Call me maybe : experimental evidence on using mobile phones to survey microenterprises

High-frequency data is useful to measure volatility, reduce recall bias, and measure dynamic treatment effects. We conduct the first experimental evaluation of high-frequency phone surveys in a developing country or with microenterprises. We randomly assign microenterprise owners to monthly in-person, weekly in-person, or weekly phone interviews. We find high-frequency phone surveys are useful and accurate. Phone and in-person surveys yield similar measurements, with few large or significant differences in reported outcome means or distributions. Neither interview frequency nor medium affects reported outcomes in a common in-person endline. Phone surveys reduce costs without increasing permanent attrition from the panel