Frontier of error minimization from copula model application: Evidence from dependence structure of BRICS's stock markets

Abstract

© Published under licence by IOP Publishing Ltd. This study is proposed to focus on the comparison between Maximum Likelihood estimation (MLE) and Maximum Entropy bootstrap testing (MEboot) in AR-GARCH model in order to seek the frontier of error minimization or the minimum error terms by employing a cross-entropy selection approach. Empirically here, this paper finds that the error terms estimated by the MEboot are smaller than those from the MLE. Therefore, the MEboot estimator is proposed to estimate time series data like financial data because it is more robust statistically than MLE, thus MEboot can be resistant to errors in the results. Consequently, the MEboot test's residuals are selected for estimating mathematically an econometric model called D-vine copula which is based on Gaussian distribution to investigate the pattern of dependence structure in BRICS countries. The results showed that Brazil was the major stock market in BRICS that investors will be interested in its financial flows. Accordingly, we had better understand the central point of capital flows inside BRICS's financial systems. In addition, the dependence structure results show that most financial flows into Brazil come from the Asian continent (India and China). On the other hand, capital out-flows from Brazil have been destined to Russia and South Africa.