RDP 8603: Risk Premia, Market Efficiency and the Exchange Rate: Some Evidence Since the Float Data Construction Appendix

Data were obtained from the Commonwealth Bank's daily exchange rate release. Three maturities were tested. These were for the 15-day, 30-day and 90-day markets. Weekly observations on bid spot and forward rates were used. Careful attention was given to the construction of the lagged forecast errors, premiums and holding period yields to ensure that each was observable at the time the forward contract was being entered into.

15 Day Data

Forward rates were sampled on Tuesday's and spot rates were sampled on Wednesday's 15 days hence. This yields the forecast error

Many studies simply lag the forecast error n times, where n is the number of weeks of the contract, to test for serial correlation in the forecast errors. However, in the 15 day market this would yield an independent variable equal to

Clearly, S_t+1 is unobservable at period t the time the forward contract is being entered into. To overcome this a series of observable lagged forecast errors were constructed by sampling spot rates on Tuesdays (the day forward contracts are being entered into) and forward rates on Mondays 15 days prior. This yields

This series was then used as the independent variable in the relevant equations.

30 Day Data

Forward rates were sampled on Wednesdays, and spot rates on Fridays 30 days hence, yielding the dependent variable

Lagging this four times would yield

Once again we would have an unobservable spot rate. Thus to construct the independent variables spot rates were sampled on Wednesdays and forward rates on Mondays 30 days prior, yielding independent variables

90 Day Data

Forward rates were sampled on Tuesdays and spot rates on Mondays 90 days hence, yielding

Lagging this by 13 weeks times yields

Clearly the spot rate is observable. Therefore this was used as the lagged forecast error in the relevant tests.