Blackwell Reference Online

Extract

Producers and consumers of empirical models take a shared interest in comparison of model forecasts. As pointed out by Granger (1990) , consumers of models care about out-of-sample model properties and consequently put weight on comparisons of model forecasts. Since producers of models in turn wish to influence the beliefs of model consumers, comparison of model forecasts provides an important interface between producers and consumers of empirical models. A comparison of forecasts from empirical models can be based on “raw” model forecasts (for example, provided by the producers) or on the published forecasts that include the effects of judgmental corrections (intercept corrections). In this chapter, the aim is to focus on the mapping from model specification to forecast properties, so we consider the raw model forecast, without the intervening corrections made by forecasters. Another issue is that any comparison will inevitably consider only a subset of macroeconomic variables, and the choice of variables will influence the outcome of the comparison. This problem extends to whether it is levels variables that are forecasted or their growth rates, since the ranking of forecast accuracy may depend on which linear transformation one uses: see Clements and Hendry (1993) and Clements and Hendry (1998 , ch. 3). In the following, these issues are pushed somewhat in the background ... log in or subscribe to read full text