star twitter facebook envelope linkedin instagram youtube alert-red alert home left-quote chevron hamburger minus plus search triangle x

Study of Taxi Drivers Offers Insights into Behavior, Decisions, Ideas to Improve Market Efficiency

A new study used information from Global Positioning System (GPS) trace data to analyze the decision-making behavior of taxi drivers to understand the supply side of urban mobility markets. The study found significant differences between new and experienced drivers in learning behavior and driving decisions. The findings suggest that efficient information sharing can lead to additional income-generating opportunities for drivers, providing taxi companies with a strategy to improve market efficiency.

The study, by researchers at Carnegie Mellon University (CMU) and the University of Texas at Dallas, is published in Information Systems Research.

“The taxi industry generates $16 billion annually in revenues, but its inefficiency is well-documented, leading to low use by and high wait times for customers, wasted time for drivers, and a surge in ride-hailing platforms such as Uber and Lyft,” notes Beibei Li, associate professor of IT and management at CMU’s Heinz College, who coauthored the study. “To address this issue, we focused on the different types of information available to drivers to help them find customers in between passengers.”

The use of mobile and sensor technologies allows researchers to observe and record behavior in physical (offline) settings. In this study, researchers used a dataset of approximately two million GPS observations to analyze the decision-making behavior of nearly 2,500 single-shift taxi drivers in a large Asian city. The study sought to understand the underlying mechanism of drivers’ behavioral changes and what factors affect the supply side of urban mobility markets.

Specifically, the authors wanted to determine how taxi drivers learned about demand from different types of temporal, spatial, and contextual information (e.g., popularity of a location, time of day), trip information (e.g., trip income, travel speed, waiting time), and social information (e.g., nearby taxi drivers’ pickups and drop-offs).

The study’s dataset included detailed taxi GPS trajectories, taxi occupancy data (i.e., whether the taxi was occupied or not), and taxi drivers’ daily incomes. Researchers focused on what actions drivers take to find new passengers after dropping off current passengers. In particular, they looked at the role of information derivable from GPS trace data (e.g., where passengers were dropped off, where they were picked up, longitudinal taxicab travel history with time stamps) that were available to drivers to help them learn the distribution of demand for their services over space and time.

Among the study’s findings:

  • On average, new drivers were much more active in learning from information than experienced drivers. This indicates that experienced drivers have more precise prior knowledge on demand and that additional information is less valuable to them.
  • New drivers were much better at learning from simple and straightforward information, while experienced drivers tended to ignore this simple information and instead focused more on learning from more complicated information.
  • Drivers benefitted significantly from their ability to aggregate information on demand flows. This information became more valuable when aggregated across relevant dimensions.
  • Efficient information sharing could boost the welfare of drivers—especially those who are low income—because of potential market expansion. In the taxi market, sharing information efficiently could bring additional income-generating and customer-service opportunities for drivers.

The study’s authors point to several limitations of their work, noting that they did not consider individual driver-level characteristics such as past driving experience, family background, and company resources. Nor did they consider drivers’ strategic behavior. Finally, the study covered just two months and did not consider how market changes such as the rise of ride-hailing services might affect taxi markets.

“Our results show that drivers who are exposed to different information differ in the way they evaluate their choices when seeking new customers,” explains Ramayya Krishnan, dean of CMU’s Heinz College, who coauthored the study. “By aggregating the information extracted from drivers’ GPS traces on a large scale, we can significantly improve the quality of individual drivers’ decision making and, in doing so, boost industry efficiency.”


Summarized from an article in Information Systems Research, Learning Individual Behavior Using Sensor Data: The Case of Global Positioning System Traces and Taxi Drivers by Zhang, Y (University of Texas at Dallas), Li, B (Carnegie Mellon University), and Krishnan, R (Carnegie Mellon University. Copyright 2020 Informs. All rights reserved.

About Heinz College of Information Systems and Public Policy
The Heinz College of Information Systems and Public Policy is home to two internationally recognized graduate-level institutions at Carnegie Mellon University: the School of Information Systems and Management and the School of Public Policy and Management. This unique colocation combined with its expertise in analytics set Heinz College apart in the areas of cybersecurity, health care, the future of work, smart cities, and arts & entertainment. In 2016, INFORMS named Heinz College the #1 academic program for Analytics Education. For more information, please visit