Measuring Commuting and Economic Activity inside Cities with Cell Phone Records
We show how to use commuting flows to infer the spatial distribution of income within a city. A simple workplace choice model predicts a gravity equation for commuting flows whose destination fixed effects correspond to wages. We implement this method with cell phone transaction data from Dhaka and Colombo. Model-predicted income predicts separate income data, at the workplace and residential level, and by skill group. Unlike machine learning approaches, our method does not require training data, yet achieves comparable predictive power. We show that hartals (transportation strikes) in Dhaka reduce commuting more for high model-predicted wage and high-skill commuters.
The authors are grateful to the LIRNEasia organization for providing access to Sri Lanka cell phone data, and especially to Sriganesh Lokanathan, Senior Research Manager at LIRNEasia. The authors are also grateful to Ryosuke Shibasaki for navigating us through the cell phone data in Bangladesh, to Anisur Rahman and Takashi Hiramatsu for the access to the DHUTS survey data, and International Growth Center (IGC) Bangladesh for hartals data. The cell phone data for Bangladesh is prepared by the Asian Development Bank for the project (A-8074REG: "Applying Remote Sensing Technology in River Basin Management"), a joint initiative between ADB and the University of Tokyo. We are grateful to Lauren Li, Akira Matsushita and Zhongyi Tang, who provided excellent research assistance. We sincerely thank David Atkin, Alexander Bartik, Abhijit Banerjee, Sam Bazzi, Arnaud Costinot, Dave Donaldson, Esther Duflo, Gilles Duranton, Jean-Benoit Eymeoud, Ed Glaeser, Seema Jayachandran, Sriganesh Lokanathan, Danaja Maldeniya, Melanie Morten, Ben Olken, Steve Redding, members of the LIRNEasia BD4D team, and seminar participants at MIT, LIRNEasia, NEUDC 2016, the Harvard Urban Development Mini-Conference, ADB Urban Development, and Economics Conference, UEA 2019, NBER Cities and Global Economy Conference, for constructive comments and feedback. We thank Dedunu Dhananjaya, Danaja Maldeniya, Laleema Senanayake, Nisansa de Silva, and Thushan Dodanwala for help with Hadoop code and GIS data in Sri Lanka. We gratefully acknowledge funding from the International Development Research Centre (IDRC) and The Weiss Fund for the analysis of Sri Lanka data, and from the International Growth Center (IGC) for the analysis of Bangladesh data. We also acknowledge Darin Christensen and Thiemo Fetzer's R code to compute Conley standard errors (http://www.trfetzer.com/using-r-to-estimate-spatial-hac-errors-per-conley/), on which we built our code. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.