This paper proposes the metafrontier non-radial Malmquist CO2 emission performance index (MNMCPI) for measuring dynamic changes in total-factor CO2 emission performance over time. The MNMCPI method allows for the incorporation of group heterogeneity and non-radial slack into the previously introduced Malmquist CO2 emission performance index (MCPI). We derive the MNMCPI by solving several non-radial data envelopment analysis (DEA) models. We decompose the MNMCPI into an efficiency change (EC) index, a best-practice gap change (BPC) index, and a technology gap change (TGC) index, and based on the proposed indices, we examine the dynamic changes in CO2 emission performance and its decomposition of fossil fuel power plants in China for the 2005–2010 period. The empirical results show a 0.38% increase in total-factor CO2 emission performance as a whole and a U-shaped MNMCPI curve for the sample period. Because companies owned by the central government lack innovation and technological leadership, the results suggest a missing link in the role of the central government in promoting CO2 emission performance. •The metafrontier non-radial Malmquist CO2 emission performance index is proposed.•The dynamic changes in CO2 emission performance and its decomposition can be studied using the proposed model.•An empirical study of fossil fuel power plants in China are conducted for the 2005–2010 period.