Presentation Information
[2Yin-B-54]Synthetic Control Method Using Dynamic Time Warping and Temporal Distortion Index
〇Hayato Ishiyama1, Daigo Fujiwara1,2, Tomonori Izumitani1 (1. NTT DOCOMO BUSINESS, Inc., 2. Shiga University)
Keywords:
Synthetic Control Method,Causal Inference,Dynamic Time Warping,Temporal Distortion Index,Time Series
This study examines a causal inference approach that uses Dynamic Time Warping (DTW) and the Temporal Distortion Index (TDI) for donor selection in the synthetic control method. By doing so, it becomes possible to absorb differences in the speed of changes across units and to incorporate prior knowledge about such differences.
A method called Dynamic Synthetic Control, which uses the DTW warping path within the synthetic control framework, has been proposed. In this study, we aim to improve accuracy by incorporating TDI—an index of temporal distortion—into Dynamic Synthetic Control as a penalty for lags or leads between units.
Using Monte Carlo simulations, we compare Dynamic Synthetic Control with the proposed method. The results show that the proposed method achieves better accuracy in 74% of cases and reduces the mean squared error (MSE) by 33.6% on average, indicating that the proposed method is effective.
A method called Dynamic Synthetic Control, which uses the DTW warping path within the synthetic control framework, has been proposed. In this study, we aim to improve accuracy by incorporating TDI—an index of temporal distortion—into Dynamic Synthetic Control as a penalty for lags or leads between units.
Using Monte Carlo simulations, we compare Dynamic Synthetic Control with the proposed method. The results show that the proposed method achieves better accuracy in 74% of cases and reduces the mean squared error (MSE) by 33.6% on average, indicating that the proposed method is effective.