I evaluate the out-of-sample forecasting performance of five models of Chinese and Indian energy consumption. The results are mixed, but in general the auto-regressive distributed lag and unobserved components models perform the best over multiple evaluation criteria. I then use these two models...