What type of forecasting system is desirable

Often forecasting is performed in different sections of a policy-making institution, each with responsibilities for specific parts of the forecast horizon. In particular, it is very common to have two sections. One section concentrates upon modeling of the whole system with a "core" model, adopting an horizon that is longer than six months, while the other focuses upon the initial six months. This latter group generally employs either an "indicator" or "spreadsheet" model, but also uses a good deal of judgment in forming the short-term forecast. The knowledge of sector specialists is relied upon quite heavily in this regard. Models may be involved but they are mostly statistical in nature, although economic ideas may guide and inform the choices. It is rare for much economic structure to be formally imposed at this level, and the methods do not lend themselves to producing a consistent statement about the evolution of the complete system. Whilst it is certainly possible to preserve national account identities by sufficient iteration between the different sector specialists, having a large number of different "local" opinions can sometimes lead to seemingly inconsistent behavioral relations. For example, if a Phillips curve connects wage and price inflation and unemployment, each person producing forecasts for their own series may have a different view of this relation in their own minds (or none at all). In the case of a specific series such as the consumer price index (CPI) a great deal of disaggregation may be undertaken in order to enable "sector specialists" to incorporate their knowledge about events such as tax changes, modifications of statistical procedures for compiling the data, etc.

Perhaps the most important issue to be resolved when faced with such a separation in responsibility is how to make the forecasts of the two groups cohere over the total forecast horizon. In many instances the two strands coexist rather uneasily and with little interaction. In that case it is generally those receiving the forecasts who must resolve any conflicts. Recently, some central banks have tried to fully integrate the two groups within a comprehensive forecasting system. The leader in this endeavor has been the Bank of Canada in their Quarterly Projection System (QPS), and their techniques have been transplanted to at least one other site, the Reserve Bank of New Zealand through its Forecasting and Policy System (FPS). There is a strong case for having a seamless approach to forecasting and the appropriate way of doing this is likely to receive quite a lot of attention in the years ahead.

0 0

Post a comment