- Position: PhD Student - Department of Economics
- Load: PhD Student Full Time
- Principal supervisor: Associate Professor Roselyne Joyeux
- Associate supervisor: Associate Professor Stefan Trueck
- Date of submission: 01/01/2003
- Thesis title: An Empirical Investigation into Dynamic Allocation Strategies.
- Purpose: To empirically investigate the aptness of static portfolio optimization rules in the Australian funds management context.
- Originality: This study uses Australian data to compare performance of the naïve optimization rule to those of sophisticated nature such as Bayes-Stein Shrinkage model, Ledoit-Wolf Covariance Shrinkage model, Michaud-Michaud Resampled Efficiency, Minimum Variance model, and the Mean-Variance model.
- Key Literature/theoretical perspective: Research findings differ on the topic of which portfolio optimization methods perform better. Some researchers, like DeMiguel et.al (2009), Duchin and Levy (2009), and Fugazza (2010), argue that the naïve allocation strategy outperform sophisticated strategies as it does not require the estimation of moments, thus avoiding estimation error. Others, on the other hand, like Kritzman (2010), advocate the superiority of sophisticated quantitative strategies that correct for estimation error highlighting the fact that the naïve strategy assumes lack of knowledge of asset behavior.
- Design/methodology/approach: The approach followed by this study is to evaluate the out-of-sample performance of the sample-based MV portfolio rule – and a variety of its extensions designed to reduce the sample estimation error – relative to the performance of the naïve optimization rule.
The main outcomes from this research are:
- Using various combinations of rolling windows, risk aversion, and holding period returns, we find that, over all, the sophisticated optimization strategies dominate the naïve strategy.
- There is an improvement in the performance of the naïve allocation strategy when the financial crisis, i.e. market volatility, is excluded.
- Advances in portfolio optimization theory proved to be instrumental in creating value for asset managers.
- Research limitations/implications: Despite the limitation of a relatively short historical time series for the asset classes, this research provides Australian asset managers with evidence of the superiority of sophisticated optimization models that correct for estimation errors.
- Keywords: Optimization, Bayesian, Resampled Efficiency, Estimation Error, Shrinkage Estimate