Welcome to the third issue of The Journal of Investment Strategies. You will find the same diversity of topics in this issue as in the last two, ranging from methodology to practical applications, from portfolio management frameworks to specific strategy design, and incorporating cutting edge research concepts from both academia and the investment management industry. Indeed, these are the features that our Editorial Board is determined to maintain as the hallmarks of this journal.
For this issue we have selected four papers, each very different from the others, and each offering a unique insight into a particular topic of significant interest for investment managers, strategists and research analysts. I am certain that you will find them as illuminating and useful as I have.
In the first paper, "Rationalization of investment preference criteria", Jacques Pézier tackles the extremely important task of consistently and rationally defining riskadjusted performance metrics. Given the central role that these metrics play in topdown capital allocation throughout the investment management industry, the stakes in this investigation are extremely high: we are talking about a change in methodology that could affect investment decisions involving trillions of dollars.
Pézier begins by giving a comprehensive overview of the risk-adjusted performance metrics that have previously been suggested and that are, to a greater or lesser extent, used in portfolio management applications. A range of metrics - from the familiar Sharpe and Treynor ratios to the somewhat more obscure Omega index, returns on value-at-risk, etc - are thoroughly investigated to determine whether they are indeed fit to be used in practice. This is a monumental task, and it is accomplished in a very pedagogical fashion, revealing not only the statistical properties of these metrics but also the pitfalls of interpretation and hidden assumptions associated with them.
Having shown that all of the surveyed measures fall short of the requirements, Pézier proceeds to suggest a new metric - certainty equivalent return - which he contends is much more rational and gives a much more robust preference ranking in a variety of market situations. Furthermore, this metric paves the way for an appleswith- apples comparison of heterogeneous investment strategies, for better design of structured products and even for helping to explain longstanding financial puzzles such as the origins of credit risk premium.
I fully expect the insights in this paper to eventually take their place alongside (or even ahead of) the mainstays of the financial literature and methodology such as the Sharpe ratio. It behooves anyone that is involved with investment selection and evaluation to study this work thoroughly.
In the second paper, "Momentum strategies for style and sector indexes", Linda H. Chen, George J. Jiang andKevin X. Zhu give a relatively simple but quite illuminating new analysis of the longstanding problem of detecting the momentum component in expected equity returns. Momentum is another word for "persistence", and as such can be applied to many different characteristics. The two most important such characteristics are price returns and earnings. Since Carhart extended the Fama-French original three-factor model to include the technical momentum factor alongside fundamental earnings, valuation and size factors, many researchers have found both price and earnings momentum to be very important for explaining future stock returns.
However, given their technical nature, the results do strongly depend on the specific definition of momentum that one uses. Momentum, in my view, should not be considered as a factor per se, but rather as a family of factors with similar features. Then, depending on whether one uses a straight past return (as in the simplest definitions), an acceleration of the past return (as in some Barra models, which use the difference between short-term and long-term trailing returns) or some risk-adjusted flavor of the past return, one obtains quite different results.
In addition to the mathematical definition of the momentum, another equally important part of the definition is the selection of the universe of securities and the crosssectional procedure for ranking. Both of these seemingly technical parts of the modelbuilding framework have a very strong influence on the results. It is in this direction that the paper by Chen et al makes its most important advances.
Specifically, by using the Morningstar style index construction rules, which allow the universe of stocks to be divided into nine style portfolios across the value/growth and size dimensions, they discover that the indexes corresponding to these style portfolios exhibit strong momentum but show no earnings momentum. At the same time, they present evidence that, in sector indexes built around the industry classification, the observed price momentum is driven by the earnings momentum. I think this result is quite intuitive, since one can indeed expect to see common movement in the earnings of all companies in a given industry sector, which in turn would translate into a common return. In contrast, in the much more diversified style indexes, the returns can be expected to be driven by changes in risk premia rather than fundamentals.
In summary, readers will find that there are many very useful observations in this paper and they will be able, if they are so inclined, to verify and replicate the methodology themselves, given its transparent and accessible formulation.
In the third paper, "Optimal trading with linear costs", Joachim De Lataillade, Cyril Deremble, Marc Potters and Jean-Philippe Bouchaud, a team of researchers from Capital Fund Management, share their latest insights into the optimization of algorithmic trading strategies.
Given the prominence of such strategies in current markets (where, by some estimates, up to 70% of trading in US stocks is done using an algorithmic approach), one cannot overestimate the importance of their analysis. The vast majority of algorithms that are used are actually made for optimal execution, with an objective of minimizing the market impact and reducing the cost of trading. However, some algorithms are also designed for active trading strategies, which complement unidirectional execution trades and allow the market to function in a more liquid manner.
The model presented in this paper addresses a trading strategy with a cap in the maximum (long or short) size of the position allowed, under an assumption of linear costs of trading. Using Bellman's method, they find an explicit solution to the optimization problem and demonstrate that, instead of trading incrementally, one must switch between maximum long and short amounts once the predictor exceeds a certain threshold value. Moreover, this threshold value is further reinterpreted as the optimal "no-trading" band size for the case when a quadratic risk penalty is imposed.
Despite this seeming to be a narrowly focused mathematical study, the results obtained are actually extremely valuable for many practitioners. Trading with maximum size allocation is one of the most common ways of managing many quantitative portfolios, especially ones that frequently change the sign of the trade, such as CTA strategies with futures or foreign exchange. In these strategies, portfolio managers often apply their trading signals and predictors to individual instruments, and manage the capital allocation by assigning maximum positions. Although stocks can also be traded in this fashion, it is far less common, and portfolio managers rely instead on portfolio optimization methods. Thus, if the portfolio manager believes that his/her trading costs can indeed be well approximated by a linear cost function, this paper will provide a formula for trading any CTA strategies. I am sure that many readers will recognize this powerful insight and use the presented methodology in their own practice.
The discussion paper in the Investment Strategy Forum, "When games meet reality: is Zynga overvalued?" by Zalán Forró, Peter Cauwels and Didier Sornette, is extremely timely and topical. It focuses on the valuation of Zynga - a social gaming network closely associated with Facebook - and claims that its share price (as of the time of writing) represents a bubble.
The authors do not make such an assertion lightly. Their analysis, in contrast with the vast majority of "valuation as metaphor" analyses carried out in the era of trendy social media companies, is unusually deep and thorough. They do not take the claims of having vast numbers of users at face value, nor do they extrapolate the potential revenues from these users "to infinity and beyond" as is customary in many sell-side analyst projections. Instead, they develop a detailed dynamic model of a growth/decline cycle of each new game, and show how to calibrate such models on specific game usage numbers. They then value the company using revenue streams from known and potential future games, and eventually come to the conclusion that even the most optimistic estimate of the share price is well below the trading price of the Zynga stock. Finally, they identify the critical timing of events that could serve as catalysts for a revaluation of the share price.
As we reviewed the paper (which was originally sent to us in mid-April 2012), many of its predictions came true. In fact, the authors had to return to the text and add a "postmortem" section to the paper to discuss the events that had occurred since April 17, and how their specific predictions had fared - and they had fared very well.
While I think the magnitude of the price correction in Zynga's stock has surely been influenced by the coincidental flare-up in overall risk aversion due to the ongoing European crisis, I do believe that the authors' analysis was very much on the mark, and that their methodology gives a much more solid basis for valuation of social stocks like Zynga and Facebook than anything I have seen previously. I think this paper raises the bar for the level of serious fundamental valuation in this area of interest, and for the level of detail that a strategy paper should contain.
On behalf of the Editorial Board I would like to thank our contributing authors for their excellent papers, and our readers for their keen interest and feedback. I look forward to receiving more insightful contributions and to continuing to share them with eager audiences worldwide.