Factor in risk
Risk management is an integral part of our investment process. From the research of asset classes that could improve diversification to identifying tactical opportunities and alternatives sources of return, there’s always a focus on detecting potential risks.
While understanding the risk of each portfolio building block is important, we like to focus on managing portfolio risk holistically. We believe a holistic approach is more efficient, fostering alpha idea generation while keeping portfolios aligned to investors’ objectives and constraints.
We use several quantitative and qualitative techniques to manage portfolio risk across the team. But one approach we find particularly helpful when managing our unconstrained portfolios (where long and short positions make traditional asset-class risk models less intuitive) is risk-factor analysis.
What are risk factors?
A risk factor is an underlying exposure that explains the return profile of an asset – a little bit like the relationship between food and its underlying nutrients. Several types of model can be used to measure this exposure by using assets and factor returns. Factors can be pre-defined, like macroeconomic variables or stocks’ fundamentals, or estimated from the asset returns using statistical factor models such as principal component analysis.
In our holistic risk-management approach, we pay most attention to sensitivities to financial markets (also known as betas) and use a linear factor model to measure the sensitivity of our portfolios to markets such as equities, interest rates and currencies, with the equity beta the most well known.
Why use risk factors?
In our view, there are several benefits of looking at portfolio risk through the factor lenses.
We’re always striving to add new asset classes to our portfolios with the aim of improving diversification. However, the fact that a portfolio is diversified across asset classes does not mean it’s necessarily diversified across markets. The problem is that new asset classes are not completely independent from traditional sources of return; they’re actually exposed to common market risk factors that cross asset-class boundaries. Factor-risk analysis helps us identify those common loadings and maintain the portfolio exposure to market factors at a level that’s in line with the portfolio’s risk objectives.
Having this risk-factor perspective of the portfolio is even more important when managing our unconstrained portfolios, in our view, where we aim for a low reliance on market direction (low equity beta). These portfolios combine multiple investment strategies that focus on different sources of return, such as market returns, alternative risk premia and tactical opportunities.
Taking a holistic portfolio view that includes factors, can help us to mitigate any unwanted market exposures at a portfolio level while allowing the different return sources to perform independently. This way, strategies that are systematically defined or short-term tactical ideas, can still be explored even if they’re loading onto the same market factors. Overall portfolio risk management can then address potential deviations from our core market views by offsetting those sensitivities with risk-management strategies aiming to reduce factor exposure.
It’s not all about risks, though. Even though it’s not always possible to express views using risk factors directly, as they’re not necessarily investable, understanding the sensitivities of different assets to market factors can help us identify the most efficient way to express our views. For instance, we could express a Chinese growth view using bonds, equities or currencies, depending on which asset our analysis indicates provides the most beneficial valuation – potentially adding value to the portfolio over time.
Assets’ sensitivities to market factors are not constant. We try to account for that by giving higher weights to more recent returns when estimating factor betas, aiming this way to capture the current market dynamics.
We also observed that correlations tend to increase in market downturns. Correlations and betas are closely linked. All else equal, as the correlation of an asset to a factor increases, so does the sensitivity to that factor. For that reason, we also monitor stressed factors. We measure stressed factors as the sensitivity of assets to factors in periods when equity markets had high volatility and negative returns.
By also controlling for stressed factor levels in this way, we believe we can better prepare the portfolio for significant market drawdowns using tail hedges, i.e. strategies that are designed to perform particularly well in that type of environment.
The chart below demonstrates the increase in sensitivity to equity markets during stressed periods. Note in particular the higher equity beta of property in downturns, when less-liquid asset classes tend to underperform.
As with most quantitative measures of risk, factors are calculated using past data. We can’t be certain that these relationships from the past will persist in the future. Statistical-significance tests help to assure us that sensitivities are reliable, but they’re not bulletproof.
To avoid relying solely on historic outcomes to measure risk, we therefore also conduct forward-looking stress tests and scenario analysis. For that, we use deterministic scenarios constructed by a panel of investment experts across multiple teams at LGIM.