<![CDATA[Fluctuations in exchange rates and foreign stock indices strongly influence domestic stock performance, particularly in the banking sector, which is highly sensitive to global economic dynamics. Traditional financial models often fail to capture the complex, non-linear dependencies between these variables, underscoring the need for more advanced approaches. This study examines the effectiveness of copula-based regression models in predicting Bank Mandiri’s (BMRI) stock price using exchange rates and the Nikkei 225 Index as predictors. Conventional regression methods, such as Linear Regression, cannot adequately capture nonlinear relationships and tail dependencies in financial time series. To address this, we compare Elliptical Copula, Symmetric Archimedean Copula, Asymmetric Archimedean Copula, and Generalized Nested Copula models. Results show that the Generalized Nested Copula Regression achieves the lowest Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Weighted MAPE (wMAPE), effectively modeling asymmetric and tail dependencies that are crucial in financial forecasting. While Elliptical Copula (t-Copula) also provides strong predictive accuracy, Archimedean copulas perform poorly, failing to improve upon linear regression. These findings highlight the importance of flexible statistical models in financial prediction, demonstrating that copula-based regression offers a superior alternative to traditional methods. Unlike prior research that often relied on simpler copula families or linear models, this study introduces a Generalized Nested Copula Regression in the context of the Indonesian banking sector, addressing a gap in emerging market literature. The study assumes correctly specified marginal distributions and a stable dependency structure, which may limit applicability under rapidly changing market conditions. Future work should consider dynamic copula structures and additional economic indicators to further enhance predictive accuracy.]]>
Copyrights © 2026
