Can Consensus Exchange Rate Forecasts Be Profitably Traded?

|
Includes: FXA, FXB, FXC, FXE, FXF, FXY, UDN, UUP
by: Philip Baker

Summary

Given the well known difficulty of forecasting exchange rates, we find that the risk-adjusted performance of a consensus of experts can be in line with the Barclays Currency Traders Index.

Tests of two years of weekly consensus forecasts for ten major currency pairs show that forecasts tend to perform better, the longer the term of the forecast.

The testing also reveals that the longer term forecasts tend to perform better than shorter term forecasts, for the same investment holding period.

For just on two years, forecasts of exchange rates for major currency pairs by a select group of market analysts have been made available on fxstreet.com each Friday. Forecasts are provided for one week, one month and one quarter ahead for EURUSD, GBPUSD, USDJPY, USDCHF, AUDUSD, USDCAD, NZDUSD, GBPJPY, EURJPY and EURGBP. Last week for EURUSD 16 analysts provided their views one week ahead, 28 forecast one month ahead and 34 provided their forecast for one quarter ahead.

There is a growing body of research across a number of disciplines that a consensus of experts will give the most accurate estimate of a variable for which imperfect knowledge exists. Given the number and caliber of analysts polled by fxstreet.com, it seems reasonable to think that the consensus of their forecasts could be a good a source on which to base a profitable forex trading strategy.

We tested a strategy of trading each currency pair at closing prices on the day (Friday) the forecasts are published. The trading rule was: if the forecast was higher than the current close, go long and go short if the converse applied. Initially, the holding period for the trades was aligned with the period of the forecast. The p&l for each trade (before trading costs) was tracked and the summary results for the period of 104 weeks from w/e 7/6/2013 are shown in the table below -

EURUSD

GBPUSD

USDJPY

USDCHF

AUDUSD

USDCAD

NZDUSD

GBPJPY

EURJPY

EURGBP

Portfolio

Forecast one week ahead

Return/volatility

-45.5%

-2.5%

-58.5%

-60.6%

60.4%

11.0%

-3.0%

25.6%

62.3%

159.5%

18.1%

% profitable

47.1%

56.3%

42.2%

53.4%

52.0%

47.6%

53.4%

55.3%

50.5%

64.7%

50.5%

Forecast one month ahead

Return/volatility

57.5%

-21.2%

94.6%

32.7%

39.7%

3.3%

41.0%

-8.1%

24.0%

133.2%

90.2%

% profitable

58.3%

39.8%

59.2%

55.3%

53.4%

48.0%

47.6%

46.6%

59.2%

61.2%

56.3%

Forecast one quarter ahead

Return/volatility

110.6%

-63.1%

114.1%

22.3%

99.4%

19.6%

59.7%

45.6%

48.6%

133.8%

129.2%

% profitable

55.9%

22.3%

58.3%

44.7%

58.3%

44.7%

50.5%

51.5%

52.4%

64.1%

63.1%

The "portfolio" column shows results for an equally-weighted portfolio of the positions taken each week on the basis of the rule described above. (Because of imperfect correlation, the return/volatility for the portfolio will be higher than the arithmetic average of the return/volatility of the individual currency pairs).

The best benchmark we can find for the return/volatility results is the Barclays Currency Traders Index which indicatively showed a return/volatility of 113% over the period covered by the above table. The portfolio results, at least for quarterly forecasts, appear to be in the vicinity of this benchmark.

Of significant interest is the improvement in results as the period of the forecast increases - portfolio return/volatility improves from 18.1% for forecasts for one week ahead to 90.2% for one month forecasts to 129.2% for forecasts one quarter ahead. The % of profitable trades similarly improves.

For one month forecasts, all but two currencies (GBPUSD and GBPJPY) show positive trading returns and all but GBPUSD show positive returns for one quarter periods. EURGBP comes out the best, with consistent results, for all forecast periods. However, to minimize the risk of over-optimization, we suggest trading the portfolio rather than "cherry picking" on the basis of past results.

We also tested whether there was any correlation between the "conviction" of the forecast (increasing with the size of the gap between the forecast and then-prevailing spot rates), and the trading outcomes.

When trades were only put on when the gap exceeded 1.1%, the return/volatility for forecasts one week ahead improved from 18% (all trades taken irrespective of gap as above) to 161% and was generally above 100% for higher percentage gaps. For one month forecasts, the best results were achieved when all trades were taken i.e. no gap filter was applied. For one quarter forecasts, some modest improvement in results was achieved when trades were taken at gap filters of between 0.2% and 0.7%.

Again, care must be taken not to over-optimize in this area.

Finally, we found that holding trades for the one month and one quarter forecasts for shorter periods than the forecasts produced good results. The table below shows the outcomes from holding the trades based on the one month and one quarter forecasts for one week -

EURUSD

GBPUSD

USDJPY

USDCHF

AUDUSD

USDCAD

NZDUSD

GBPJPY

EURJPY

EURGBP

Portfolio

Forecast one week ahead

Return/volatility

-45.5%

-2.5%

-58.5%

-60.6%

60.4%

11.0%

-3.0%

25.6%

62.3%

159.5%

18.1%

% profitable

47.1%

56.3%

42.2%

53.4%

52.0%

47.6%

53.4%

55.3%

50.5%

64.7%

50.5%

Forecast one month ahead

Return/volatility

48.0%

-14.4%

122.2%

-24.9%

57.3%

46.8%

100.8%

32.0%

43.6%

186.1%

120.1%

% profitable

47.1%

50.5%

61.8%

51.5%

46.1%

47.6%

52.4%

45.6%

52.4%

55.9%

56.3%

Forecast one quarter ahead

Return/volatility

136.2%

-83.2%

53.3%

-7.3%

63.1%

112.5%

90.5%

32.5%

49.7%

170.1%

120.4%

% profitable

52.0%

45.6%

58.8%

52.4%

49.0%

51.5%

53.4%

56.3%

54.4%

60.8%

58.3%

As can be seen, trading for one week periods basing the trading rule on the one month and one quarter forecasts produced similar results at the portfolio level, which were substantially better than using the one week forecasts.

Trading for one month periods using the one quarter forecasts in the trading rule would have improved the portfolio return/volatility from 90.2% to 116.5%.

Disclosure: The author has no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it. The author has no business relationship with any company whose stock is mentioned in this article.