This paper describes a dynamic factor model for the Maltese economy. The model mainly serves as a tool to timely provide the Central Bank of Malta with nowcasts as well as short-term forecasts of the growth rate of the real gross domestic product, which in turn are used as an input in the forecasting process. Such forecasts reflect and incorporate the flow of information that periodically becomes available. Furthermore, the model can handle mixed frequencies that are likely to exist in large datasets used to summarise the Maltese economy and, as an additional advantage, it is able to deal with any path of missing data. This last feature is of crucial importance as data releases that are used to update the model do not take place in a synchronous way. The forecasting power of the dynamic factor model is compared with those of several other models available at the Central Bank of Malta. Overall, the results point towards a higher forecast accuracy of the dynamic factor model at very short horizons while, at longer ones, bayesian vector autoregressions appear to be more reliable.
dynamic factor models missing data nowcasting forecasting backdating forecasting horse-race