Markov Switching Multifractal 모형을이용한 ESG ETF의 변동성 추정과 예측

In this study, volatility dynamics were identified and predicted performance was evaluated using the Markov switching multifractal model for six ETFs (5 ESG ETFs, 1 global stock market ETF) traded in the United States and Europe from January 2, 2017 to January 31, 2023. The results of empirical analysis during the sample period are as follows. First, based on the likelihood, the two-state MSM (1) model was the most suitable, but it was confirmed that there was no significant difference between the models. Second, when comparing ESG ETFs and ETFs that track market indexes based on the highly fitted MSM(1) model, ESG ETFs were reported to have a longer fluctuation cycle than market ETFs. Third, in the performance evaluation of the 1-day and 22-day forecasts, ESGU showed excellent MSM(2) model, ESGD, SUAS, and XSER, but in the case of XAXJ, the MSM(6) model in the 1-day forecast and the MSM(4) model in the 22-day forecast showed high performance. However, in terms of predictability, there were no characteristics that the ESG ETF's prediction had significantly larger or smaller losses than the market ETF's prediction. Fourth, when comparing the loss functions of the 1-day prediction and the 22-day prediction for a model with excellent prediction performance, the case with a long prediction period was relatively high.