How Market Volatility Affects Asset Class Behavior

With contribution from Joe Shankland.

The global market over March saw historic levels of volatility comparable to that of the 2008 Financial Crisis. The instability in the market, although still quite high, seems to have dialed back and priced itself in throughout April. In recent months, the market has seen a massive uptick in volatility. When just looking at the VIX, which is a measure of expected price fluctuations in the S&P 500 Index options over the next 30 days, we have seen the index reach levels not seen since 2008.

March 2020 was where the greatest amount of increase in the VIX occurred, and following this, we have seen a significant decline. The VIX price level is only one indicator of volatility in the markets. The question is, how does market volatility translate into asset class behavior and to what degree does diversifying across asset classes minimize our risk and return profile?

For all major asset classes, across developed and emerging markets, correlations have been historically low. We’ve heard from multiple sources that there are benefits to holding fixed income securities in your portfolio given the low correlations with equity. But when market volatility is high, there is evidence that the correlation between equity and fixed income begin to converge. Using the S&P 500 and the Bloomberg Barclays US Aggregate Index returns, over the past three years, the correlation across asset classes has been a mere 0.08 (a very weak positive linear relationship).

If we focus only on the month of March, that value increases dramatically to 0.79 (a strong positive linear relationship).

In a global context, taking the same approach with the MSCI All Country World Index and the Bloomberg Barclays Global Aggregate, you would see a correlation of 0.43 over the past two years and 0.85 over the month of March 2020. Whether a U.S. focus or global focus is measured, the message is clear that equity and fixed income are converging. What does that mean? Historically if you suffered negative returns from equity investments, the income and low correlation of fixed income would act as a cushion to offset some of those losses. When both equity and fixed income securities begin to diminish in value, returns from the broader fixed income universe begin to decline, the next step is to determine if your portfolio’s asset allocation is enough to withstand the headwinds. With FactSet’s unique risk solutions and capabilities with our standard and Fat-Tail Multi-Asset Class risk models, we can measure the connections of asset class allocation, volatility, and returns.

For this analysis, we focused on iShares Multi-Asset ETFs, as they provide a range of allocations across similar ETFs. Four ETFs span a range of allocations: Aggressive (AOA), Growth (AOR), Moderate (AOM), and Conservative (AOK). The allocation to Equity is highest in the Aggressive allocation and then gradually increases to Fixed Income the closer you get to Conservative. The fifth ETF, the iShares Morningstar Multi-Asset Income ETF (IYLD), focuses on a different subset of ETFs that contain coupon and dividend-paying instruments. We can leverage the capabilities of FactSet’s Ownership database to drill down into the underlying individual holdings of these Fund-of-Fund ETFs, generate each asset’s risk statistics and roll the results back up to the group-, factor-, and portfolio-level. We can decompose each to any desired level of granularity in FactSet’s Portfolio Analysis suite to understand the underlying risk characteristics of each allocation and how it translates to performance.

FactSet employs a variety of unique risk solutions to address market headwinds. The uncertainty of the impact that COVID-19 will have across the globe has led to a new period of turbulence within the financial markets. Throughout this period, a historic number of extreme events have been observed. This turbulence has exposed some of the limitations typical risk models following a normal distribution have, namely that they underestimate the probability of these extreme events occurring. FactSet's patented Fat-Tail methodology captures the dynamic probability of extreme events, tail dependencies, and asymmetry within the distribution. These enhancements allow users to better estimate the risk they are exposed to in extreme market environments. If we look at Value at Risk (VaR) across different confidence intervals in a histogram, we see that as we get higher out in the confidence interval, the difference in VaR under the Fat-Tail methodology increases significantly.

Running a VaR breach backtest across both Conservative and Aggressive portfolios, we can see that in the turbulent market conditions from February 19 onwards, the VaR calculated using a normal distribution has been breached consistently throughout the period of turbulence.

However, when utilizing FactSet’s fat-tailed approach, during this period of turbulence the tail ends of our return distribution track the observed realized returns to a much higher degree.

Clearly in the current market climate, using the Fat-Tail methodology provides greater insight into how portfolios behave within extreme market conditions. An important point to note is that the Fat-Tail methodology does not simply inflate risk numbers irrespective of market dynamics. Instead it becomes activated in more turbulent times, as seen where the Fat-Tail VaR values are both consistent and aligned with the normal distribution prior to the turbulence. Looking at the spread between the Fat-Tail and Normal methodology also demonstrates this, as it ramps up during the more volatile periods and shows how this patented model becomes activated within turbulent times.

Ultimately, this model not only captures extreme volatility more rigorously but also doesn't overvalue risk values in stable conditions; allowing users to adhere to their optimal risk-reward return profile.

When we ran a Year-to-Date performance in our Style, Performance, and Risk (SPAR) application, we see that the Conservative (AOK) Allocation is the top performer.

The Aggressive (AOA) and Multi-Asset Income (IYLD) allocations trail significantly in performance. Across the board, March was a major drawdown and we are seeing an uptick in April. Referring to the general market and the asset-class convergence for March between the major indices, the correlation across each underlying asset class ETF also converged. We expanded the ETF allocations to their first level of underlying; the individual ETFs that make up each allocation. We then ran the correlation calculation of returns for the portfolio using the Multi-Asset Class Risk Model within Portfolio Analysis. Measuring correlation monthly year to date, we focused on the following descriptions of three asset classes:

With the Conservative Allocation, we observed an already positive correlation across asset classes within Equity and Fixed Income. However, they all increased significantly during March. Fixed Income had the most profound increase and Emerging Equity also converged on Developed Equity.

The Aggressive Allocation over the months of December, January, and February had a negative and near-zero correlation with equity. In March of 2020, that correlation shifted from negative to positive.

The multi-asset income ETF consisted of mainly Fixed Income coupon-paying ETFs, along with Equity dividend ETFs, so it is no surprise there is a history of positive correlation across asset classes. What is interesting is that during March, there too was a dramatic shift upward in correlations across all ETFs with the addition of two Fixed Income ETFs. Much of IYLD’s ETFs are fixed income (weighted around 70 during the end of March). Across each allocation, this key takeaway here is that a larger allocation to fixed income has resulted in a more positive correlation across asset classes.

We get it, the correlations have been increasing, and the returns in each allocation have taken a hit in March. Why does this matter? It matters because having better anticipation of this convergence is key to strategic decision-making. With the unique capabilities of FactSet Monte Carlo MAC Risk suite, we can use both the Normal and Fat-Tail methodology in a daily time-series to highlight a significant analysis. We have seen how the Fat-Tail methodology picks up a more accurate VaR/ETL resulting in fewer breaches—now let’s compare the models in a standardized format. We use a VaR Model Spread calculation to determine how wide the spread is for each allocation. The calculation is as follows: VaR Model Spread = VaR 1D 99% (Normal) - VaR 1D 99% (Fat Tail). Looking at the chart, the widest VaR Spread was in the Aggressive Allocation (one of the largest drawdowns), and narrowest in the Conservative Allocation (smallest drawdown). An interesting note here, the spread prior to March across all allocation was consistently and significantly close to zero.

Following the intention and nature of the Fat-Tail model which has been back-tested rigorously, in times of market stability, the spread is minimal. When market uncertainty picks up in the market, the Model Spread begins to widen and that is exactly the case we see across the iShares allocations.

As the VaR Model Spread has continued to grow and as total return has continued to decline, there have been many discussions on whether the market has truly hit the bottom. Although there is no way to be sure, we can leverage stress testing to measure the strength of recovery under different hypothetical shocks across each allocation. When we think of stress testing across a multi-asset portfolio, the tests that are conducted are usually in the form of a worst-case scenario. What about in the case of the market returning to health? Can we measure the expected outperformance in a multi-asset portfolio across different indices, industries, or spreads? Absolutely.

Across the Aggressive, Conservative, and Multi-Asset Income ETFs, we can see that a market rebound nets a positive return for the portfolio. Relative to the benchmark, as is the case with the Active Return, we only experience positive outperformance across the board in the event of Spreads Tightening. When we use Event Weighted Stress testing, we are placing a heavier weight on periods throughout the history of the model in which the value we set on the shock most resembles the value on that day in time. An example is the VIX decreasing by 20%. The days throughout history where a drop of 20% or close to 20% occurred would receive a higher weight, thus making the stress test better represent how the factor shock would cause the portfolio to behave. The Aggressive Allocation gets the highest rebound across all shocks (except for Spreads Widening) and also has the highest relative underperformance. In a rebound, the Conservative Allocation returns a positive value for the portfolio and tracks the benchmark’s expected return much more closely with the Income ETF being a mix of the two.

The allocation that we choose, when looking at volatility, correlation, and stress testing through the lens of a multi-asset class risk model influences the severity of both drawdown and recovery and there are several ways that users can leverage FactSet to derive this information.

What is next from here? As we have seen, there are several ways we can measure the increase in risk and uncertainty across asset allocations. Choosing a proper allocation is going to cause the portfolio to react differently to both the downturn and recovery in the market and having a unique set of risk management capabilities can help determine what the sweet spot is in your ideal allocation. The Fat-Tail methodology can provide an extra layer of discussion to the analysis and key analytics (such as the VaR Model Spread) to help anticipate potential headwinds.