Finding the Missing Link

One of the big challenges in risk analytics today is dealing with the different results portfolio and risk managers receive from multiple risk models that do not interact. If you’re a risk manager who needs to aggregate risk across multiple desks investing in different regions, you most likely are unable to analyze the risk/return profile of the aggregated portfolio without inconsistencies. Risk models currently on the market may offer you a predefined and broad combination of regional models into one large model, but this is often at the expense of clarity and usability. On the other hand, if you’re a portfolio manager, you may have difficulty understanding your full exposures when you’re focused on a single region strategy, but want to take positions outside of that region.

We’ve solved these problems with Linked Models – a new feature in Risk Model MachineTM that enables you to combine fundamental factor models to gain a clearer and more focused view of your portfolio risk. In this blog post, we present two use cases to explain how Linked Models can enhance your investment process.

Case #1: Alignment and Model Coverage with Mandate

Linked Models are designed to handle strategies that take positions outside of a standard mandate. For example, say you are focused on a European strategy that takes positions in US securities for hedging or liquidity reasons. Without Linked Models, this approach cannot be easily modeled, and your only options are to use either a European (EU) model (with the US portion of the portfolio not covered) or a global model where your specific EU bets are diluted.

Or imagine your mandate is picking mostly Canadian securities but also looking for opportunistic bets outside of Canada. Again, your choices are limited. Using a World-Wide (WW) or global model would dilute your Canadian positions, but using a Canadian (CA) model would exclude any of your opportunistic bets.

Using Linked Models, you can create a risk model that is mostly region specific, but that would also accommodate out-of-region bets that better reflects your true investment process. For the above examples, you could build a hierarchical EU + US Linked Model and a hierarchical CA + WW Linked Model. This would provide full coverage without diluting regional or country-specific exposures.

Case #2: Consistency from Front-to-Middle Office

Within your organization, there may be separate desks managing different regional mandates, such as North America, Europe and Asia-Pacific using regional models for risk and return attribution. The middle office has to use a worldwide or global model to analyze the risk/return profile of the aggregated portfolio, but the underlying factor exposures taken by your colleagues in different regions can neutralize each other when evaluating at a combined multi-manager pool level using a global model.

For instance, you might be overexposed to value in Asia-Pacific, and your colleague might be underexposed to value in Europe. When evaluated at the top level with a model that has a single global value factor, these exposures could offset and these over/underexposures would not be visible. These inconsistencies make it very difficult for the middle office to have a clear view of risk across the front office.

Linked Models can take different regional models and link them to form a custom model that accurately captures all exposures. The middle office custom model could combine North America, Europe, and Asia-Pacific models and, in addition, have the World-Wide model as a catch-all for any assets not in those regions. This results in consistent risk analysis for both the front and middle office.

In addition to linking off-the-shelf Axioma risk models, you can also link custom models (developed using RMM) to create a custom linked model that better aligns with your investment process. For example, you can create custom US or global sector models and link them to develop a custom linked sector model for the US or global market segment.

The new Linked Models in Risk Model Machine are suitable for risk and attribution analysis and work seamlessly with Axioma Portfolio Analytics and Axioma Portfolio Optimizer applications.

Learn more about Linked Models here.

Questions? Contact us.

Arnab Banerjee, PhD

Arnab is involved in research and product development efforts for Axioma and leads product development initiatives involving Axioma’s new and existing risk models.