Abstract
It is well known that volatilities and correlations of international stock markets tend to increase in times of financial instability. A dynamic rebalancing scheme is proposed where the underlying market volatility functions as a timing device and portfolio is only rebalanced when the underlying volatility regime changes. In addition, the traditional Markowitz mean variance (MV) optimization can lead to an ‘inefficient frontier’ with wrong expected returns. A risk-adjusted expected return (RAER) approach is proposed where expected returns are expressed as a linear function of the risk incurred through a risk-aversion coefficient. The results show that the addition of volatility filters adds value to the portfolio performance in terms of annualized return, maximum drawdown, risk-adjusted Sharpe ratio in the whole out-of-sample period as well as all the sub-periods. Moreover, the proposed RAER approach produces most consistent performance with and without the constraint on short-selling compared to other dynamic rebalancing approaches and a constant equally weighted portfolio.
Notes
See, for instance, Erb et al. (Citation1994), Solnik et al. (Citation1996), Ramchmand and Susmel (Citation1998) and Longin and Solnik (Citation2001).
The volatility threshold is determined using in-sample data (02/01/1989 – 25/12/1991). We calculate the mean and standard deviation of the RiskMetrics volatility forecasts of MSCI USA during the in-sample period, and the volatility threshold is set as the mean plus one standard deviation.
Except the minimum variance strategy, where the optimal weights are those with the minimum expected portfolio variance, the other 3 dynamic strategies calculate optimal maximizing the expected Sharpe ratios.
For simplicity, transaction costs for commodities are set the same level as that of stocks at 0.5% per round trip, while in reality much lower transaction costs can be obtained in the futures markets.