DOES MRS-GARCH OUTPERFORMS THE SINGLE GARCH-TYPE MODELS FOR FORECASTING?

• Main Author: Pereira Conte, Bruno; Universidade Federal de Santa Maria
• Publication Date: 2019
• Other Authors: Ceretta, Paulo Sérgio; UFSM
• Language: eng
• Source: Revista de Desenvolvimento Econômico (Online) - RDE
• Download full: https://revistas.unifacs.br/index.php/rde/article/view/5726
• topic: Forecasting; Value-at-Risk; Volatility; MRS-GARCH; Stock market; Previsão; Value-at-risk; Volatilidade; MRS-GARCH; Mercado acionário

This paper evaluated if the Markov switching regime GARCH models can outperforms the single GARCH-type models on forecasting. To achieve this objective, it was used the Brazilian stock market – Ibovespa –  between 2012 and 2017 as study scope. The major difference of our work is the definition of the in-sample and out-of-sample pattern. This paper determined a 200 steps-ahead (10 months) forecast, with model re-estimation at each step-ahead, in order to find conclusive evidence and robust results of the model which has better predictive ability. The results showed that the single GARCH-type models show a slightly better performance for VaR at 5% and switching regime models had better performance at 1% and predictive accuracy considering the most of statistical criteria. Besides that, no model could be determined as benchmark by statistical criteria, which displays that there´s no way to determine a model that outperforms for forecasting on the Brazilian stock market.