Financial risk management traces back to maritime trading in the Mediterranean Sea: merchants knew not to send all of their goods aboard one ship. Using modern terminology, they reduced their risk by diversifying their investments across uncorrelated journeys. Different crews, different weather conditions and different boats did the diversification work of ensuring that risks were lowered.
The same approach – of limiting exposure to individual entities – remained the state of the art in risk management until very recently, when modern finance allowed for detailed calculations of risk with a bewildering array of choices like exponential weighting of look-back periods, volatility surface splining, and interest rate bootstrapping. Since the global financial crisis, we’ve also seen a growing number of boutique specialist firms offering their particular selection of available risk methodologies. Such specialisation and fragmentation has prevented them from offering a compelling combination that almost every buy-side investment manager has sought – performance attribution coupled with comprehensive risk.
Comprehensive risk analytics include a wide range of different methodologies, approaches and types of risk measures including variance-covariance, Monte Carlo, historical simulation, stress testing and revaluation simulation. Unfortunately, most of the recent innovations in risk have been limited, by and large, to improvements in existing calculations – fulfilling Lord Kelvin’s admonition that future practitioners would be limited to calculating the same quantities to the next decimal place. But there are three notable exceptions where risk innovation has recently occurred: liquidity risk, hybrid factor modeling and the combination of performance and risk.
Volume based liquidity risk measures, despite their ubiquity, are useful only when markets are operating normally – precisely when liquidity is not an issue. Some firms have taken to survey-based liquidity measures, relying on market participants to honestly and accurately share their liquidity issues in order to identify them for others. But a more reliable approach is to use bid-ask spreads, calibrated and treated as traders would for their own buy/sell decisions. By considering the expected loss under normal, stressed and highly stresses markets, clients can begin to see the dynamics that a future liquidity crisis might impose on them. The three scenarios – normal, stressed and highly stressed – are defined by recent bid-ask spreads, those during the first six months of 2008, and those between September 2008 and March 2009, respectively. By computing these measures for each security through a proprietary pricing function, we are able to share a deeper understanding of how liquidity effects can impact a portfolio. The highly non-linear impact of the three chosen scenarios’ severity is also evident in the results – demonstrating that liquidity risk needs to be managed well before the markets start behaving that way.
Hybrid Factor Models
Equity managers are accustomed to either fundamental factor models, which rely on just a handful of applicable factors, or on statistical models which rely on uninterpretable factors built from the portfolio constituents. But with the introduction of hedge funds, FX, commodities and other asset classes into traditional asset owner portfolios, neither of these approaches provides accurate and useful results. Fundamental equity factor models offers no explanatory power for a derivative portfolio and statistical models offer no meaningful interpretation. Rather than relying on a one-size fits all solution (fundamental models) or on uninterpretable factos (statistical models), the ‘hybrid’ approach uses thousands of tradable and understandable indices as potential factors from which an algorithm selects the best subset for each sub-portfolio. The result is a statistically meaningful factor model made to order for each portfolio made up of a manageable number of interpretable and even tradable factors.
The hybrid approach offers the best of both older methods – a small number of statistically meaningful factors, but this time applied to diversified portfolios.
Advanced Performance Attribution and Comprehensive Risk together
All portfolio managers and investors track the performance of their investments and look for attribution analysis so they can identify the best and worst picks they’ve made – the goal of which is to identify winning and losing strategies so they can repeat what is shown to work. Those same managers and investors have come to value a robust risk analysis on the same portfolios. Since before the global financial crisis, managers and asset owners have sought a single source of performance/risk analysis but until recently, they’ve had no choice but to use different systems and reports for these two very closely related analyses.
In August 2008, a prominent hedge fund decided to de-lever their portfolio to ride out the storm of market volatility. After one of their analysts suggested cutting their worst losers, the portfolio manager decided to use our risk services to see how such trades would impact their risk profile and found that risk would actually soar. It turned out that their losing positions, in aggregate, formed a significant hedge of their wining positions. How much easier would it have been to identify such a pattern if we had had a performance engine integrated with our risk service? Imagine a report that shows multiple attributions simultaneously: attribution of returns and risk into the same underlying causal factors for each set of investments in a portfolio. Simultaneously learning where you made your money and how much risk it cost you to make that money would be instantaneous. Such insight is not far away.
It’s an exciting time to be in the fintech analytics business – innovations and disruptive improvements are changing the game for everyone, and I’m excited to play my part in helping people make better investment decisions.
By Damian Handzy, Global Head of Risk, StatPro.