Covariance matrix estimation impact in risk-budgeting equity portfolios
Carregando...
Arquivos
Data
2025-10
Autores
Orientador(res)
Maialy, André Cury
Marques, Alessandro Martim
Métricas
Título da Revista
ISSN da Revista
Título de Volume
Resumo
This study examines how different covariance-matrix estimators affect the construction, returns and risk characteristics of equity risk budgeting oriented portfolios. Using daily return data for S&P 500 constituents over 2004–2024, three estimation methods, maximumlikelihood, exponentially weighted, and Ledoit–Wolf shrinkage are applied weekly to construct covariance matrices estimates. These matrices are used with three allocation strategies, inverse-variance, risk parity, and hierarchical risk parity to generate a total of 9 portfolios that are compared against an equal-weight benchmark. Portfolios are evaluated on return, volatility, risk-adjusted performance, and concentration metrics. Results show that while estimation method choice has no clear impact on returns or traditional performance ratios, it significantly alters portfolio concentration: shrinkage estimators yield more diversified allocations, whereas exponentially weighted covariance matrices increase concentration. Forecast errors correlate strongly with realized volatility, especially during market turmoil, but do not vary significantly regarding estimation method selection.
Descrição
Palavras-chave
Covariance matrix estimation Risk budgeting Portfolio selection Shrinkage estimator Risk parity Hierarchical risk parity Inverse volatility Estimação de matriz de covariância Orçamento de risco Alocação de carteira Estimador de encolhimento Paridade de risco Paridade de risco hierárquica Variância inversa