This paper evaluates the performance, bias, and the efficiency of option-implied and return-based correlation measures using 12 years of daily data on foreign exchange and over-the-counter (OTC) currency option. The sample includes five years of rates for the Polish zloty and the Czech koruna with respect to the euro and the U.S. dollar. The results show that implied correlation is a good predictor of realized correlation and is, generally, unbiased and efficient.