Aplicação dos modelos auto regressivos com heteroscedasticidade condicional para análise de volatilidade do preço das acções da Cerveja de Moçambique no período de 2019-2023

Abstract

This study investigates the volatility of Cerveja de Mo¸cambique share prices over the p erio d 2019 to 2023, using autoregressive mo dels with conditional heteroskedasticity. The initial analysis identifid an ARMA (1,1) mo del as p otential, based on the auto correlation and partial auto correlation functions and the Bayesian information criterion. However, the presence of conditional heteroskedasticity in the residuals indicated the need for a more complex mo del. Subsequently, the ARMA (3,2) mo del was selected, following Akaike’s information criterion, as it satisfid the condition of no conditional heteroskedasticity in the residuals. However, diagnostic tests such as Ljung-Box and the Lagrange Multiplier confimed the presence of ARCH. To capture the p ersistent volatility observed, various ARCH mo dels were estimated, including GARCH, EGARCH and GJRGARCH. The ARMA (3,2)-EGARCH (1,1) mo del proved to b e the most suitable for fiting his torical data. Despite this, all the mo dels presented problems related to the normality of the residuals and risk management, es p ecially with regard to the value at risk. The presence of conditional heteroskedasticity in the time series of Cerveja de Mo¸cambique share prices justifis the application of ARCH mo dels for a more accurate analysis of volatility