The China Education Panel Survey (CEPS) in 2014 adopted a multi-stage probability proportional to size (PPS) sampling method, which is divided into four stages of sampling.In the first stage of sampling, administrative units at the county level are selected as Primary Sampling Units (PSUs). Based on the data from the 6th National Census conducted in 2010, a total of 2,870 county-level administrative units with a permanent resident population in the 31 provinces, autonomous regions, and municipalities directly under the central government (excluding Hong Kong, Macau, and Taiwan) are grouped into three sampling frames 2 . From these sampling frames, 28 counties (or districts) are sampled, including 15 core samples and 13 supplementary samples.In the second stage of sampling, schools are selected as Secondary Sampling Units (SSUs). Local cooperating agencies collect the latest statistics of school lists, school types, and school sizes for the academic year 2013-2014 from the education departments of the sampled counties (districts). Based on this information, the research team constructs the second-stage sampling frame and adopts a Probability Proportional 1 The manual of sampling design (written in Chinese) of the CEPS can be retrieved at: http://ceps.ruc.edu.cn/__local/B/2F/1C/3EB9F842184845AC685C3A250FF_870BE0DA_56B46.pdf?e=.pdf. For more information, please visit the CEPS website: http://ceps.ruc.edu.cn/English/Home.htm. 2 Sampling Frame 1: All 2,870 county-level administrative units in the entire country. From this frame, 15 districts (counties) are selected as the core samples. Sampling Frame 2: The 18 districts (counties) under Shanghai municipality. In order to fully reflect the unique characteristics of the mega-city Shanghai, this survey treats the 18 districts as a separate sampling frame. From this frame, 3 districts (counties) are selected as supplementary samples. Sampling Frame 3: The 120 districts (counties) nationwide with a significant number of migrant populations. In order to include more migrant children and children of migrant workers in the sample and thus better reflect the impact of the special characteristics of migrant populations on the educational process and educational inequalities, this survey selects 10 districts (counties) from the 120 districts (counties) with a large migrant population as supplementary samples.
