A Study on KOSPI Return Volatility Estimation Using Heterscedastic Time Series Model

Abstract

In this study, return volatility prediction model was estimated by converting composite stock index into return variables. As for the heteroscedasticity test for the return, the Q-test and the LM-test showed that heteroscedasticity existed at the 2nd and 7th lags. Therefore, the order was determined using the SBC statistic, the parameters were estimated using the ARCH (2) model, and the model fit test was conducted. The parameter estimates of the ARCH(2) model were statistically significant, but the residual analysis showed that autocorrelation existed and did not satisfy the normality test. In the results of applying the GARCH (1,1) model, the parameter estimates, residual analysis, and normality were found to be satisfactory. Therefore, the GARCH (1,1) model was determined as the KOSPI return volatility estimation model and volatility was predicted. Volatility was predicted to show high volatility in the period of January and February 2023, and to stay calm after March for a long-term. Continuing a calm state means that the probability of large volatility is high. Therefore, those in charge of government agencies need to check system improvement or policy establishment and make efforts to identify trends in volatility by market.