Abstract:      Technological innovation in high-tech industries is an important driver for adapting and leading the new normal of economic development. In view of the low efficiency of technological innovation in Chinese high-tech industries, the government will inevitably solve the problem by increasing the number and intensity of innovation strategies and policies. Therefore, identifying the key factors affecting the efficiency of technological innovation in high-tech industries is of great significance. It is crucial to the effective formulation of strategies, the full release of policy dividend, and the sound and rapid development of industry.
     However, although the existing literature has conducted in-depth research on the influencing factors of technological innovation efficiency in high-tech industries, there is still room for improvement. On one hand, scholars usually select one or a few influencing factors subjectively based on existing literature. It is easy to cause mission and omission of factors. On the other hand, most studies apply the traditional single equation regression model to examine the influence of selected factors. Because of inherent uncertainty problems existing in model, the empirical results will vary considerably depending on different influencing factors chosen and different model settings, resulting in reduction in the guiding value of research conclusions.  Bayesian Model Averaging (BMA) is a method that is suitable for large information sets, and thus can be used to screen explanatory variables and solve the uncertainty problem effectively. Based on above all, this paper adopts the method, trying to make up for the lack of the existing literature. 
     First, this paper uses the DEA-Malmquist method to measure the technological innovation efficiency of 29 provincial high-tech industries in China from 2011 to 2016. The results show that EFFCH has risen in the fluctuation and TECH has decreased slightly in the fluctuation. When EFFCH rises from the previous year, TECH decreases from the previous year, and vice versa. In other words, there is no good balance between the improvement of technological efficiency and the promotion of technological progress. As a result of the unbalanced innovation status, the overall technological innovation efficiency of Chinese high-tech industries has increased slightly in the six years, not achieving significant improvement. From the results of the provincial efficiency estimation, it can be seen that the level of technological progress of most provinces has declined slightly, but due to the improvement of technological efficiency, the efficiency of technological innovation has improved. Therefore, the slight decline in the level of technological progress is the main reason hindering the efficiency of technological innovation.
    Then, based on a comprehensive review of relevant empirical literature in recent years, this paper selected 22 potential factors that may affect the efficiency of technological innovation from the perspective of direct and indirect participants in technological innovation, regional internal innovation environment and external international environment. To find out the precise influence of potential factors, taking the Malmquist index, EFFCH and TECH as explanatory variables respectively while establishing regression equation. 
    After estimation using the Bayesian Model Averaging method, it can be concluded that: (1) The key influencing factors in the direct participants in technological innovation include innovation atmosphere, the tendency of independent innovation and ownership structure. The innovation atmosphere is the key negative influencing factor of Malmquist index, also has negative impact on EFFCH and TECH as well. The tendency of independent is the key negative influencing factor of EFFCH, also has negative impact on Malmquist index and positive impact on TECH. The ownership structure is the key factor affecting the three indices. It has negative impact on EFFCH and positive impact on Malmquist index and TECH. (2) The key influencing factors in the indirect participants in technological innovation include the level of cooperation between enterprises and universities and scientific research institutes, government support and R&D tax. The level of cooperation between enterprises and universities and scientific research institutes is the key positive influencing factor of EFFCH, also has negative impact on Malmquist index and positive impact on TECH. Government support is the key positive influencing factor of Malmquist index and EFFCH, also has the negative impact on TECH. R&D tax is the key positive influencing factor of Malmquist index and TECH, also has the positive impact on EFFCH. (3) The key influencing factors in regional internal innovation environment include the level of economic development and industrial structure. The level of economic development is the key positive influencing factor of Malmquist index, also has negative impact on EFFCH and positive impact on TECH. Industrial structure is the key positive influencing factor of Malmquist index and EFFCH, also has positive impact on TECH. (4) The key influencing factor in external international environment is the degree of openness. It is the key positive influencing factor of Malmquist index and TECH, also has positive impact on EFFCH. (5) Ranking the key influencing factors of the three indices according to posterior probability. The results show that: the key influencing factors of Malmquist index are innovation atmosphere, the degree of openness, industrial structure, the level of economic development, R&D tax, ownership structure and government support; the key influencing factors of EFFCH are ownership structure, government support, the tendency of independent innovation, the level of cooperation between enterprises and universities and scientific research institutes and industrial structure; the key influencing factors of TECH are ownership structure, the degree of openness and R&D tax. When formulating relevant strategies and policies to promote technological innovation in high-tech industries, the government should not simply treat all the key factors equally, nor pay full attention to the important factors while ignoring the minor ones. It is well advised to focus separately on the key influencing factors according to their ranking. Since the level of technological progress is the main reason hindering the efficiency of technological innovation, the key influencing factors of TECH should be given priority control. In particular, since each key factor has different influence directions and degrees on the three indexes, the different influence should be considered comprehensively when formulating targeted policies. Otherwise, it is easy to get half the results with double the effort.

Key words:  technological innovation efficiency, key influencing factors, Bayesian Model Averaging, DEA-Malmquist, high-tech industries