Forecasting Private Consumption with Digital Payment Data: A Mixed Frequency Analysis
High-frequency data on digital payment and personal credit could be valuable input for predicting macroeconomic variables.
The recent increase in the use of a digital payment system is an interesting prospect in predicting macroeconomic activity. A digital payment system comprising credit cards, debit cards, automated teller machines, and mobile banking represents a broad spectrum of spending activity. The data on these indicators are also available at a higher frequency and therefore suitable for predicting lower-frequency macroeconomic variables. We use mixed data sampling (MIDAS) regressions to forecast quarterly consumption using monthly data on the digital payment system and other macroeconomic variables such as personal credit and the Index of Industrial Production. Empirical results show that the digital payment system data have better predicting power in forecasting private consumption in India. Both in-sample and out-of-sample forecast evaluation confirm that MIDAS provides superior forecast evaluation to the standard time series models using a single frequency. This finding reinforces the potential usefulness of such novel data among policy makers and practitioners.
WORKING PAPER NO: 1249