Measuring Commuting and Economic Activity inside Cities with Cell Phone Records

We show that commuting flows constructed from cell phone transaction data predict the spatial distribution of wages and income in cities. In a simple workplace choice model, commuting flows follow a gravity equation whose destination fixed effects correspond to wages. We use cell phone data from Dhaka and Colombo, covering hundreds of millions of commuter-day observations, to invert this relationship. Model-predicted income at the workplace level predicts self-reported survey workplace income, and model-predicted residential income predicts nighttime lights. In an application, we estimate that predicted commuter income is 4-5% lower on days with hartals (transportation strikes) in Dhaka.