How we built a risk management system
"As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain they do not refer to reality."
It's important to remind oneself of Einstein's insight when using a mathematical model to evaluate investment portfolios. Modeling investment risk is an exercise in mathematical complexity, perspective, and perhaps just as importantly, humility. Anyone who has ever had more than a five-minute conversation with Stas Melnikov, our managing director of investment risk, quickly appreciates all this, even if lost in a fog of mathematical complexity.
Stas recently penned a paper that describes Russell Investments' framework for modelling investment risk and the rationale for the development of our risk management system. As he acknowledges at the start, the global financial crisis was a pivotal moment for risk management. In its aftermath there was the widely held belief that risk models had failed to quantify market and liquidity risk, leading to losses far beyond expectation. That last word of the sentence, expectation, is curious in the sense that seldom do investors expect to lose, let alone spend much time estimating expected losses. In our view, they could greatly benefit from using imagination rather than expectation to consider what could go wrong--and how far wrong things could go--and then envision how to respond. To address this problem, over the last several years Russell Investments has implemented the Enterprise Risk Management System (ERMS) to help our portfolio managers and clients engage our understanding, measurement and imagination of potential financial peril across many dimensions.
Stas' paper describes the considerations which motivated the development of a broad set of tools, constantly evolving with our understanding, that Russell Investments uses to evaluate our portfolios and those of our clients. Much more important than expectations of return or of outcome is an understanding of the range of uncertainty surrounding those expectations and outcomes. It doesn't matter much where the road eventually leads if you can't tolerate the path from here to there, or don't acknowledge that there isn't actually any single path, but many--and you don't know which one you're on, or that you're only (on average) going to get to the destination or that the terrain is going to change en route. The risk model is the investor's representation of the changing terrain immediately ahead, but it's a model, and representation of reality, not reality itself. Therefore, it's critical to understand a risk model's assumptions and mechanics, which leads to difficult choices between using off-the-shelf tools from vendors or internally developed products.
Russell Investments' approach was to define the principles and objectives that a risk system should achieve--total portfolio coverage, key sources of risk, transparency (no black boxes), actionable insights and high frequency. As Stas and team tried to align those objectives with the available offerings, they determined that a flexible approach which blended the best of each resource was desired. As this combination of resources was being developed, several other insights and principles became important. Viewing the portfolio from different angles or perspectives (multiple lenses) is valuable since all models are wrong--albeit hopefully not all at the same time. Of course, when comparing different models, it's important to assure consistent assumptions. In order to achieve this flexibility, it became important to the team to include best-of-breed components which demanded modularity and open architecture, leveraging some vendor tools while dismissing others.
The result? A comprehensive system used for regulatory, risk oversight and portfolio management without sacrificing accuracy, flexibility or transparency. Stas' team continues to evaluate Russell Investments' portfolios as well as to evolve the system components and overall design as markets and technology develop. This set of capabilities has become a vital component not only for our portfolios, but also--with appropriate extensions--to risk evaluation and management for our clients, in addition to serving as a critical part of our asset / liability and asset allocation analysis.