As alternative investment firms and traditional asset managers chase converging business opportunities, hedge funds may take a second look at risk, how they use it, how they talk about it. The development of their "Alpha Engines" may benefit from a different market perspective.
In recent years, more hedge fund managers have put an emphasis on building systematic return-generation models. The business motivations for developing formal "Alpha Engines" are manifold. On the one hand, questions come from process-keen institutional investors. They want transparency and investment philosophy explained.
On the other, introspective questions have pointed towards the limits of some investment ideas. Persistent Beta footmarks have appeared on some supposed Alpha territories. Another reason for building Alpha Engines is an operational one: information overflow. Consistently chasing new Alpha opportunities in a scalable manner requires ongoing manipulation of daily data. Only streamlined processes can now support this exercise.
The importance of "Alpha Engines" will continue to grow over time. In their ongoing efforts to make Alpha-generation processes robust and scalable, hedge fund managersare rediscovering the charms of fundamental multi-factor models (MFMs). Widely used in risk management to forecast market volatility and identify its drivers, the origins of fundamental MFM are in explaining securities' returns.
As such, these models contain structured, ongoing return-based market information. It ranges from the definition of intuitive systematic factors, their performances and volatilities, to how individual assets are dynamically exposed to these factors.
Risk professionals have been aware of the advantages of using this multi-factor approach for many years. First, MFMs represent an objective communication tool between teams with different business interests, and funds' investors. Whether a portfolio manager wants to short a particular stock, or runs a long/short equity book with undesired exposures, risk managers have occasional reasons to disagree with their investment colleagues. These healthy divergences are easier to settle when they emerge from applying the company's "rules book", written in a language understood by all parties. A detailed MFM often provides this language layer. The "rules book" itself remains the ongoing work of investment and risk teams, together with managing partners.
The second aspect is the consistent market picture that MFMs produce over time. When built around stable, intuitive dimensions, these models provide insight into market sources of ex-ante volatility, and changes in correlations. The ongoing visualization of risk through understandable factors provides structured information. For example, bucketing Japanese equity factors by risk level enables analysts to distinguish high from low volatility market characteristics over time. Some risk managers may prefer to display heat maps, showing ex-ante risk and correlation within markets at any one point in time. In both cases, the multi-factor framework simplifies the complex matter of risk analysis. The underlying model can be used to map market volatilities and covariances. Each individual strategy can then be overlaid as an independent layer of exposures. This framework of volatility analysis can be extended to include factor returns too.
The third motivation is operational efficiency. As their asset base expands, many hedge funds operate as business umbrellas. They host different investment strategies and teams, preferably with positive uncorrelated performances. With the diversification of these strategies across asset classes and regions, risk managers face sizable operational challenges. On the one hand, they need to preserve the firm-wide consistency of their risk analysis framework, including automation flows. On the other, they need to use granular information to "slice-and-dice" the risk of individual strategies down to asset/factor level, on a daily basis. With an ongoing need for broader coverage of assets, valuation libraries and advanced risk models, expert teams often reach a natural operational burden. The frequent choice to outsource the development of valuation tools and risk models is not solely driven by hard economics. Now closer to the front office, risk managers are giving increasing priority to actively supporting core investment activities, on top of risk reporting. The leaner their own operations remain, the easier it is for them to address custom risk analysis demands from across the company.
For many hedge fund managers, the alignment of Alpha generation processes and risk management is another area of development.
At the strategy level, consistency between return and risk forecasts helps investment teams identify skills – like high Sharpe Ratios – to leverage, and investment areas they may prefer to hedge. For example, a manager running a concentrated, international long/short equity book may spend more risk budget on stock selection, his or her proven expertise, and systematically neutralize currency exposures.
With risk-adjusted performance in mind, this alignment can be either of a time-dimensional or directional nature.
The time horizon of a strategy has an important impact on the choice of risk model. A long term risk model used to support a process with a short term Alpha model would most certainly give an inaccurate investment picture. In this case, a risk model calibrated for shorter horizons, correcting for using daily data, would give more sensible volatility forecasts. When used in conjunction with portfolio optimization tools, the calibration of the risk model horizon can also have a noticeable impact on the strategy turnover. For equity long/short funds driven by expected mean-reversion effects in asset prices, testing risk forecasts at various horizons can be an ongoing part of the strategy setting.
The directional alignment of risk and Alpha models is another area in which many investment teams are active. This alignment is useful when a factor-based model used to explain risk does not fully match the portfolio construction process. Investment decisions, risk management and risk-adjusted performance analysis are somewhat independent. With risk now embedded into most portfolio construction processes, the misalignment is of a different kind: while risk and Alpha time horizons often match, the same cannot always be said of Alpha-seeking directions and risk factors.
Inherent to time-varying investment themes, the alignment of risk and Alpha models can be viewed in different ways: full customization, flexible grouping, or remapping of factors. Full customization implies developing a unique risk model, exactly matching the chosen investment dimensions. The costly maintenance and enhancement can only be justified by a stable investment process over time. This often comes with a strategic motivation for a firm to gather assets around few, scalable, core investment ideas.
Factor grouping is a more accepted and simple approach. A manager following different Alpha themes through time can choose to bucket risk factors to simplify visualization, while still benefiting from a very granular volatility model. It is like reassembling all the pieces of a puzzle in different groups, over time. The operational flexibility bucketing provides makes it very popular. Teams using different Alpha ideas can all use a central risk model, and develop their own bucketing schemes. These often include regions, thresholds of valuation ratios, or aggregations of styles.
The remapping of risk factors addresses a more technical issue. Unlike the grouping approach, remapping involves projecting variance information, captured by existing risk factors, onto user-specified factors that reflect the investment decision-making process. For example, the information captured by a "Momentum" risk factor can be remapped onto an individualized "My Momentum" factor, used by Alpha managers to construct expected returns in their own investment process.
As explained by J.Menchero in a recent working paper, this transformation – when performed using a risk model with high explanatory power – preserves the integrity of the original volatility forecast, and redistributes it onto a different, correlated Alpha vector grid. While the overall risk and return forecasts remain unchanged, user-defined custom factors can overlay the original risk pieces. Benefits of this method may include high consistency in risk and performance attribution, along the same flexible factor-based dimensions.
However one chooses to align risk forecast and Alpha generation, some institutional investors still fail to address the risk pricing question. For these few funds, a misalignment of Alpha generation and risk models can highlight important process mismatches. A long/short equity strategy built upon a stock-picking process in Japan can, for example, be affected by undesired stylebiases. Even with sector-neutrality in mind, these strategies often end up exposed to domestic styles, whose volatility can exceed 6% pa, according to our estimates. If such an unwanted bet generates an occasional profit, "pure luck" may be the most likely explanation.
The broad question of process alignment enables hedge fund teams to better understand the factors upon which more frequent bets should be placed. Solving this often requires open access to the underlying MFM, as well as the operational ability to manipulate data. As to how a particular hedge fund team analyzes factors' risk and returns and blends them with proprietary themes, this remains well guarded information. After all, successful chefs rarely disclose the recipes behind their finest dishes.
Financial engineers tailoring investment products understand that they can only issue structured notes if the risk can be partially hedged at an acceptable cost. For managers facing a hedging dilemma, fundamental MFMs provide information beyond the risk model: individual assets' exposures to MFM factors. Like "sell-side" financial engineers, prop traders and their hedge fund colleagues can use the multiple exposures to fine-tune their own portfolio hedging. For example, traders running a long/short equity strategy can segment hedges into the individual systematic components their books are exposed to. Some of these exposures can be steered with listed Beta instruments such as Futures, or ETFs. Others, like pure industries and styles, can be individually managed in the portfolio design. For those focusing on idiosyncratic stock returns, the use of detailed exposures to systematic factor returns enables them to isolate and measure the residuals they trade upon.
Whether a hedge fund runs absolute return strategies, or intends to beat the performance of market indices with less volatility, is a choice influenced by the Alpha generation process. In both cases though, the definition of a formal benchmark is not required to accept the presence of Beta-related returns. Restating W. Sharpe's original work, W. Barton and L. Siegel reminded us in 2006 about this fact of investment life. Unless overnight cash is the defensive portfolio of choice, Alpha-Beta separation is inherent to any strategy construction. The identification of Alpha and Beta components is helpful for fund managers. When ETFs and Futures are available, it allows them to potentially lower the cost of Beta exposures. With clearer Beta identification, skilled teams can better deploy their Alpha strengths too.
Hedge fund managers would certainly like to consistently produce positive, market-independent excess returns. Rare by most standards, the lean processes able to generate this "portable" performance also have physical limitations. The question of the scalability of a strategy over time is one that matters to both investors and hedge fund managers. When does an independent Alpha reach its scalability limit? As shown in Fig. 1, the popular transfer coefficient (TC) of a portfolio can be used to better understand a fund's capacity. In Fig.1, an active equity strategy is optimized using fixed constraints, and expected returns. Related transaction costs evolve as the size of the fund grows. While this specific simulation result shows a TC peak at the lower end of the size scale, the implied information ratio decreases slowly as the fund's size grows.
Hedge fund managers running this type of simulation may decide to fully deploy their strategy with leverage, to focus on high risk/performance ratios on a small scale, or to market a close-ended fund. This ultimately depends on their business goals.
Many investment process enhancements derived from using fundamental MFMs can simultaneously serve the interests of risk management and Alpha teams. As described earlier, these can be summarized as follows: common risk language, granular factor returns information, and calibrated asset exposures for hedges. Other shared interests include the application of Extreme Value Theory (EVT) to forecasting risk and eventually, designing portfolios. As with MFMs, using EVT helps explain risk beyond traditional VaR. After yet another turbulent August, managing partners, CFOs and lenders to hedge funds are all keen to better understand left-end tail behaviours. Receptive to theory, portfolio managers may be unclear about the practical impact of extreme event modeling on their books. In the absence of a link between EVT and liquid hedging instruments, their daily priorities may well be elsewhere. Public reports of higher correlations of some hedge funds' returns to major equity indices, and the "arrival of the clones" will only add to the motivation for hedge fund managers to refocus on pure Alpha mining.
As Alpha becomes more exclusive, it will soon be more expensive too. This could be good news, but only for true Alpha wizards.
Patrick Braun is a Vice President, Product Management, at MSCI Barra