Thornburg’s Portfolio Analytics group deploys both industry-standard and proprietary tools to gain insight into the nonlinear additive nature of portfolio construction.
Don’t Reduce Risk Management to a Prohibitive, Isolated Function
- Asset managers may choose to position their risk function apart from investments, leaving risk managers to operate siloed away from the team they monitor. In this framework, risk managers are focused on establishing and tracking quantitative forecast triggers and guardrails, as well as on delivering ex-post analysis, with little or no feedback from Investments.
- At Thornburg Investment Management, we feel this approach is lacking. In our experience, a siloed structure can lead to entrenched misunderstandings between investment managers and the risk team. Yet, both groups work on behalf of clients: the former to maximize investment return, and the latter to minimize potential risks for the same investment. At Thornburg, we emphasize and encourage a collaborative approach that brings investment and risk managers together to achieve better outcomes from portfolio construction.
Do Empower Risk to Be Actively Involved with Investments on Portfolio Construction—Collaboration Works Best
- A deep, comprehensive understanding of investment portfolio risks, as well as thoughtful, effective portfolio construction to mitigate these risks, is paramount to delivering excellent client outcomes. We believe this can be achieved through open, continuous, and deliberative interaction between risk management and Investments, as well as among portfolio managers and analysts. We have found that to foster such interaction, collaboration rather than segregation works best.
- Thornburg’s risk management function lies within the Portfolio Analytics group that is a subset of the Investment team. This integration is an intentional part of our investment process, as it creates organic synergies between risk management and investment functions.
- Our process combines ongoing monitoring and communication of the risk1 exposures within each portfolio, as well as scenario analysis, to the portfolio managers. We hold monthly in-depth risk meetings between Portfolio Analytics and the individual portfolio management teams. Attended by both the corresponding client portfolio manager and heads of investments, these meetings not only ensure that risk guardrails are maintained, but also to discuss portfolio exposures in the context of the market environment and the stated strategy objectives.
Don’t Take a Myopic View on Portfolio Risk—Models Cannot Explain Everything
- Risk models for asset management are based on academic theories that are imperfect at best, and their standard output only provides a starting point. Powered by the cut-and-dry statistical analysis of past events, these financial models tend to work well only when forecasting business-as-usual outcomes, i.e., the models are unable to calculate risks attributed to events that have not occurred in the past. For instance, standard forecasts do not automatically capture such tail events as the 2008 financial crisis or, more recently, the February/March 2020 market reaction to the onset of the coronavirus pandemic.
- The common industry-standard risk models for an active asset manager are “fundamental,” i.e., based on econometric factors.2 Though based on the trading behavior and the fundamentals of individual securities, these models can only “explain” some of the portfolio risk, assigning the remainder as “unexplained” or “security selection,” i.e., the asset-specific [idiosyncratic] risk that is unique to each individual holding.
- Without the insight of a bottom-up fundamental security analyst(s) from the Investment team, this portion of the risk may remain a blind spot to the Risk team.
- No risk model can forecast where the price of an asset will end up in a year, or even tomorrow. Risk models will only predict the level of uncertainty (riskiness) of that outcome. On the other hand, successful active asset managers make it their job to project target outcomes from their fundamental research of the asset, as well as from their experience and intuition. When combined, the two imperfect outlooks can complement each other.
- To avoid these blind spots and uncover risks the models have difficulty accounting for, risk managers must work with the fundamental research professionals from the Investment team. Through this collaboration, the top-down, quantitative perspective and the bottom-up, fundamental security research combine to create not only a more comprehensive risk outlook, but also mutually agreed-upon, actionable portfolio construction guidance.
Do Combine Quantitative and Fundamental Perspectives on Risk—Integrate Bottom-up Research and Mathematical Testing
- What differentiates Thornburg’s risk management process from many of its peers is the integration of fundamental research and quantitative analytics expertise. We see risk and opportunity as highly inter-related aspects of successful active asset management; if you understand the opportunity well, you are better positioned to understand the risks associated with that opportunity and vice versa. This approach is applied from the aggregate strategy level all the way down to the individual securities.
- The fundamental aspect of our proprietary risk management process starts with bottom-up security selection. Portfolio managers, analysts, and traders evaluate fundamental and trading characteristics of individual securities, evaluate potential risks, and establish return targets as part of the fundamental security selection process and then aggregate this during portfolio construction. The quantitative risk analysis begins simultaneously at the portfolio level, continues into more granular components such as portfolio baskets or industry/country exposures, and then drills down all the way to individual securities. The discussions between members of the Portfolio Analytics group and portfolio managers happen at the intersection of the two outlooks. For instance, if risk managers’ models signal (either via automated alerts or during a risk review) an imbalance in a portfolio or portfolio basket, an inquiry into the current positioning is triggered. Alternatively, portfolio managers will reach out to the Risk team on the potential exposure outcomes from hypothetical trades. These conversations occur as new risks or opportunities are identified, as well as during monthly risk meetings that bring the team together for a formal portfolio review.
- To evaluate tail event risk, one must define relevant and precise stress test scenarios. To do so, collaboration between the fundamental research professionals and quantitative risk managers is critical. On the fundamental side, one must not only have an excellent understanding of how similar events impacted financial markets in the past, but also an experience-based intuition on how their impact may differ as future tail risk events may not be triggered by the same factors. On the quantitative side, one must rely on the intimate knowledge of the academic scope and technical limitations of the risk models, to fine-tune the forecasting tools and adequately interpret the output from the stress test.3
- Although a well-designed, insightful stress test may forecast potential outcomes well, they will not provide the likelihood of these outcomes. Thus, the decision on how or whether to act ultimately resides with portfolio managers. Portfolio Analytics makes sure that such decisions are well-informed from both the risk management and the quantitative portfolio construction perspective.
- Disagreements are healthy. Robust debate4 around portfolio holdings and weights is welcomed by portfolio managers as they seek out the best investment and portfolio construction ideas. That extends to the risk management process, for which Thornburg established a series of escalation points should disagreements occur. The initial escalation point would involve our co-heads of investments, followed by the firm’s CEO.
Don’t Rely on a Single Off-Shelf Risk Solution as It May Not Reveal All Aspects of Risk
- Industry-standard quantitative risk models, while tremendously helpful in identifying and measuring risks within portfolios, are neither perfectly accurate nor exhaustively comprehensive in isolation. Similarly, even the most flexible off-the-shelf risk solutions may need to be supplemented with proprietary analytics intelligence in order to serve the unique team- and strategy-specific needs.
- It is essential to employ the best models and practices of traditional risk management. For an active management shop, we believe it is essential to create a collaborative exchange of thought and insight between Risk and Investments. This exchange must be facilitated by a digital ecosystem where quantitative and fundamental perspectives on portfolio construction and risk are served to both Risk and Investments through a set of flexible, intuitive, self-served analytics.
Do Develop a Robust, Efficient, and Thoughtful Process to Improve Outcomes—Complement Industry Platforms with Proprietary Tools
- At Thornburg, we use multiple risk lenses and models for fixed income, equity, and multi-asset class portfolios to derive ex-ante risk, ex-post attribution, and scenario stress tests. We then synthesize and augment this array of analytics with the help of proprietary tools in a way that makes it both insightful and actionable to the Risk team and portfolio managers.
- Standard risk models we use provide a detailed, statistically accurate (to within model limitations) view into current and historic portfolio risk. Here risk is defined as both stand-alone (forecasted standard deviation of returns) and active (forecasted standard deviation of relative-to-benchmark returns, which is the same as forecasted tracking error). The model will calculate what portion of risk derives from common market themes (factors such as style, country, industry, and currency) and what portion cannot be explained by common market themes and therefore must be specific to the individual securities in the portfolio. This latter form of risk is known as idiosyncratic risk, or security selection risk. We want our risks to come from positions that hold excess upside risk versus the common trends within the aggregate market, while monitoring (and striving to limit) exposures that could lead to excessive downside risk.
- Our ongoing risk monitoring starts with a documented set of asset-class specific guardrails.5 These are tracked by a proprietary automated system as well as through the ongoing review by Portfolio Analytics. Standard risk monitoring continues with a strategy-specific set of analytics and guardrails that is laid out in the corresponding Philosophy and Process document for each strategy. This process relies heavily on a proprietary self-service Performance and Risk dashboard that facilitates daily monitoring of risk and performance across strategies, as well as the monthly risk reviews for individual portfolios.
- We supplement industry-standard risk management platform output with proprietary analytics and tools. These range from automated risk parameter guardrail alerts to sophisticated applications developed to enhance and empower the construction and management of the portfolio baskets (i.e., sub-portfolios designed to counterbalance each other through the economic cycle and provide better risk-adjusted returns over time).
- The team keeps track of and evaluates the new risk platforms and modules to make sure that what we are using is at the leading edge of the industry standard. This monitoring is extended to our current risk platform providers, with whom we maintain an ongoing conversation on risk model enhancements.