This paper identifies the causal effect of low-carbon city pilot (LCCP) policy on carbon emission efficiency (CEE). Specifically, we first develop a general nonconvex metafrontier data envelopment analysis model to calculate CEE. We also provide a quasi-experimental evidence using a unique dataset of 251 cities in China during the years 2003 to 2018. Specifically, difference-in-differences (DID) and spatial DID (SDID) estimators are used as the main empirical strategy. We find that the LCCP policy improved CEE by 1.7% which are both economically and statistically significant. Further, its impact on neighbor untreated cities is about 64% of that on the treated cities. Scenario analysis documents that the average carbon dioxide emissions should be mitigated by approximately 8.37 million tons with a CEE increase of 1%, 8.84 million tons with a 2% increase, and 9.31 million tons with a 3% increase. Our findings also indicate that a 1% increase in years relative to a city's carbon dioxide emissions peak year commitment associates with a 1.3% increase in CEE.