Inflation Forecasting in India: Are Machine Learning Techniques Useful?

Abstract

The COVID-19 pandemic and the associated supply chain disruptions have impacted not just the inflation dynamics but also the performance of inflation forecasting models. Traditional econometric models with their implicit assumption of linear as well as time-invariant relationship between the target variable and explanatory variables have been questioned for long, resulting in the emergence of alternative models and techniques to better capture the changing inflation dynamics. This paper uses machine learning (ML) based forecasting techniques to capture the possible non-linear relationships between inflation and its determinants and compare their forecasting performance with some of the popular traditional time series models for both the pre-COVID and post-COVID periods. The empirical results suggest performance gains in using ML-based techniques over traditional ones in forecasting inflation in India over different forecast horizons.

Keywords: Inflation forecasting, deep learning, time series, rolling forecast, monetary policy, COVID-19 pandemic

JEL Classification: C45, C52, E31, E37, E52, E58