Abstract: Facing the coming of the risk society, improving the government’s crisis learning ability has become an important issue. This paper focuses on the driving factors and configuration path of government crisis learning, encodes the text data of 16 major emergencies according to five variables in three dimensions: event attributes, organizational factors and external environment, and qualitatively analyzes the univariate and combined effects of conditional variables on government crisis learning through fsQCA. It is found that crisis double-loop learning cannot be determined by a single condition, but is the result of a combination of multiple factors. Organizations that occur in densely populated areas, have higher death toll and economic loss, and have a better internal drive learning culture are more likely to drive double-loop learning. For factors such as external attention, a configuration path needs to be formed with other factors to drive the occurrence of double-loop learning. Based on the results of case analysis, this paper puts forward specific suggestions in order to provide theoretical support for improving the government’s crisis learning ability.

Key words: government crisis learning, configuration analysis, major emergencies