A RGARCH-CARR-SK model: A new high-frequency volatility forecasting and risk measurement model based on dynamic higher moments and generalized realized measures

The RGARCH-CARR-SK model is developed in this paper by incorporating the characteristics of parameters that directly reflect higher moments in the Gram-Charlier expansion distribution, as well as leveraging the advantages of the RGARCH-CARR model for high-frequency volatility prediction. Simultaneously, we extend the realized volatility measure in the model to explore its efficacy in volatility forecasting and risk measurement under a variety of generalized realized measures. Additionally, we investigate the finite sample behavior of model parameter estimation using Monte Carlo simulations. The result demonstrates that the model exhibits favorable asymptotic performance in parameter estimation across various finite samples. Finally, the empirical study employs the forecasting of high-frequency volatility in the RGARCH-CARR-SK model for China’s GEM and evaluates its effectiveness using various risk methods based on the model. The result reveals that the RGARCH-CARR-SK model outperforms the benchmark models in in-sample fitting, out-of-sample volatility prediction, as well as risk measurement.