Promoting connotative development of the higher education means improving efficiency of the higher education.Based on the input-output perspective,this paper firstly established the framework of calculating the efficiency of Fujian's higher education according to the Data Envelopment Analysis(DEA) method to measure the level of Fujian's higher education.To avoid the routine of setting models and choosing variables artificially before the estimation by most previous studies which may lead to systematic deviations,this paper then introduced the model uncertainty and the use of Bayesian Model Averaging(BMA) to filter the potential influencing factors of Fujian's higher education and finally classified and sorted the chosen variables of affecting Fujian's higher education in accordance with their importance measured by Posterior Including Probability(PIP) and robustness.The results showed that inputs and environment have significant impacts on Fujian Province's higher education but the positive and negative effects co-existed.Therefore,Fujian Province should regulate the way of resource inputs into higher education in the short term;but in the long run,Fujian Province must strive to enhance the input-output efficiency.