Forecasting Inflation with Mixed Frequency Data in India

Abstract

In the standard approach of building prediction models, macroeconomic data is matched with monthly or quarterly or annual aggregates of financial series, since macroeconomic data are typically available at those frequencies. Such aggregation leads to the loss of useful forward-looking information of financial data. This is so because financial data are usually observed with higher periodicity than monthly data. Recent empirical evidence suggests that a Mixed Data Sampling (MIDAS) regression technique improves the predictive power of the models that incorporates data of different frequencies. In this article, we propose an autoregressive MIDAS model for forecasting monthly inflation in India using daily treasury yield information. Empirical results show that the proposed model has better predicting power over forecasting inflation in India.