中文摘要：

以中国家庭收入项目（cHIPs）纵向数据为样本，采用随机前沿法（sFA）和样本选择模型，分析了我国过度教育现状及其演变。研究发现：随机前沿模型设定合理，而忽略样本选择会导致严重偏误；教育回报历年来显著为正，男性、东部地区、一级劳动力市场、正规部门、固定长期工的效率更高；过度教育呈现先上升后下降的趋势，不同性别、所有制及不同地区的教育不足与过度教育并存；与发达国家相比，我国的过度教育水平不高，教育投入总量不足。

英文摘要：

Based on the longitudinal data of China Household Income Project （CHIPS）, this paper analyzes the current situation and trend of over-education in China by using stochastic frontier approach （SFA） and sample selection model. It is found that the stochastic frontier model is reasonable and ignoring the sample selection will lead to serious errors. The returns to education were sig- nificantly positive over the years; the efficiency of men, the eastern region, the primary labor market, the formal sector, long-term workers and contract workers were higher. The ratio of over-education appears rising at first and falling subsequently with the expan- sion of higher education. Over-education and inadequate education exist at the same time between different groups, such as genders, regions and ownership systems. Compared with developed countries, over-education level is not high in China, the total investment in education is insufficient.