Title: Risk parity portfolio allocation under non-Gaussian returns
Authors: Patrick Walker - University of Zurich (Switzerland) [presenting]
Marc Paolella - University of Zurich (Switzerland)
Abstract: A new method for fast computation of large-scale risk parity portfolios under heavy tailed returns is proposed. Asset returns are modeled with an elliptical multivariate generalized hyperbolic distribution, allowing for fast model estimation and a semiclosed-form solution of the risk contributions. Risk is measured by expected shortfall and conditions are given under which this risk parity portfolio coincides with the one when the variance is used. Exploiting several numerical shortcuts, we can compute the risk parity portfolio exceptionally fast and with high precision. An empirical out-of-sample analysis shows that accounting for heavy tails in risk parity allocations leads to improved portfolio performance and lower drawdown. However, the risk parity strategy is dominated by the minimum expected shortfall portfolio in terms of lower risk and higher Sharpe ratio. The portfolio turnover and proportional transaction costs of the competing strategies are investigated. Regularisation of the objective functions is shown to decrease the impact of transaction costs on the net Sharpe ratios of both strategies. The popular equally weighted portfolio is outperformed even under extreme levels of transaction fees. Additionally, we consider using a GARCH-CCC model, instead of the i.i.d. assumption, in order to investigate the impact of heteroskedasticity on the risk parity portfolio, both under Gaussian and heavy tailed returns.