Time-varying copula modeling: a Cholesky approach

• Ano de defesa: 2025
• Autor(a) principal: Tavanielli, Renata
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertacoes da USP; Universidade de São Paulo; Instituto de Matemática, Estatística e Ciência da Computação
• Palavras-chave em Português: Cópulas; Decomposição de Cholesky; Modelo linear dinâmico; Cholesky decomposition; Copulas; Dynamic linear model
• Link de acesso: https://teses.usp.br/teses/disponiveis/45/45133/tde-10022026-223709/

This thesis proposes a new Cholesky Dynamic Copula model to address the challenge of modeling time-varying dependence structures in financial data. Our framework allows for sequential and separate estimation of dynamic dependence across multiple assets, making it scalable to high-dimensional settings. By integrating a discount factor specification within a state-space formulation, our model offers flexibility and efficiency, which makes it a viable alternative to factor models and copula dynamic conditional correlation models. We evaluated the model through a series of in-sample and out-of-sample experiments, comparing its performance against established benchmarks such as the Factor Stochastic Volatility (FSV) model and copula Dynamic Conditional Correlation (cDCC) models. Empirical results show that the proposed model delivers competitive predictive accuracy in covariances, correlations and constructing Global Minimum Variance Portfolios (GMVPs). It also performs well in Value at Risk (VaR) backtesting, across diverse datasets. Our findings highlight the practical and theoretical value of the Cholesky Dynamic Copula approach, offering a flexible and robust tool for financial risk management and portfolio optimization, while advancing the literature on dynamic dependence modeling.