This study analyzed the hedging performance of copper futures using return data from Chinese copper spot and copper futures markets from January 2, 2001 to June 30, 2006. Unit root test results showed that copper spot and copper futures prices contained unit roots and were non-stationary time series. First-differenced variables were stationary time series. However, copper spot and futures prices were cointegrated, indicating that the long-term relationship between the two time series was stable. We compared hedge portfolio return variances and hedging performance obtained from the OLS model, VECM, and bivariate GARCH(1,1) model. For in-sample results, the hedge portfolio return variance was smallest in the OLS model and highest in the naive hedge model, and the differences in return variance between the naive hedge portfolio and other hedge portfolios were significant. For out-of-sample results, hedging performance was highest in the bivariate GARCH(1,1) model and lowest in the naive hedge model, and the differences in return variance among hedge portfolios were not significant. When foreign exchange gains and losses based on the Korean won to Chinese yuan exchange rate were present, in-sample hedge portfolio return variance was smallest in the VECM, and the differences in return variance between the naive hedge portfolio and other hedge portfolios were significant. For out-of-sample results, the bivariate GARCH(1,1) model showed the highest hedging performance, and the differences in hedge portfolio return variance between models assuming time-invariant variance and models assuming time-varying variance were significant. Interestingly, when accounting for foreign exchange gains and losses, in-sample hedge portfolio return variance increased, but out-of-sample hedging performance improved. In conclusion, when hedging in the Chinese copper futures market, simply hedging spot and futures one-to-one shows marginal efficiency compared to using complex models when considering model-related costs; however, when Korean firms need to account for foreign exchange gains and losses, it is advantageous to use models that assume time-varying variance.