Macroeconomic Forecasting in India: Does Machine Learning Hold the Key to Better Forecasts?

Abstract

Forecasting of macroeconomic indicators is a challenging task, compounded by complex processes and dynamic nature of the macroeconomy. With recent advancements in computing power and the advent of data, machine learning methods have been explored as an alternative to traditional forecasting methods. We review the paradigm of machine learning and apply it to forecast inflation for India. We train various machine learning algorithms and test their forecasting accuracy against standard statistical methods. Our findings suggest that machine learning methods are generally able to outperform standard statistical models. Further, we find that combining forecasts from different competing models improves forecasting accuracy when compared to individual model forecasts. Also, direct forecast of headline inflation provides better forecast than the forecast based on different components of inflation. Lastly, our analysis also finds preliminary evidence for stochastic seasonality in the inflation series for India.