Rapid assessment of earthquake economic loss is an important basis for disaster reduction and relief. In this article we choose magnitude, focal depth, intensity of the epicenter, seismic fortification intensity, the difference between the intensity of earthquake epicenter and the intensity of earthquake fortification, per capita GDP as the input index, which are used as input layer nodes. the direct economic loss of earthquake is used as output layer node. After the training and simulation analysis of earthquake disaster loss data from 2000 to 2013, the Elman neural network earthquake economic loss assessment model was constructed. By using this model, the direct economic losses of six major destructive earthquakes in recent years are calculated and analyzed, and the results are in good consistent with the actual direct economic losses from field investigation. This model provides a new way to quickly assess the economic losses after the earthquake.