Inflation Forecasting under Different Macroeconomic Conditions: A Case Study of Pakistan

Abstract

Inflation forecasting is of primary importance not only for the conduct of monetary policy, but also for individuals to make choices. Forecasting inflation provides the precise image of how the economy is expected to accomplish in the future. For forecasting inflation, personal consumption expenditure is used to measure inflation because of its superiority of less sensitivity of price shock and its revision in subsequent years. For inflation forecasting, naive model, ARIMA model, Philips curve model, and Philips curve threshold autoregressive model are applied under different macroeconomic conditions with real-time, revised and final data from 1974 to 2016. The result shows that the naive model is superior to other models because RMSE and MAE of naive model are smaller than other models by using real-time, revised and final data for one year-ahead out-of-sample inflation forecasting. However, for two years ahead out of the sample inflation forecast, the real-time data RMSE shows that the naive model outperforms the other models, whereas the MAE shows that Philips curve threshold autoregressive model is superior than other models. For revised and final data for two years ahead out-of-sample inflation forecasting both forecasting accuracy measures show the naive model performance is the best.

JEL Classification Codes: C53