Over the past decade, there has been a huge demand for risk management as well as a significant increase in demand for risk modeling and risk analytics in the financial industry.
We’ve seen disruption in the industry over the past couple of years, particularly around massive consolidation, analytics providers in the fixed income space, and in the index solution space.
All of this industry and market disruption has left voids in many professionals’ workflows. For others, it offers an opportunity to take a fresh look at these different workflows that have been ingrained in their processes for many years to see where there is room for change.
We have looked to a panel of risk experts during the latest FactSet Americas Symposium in October 2018 to help shed light on these topics and more. The panelists were:
Here are some of the key insights, findings, and tips from these risk providers.
The general consensus across all providers is that fixed income is a challenge. All of them are building solutions to push the limits and continually increase instrument coverage.
According to Peter Shepard of MSCI, “there is probably the most opportunity for innovation around liquidity.” Indeed, today, liquidity models are looking at liquidity services and considering true liquidity risk to understand not just what the market looks like today, but how the market can respond to a liquidity event, specifically the tail risk around a crisis.
Sebastián Ceria from Axioma added that “because of that liquidity issue, there has been a lot of innovation lately around estimating curves and ‘returns’ of bonds by using peer analysis, which is what traders have used consistently for pricing fixed income securities.” This allows providers to build fixed income models with similar characteristics to equity models and style factors that are much more consistent with the style factors used in equities, adding much statistical significance.
Fixed income modeling and coverage rely heavily on data, and gathering, preparing, and maintaining data to build quantitative models can be extremely challenging. Additionally, high-quality fixed income analytics are critical to the process, as are a reliable pricing and modeling library. Providing fundamental fixed income models with scalable, flexible extensions to valuation and risk into an enterprise risk system can help with consistent approaches towards risk for all type of fund managers.
Alignment is important for multiple reasons, especially if risk models are used for portfolio construction with an optimizer or in a performance attribution framework. As Ceria explained, “a misalignment will cause the optimizer to tend to underestimate the risk of alpha factors which are not present in the risk model, and hence [create] an underestimation of risk and potential bad risk budgeting.” Additionally, the consistent story between risk decomposition and performance attribution might happen as the systematic factors are not exactly represented in the risk model or in the model used for factor attribution.
Ceria added that “the notion of custom models is essential both to deal with the concept of alignment in portfolio construction, and also to build that consistency in the context of performance attribution.”
It is absolutely essential for the investment manager to be able to incorporate her or his own views on the factor construction and the drivers of future return coming from those factors. Factor choice becomes crucial because factors have different tail behavior and different probability of extreme events, both on the positive side and on the negative side. This adds a layer of complexity to an already complicated picture when we start thinking about extreme risk and tail dependencies in the context of factor selection. Boryana Racheva-Iotova of FactSet stated that “when you want to establish a new methodology, you need to allow financial institutions the freedom to both customize the models, play with the different settings, try different components of the models and back-test them, so that they can really prove for themselves the value out of those models.” Standard models are still very essential so that there is rapid time to productivity and a medium for viewing risk through a consistent lens across an organization and within the industry. But, for those who are embedding risk deeply into their portfolio construction and evaluation process, the option for customization, model validation, and back testing needs to be available to highlight relevant insights.
Setting up a long-term asset allocation mix, while simultaneously avoiding short-term turbulence and capturing possible alpha on the journey, is not an easy task. Risk models associated with these goals can vary in terms of structure for different horizons, but all need to be underpinned with a holistic, consistent, and multi-step Monte Carlo engine to cater to all horizons. Selecting the appropriate factors for various horizons and mandates is tremendously important for capturing the different macro environment regimes. “It’s not a one-size-fits-all solution” said Racheva-Iotova. “It is very important to have this wisdom in terms of how this particular problem is being approached.”
Ceria said that “the market is pushing for risk providers to look at stress tests for the long run, with a good engine for economic scenario generation. Then, the challenge of doing stress testing, especially on such long horizons, is that not only you have to have a robust Monte Carlo engine to do that, but you have to worry about reinvestment strategies. We're seeing more sort of the classical approach of factor models or full repricing for short-term scenarios and stress testing for the longer ones.”
Beyond the explicit investment mandate or the choice of factors, the behavior of the portfolio itself is a major consideration in long- vs. short-term risk modeling. “The horizon of the risk model should be something to do with the annual turnover of the portfolio,” added Jason McQueen of Northfield Information System. “You may have people that turn up very quickly, hedge funds for example, for which, your short-term model would be appropriate. You may have people that only turn over 50% a year perhaps, so the average holding period is two years, and for those, you need a longer-term model.”
Long- vs. short-term risk modeling isn’t necessarily a black-and-white decision. Especially for portfolios with long investment mandates, managing risk for the short term and keeping an eye on interim market events is key. “Long horizon investors are also shorter horizon investors. Some have had to sell illiquid assets in the depths of the crisis to meet the liquidity needs of their private equity book,” Shepard pointed out. “So, you may think of private equity as a long horizon investment, but you also need to manage liquidity on a short horizon.”
When working in the illiquid assets space—private equity and private real estate, in particular—access to data is the biggest challenge. Beyond that, understanding how to use that data in a sensible way to garner a holistic understanding of a multi-asset class portfolio poses additional complexities. There is a responsibility among investors who use private assets to understand the risk and return profile of those assets, which is certainly no easy task.
For illiquid and private assets, there is a responsibility to make a best effort to understand the risk/return profile, but the approach needs to be sensible. Shepard observed that “private real estate has cyclical dependence in the capital markets, and an asset owner making an asset allocation decision thinking that it's going to behave like a treasury is really dangerous.”
McQueen noted that it’s important to ensure that if you’re using proxies, the proxy adds some intuitive sense of value. “Developing proxies for the individual components of the illiquid asset and then putting them together forms a better overall proxy for each asset,” he said.
At the end of the day, the approach used for modeling private assets, whether extremely granular or based on high level proxies, needs to align with the investment mix and mandate of the portfolio. Racheva-Iotova suggests a pragmatic approach to the problem. “When we see illiquid assets with significant allocations, we suggest modeling them having in mind an in-depth understanding of the drivers and characteristics for risk. For similar assets with a very small allocation within the portfolio, it may make sense to use a sophisticated proxy-based approach. With regard to private assets, practicality is extremely important.”
The challenge with smart beta and factor strategies is that the models used to analyze those portfolios can turn out to be inaccurate because of a mismatch between factor ETFs and the analysis from a factor risk model. There is an innate tension between intuition and accuracy. Using factor models is the only reliable way to analyze an ETF portfolio, with a deeper look on what real factor bets those portfolios are taking rather than what the name is.
“In some instances, marketing teams have taken over the industry of smart beta and the general public, in particular, who invest in these portfolios, are not being properly educated about what these portfolios really are,” explained Ceria.
McQueen added that “people that run ETFs should take portfolio construction seriously, and they should get a risk model and try and build a portfolio that does have a high exposure to the style they're trying to target and minimal exposure as far as possible, subject to the long-only constraint to all other exposures.”
For Racheva-Iotova, “ETF providers are now well equipped with factor models. We mainly see two approaches depending on the purpose of the analysis. In some cases, it is necessary to apply the regular type of risk models like multi-asset class risk models or if the ETFs are equity focused, just the fundamental type of equity factor models in the same way as for any other portfolio, with full holdings transparency. In other cases, there is a different perspective that if you want to manage that portfolio for the long run, the holdings will not tell you enough because they change from period to period; so what we do for the longer term risk management and portfolio construction is to apply, let's say, returns based approach i.e., to take the return levels, and based on the information that we have with regard to the nature of the ETF, to construct returns based analysis model in order to look at the fact that it's somewhat higher level.”
Finally, Shepard shared that “the rise of passive has actually made the role of factors all the more important for active managers. If there's more money tracking these strategies, you need to be aware of them, whether you're trying to track them or whether you're trying to avoid them.”
There is no perfect answer when it comes to the number of factors necessary for analysis. The more factors, the more dimensions, and that collinearity can make the explanations shift. “That's why the notion of custom models makes a lot of sense”, according to Ceria, “because you have a basic structure of factors that are the ones that ‘everybody’ agrees to. And then, you have the set of factors that are very much aligned with the model.” Factor modeling not only brings intuition into volatility measurement but also ensures that there is statistical significance to that intuition. As markets, economies, and portfolio management all evolve, that intuition becomes dynamic and risk management must keep up.
Risk is a never-ending story and the types of risk factors change because there is always evolution of the financial markets. This is one reason that translating granular factors into high level, intuitive drivers of risk and performance can be extremely powerful. Indeed, the numbers of factors in multi-asset class portfolios can become quite large. So, for representation and communication purposes, the ability to build it up by relevant groups of factors is key. “Maybe five years ago, the standard fundamental type factors were indeed the major drivers of risk”, Racheva-Iotova explained. “And probably, you could have constructed a model portfolio, where you have 90% or 95% of the risk explained by those factors if you mimic a well-diversified index. Today, this becomes much harder, and this is because of the structural changes in how the markets operate that slowly, but very persistently enter the markets maker structure.”