress trips in Manhattan.
• To estimate the revenue potential of e-scooter sharing by quantifying trip displacement across different modes and distances.
• To model the distance-dependent revenue structure of access/egress trips versus direct taxi trips.
• To validate the model using data from Portland, OR, and apply it to New York City’s Manhattan market.

Proposed method

• A log-log regression model is estimated using user-level data on age, income, labor force participation, and health insurance coverage.
• The model is calibrated on Portland, OR data and extrapolated to New York City’s Manhattan market.
• A nonlinear, multifactor model decomposes daily e-scooter trips by alternate transportation modes, including taxis and transit access/egress trips.
• The model quantifies substitution effects by comparing e-scooter trips to historical trip patterns in competing modes.
• Revenue is projected using average trip duration (12 minutes) and historical fare pricing models.
• Distance-based revenue structures are analyzed separately for access/egress trips and substituted taxi trips.

Experimental results

Research questions

• RQ1 How do e-scooter trips in Manhattan compare to those in Portland, OR, in terms of demographic and socioeconomic drivers?
• RQ2 To what extent do e-scooters substitute for taxi trips in Manhattan?
• RQ3 How does the revenue potential of e-scooters differ between access/egress trips and direct taxi trips?
• RQ4 What is the contribution of e-scooter competition with public transit access/egress trips to total annual revenue?
• RQ5 How does the distance distribution of e-scooter trips affect revenue generation when replacing different modes?

Key findings

• The log-log regression model achieves an adjusted R² of 0.663, indicating strong explanatory power for e-scooter trip demand.
• The model predicts 66,000 daily e-scooter trips in Manhattan, generating an estimated $67 million in annual revenue.
• E-scooters would replace at most 1% of taxi trips in Manhattan, contributing $800,000 to annual revenue.
• The distance structure of revenue from access/egress trips differs significantly from that of substituted taxi trips.
• The model identifies a measurable but limited substitution effect for taxi trips, suggesting e-scooters primarily serve short-distance, last-mile mobility.
• Demographic factors such as age, income, labor force participation, and health insurance coverage significantly influence e-scooter trip demand.