Do exports increase the firm's productivity causally? Focusing on the matched information of highly disaggregated transaction and firm‐level data from 2000 to 2006 in China, we construct a new measure of firm‐specific demand shock as an instrument for firm exports, based on the GDP growth rate of destination countries. We find that a one percentage point expansion in exports raises firm total factor productivity (estimated by the Olley–Pakes method) by approximately 0.224 percentage points on average. Moreover, we find that exports to high‐income countries, more processing exports and scope expansion about variety contribute to the learning effect.