Welcome to the third issue of the third volume of The Journal of Investment Strategies. In this issue you will find four papers that cover topics on asset allocation and risk control techniques, and give an overview of a broad set of systematic strategies.
In the first paper of the issue, "Risk-return-efficient target-volatility strategies", Guido Giese explores a broad set of risk-controlled systematic strategies that dynamically reallocate capital between an equity account and a money-market account based on the expected volatility of the equity strategy. This seems like a very natural setting for risk-control strategies, as it covers the case of volatility targeting (a very commonly used technique) as well as a broader set of strategies that can be described according to their choice of volatility response function. Given that the expected rebalancing costs are also proportional to the variance of the risky portfolio component, the definition of the allocation strategy that depends only on volatility remains consistent even when transaction costs are taken into account. Giese not only focuses on the particular cases of specific risk-control strategies, but also derives an optimal response function for both optimal return and optimal Sharpe ratio risk-control algorithms, and proves that they are indeed superior to conventional methods. He then illustrates these results with numerical simulations for an equity index investment.
I like both the premise and the results of the paper. It shows that new and nontrivial results can still be obtained even in an area that has been well-studied both academically and by practitioners. The applicability and efficacy of the results hinges on the quality of the volatility predictions employed by the portfolio managers, and whether they satisfy certain unbiasedness conditions laid out in the paper.
In the issue's second paper, "The Bayesian roots of risk balancing", Hakan Kaya gives yet another view of risk parity, this time from the perspective of Bayesian estimation of the optimal portfolio within the general mean-variance framework. The author's view is complementary to initial approaches to risk parity that emphasized diversification and risk-balance efficiency, as well as more recent studies that have identified leverage aversion as a possible source of abnormal returns. By considering risk parity within a broader class of risk-constrained optimal portfolios, the author identifies the conditions under which risk parity emerges as an effective portfolio choice. The emphasis on managing the uncertainty of parameters in the context of optimization of future risks and returns brings in the Bayesian interpretation. Interestingly, the Bayesian shrinkage that appears in this study is strictly related to estimates of covariance, while the better known Black Litterman Bayesian approach is related to the shrinkage toward equilibrium mean estimates. Thus, it appears that these two approaches to regularizing the mean-variance optimization problem are complementary and there is potentially an interesting combination of the two whereby both risk constraints and equilibrium are considered together.
In the third paper in the issue, "Two centuries of trend following" by Yves Lempérière, Cyril Deremble, Philip Seager, Marc Potters and Jean-Philippe Bouchaud, a team of Capital Fund Management researchers expand the studies in trend following to over two centuries and demonstrate that the trend return anomaly has always been very strong and has been statistically very significant across commodities, currencies, stock indexes and bonds. Despite the fact that trend following appears to be completely contrary to the notions of financial equilibrium, it is just as pervasive now, when markets are thoroughly researched and arbitraged, as it was decades ago when this was not the case. The underlying question that the study addresses is whether the recent underperformance of traditional trend-following strategies is a precursor to the "end of trends" market transition or just a temporary fluctuation within the expected norm. The authors argue that when it comes to long-term trends they do not see any reason to doubt the continued strength of the return anomaly, but when it comes to short-term trends recent years appear to have shown a marked weakening of the expected alpha from this strategy. These conclusions should be especially relevant for investors considering allocation to such strategies in a longer-term framework.
Our fourth paper, "Confidence intervals for the Kelly criterion" by Euan C. Sinclair, continues our Investment Strategy Forum series, in which we publish practical, academic and pedagogical insights into topics in strategy research. This paper on the Kelly criterion certainly demonstrates all of these hallmarks and it will be interesting to readers who wish to better understand this information-theoretic risk-control approach, which despite predating modern portfolio management is still not as widely known or used as it could be. Sinclair covers an important gap in this area by providing additional uncertainty estimates for the Kelly criterion, which in its original form presumes exact knowledge of expected returns and risk parameters. While this assumption is accurate in the game-of-chance setting for which the Kelly criterion was originally derived, it is much less reliable in financial markets. Therefore, knowing the confidence intervals will allow traders to apply the Kelly criterion in a more consistent manner.
Arthur M. Berd
General Quantitative LLC
Volume 3, Issue 3 (2014)
Volume 3, Issue 3 (2014)
Risk balancing has been considered a heuristic asset allocation method. In this paper, the authors show that, on the contrary, risk balancing is a special case of a utility optimization problem with log regularization that constrains risk concentration.