Abstract: Abstract: The Sloping Land Conversion Program (SLCP) was launched by the Chinese government in 1999 with environmental goals of reducing soil erosion and desertification and increasing the country's forest and grassland cover by retiring steeply sloping and marginal land from agricultural production. The SLCP has converted 8.20 million hectares of cropland to forest, and more than 32 million households participated in SLCP through the end of 2007. According to the article 4 in SLCP Regulation, the ecological service is the first principle of SLCP. Therefore, the non-timber forest products (NTFPs) management is one of the land-use options that SLCP households might choose. Generally, there are two alternatives of NTFPs management, one is the commercial forests management, including fruit trees, nut fruit trees, tea, woody medicine, etc.; the other is understory planting and livestock breeding in the understory, including grains, vegetables, grass, fungus, medicine, chicks, etc. The technical efficiency will be different if SLCP households adopt different NTFPs management options. The objective of this paper is to analyze the technical efficiency of households' NTFPs management considering the different management options. Stochastic Frontier Analysis (SFA) is one of the approaches for technical efficiency analysis, which assumes that the production technology is the same for all production units. However, there are some unobserved differences in technology that might be inappropriately labeled as inefficiency if such variations in technology are not taken into account. Therefore, we use a latent class stochastic frontier model (LCSFM) combined with the SFA and a latent class structure to account for this heterogeneity. The Trans-log function is adopted in LCSFM finally through a series test, and the seedling and labor input are considered as the criteria of technology class. The empirical analysis is conducted by using 235 households' survey data from Ya'an City Sichuan Province, and some important results are revealed. First, there are four technical classes on the different technology frontier and the technical efficiency of them can be listed in the descending order, i.e. class of livestock breeding and medicine in the understory, class of fruit and woody medicine, class of grains and vegetables in the understory, and class of nut fruit,. Secondly, the technical efficiency of all households' NTFPs management increased from 0.5393 in 2008 to 0.5572 in 2011. Except for the class of grains and vegetables in the understory, all other classes showed the same increase trend. To that end, we estimated the factors' elasticity of different classes, and held that it is the elasticity different that explains why SLCP households are divided into different classes. Therefore, some policy implications are given in order to improve the technical efficiency of SLCP households' NTFPs management. First, the government should actively disseminate forestry technology information and train the farmers through different channels. Secondly, the R&D activities and forestry machinery popularization should be promoted, because mechanization greatly contributes to the forestland productivity. It is hoped that these efforts will be contributed to improve the SLCP.