Portfolio optimisation using alternative risk measures

We use a numerical methods algorithm based on gradient descent to optimise investment portfolios of global indices using raw and forecasted risk measures at differing frequencies. The results permit a comparison of how the characteristics of risk measures other than the variance and standard deviation impact portfolio performance. Asymmetric risk measures result in superior portfolio returns, while risk measures incorporating unsquared deviations outperform those incorporating squared deviations. Risk measures forecasted using the exponentially weighted moving average (EWMA) methodology do not yield significant increases in portfolio returns. Semi-absolute deviation, mean absolute deviation and downside semi-deviation perform favourably in producing higher returns.