Many studies have estimated the optimal time‐varying hedge ratio using futures, with most employing a bivariate generalized autoregressive conditional heteroscedasticity (BGARCH) model or a random coefficient model to estimate the time‐varying hedge ratio. However, it has been argued that when the variability of the estimated time‐varying hedge ratio is large, this ratio's hedging performance is not as good as that of the unconditional (constant) hedge ratio. This study proposes a nonparametric estimation approach to estimate and evaluate the optimal conditional hedge ratio. This method produces a time‐varying hedge ratio with less volatility than those obtained from the BGARCH and random coefficient models. We evaluate the hedging performance of the various models using soybean oil, corn, S&P 500, and Hang Seng futures indices. The empirical results support the proposed nonparametric approach in terms of both in‐sample and out‐of‐sample performance. © 2015 Wiley Periodicals, Inc. Jrl Fut Mark 36:968–991, 2016