Journal of Investment Strategies
Editor-in-chief: Ali Hirsa
Volume1, Number 1 (December 2011)
Both the science and the practice of investment management have taken huge leaps in the past decade. The explosive growth of the alternative investment industry has reflected the similar growth in the new methodologies and approaches to managing investments, which have gone far beyond the classical portfolio management techniques that allocate between stocks and bonds. Correspondingly, much of the research efforts and innovation coming from both the academic community and industry practitioners has been concentrated in these new areas.
This development has produced a niche in the existing journal publication space, with the established magazines still being mostly targeted toward either a specific product area (such as equities, fixed income or commodities), a specific aspect of quantitative modeling (such as pricing of derivatives, computational methods or risk management), or the "classical" asset management style, with a significant underrepresentation of the more modern styles of investment management that often incorporate many of the above-mentioned topics.
With this in mind, Risk Journals has decided to launch a new journal: The Journal of Investment Strategies. Following the format of others in the Risk Journals series, it will be a quarterly publication, containing in-depth research papers and discussion papers on technical and market subjects. We also envision regular thematic issues focusing on specific areas of investment strategy.
As its founding Editor-in-Chief, I believe that this journal, dedicated to the rigorous treatment of modern investment strategies - and going well beyond the "classical" approaches both in its subject instruments and in its methodologies - will be a timely and much-appreciated addition to the field. Given the particularly challenging times in the market and the diversity of opinions about what has been happening and why, I believe that plenty of material will be available to start the discussion on the pages of our journal, and that such material will prove to be very useful to a wide audience among investment professionals and researchers everywhere.
The Journal of Investment Strategies is led by an all-star Editorial Board that includes highly respected experts both from academia and from different sectors of the industry where strategy ideas are generated and applied. The academic advisory board features Professors Robert Engle (New York University), Kenneth Froot (Harvard University) and Robert Jarrow (Cornell University). The Editorial Board members bring insights from leading universities, major sell-side banks and independent research institutions, some of the largest traditional asset management firms and pension funds, highly respected hedge funds and sovereign wealth funds.
This journal will treat investment strategies in an integrated fashion, promoting the cross-pollination of ideas between researchers and practitioners in the hope of achieving a unique place in the nexus between academia and industry on the one hand and between theoretical and applied models on the other. It is my hope and belief that our journal will set a new standard for the quality of strategy publications and that it will attain a benchmark status as an independent platform in which rising stars in this subject area will find their footing and their following.
To aid our contributors, the Editorial Board has devised a subject set for prospective papers that it considers to be most relevant for the purposes of our journal (please see further down the page). It is not an exhaustive list, and there will certainly be interesting papers that do not fit neatly into this classification. But if an article fits in one of these categories, then we believe it to be a good candidate for our journal.
We do not limit the journal to a specific market or style of investment, but we also do not plan to accept submissions on topics that only have marginal relevance to investment strategy, even if they are otherwise of excellent research quality. The ideal contribution to our journal would not only answer the questions "what?" (ie, what market behavior or type of strategy is being discussed?) and "how?" (ie, how is the prototypical method or strategy constructed?), but also "why?" (ie, why is this methodology or strategy interesting and what was the intellectual motivation for focusing on it?) and "wherefore?" (ie, what are the reasons that the market and/or the instruments behave in the way it supposes that they do?).
Ours is a quantitative journal, and we expect the scientific level of the accepted articles to be very high, both in terms of sophistication of the computational and statistical methods and in terms of clarity of the exposition of technical details. In this area, the most interesting and influential research papers are often not those that propose some specific strategies, however impressive they might be, but those that analyze the behavior of certain broad types of strategies and help to elucidate hidden risks and particular conditions under which they would be successful. The Editorial Board is also interested in more general methodology papers that introduce either new portfolio management concepts or specific novel methods with direct relevance for investment strategy construction, such as statistical estimation and forecasting methods or valuation frameworks that can lead to fundamental or relative value strategies.
In addition to research papers, the journal will also include an Investment Strategy Forum featuring short articles of a topical nature, opinion pieces and discussion and educational articles helping the audience to orient themselves in the flow of market events and research and industry trends.
The inaugural issue of The Journal of Investment Strategies includes three very interesting and diverse research papers and a discussion paper that span several of the themes highlighted in the introductory notes above.
In the first paper of the issue, "Option strategies based on semiparametric implied volatility surface prediction", Francesco Audrino and Dominik Colangelo present a sophisticated statistical machine learning framework for predicting short-term changes in the implied volatility surface in the equity index and stock options markets, and they construct a relative value investment strategy based on these forecasts. The method utilizes the nonparametric regression tree approach to improve the forecast of the volatility surface for up to ten days into the future, which in turn allows them to come up with forecasts of the expected returns for long and short positions in (almost) at-the-money put and call options. Taking the S&P 100 index and its constituents as their coverage universe, the authors then specify a relative value strategy that goes long options with highest predicted returns and short options with lowest predicted returns, while maintaining equal notionals across the positions and dollar neutrality between longs and shorts. They demonstrate a positive performance from their forecasting model and from the proposed investment strategy over a two-year out-of-sample period.
While discussing the robustness of the results, the authors note that their sophisticated forecasting methodology does not significantly outperform a simpler sorting method based on comparing the implied volatilities of the options with past realized volatility of the underlying. Unfortunately, the presented test period (2004-6) is too short to be able to tell whether this is a temporary effect (as I suspect it might be) due to an unusually good performance of the simple method in that relatively quiet historical period. Only a much bigger test, preferably encompassing the experience of 2008-10 as well as previous market downturns, such as those in 1998 or in 2001-2, would enable one to truly distinguish the benefits accruing to the strategy from the better forecasts of the implied volatility surface.
Another important side note is that, all else being equal, having a good implied volatility surface forecast usually benefits the option strategies with longer expiries, while the short-term options strategies benefit more from the predictions of the realized volatility (or the spread between the implied and realized volatilities). Therefore, I would expect an even better (and perhaps far more competitive, compared with naïve strategies) result from applying the methodology presented in this paper to trading of six- or twelve-month options as opposed to the one-month contracts covered by the authors (although that would require a relaxation of the authors' simplifying assumption of holding the options to expiry and would complicate the profit and loss computations).
In the second paper, "Constructing the best trading strategy: a new general framework", Philip Maymin and Zakhar Maymin present a general framework for evaluating the performance of market-timing strategies based on a historical signal that has some predictive power, characterized by a known conditional probability distribution of asset returns given the value of the signal. The main insight of this paper is that it changes the perspective of the conventional approach to such problems and, instead of studying the given strategy (which can be defined as the signal itself plus the portfolio construction rules) for robustness and attractive performance, it asks the more general question, what is the best possible strategy that can be constructed given a particular signal? Two alternative definitions of the "best strategy" are considered: the one that corresponds to the highest expected return and the one that corresponds to the highest information ratio. Posing a question in this way, the authors are able to derive quite general characteristics of the optimal strategies and, as a result, to judge all other strategies in terms of their relative performance with respect to the optimal benchmark. Even when the optimal benchmark is not feasible for various pragmatic reasons, the explicit construction of the benchmark is still helpful as it can guide portfolio managers in selecting the tradeoffs between various risk, turnover and other restrictions that they make in the course of building tradable strategy portfolios.
A desirable next step in this direction of research would be to generalize the results of the presented framework by taking into account the uncertainty of the parameter estimation for the conditional probability of returns given the signal (parameters such as the correlation and volatility of the conditional returns). Once this uncertainty is taken into account, the specification of the best strategy will probably acquire closer semblance with more conventional rules of thumb, such as the smoothing of historical signals and gradual switching between the maximum long and short positions for intermediate values of the signal. It would also allow the characterization of a "distance" of a given strategy from the optimal one in universal units, such as the return differential normalized by the uncertainty of the return prediction.
In the third paper, "On a multi-timescale statistical feedback model for volatility fluctuations", Lisa Borland and Jean-Philippe Bouchaud present a novel timescale statistical model for predicting the volatility of asset returns. They start by presenting an agent-based framework that explains the origins of the nonlinear statistical feedback model of asset returns (Borland (2002), cited in the paper) and which neatly incorporates the power-lawfat-tailed distributions of returns that are frequently observed in practice. The extension of the single-period agent-based framework to a multiperiod setting leads to a natural generalization of this model to multiple timescales in the spirit of quadratic autoregressive conditional heteroskedasticity (QARCH) models and allows Borland and Bouchaud to specify a relatively parsimonious model that is capable of reproducing many important statistical features of the financial time series.
The model shows excellent correspondence with most of the stylized facts and empirical universalities observed in financial returns, including clustering and persistence (long-range correlations) in volatility, the asymmetric response of the volatility to past negative and positive returns (leverage effect), and the slow convergence to normality of the aggregated returns, particularly in cases exhibiting large asymmetry when the skewness of the aggregate returns distribution grows over extremely long time horizons before reverting back to the normal (zero) limit. In addition, the model is even capable of explaining such complex observations as the near-criticality and multifractal characteristics of the returns process, anomalous volatility relaxation after shocks, and the time-reversal asymmetry of the returns process - and to do all this with just a handful of parameters. Moreover, the authors are able to significantly narrow down the range of acceptable model parameter values based on a "soft calibration" methodology, which allows them to circumvent the otherwise unwieldy task of exact calibration of a process with a very long memory. They also demonstrate an alternative method for estimating some of the model's parameters based on direct volatility forecasting, finding good agreement with earlier results.
While there are many well-known models of time-varying (or stochastic) volatility, it is difficult to cite one that can simultaneously capture so many stylized facts with so few model parameters over such a great range of time horizons (spanning from minutes to years). In my view, the Borland-Bouchaud model is one of the best there is for predicting future volatilities, especially in cases when the nature of the time series is likely to have been unchanged over many years, such as for broad market indexes, currencies or commodities. In the case of individual company stock returns one must always be careful not to mix periods when a particular company has changed its relevant characteristics substantially. In particular, this applies to cases when it has rapidly grown from a small or medium size to become a dominant large-cap leader, when it has undergone capital structure changes that have resulted in significantly higher or lower balance sheet leverage, or when it has changed its mix of businesses and industry focus over the years, whether due to acquisitions or internal evolution.
The paper does not actually delve into specific applications of the model to particular investment management problems, but the authors suggest a broad range of such problems for which the model can clearly add value: from risk management to asset allocation to option pricing, and from relative value estimation to defining the optimal hedging strategies for derivatives trading. I would also add that such a model should be very valuable for a growing number of volatility investment strategies, where an accurate forecast of future volatility (as well as volatility of volatility and its distribution) is obviously critical. (An application of the model to each of these areas is a formidable research task in itself, and we hope that the authors will undertake some of these tasks in the future: we would be very happy if they return to the pages of our journal to share the results.) Perhaps the only drawback of the model is that it lacks the analytical (or even easy numerical) tractability that would make its use practical in some cases - but who said life was supposed to be easy?
In their brief paper "Perspectives on systemic risk" in the Investment Strategy Forum, Dean Curnutt and George Lam discuss the practical challenges faced by investors who wish to implement tail-risk hedging strategies. The authors begin by illustrating the historical facts regarding instances of systemic risk flare-ups in the past two decades: notably the 1997-8 period, which started with the Asian currency crisis and ended with the Russian default followed by the liquidity crisis related to the fall of Long-Term Capital Management; the 2000-2002 period, which started with the bursting of the technology bubble and ended with a US recession; and finally the 2007-to-present period, which started with the subprime credit crisis in the US and has gone through several phases, culminating in the current European banking and sovereign debt crisis. They emphasize the importance of understanding the linkages between the apparently different parts of the financial markets and the economic system and of recognizing the potential for the feedback effects to amplify the risks across all markets.
They then proceed to offer a relatively simple and easy to understand (even if not so easy to implement in practice) "tail-risk action plan". They argue for adherence to simple solutions, liquid exchange-traded securities and easily comprehensible payout patterns, while always keeping in mind the cost of hedging versus the projected losses in the main investment program. The latter is no trivial matter, given the currently high levels of volatility and even higher levels of volatility of volatility, making the carry costs of an inefficiently executed tail-risk hedging program prohibitively expensive. As their chart of cumulative volatility premium illustrates, timing is also a very important factor in building a good tail hedge.
With its pragmatic advice, the paper illustrates well the objectives of our Investment Strategy Forum, which focuses on such topical discussions and simultaneously serves to educate our broad readership on important advances in the theory and practice of investment management.
On behalf of the Editorial Board I welcome you to our inaugural issue and look forward to receiving your insightful comments and contributions and to sharing them with eager audiences worldwide.
Papers in this issue
Option strategies based on semiparametric implied volatility surface prediction
Constructing the best trading strategy: a new general framework
On a multi-timescale statistical feedback model for volatility fluctuations
Perspectives on systemic risk