Based on data of China Migrants Dynamic Survey (CMDS) conducted in the origin and destination in 2015, this paper compares the structural differences of the floating population under the two survey perspectives, with the mini-census in 2015 as the benchmark. The results suggest that the survey conducted in origin is closer to the results of the mini-census in age structure, marriage, and distance of movement. The survey conducted in destination is closer to the mini-census in terms of residence time. Women in the destination areas and short-term (1 year or less) migrants in the origin areas are over-sampled. Both surveys focus on the population of low education level while neglecting the population with college degree or above. From the perspective of survey methodology, recruiting different respondents is one of the most important sources of survey errors, especially measurement errors. In addition to coverage error, non-response error is also one of the important factors affecting the survey results and sample structures.