이중차분분석의 ‘금지된 비교’ 넘어서기: 이중차분분석의 최신 동향을 반영한 산정특례제도 효과 분석 = Beyond the ‘Forbidden Comparison’ of Difference-in-difference Analysis - An Analysis of the Effectiveness of the Relieved Co-payment Policy Reflecting the Latest Trends in Difference-in-difference Analysis

다국어 초록 (Multilingual Abstract)

Difference-in-difference (DID) analysis is the most widely used research design to evaluate policy effects on public health. Recent literature poses the risk of forbidden comparison, which occurs when heterogeneities of treatment effects between time and population groups. The estimates of DID using a Two-way fixed effect estimator (TWFE) tend to be biased. The TWFE estimate is the sum of the variance-weighted average of the average treatment effect of the treated (VWATT), the variance-weighted average of parallel trend (VWPT), and the variation within the treatment effect (△ATT) over time. Research design, which includes a long post-treatment period as a treatment period, can capture the lag of policy effects and increase the data. Because of its usefulness, it is necessary to review the research design in a way that captures the treatment effect at multiple time points with a single coefficient and to identify the estimation problem in the process. In this study, the recent development in the DID was briefly reviewed, and the research design of the previous study was modified to explain the biasing of the estimates generated in the forbidden comparison. The estimates from TWFE should be interpreted with caution because the DID is not the same as TWFE. It is necessary to accurately identify the estimand to be estimated and connect it with the appropriate estimator. Efforts should be made to reevaluate the design and estimates of previously conducted studies based on the rapidly developing discussions on the DID and to improve the quality of future policy analysis studies.