Journal of Investment Strategies

Welcome to the latest issue of The Journal of Investment Strategies. We are pleased to present you with four papers devoted to analyzing the design and performance of portfolio optimization methodologies, the construction of trend-following strategies, and multi-asset indexing solutions.

In the first paper of the issue, Edwin O. Fischer and Michael Murg aim to test empirically the performance of several optimization algorithms that exist in the literature and then compare them, in both a single-regime market and a two-regime market. Traditionally, portfolios are optimized with a single-regime Markowitz model, using volatility as the risk measure and historical return as the expected return. This paper shows how the combination of alternative risk and return measures within the regimeswitching framework gives significantly better results in terms of both nominal and risk-adjusted returns (modified Sharpe ratios).

The paper addresses a very real problem facing many investors, who clearly observe the variability of market risk conditions and yet often follow traditional academic portfolio construction techniques that ignore this variability. I am sure that many portfolio managers throughout the industry will find useful insights in this paper.

In our second paper, Zura Kakushadze discusses aspects of optimizing weights for alpha streams (by alpha streams the author means a sequence of predictions of expected returns for each asset given by different models employed by portfolio managers). The paper begins with the simpler case of independent alpha forecasts with a diagonal covariance matrix, then moves on to the nondiagonal case using a factor model, and ends with the inclusion of linear and nonlinear transaction costs. A novel feature of this paper is that the author reduces the multivariate optimization problem to a one-dimensional search problem that is much more efficient to solve numerically.

While traditional long-only portfolio managers rarely face such a problem, it is nevertheless a very common situation confronted by quantitative long-short hedge fund managers, who routinely have to combine many different models to come up with the final "blend" portfolio. The common naive approach of running different substrategies, each of which follows a specific single model is clearly inefficient. Kakushadze's approach offers a consistent way of running a combined portfolio in a more efficient manner.

In the issue's third paper, Guillaume Bernis and Simone Scotti show how to handle the problem of trend detection in the context of trend-following trading strategies, when the data is potentially erroneous. They focus on the case of a filtering method based on wavelets. This is used, for instance, to build an estimator of a given security at a future time horizon, or to construct trading signals based on extreme variations from the trend. The authors study how the erroneous observation of past data is incorporated into the filter method and, therefore, into the estimator built with it. The techniques of error calculus with Dirichlet forms are applied to see how the errors affect the estimation: they define an expansion of the estimator in terms of its first and second-order moments, interpreted as statistical variance/covariance and bias.

The questions raised in this paper are very important for many commodity trading advisors, and more broadly for systematic trend-following strategies, which represent a sizable chunk of global hedge fund investments. While some of these strategies are built on rather simplistic estimators (such as slow and fast moving averages), many of the advanced ones follow more intricate trend-detection techniques. Although the authors address the issue of error influence in a specific algorithm, I believe it is a wider problem and that practitioners can take valuable insights from the author's investigation.

Finally, in our forum section, Xiaowei Kang, Aye Soe and Keith Loggie from the Index Research & Design team at S&P Dow Jones Indices explore the potential role of multi-asset solutions in the indexing landscape as well as challenges in constructing multi-asset indexes. Due to their traditionally bespoke origins, the indexation of multi-asset strategies takes on an additional layer of complexity as there is very little consensus on the asset classes that underlie the strategies or on portfolio construction. Depending on the index construction methodology, a multi-asset index targeting a specific outcome may yield different risk and return characteristics. They demonstrate this using a stylized example of a multi-asset income strategy.

This paper provides an interesting perspective from a leading index provider and I am sure it will attract the attention of institutional investors as it is likely to be an area of future product development and innovations in the index-based investment industry.

I would like to thank our readers for their continued support and interest, and hope that they will find something useful in this issue of The Journal of Investment Strategies.

Arthur M. Berd
General Quantitative LLC