Bayesian non-linear quantile effects on modelling realized kernels

This article examines the non-linear responses of a stock market's realized measure of volatility to its potential factors across different quantile levels. Specifically, we apply the threshold quantile autoregressive model with exogenous variables and GARCH specification to model the realized kernel series of the Nikkei 225 stock market. Using this model, we investigate the relationship between the realized kernel series and its lagged one autoregressive effect, intraday returns, and abnormal trading volume. Applying an adaptive Markov chain Monte Carlo sampling scheme, our results confirm the existence of volatility clustering, a leverage effect, and a negative and asymmetric impact of trading volume on market volatility. We discover that the asymmetric characteristic of the news impact curve of the stock market varies over different quantile levels. This finding provides an in-depth understanding about how stock volatility reacts to its determinants and how stock markets operate under different market conditions.

Automated summary

This article examines the non-linear responses of a stock market's realized measure of volatility to its potential factors across different quantile levels. Using a threshold quantile autoregressive model and GARCH specification, the study finds that volatility clustering, leverage effect, and negative and asymmetric impact of trading volume on market volatility exist. The asymmetric characteristic of the news impact curve of the stock market varies over different quantile levels.