Time?varying trend models for forecasting inflation in Australia

We investigate whether a class of trend models, which decompose time series into an underlying and transitory component, with various error term structures can improve upon the forecast performance commonly used models when forecasting consumer price index (CPI) inflation in Australia. The main result is that tend to provide more accurate point density forecasts at medium long horizons compared conventional autoregressive Phillips curve models. best medium-term come from model stochastic volatility component moving average whereas long-run are better made by volatilities component. In full sample study, we also find capture dynamics periods significance Australian economy cannot. This includes dramatic reduction RBA adopted targeting, one-off 10% Goods Services Tax inflationary episode 2000, then gradually decline since 2014.