The Use of Information in Balance of Payments Forecasting

Abstract

Judgmental forecasting is most common approach adopted within organizations for producing Casual evidence suggests that practitioners believe various approaches adopted under this heading are cheap and flexible, incorporate subtleties that quantitative models cannot, and, most important, are more accurate than simple extrapolative models. Researchers associated with area of forecasting agree only to flexibility of approach. Even question of cost is disputed, with Mabert (1976), for example, showing that a formalized judgmental approach may have substantially higher costs than various alternative extrapolative models he considered. Hogarth and Makridakis (1981) offer a good summary of many sources of bias that can undermine judgmental forecasting performance. Such judgmental forecasts or expectations play an important theoretical role in economic arguments. However, words of Cragg and Malkiel (1968) still bear repeating some 13 years after they wrote: the extent of agreement of significance of expectations is almost matched, however, by paucity of data that can even be considered reasonable proxies for those forecasts. Most research has considered aggregate economic expectations. Study of disaggregated judgmental economic forecasts has been concentrated largely in accounting literature, with thrust of research directed towards comparing comparative efficiency of stock market analysts, management forecasts and extrapolative models in forecasting future earnings per share. Typical papers are Green and Segall (1967), Brown and Rozeff (1978) and Ruland (1978). However, a recent working paper from National Bureau of Economic Research (Zarnowitz, 1982) considers disaggregated macroeconomic forecasts that were produced for American Statistical Association Business Outlook Surveys. Given conflicting results reported in these and other papers, it is difficult to argue that one particular forecasting model or type of institution consistently outperforms its competitors. In areas outside accounting and economics such as those discussed in Dawes (1977) and Armstrong (1978), evidence seems to suggest that judgmental methods are worse than their quantitative alternatives. In this paper we examine three judgmental forecasters working for City of London institutions who produce monthly balance of payments The market is competitive in that degree of accuracy of such forecasts is seen as a determinant of exchange dealings and commission income. We attempt to determine comparative effectiveness of these judgmental forecasters, both between themselves and compared with advanced time series forecasting methods such as Box-Jenkins ARIMA class of model. As we have mentioned, issue of whether such judgmental forecasts can improve