Using advanced risk modeling to unlock critical insights and optimize portfolios

This article was written by Antonios Lazanas, Head of Portfolio and Index Research at Bloomberg.

Modern risk modelling is not just about monitoring risk. Sure, the specialists who manage risk are using sophisticated modelling techniques to ensure portfolios aren’t vulnerable to unanticipated losses when markets change. But these models can also be a fundamental component of a portfolio manager’s armory across the investing process, providing the tools that enable careful portfolio construction with an eye on what risks can be taken to achieve targeted expected returns.

The portfolio construction process requires precise risk control at all times and under different circumstances, which creates a number of challenges. Portfolio managers are typically leveraging traditional risk models that do not account for their unique investment horizons and don’t allow them to accurately respond to market changes. Additionally, these generic models tend to leave portfolio managers underestimating risk of their optimized portfolios and issuer-specific risk, while overestimated factor risk.

As a result of these issues, portfolio managers tend to utilize specialized custom risk models for portfolio construction, often inconsistent with the risk models used for risk management, ultimately leading to operational inefficiency.

Investment horizon: One-size doesn’t fit all

It should not be surprising that risk varies with the investment horizon. In the middle of a market crisis, short-term risk is quite high, but long-term investors know that ripples of any crises will largely peter out. Hedge fund managers that rebalance their portfolios daily (or even intra-day) need a risk model that represents the current state of risk in the market. On the other hand, pension fund managers need a risk model that can tell them both the short-term risks for risk management purposes, as well as long-term risks for portfolio construction.

Long-term managers that fail to account for the mean-reversion property of risk over long horizons may overreact and reduce portfolio risk too quickly, potentially undermining their long-term returns.

Responsiveness to market changes

While the premise of any risk model is an understanding that past volatility is a good predictor of future volatility, there are many situations where market volatility changes rapidly. For a risk model to be sufficiently responsive in such situations, it needs to use a short lookback window to estimate past volatility. Unfortunately, short lookback windows pick up a lot of noise that can overwhelm risk estimates and produce misleading intel.

Traditional risk models try to solve this problem by using lookback windows of intermediate length, striking a compromise between noisy estimates and responsiveness. These models tend to under-estimate short-term risk at the onset of a market crisis.

Underestimating the risks of optimized portfolios

Another common shortcoming surfaces during portfolio construction using optimization, a process that seeks to hedge out unwanted factors and expose factors with positive expected returns. This exercise requires an accurate measure of the risk relationships between factor pairs. However, there are more than a million factor pairs in a multi-asset risk model making the estimation of risk relationships complex.

Accurate estimates are key for risk management, but the noise this process introduces distorts the risk relationship between certain factor pairs, with some factor combinations erroneously showing very little or even zero risk. This noise can then lead to serious risk under-estimation.

Separating systematic from issuer-specific risk

While macro managers focus on factor selection, fundamental portfolio managers aim to select issuers they expect to do better than the market. They seek to minimize factor risk and maximize issuer-specific risk in their portfolios. Traditional risk models tend to overestimate factor risk and underestimate issuer-specific risk.

As a result, fundamental managers may hedge out more risk than necessary, compromising their returns.  Conversely, macro managers may be exposed to higher issuer-specific risk than their risk model is indicating.

Regaining the trust of investment professionals

The upshot of these shortcomings is that many managers do not put their full faith on their risk model, especially during times and situations they need it the most. Creating a model that addresses all these issues is a difficult task, but Bloomberg has delivered.

The MAC3 model now available on the Bloomberg Terminal uses Bloomberg’s industry-leading data and insights to find signals from short-term market movements and adjusts models accordingly, in line with an investor’s chosen model horizon. MAC3 increases risk estimates correctly in a crisis, but also appropriately decreases them as events dissipate, allowing investors to take an appropriate amount of risk, and therefore optimize returns. MAC3 also properly separates factor from issuer-specific risk catering both to macro and fundamental investors. Finally the model blends accuracy with robustness making it appropriate to be used across the investment cycle, from portfolio construction to risk management.

At a time of increasing competition and compressed fees, a trustworthy model offers more than just an accurate risk forecast, it also provides critical insights that enable managers to target optimal portfolios in line with the investment horizon and views of the market.

When managers can trust their models, they can also delegate decisions to them, setting operations on the path to automation, achieving significant cost savings, and ensuring consistent approaches for portfolio decision making, risk monitoring, and client communication.

MAC3 is an integral part of Bloomberg’s offerings for the buy-side. Bloomberg Buy-Side Solutions support the entire investment lifecycle – from research and idea generation, portfolio, risk management, order and execution management, to post-trade and operations, compliance and risk oversight, as well as technology and data management