RDP 8609: The Performance of Exchange Rate Forecasts 2. The Forecasts

The AFR commenced its weekly survey of the forecasts of foreign exchange dealers on 13 March 1985. The survey continued until 18 December 1985 when the AFR ceased publishing the forecasts.[6] Over this period of 41 weeks a total of 406 individual forecasts were provided on various exchange rates. Under the terms of the AFR survey a number of foreign exchange dealers were asked each Wednesday, to predict four pieces of currency-related data for the coming week. The survey participants were asked to estimate:

  1. the range of the $A/US$ exchange rate over the following week;
  2. the $A/US$ hedge settlement rate[7] (HSR) current on the following Wednesday;
  3. the US$/Yen exchange rate applying at the close of trading in Tokyo on the following Wednesday; and
  4. the $A/US$ three-month forward margin applying on the following Wednesday.

The results of the survey were published in the AFR each Thursday. The maximum number of forecasters included in the survey on any one week was twelve. Participation in the survey was on a revolving basis, with the two forecasters providing the week's poorest forecasts of the hedge settlement rate being dropped from the following week's survey and replaced by two forecasters from the AFR's master list of forecasters. This master list consisted of all dealers with a licence to deal in the domestic foreign exchange market. A total of 49 different forecasters provided at least one forecast over the sample period with the maximum number of forecasts provided by any one forecaster being 35. Only two dealers supplied more than 20 forecasts.

The data on the forecasts for the $A/US$ hedge settlement rate and the US$/Yen exchange rate will be used in the current analysis.[8] Figure 1 is a plot of the actual (solid line) and forecasted (broken line) $A/US$ exchange rate. It shows that there was a small net depreciation of the Australian currency over the sample period. The graph also shows that there were frequent changes in the direction of movement. As a result, a forecaster who was consistently pessimistic about the value of the $A need not have supplied forecasts superior to those generated from a simple no change model.

Figure 1
Figure 1

The graph of the US$/Yen exchange rate, presented in Figure 2, shows that the Yen appreciated against the US$ over the sample period. The bulk of the depreciation of the US$, however, occurred in late September 1985. Notwithstanding this overall weakening of the US$ there were many weeks over which the US$ strengthened against the Yen.

Figure 2
Figure 2

Footnotes

The results of a similar survey are now published on the “Reuters screen”, just after page FEWV. [6]

The hedge settlement rate is fixed each day. It is calculated by taking the average of the midpoints of the Australian dollar/U.S. dollar offer/bid rates of the ten contributing banks on the OZZU Reuters page at 9.45 a.m. The rate is used to close off all hedge contracts maturing on that day. It is quoted as 1$A=xUS$. A fall in this rate is thus a depreciation of the Australian dollar against the U.S. dollar. [7]

The responses to the other two questions will not be used in this study. It is difficult to evaluate forecasts of trading ranges (e.g., is a range that is too narrow preferable to one that is too wide), and there is little weekly movement in forward margins. [8]