Spatiotemporal Analysis of Ridesourcing and Taxi Demand by Taxi zones in New York City

TL;DR

This study analyzes the spatiotemporal patterns of taxi and ridesourcing demand in NYC, revealing significant spatial and temporal dependencies that impact predictive modeling of transportation usage.

Contribution

It introduces a comprehensive approach combining time series and spatial analysis to understand ridesourcing and taxi demand patterns at the taxi zone level in NYC.

Findings

Identified significant spatial correlations between taxi zones.

Demonstrated the importance of accounting for both temporal and spatial dependence in demand modeling.

Revealed how demographic and land use factors influence spatial patterns.

Abstract

The burst of demand for TNCs has significantly changed the transportation landscape and dramatically disrupted the Vehicle for Hire (VFH) market that used to be dominated by taxicabs for many years. Since first being introduced by Uber in 2009, ridesourcing companies have rapidly penetrated the market. This paper aims to investigate temporal and spatial patterns in taxi and TNC usage based on data at the taxi zone level in New York City. Specifically, we fit suitable time series models to estimate the temporal patterns. Next, we filter out the temporal effects and investigate spatial dependence in the residuals using global and local Moran's I statistics. We discuss the relationship between the spatial correlations and the demographic and land use effects at the taxi zone level. Estimating and removing these effects via a multiple linear regression (MLR) model and recomputing the Moran's I statistics on the resulting residuals enables us to investigate spatial dependence after accounting for these effects. Our analysis indicates interesting patterns in spatial correlations between taxi zones in NYC and over time, indicating that predictive modeling of ridesourcing usage must incorporate both temporal and spatial dependence.