The fluctuations of China's energy intensity have attracted the attention of many scholars, but fewer studies consider the data quality of official input-output tables. This paper conducts a decomposition model by using the Divisia method based on the input-output tables. Because of the problems with input-output tables and price deflators, we first produce constant prices to deflate the input-output tables. And then we consider different levels of biased technical change for different sectors in the adjusting the input-output table. Finally, we use RAS technique to adjust input-output matrix. Then the decomposition model is employed to empirically analyze the change of China's energy intensity. We compare the decomposition results with and without biased technical change and do sensitive analysis on the level of biased technical change. The decomposition results are that during 2002-2007, the energy intensity of coal and electricity increased, the changes were mostly attributed to the structural change and the contribution was 594.08%, 73.88%, respectively; as for crude oil and refined oil, the energy intensity decreased, the changes were mostly attributed to the changes in the production technology and the contribution was 978.89%, 246.95%, respectively. And the results of sensitive analysis shows that 1% variation of the level of biased technical change will cause at most 0.6% change of decomposition results. Therefore, we can draw our conclusions: compared to the decomposition without biased technical change, decomposition results are sensitive to the level of biased technical change; the level of biased technical change can be determined by the difference in the change rate of total factor productivity and energy efficiency. (C) 2014 Elsevier Ltd. All rights reserved.