This paper proposes a formal statistical framework for the real-time monitoring of current economic conditions in Japan. We identify the ‘market moving’ indicators that are monitored constantly by market participants, statistical offices, newspapers, and policy makers. This results in the selection of around 30 variables. We track the release calendar and use vintages of real-time data in order to reconstruct the exact same information set that was available at the time when the forecasts were made. These variables are used to estimate a dynamic factor model (DFM) which is updated continuously at each new data release over a historical period of 11 years. Our results show that the proposed now-casting model tracks GDP realizations well throughout the evaluation period. The forecasts produced by the sophisticated yet transparent model are comparable with both the markets and the professional forecasts.
- Dynamic Factor Model