Rural-Urban Risk Sharing Networks : Uber Driver Labor Supply Responses to Shocks in Uganda

This paper examines the effect of flexible labor arrangements on insuring against rural weather shocks in developing countries, using novel data from the rideshare platform, Uber. I use machine learning classification to match drivers in Kampala to rural weather events. An econometric estimator using the machine learning probabilities is developed to account for the misclassification bias in impacts. A one standard deviation increase in the intensity of a shock during the growing period of three main crops is associated with working 20% more in the week of a shock and then hours decline steeply the following week. Dynamics are driven by extensive margin changes and are stronger for poorer parts of Uganda. Driver behavior changes based on the type of shock; drivers tend to drive more intensively and increase income if a drought event occurs, but leave the platform in the case of a flood