โ Scribed by H. O. Stekler
No coin nor oath required. For personal study only.
A previous study questioned whether the published forecasts of one forecasting organization were valuable to users. A forecast was considered valuable if it differed significantly from a naive model in the sense of predicting the direction of change. The results indicated that it was not possible to show that the one-quarter-ahead predictions were valuable. However, the current-quarter forecasts were valuable. This paper examines whether the results are robust. The forecasts of two additional organizations were examined and the time period was extended. The results are generally robust.
A previous study (Schnader and Stekler, 1990) asked whether the published macroeconomic forecasts of one forecasting organization were valuable to users. In that context, it was indicated that a forecast had value if it could change the user's prior distribution about the direction of change of the economy. Another interpretation of the analysis was that the forecast was valuable if it differed significantly from a naive model in the sense of predicting the direction of change.
Two procedures were used to determine whether that organization's forecasts could be useful to users. The first was based on Merton's (1981) and Henriksson and Merton's (1981) method for determining the conditions under which a market-timing forecast was useful to investors. The second procedure was to construct a 2 x 2 contingency table relating the direction of predicted and actual changes. This table could be used to determine whether the predicted changes are independent of the actual changes.
The Schnader-Stekler study showed that, using either procedure, it was not possible to reject the nu11 hypothesis and, thus, show that the organization's one-quarter-ahead predictions of the direction of change were useful. However, the predictions for the current quarter, which are a combination of pure forecasts and estimates using current data, showed the opposite relationship. All those forecasts were judged to be valuable.
Even though these forecasts had significant turning-point errors in predicting quarterly changes, they were highly successful in distinguishing between periods of high (> 2%) and low ( <2%) growth. This result holds for both one-quarter-ahead and current-quarter predictions. This paper will examine whether these results are robust. First, the forecasts of two other forecasting organizations are examined and compared for the same time period with those in
๐ SIMILAR VOLUMES