RDP 2004-07: The Profitability of Speculators in Currency Futures Markets Appendix A: Stationarity Tests

Jonathan Kearns and Phil Manners

September 2004

Tables A1 and A2 present stationarity tests for net speculator position and futures prices. Net speculator positions, except for the German mark, are stationary according to these tests. Due to our economic priors, and the results for the other series, for consistency we treat the German mark net speculator position as being stationary. Consequently we use both the level and difference of net speculator positions in our analysis. The stationarity tests on the currency futures prices confirm that, like exchange rates, these are I(1) variables. The results for the British pound suggest it may be fractionally integrated, though for consistency we treat all futures prices as being non-stationary, and so use the log difference of the futures price (in percentage terms) in our statistical analysis.

Table A1: Persistence of Net Speculator Positions
  Unit root tests Autoregressive coefficient
  DFGLS KPSS  
Australian dollar −4.15*** 0.37* 0.95
British pound −6.53*** 0.10 0.85
Canadian dollar −4.96*** 0.13 0.95
German mark/euro −3.70*** 1.32*** 0.92
Swiss franc −4.71*** 0.23 0.92
Japanese yen −5.71*** 0.27 0.93
Notes: Sample 05/01/1993 to 18/02/2003 (weekly)
DFGLS is a detrended version of the Augmented Dickey-Fuller test. The null is that there is a unit root.
DFGLS critical values: 1% = −2.57, 5% = −1.94, 10% = −1.62.
Lags are chosen using the Schwarz information criterion.
The null of the KPSS test is that the series is stationary.
KPSS critical values: 1% = 0.74, 5% = 0.46, 10% = 0.35.
* and *** mean that the null is rejected at the 10 per cent and 1 per cent level, respectively.
The autoregressive coefficient is the AR(1) coefficient, corrected for OLS bias using the Andrews median unbiased estimator (Andrews 1993).
Table A2: Futures Prices
  Descriptive statistics   Unit root tests Autoregressive coefficient
  Observations Mean return(a) DFGLS KPSS  
Notes: (a) return = |Δlog f| in per cent.
Returns are for being long in foreign currency.
Sample 05/01/1993 to 18/02/2003 (weekly)
See notes under Table A1 for unit root test and autoregressive information.
*,** and *** mean that the null is rejected at the 10 per cent, 5 per cent and 1 per cent level, respectively.
Australian dollar 529 1.03   −0.78 2.09*** 1.00
British pound 529 0.86   −2.03** 0.55** 0.98
Canadian dollar 529 0.58   0.15 2.67*** 1.00
German mark/euro 529 1.12   −0.97 2.10*** 1.00
Swiss franc 529 1.22   −1.47 1.47*** 1.00
Japanese yen 529 1.25   −1.36 1.02*** 1.00