Forecasting the Non-Rental Component of Hong Kong's CCPI Inflation - an Indicator Approach

This paper develops both a bivariate and a multivariate indicator model using a large group of high-frequency economic indicators to forecast Hong Kong's non-rental component inflation. Indicator models can offer timely forecasts on future inflation developments because monthly indicators are often employed, thus allowing more frequent updates of forecasts. We first apply the bivariate model to investigate the predictive content of 66 indicators and find that quite a number of them have high predictive content for inflation forecast. In particular, indicators from the real and financial sector have more predictive power than those from the monetary sector, partly owing to Hong Kong's unique monetary arrangement. We then apply the multivariate model to examine the predictive content of groups of combined forecasts and indicators. Our results suggest that combining individual forecasts or individual indicators adds additional information and can help improve the forecast accuracy of Hong Kong's inflation by a considerable margin. Three preferred indicator models are employed to forecast the near-term (3 to 6 months) and short-term (12 months) inflation for non-rental component of the CCPI in Hong Kong. These models generate a range of averaged year-on-year inflation forecasts from 1.6% to 2.4%, which is in line with our assessments of the prevailing economic conditions. Though at an early stage of development, the performance of these models suggests that they are quite promising tools to help improve the accuracy of our inflation forecast. Specifically, the forecasts derived from these indicator models can be used as priors for formulating our view on future inflation developments in Hong Kong.