Endogenous Threshold Stochastic Volatility Model: An Outlook Across the Globe for Stock Market Indices

Abstract

Asymmetries and heavy tails are well-known characteristics on compound daily returns stock market in dices. The THSV-SMN–Threshold Stochastic Volatility Modelwith Scale Mixture of Normal Distributions– model has become an important tool for analysis regarding forecasting asset returns and Value at Risk and Expected Shortfall portfolio estimations in order to assess marketrisk.Therefore, under a Bayesian approach,we develop an extensionon the model proposed by Abanto & Garrafa(2019).This extension allows for an endogenous threshold and will be studied under two theoretical frameworks: the use of order statistics and a random walk Metropolis–Hasting algorithm(RWMH). We test themodel extension upon stock market indices across the globe along four regions (NorthAmerica, LATAM,EuropeandAsia) withour proposed RWMH algorithm and compare the results with the original (fixedthreshold) model using goodness-of-fit and error prediction criteria. Evidence shows that stock markets indices differ both within and across regions,yet in most cases the extended model outperforms the original THSV-SMN.Thus,prudence and a personalized analysis per index are strongly recommended.