Abstract: Quality and efficiency improvement of tourism industry is one of the important issues in tourism science researches under the new normal of China’s economy. Based on DEA-Malmquist index secondary decomposition model and variable coefficient fixed effect regression model, this paper explores the spatio-temporal dynamics, influencing factors and formation mechanism of tourism industry’s efficiency from 2001 to 2016 in Hunan Province. The main conclusions are as follows: (1) The efficiency of tourism industry in 14 cities and municipalities in Hunan Province was improving gradually. The main performance is that the scale efficiency drove comprehensive efficiency to realize optimized development, the total factor productivity increased at an average annual rate of 10.9%, and the intensive management level of tourism industry has been improved rapidly. (2) The efficiency and its decomposition of tourism industry present obvious characteristics of concentric agglomeration feature and distance attenuation phenomenon; the pure technical efficiency presents more significant spatial structure of “core-periphery”. (3) The five factors such as government macro-control, informatization level and opening degree have more extensive and significant positive effects on the efficiency of tourism industry than other elements. The influencing factors exist with obvious intensity difference and spatial heterogeneity as well as different compound mode and complexity level. Consequently, there is a significant spatial differentiation characteristic of tourism industry’s efficiency in Hunan province. Generally, the low-gradient districts of the tourism industry’s efficiency are relatively less interacted by complicated elements while the density and intensity of complex factors in the high-gradient districts are more complicated.