We present empirical evidence supporting the conjecture that rejection
s of the simple foreign exchange market efficiency hypothesis could re
sult from the use of statistical tests that, while asymptotically vali
d, may have poor finite sample properties.