Welcome to the fall 2013 issue of The Journal of Investment Strategies. In this issue you will find four research papers that cover diverse topics: from fundamental and technical strategy design to efficient portfolio construction techniques and new methods in statistical forecasting.
In the issue's first paper, "Enhancing the profitability of earnings momentum strategies: the role of price momentum, information diffusion and earnings uncertainty", Marc-Gregor Czaja, Philipp Kaufmann and Hendrik Scholz analyze earnings momentum strategies from the perspective of behavioral finance. This is in line with a theory that behavioral biases reinforce this abnormal source of returns. The authors focus on liquid stocks in the German market and conclude that it is possible to find very high alphas after correcting for risk adjustments and trading costs. They show that the rate of information diffusion has a significant impact on their findings, reinforcing the behavioral underpinning of their model.
I would like to comment here on the practical applications of the study. I find it very illuminating, especially regarding the manner in which the authors address the "theoretical biases" of the existing literature, which often make it difficult to use academic models in practical portfolio management. Czaja, Kaufmann and Scholz narrow their focus to the investable universe of most liquid German stocks: these have both ample free-float market capitalization and sufficient daily trading volume, as well as a high degree of sell-side analyst coverage. Moreover, their portfolio construction, based on two-way sorting with earnings momentum and a secondary criterion (such as market cap, turnover volume, dispersion of return forecasts or stock volatility), is quite easy to follow and to implement in practice. The fact that they find a sizeable excess alpha in such a setting tells us that the effects they seek are not phantoms and that they are not subsumed by illiquidity premiums. The fact that these effects persist even beyond expanded efficient-market models, such as Carhart's four-factor framework, gives further weight to the behavioral origins story of the paper.
In our second paper, "The impact of stop losses on short-term countertrend trading strategies", Nicholas Libertini addresses another very practical topic: the impact of stop losses on trading strategy performance. Stop losses are commonly found in many systematic trading strategies, and particularly in the more traditional ones that are driven by technical signals and indicators. The author takes a hard look at the viability of this risk-mitigation technique and finds that it not effective in the case of countertrend strategies.
I find this paper particularly interesting due to the sound analytical framework that it brings into an area that has long been dominated by rules of thumb. While such rules do often work (otherwise they would not have achieved the practical dominance that they have over the years), it is far from certain that they are optimal.We have seen many examples of revisions, with new insights coming from both a more comprehensive statistical framework and a better understanding of the second-order effects of the application of such rules (such as their market price impact). Libertini's paper contributes to the topic by articulating a clear definition of stop loss rules in terms of optimal leverage ratio, related to Kelly optimal sizing of trades, and proceeds to show that the consistently defined stop loss rules do not improve the performance of countertrend strategies. My one side comment would be that the area in which the use of stop loss rules is most dominant is in traditional commodity trading advisor strategies, which are mostly trend-following. Whether these rules add value in that area is therefore still an open question.
The third paper in the issue, "Market crises and the 1=N asset-allocation strategy" by Marcos Escobar, Michael Mitterreiter, David Saunders, Luis Seco and Rudi Zagst, takes another look at the question of which of several risk-based portfolio construction techniques is most efficient. It does this by recasting the question in a conditional form that is dependent on the value of several possible crisis indicators.
The authors consider three different types of crisis indicators, including the socalled heuristic indicator, which is based on identification of patterns in past time series of market returns, the recession indicator, which is based on economic activity metrics as published by the Federal Reserve Bank of St. Louis, and finally the turbulent times indicator, which is built on the basis of a Markov chain regime switching model using market observables. After a comprehensive study encompassing all three crisis indicator types and using several different baseline investment strategies, the authors find, quite intuitively, that during benign market periods portfolios that are more aggressive, such as those based on equal value or equal-risk contribution, outperform others, while during periods when markets are turbulent more defensive portfolios, namely the minimum variance portfolio, tend to perform best.
While the specific recommendations of the paper may or may not be sufficient to build a full investment strategy, I believe that the authors are absolutely right to emphasize the conditional nature of the preference between different portfolio construction strategies. Each strategist can decide for themselves which crisis indicator they prefer, and which baseline strategy they are interested in, but rather than trying to find a single best portfolio construction rule, they would do better to follow the line of thinking in this paper and determine a more dynamic rule for portfolio construction. Of course, any such more complex portfolio construction method might result in additional turnover, and should therefore be evaluated bearing in mind possibly higher transaction costs, but it seems likely that there will be situations in which it is more efficient to switch rules than to stay with the old one.
Our fourth paper is "Efficient high-frequency variance estimators" by Alexander Saichev, Didier Sornette and Vladimir Filimonov. The authors follow a recent Journal of Investment Strategies paper on empirical estimators of integrated variance using the time-bridge technique (Volume 2, Issue 2) with a new paper on high-frequency variance estimators. The authors construct a theory of such estimators, with a special emphasis on homogeneous ones that are built on high and low values in the interval rather than quadratic moments, and show that their estimator is both more efficient than the widely followed Garman-Klass estimator and is (almost) unbiased with respect to drift of the underlying process.
The paper is written in an academically rigorous manner and I would specifically like to encourage our readers not to be afraid to follow the complexities of the proofs and the case studies given: I think that great benefit can be derived from what I believe to be a very good alternative to commonly used volatility metrics. Whatever difficulties are presented by trying to understand its theoretical underpinnings will be more than compensated by the relative ease of its practical implementation and its possible impact on the accuracy of risk estimates in real time.
On behalf of the Editorial Board Iwould like to thank our readers for their continued support and keen interest in our journal. Our pipeline of submitted papers grows stronger with each issue, and I look forward to sharing with you more of the excellent papers on a broad variety of topics related to modern investment strategies that we continuously receive from both academia and practitioners.
Arthur M. Berd
General Quantitative LLC
Enhancing the profitability of earnings momentum strategies: the role of price momentum, information diffusion and earnings uncertainty