On a Multivariate Extension for Copula-Based Conditional Value at Risk

Copula-based Conditional Value at Risk ( $$\textrm{CCVaR}$$ ) is a real-valued tail risk measure for multivariate random vectors defined through conditioning on a copula level set. This paper considers the extension of $$\textrm{CCVaR}$$ to dimensions $$d\ge 2$$ under Archimedean dependence. While existing work has primarily treated the bivariate case, the multivariate setting involves a d-dimensional conditioning region and does not admit the same low-dimensional representations. For d-dimensional Archimedean copulas, we derive an explicit representation that reduces the defining conditional expectation to a one-dimensional form involving the multivariate Kendall distribution function and a dimension-dependent correction term, enabling tractable computation and calibration in moderate dimensions. We also examine coherence-type properties. Finally, numerical experiments using real data illustrate the behavior of $$\textrm{CCVaR}$$ and compare it with classical $$\textrm{VaR}$$ and $$\textrm{CVaR}$$ under several copula families and marginal specifications.