关键词: PSM-DID, 特殊类变量, ps值序列匹配

Abstract: In the application of PSM-DID model,there is always the problem of how to apply PSM to panel data, which is originally applicable to cross-section data. Two traditional solutions are directly transforming panel data to section data and phase by phase matching. These measures are prone to produce "self-matching" or the problem of unstable control subjects. This paper makes two improvements on the basis of phase by phase matching to overcome these problems: one is to distinguish the matching variables, and the matching is controlled by special class variables and propensity score value (ps value); the other is to improve the matching mode from the original point-to-point matching of the ps value to the whole matching between ps value sequences. In the case study, it is found that when there is a big difference between the control group and the treatment group and there are special class variables in control variables, the improved PSM-DID model can identify the policy effects more effectively.

Key words: PSM-DID, Special Class Variables, ps Value Sequences Matching