The present paper focuses on the factor modelling, which has been widely used in macroeconomic forecasting in recent years. Within that approach, there are a number of technical issues that require attention. These include (i) the number of factors to include in a factor augmented forecasting equation, (ii) the speci
cation of the dynamics in the aforementioned equation, (iii) the information from which to extract the factors, including the issue of whether and how to select a subset of the available dataset on which factor analysis will be applied, and (iv) the benefits of combining factor-based forecasts. Of course those technical issues are interrelated and cannot be addressed in isolation. Providing answers to those technical issues, which have not been fully addressed in the literature, is the aim of this paper, with a focus on Euro-area data and issues (iii) and (iv).
Remarkably, we find that the use of as few as one fifth of the original variables yields the best results in terms of forecast accuracy. Interestingly, we can discern a pattern of variables that remain after pre-selection which
are common to all the individual countries considered: these include real variables like retail trade, survey variables like consumer and price expectations, international variables like U.S. consumer expectations and financial variables like the price of raw materials.
by Giovanni Caggiano, George Kapetanios, and Vincent Labhard