Technical research on energy productivity can support government officials as they evaluate practical energy policies for the future. This study proposed a parametric method to decompose China's energy productivity rate of change into six factors based on a theoretical stochastic frontier analysis. The method was applied to conduct an empirical study using inter-provincial panel data in China from 1995 to 2012. The results highlighted three key points. First, the general rate of change in energy productivity was mainly influenced by a steady positive rate of change in technical progress, combined with a steady negative rate of change in technical efficiency. The core factors causing fluctuations in energy productivity included: a positive rate of change in the substitution of capital and energy, and a negative rate of change in the substitution of labor and energy. Second, from a geographic perspective, provinces with a high rate of change in technical progress experienced a weaker deterioration in technical efficiency. However, the rate of change in technical efficiency tends to decline as the rate of change in technical progress increases. Third, there is a similar changing trend between the substitution of capital and energy and the substitution of labor and energy.