This study investigates the effect of external shocks on stock market volatility, focusing on improving the explanatory power and prediction ability of the heterogeneous autoregressive model of realized volatility (HAR-RV) using intraday high-frequency 5-minute return data from the KOSPI market index over the period from January 2004 to June 2016. Based on previous studies, we use improved methods in our empirical design to enhance the reliability of the empirical results: a method extracting jump components from measurements of realized volatility based on statistical significance evaluation; a method incorporating nighttime market information (without trading) into the measurements of realized volatility; a method assessing whether to improve the prediction ability of a proposed model through statistical significance evaluation; and a robustness test comparing the proposed model with models containing well-known explanatory variables of the volatility leverage effect and realized skewness and kurtosis. This study creates time series data on the external shock variable (ES) using principal components analysis by combining the selected 10-type variables that have the property of external shocks in the international financial markets, raw material markets, and commodity markets. It then uses the ES variable in empirical tests. The main results are as follows. First, the ES variable created from the principal components analysis does well at reflecting large changes in international markets. Second, the ES variable has a significant Granger causality relationship to the realized volatility over the whole period, while each of the selected 10-type variables has a significant causality relationship only for the specific periods of interest. Third, from the perspective of in-sample analysis, the HAR-RV model with the ES variable significantly improves the explanatory power of changes in the future realized volatility. In the out-of-sample analysis, the ES variable has the significant information value of enhancing the predictive power of the model in future periods. Finally, these are robust results regardless of whether the volatility leverage effect and realized skewness/kurtosis variables are included in the HAR-RV model. Our results show robust evidence that the external shocks have meaningful information value for explaining and predicting changes in future volatility in the HAR-RV model, and imply that the methodology of constructing the ES variable may provide new directions for future research using the HAR-RV model.