In this paper, we show that the price spreads of copper and aluminium's futures in the London Metal Exchange (LME) and the Shanghai Futures Exchange (SHFE) have strong predictive power of China's net imports of the corresponding commodities 2 months later. The modified Diebold–Mariano test of Harvey et al. (International Journal of Forecasting, 13, 1997, 281) indicates that our futures-based model significantly outperforms the ARIMA model and static model in terms of 1-month-ahead out-of-sample forecasts of net imports. Moreover, the market timing test of Clements (Evaluating Econometric Forecasts of Economic and Financial Variables, 2005, Palgrave Macmillan, Xxxxx) indicates that there is a significantly high correlation between the predicted signs of our model and the actual outcomes. When sorting all sign predictions according to the absolute magnitude of the out-of-sample movements, our proposed model achieves a 100 per cent forecasting accuracy for both copper and aluminium when predicting the largest movements at the top 10 percentile.