Abstract
Vine copulas are well-established and commonly used tools for modelling and forecasting financial time series, largely, favoured for their flexibility in capturing complex dependence structures among multiple time series. However, despite their strong empirical performance, the construction of vine copulas can become computationally demanding in moderate and high-dimensional settings, partly, because ordering procedures are typically required to mitigate the combinatorial complexity of model specification. In this paper, while acknowledging the importance of ordering in vine copula construction, we investigate its practical impact on risk measures forecasting and portfolio allocation. Specifically, we evaluate whether the widely adopted R-vine ordering algorithm proposed by Dissmann et al. (2013) offers significant advantages over a less demanding alternative: a D-vine copula based on an arbitrary ordering. Through Monte Carlo simulations and two empirical applications, we show that, after appropriate marginal modelling and pairs copula selection, the specific ordering of series plays a limited role in forecasting risk measures, but remains relevant for portfolio construction.