Today the Reserve Bank of India placed on its website a Working Paper titled “Inflation Forecast Combinations – The Indian Experience” under the Reserve Bank of India Working Paper Series.* The Paper is authored by Joice John, Sanjay Singh and Muneesh Kapur. As India moved to a formal inflation targeting framework in 2016, inflation forecasts, as the intermediate target, have become a critical input for the conduct of monetary policy. Multiple sources of shocks to the inflation trajectory that are difficult to fully anticipate – such as sporadic volatility in food prices, movements in exchange rates and commodity prices, evolution of inflation expectations, competition from e-commerce and increased global integration – and potential empirical non-linearities in standard behavioural macroeconomic relationships, however, make inflation modelling and forecasting challenging. Under these circumstances, it is often difficult to beat forecasts from a random walk model. Given the time-varying nature of the shocks, a particular model is unlikely to be robust across all states. This may necessitate use of alternative models, posing the challenge of combining them to generate a more reliable single forecast path. This paper examines empirically the forecasting performance of different combination approaches in the Indian context relative to a wide range of forecasts generated from different modelling frameworks and the benchmark random walk model. The performance-based combination forecasts are found to significantly improve forecast performance relative to forecasts from individual models for both headline and core inflation. This suggests that the combination approach is better at handling potential model misspecifications and breaks in structural relationships. (Yogesh Dayal)
Chief General Manager Press Release: 2020-2021/385
| |