Compare Canonical stochastic volatility model of focal MSGJR-GARCH to measure the volatility of stock returns and calculating VaR

One of the most important challenges in examining the behavior of investors in financial markets is measuring the volatility of financial assets. This is because stock price volatility is a latent variable. There are two basic approaches to modeling volatility in financial economics that differ in their probabilistic structure. In the first approach, volatility is modeled using shocks to stock returns, and in the second approach, volatility is transformed based on a stochastic process that can be independent of stock return dynamics over time. The models presented in the first approach of the GARCH class and in the second approach of the class constitute random volatility and Markov regime change. Despite the superiority of the probabilistic structure of these models, the calculation of model parameters and volatility prediction is very complex, which makes it necessary to use Bayesian methods and MCMC simulations. The results of this study indicate that in the period of this study, the existence of a leverage effect in the Tehran stock market is not confirmed and the MSGJR-GARCH method is more efficient in predicting fifty more active companies of Stock Exchange return volatility based on Bayesian information deviation criteria. Finally, based on the more efficient model, the out-of-sample VaR was calculated for the first seven days.