Different from the previous studies that evaluated the degrees of carbon market efficiency as constant, the novelty of this work is to capture degrees of market efficiency that are time-varying over time and its static and dynamic relationship with liquidity in China's five emissions trading system (ETS) pilots (Beijing, Guangdong, Shanghai, Shenzhen and Hubei) from January 2017 to June 2019. For this purpose, this work employs the generalized autoregressive conditional heteroscedasticity in mean (GARCH(1,1)-M) model with state space time-varying parameter (Kalman filter), Modified Amihud illiquidity ratio, fixed effect variable intercept panel regression model and panel impulse response analysis. The empirical results are as follows: (1) Shenzhen pilot is the best performer for the market efficiency, followed by Beijing, Shanghai, Guangdong and Hubei; although the other four except the Shenzhen pilot are inconsistent with weak-form efficiency, the market efficiency of Beijing pilot has improved and the Hubei pilot has the lowest market efficiency; (2) there are the positively long-term equilibrium relationship between market efficiency and market liquidity of pilots; and (3) market liquidity can positively affect the market efficiency of pilots lasting three months. These findings suggest that the market participation ability, linkages between carbon markets, speculation of low market efficiency and carbon derivatives tools can be considered by unified China's national carbon market.