JEFAS Vol. 26 Nº 52 (2021)
URI permanente para esta colecciónhttps://hdl.handle.net/20.500.12640/4140
Examinar
Ítem Solo Metadatos Value-at-risk predictive performance: a comparison between the CaViaR and GARCH models for the MILA and ASEAN-5 stock markets(Universidad ESAN. ESAN Ediciones, 2021-12-19) Serrano Bautista, Ramona; Núñez Mora, José AntonioPurpose. This paper tests the accuracies of the models that predict the Value-at-Risk (VaR) for the Market Integrated Latin America (MILA) and Association of Southeast Asian Nations (ASEAN) emerging stock markets during crisis periods. Design/methodology/approach. Many VaR estimation models have been presented in the literature. In this paper, the VaR is estimated using the Generalized Autoregressive Conditional Heteroskedasticity, EGARCH and GJR-GARCH models under normal, skewed-normal, Student-t and skewed-Student-t distributional assumptions and compared with the predictive performance of the Conditional Autoregressive Value-at-Risk (CaViaR) considering the four alternative specifications proposed by Engle and Manganelli (2004). Findings. The results support the robustness of the CaViaR model in out-sample VaR forecasting for the MILA and ASEAN-5 emerging stock markets in crisis periods. This evidence is based on the results of the backtesting approach that analyzed the predictive performance of the models according to their accuracy. Originality/value. An important issue in market risk is the inaccurate estimation of risk since different VaR models lead to different risk measures, which means that there is not yet an accepted method for all situations and markets. In particular, quantifying and forecasting the risk for the MILA and ASEAN-5 stock markets is crucial for evaluating global market risk since the MILA is the biggest stock exchange in Latin America and the ASEAN region accounted for 11% of the total global foreign direct investment inflows in 2014. Furthermore, according to the Asian Development Bank, this region is projected to average 7% annual growth by 2025.