Spikes in Market Volatility Through the Risk Model Lens

Amidst growing uncertainty from U.S. tariffs and reciprocal actions, rising inflation, and slowing economic growth, this April financial markets around the world experienced levels of volatility comparable to several historical events: the flash crash in June 2010, the European debt crisis in August 2011, and the Asia stock market sell-offs in January 2016 and August 2024. The purpose of this article is to show how important it is to have the appropriate risk model in place, especially in the current turbulent markets.

During normal market conditions, a risk model estimating ex-ante volatility might offer a close representation of portfolio risks. However, the large swings in market returns in April 2025 have shown that financial markets do not always operate in a normal manner. For example, the two-day return seen between April 3 and April 4 for a total of -10.5% over two days had a probability of 1 in more than 5,000,000. Notably, that also has happened during the Covid pandemic and during the 1987 market crash.

Financial markets have transitioned from periods of predictable stability to enduring turbulence, influenced by macroeconomic events such as Brexit, the Asian financial crisis, and most recently the global uncertainty resulting from tariffs on products entering the United States.

The quick transition from normal to abrupt market volatility necessitates a departure from traditional risk management methods. In this environment, risk managers and investors need a model that uses advanced probabilistic techniques, emphasizing the importance of accommodating time-varying disaster probabilities.

As we note below, FactSet's Fat Tail model can give investors a better real-time understanding of market dynamics. The model has robust predictive power during varying phases of volatility, offering strategic insights for asset-class reallocations and investment management. Together, these insights underline the necessity for adaptive models that can effectively anticipate and navigate the evolving risk landscapes of today's markets.

On the plots below we can see the level of volatility indices for equity markets in U.S., Europe, Japan, and Hong Kong to compare recent volatility with levels from previous market crises.

Source: FactSet

Source: FactSet

Source: FactSet

Source: FactSet

Risk Model Lens

Different risk models respond in different way to increasing market volatility. Below we focus on the spread between a normal model (Gaussian distribution, 125-day half-life) and fat-tailed short horizon model (fat-tailed distribution, 45-day half-life) with a 1-day 99% Value at Risk estimate (the spread) and its dynamics under changing market conditions.

Short-horizon models are built to be more responsive to recent market behavior, and so it is not surprising the estimate they provide reflects more rapidly and prominently the increase in market volatility.

On the first chart below we see that during more turbulent market regimes the spread rises as the fat-tailed short horizon model reflects recent volatility with higher weight compared to more distant historical observations—the further back we go in time, the lower the weight of observations.

Contrary to a common belief that a fat-tailed risk estimate is always higher compared to normal risk estimates, we see that in periods where market volatility is at its lowest the spread between the two model estimates turns negative. That indicates the fat-tailed short horizon risk estimate is lower than the normal model one when the market is calm.

The second chart below shows the dynamics of the 1-day VaR estimate at 99% confidence level, estimated by a normal and fat-tailed short-horizon model. It provides more visibility into responses of both models’ estimate to different levels of market volatility.

The last chart plots the daily returns of the index, which clearly indicates the periods with lower volatility, when returns are small (2017, 2021, and mid-2023 to mid-2024) and higher volatility, when returns spike in both directions (COVID outbreak 2020, inflation surge, and geopolitical risk increase in 2022, and the most recent period of market uncertainty in April 2025).

Source: FactSet

We can zoom into the most recent period of market distress and analyze the dynamics of the spread between the fat-tailed short-horizon and normal VaR estimate. Interestingly, the spread starts to pick up increasing volatility quite early—around March 5 when it starts growing and then turns positive on March 9. Afterward, it steadily rises to a peak in mid-April when volatility remains clustered for a couple of weeks.

Although in early March the most recent daily returns of the index are not largely negative, the initial signs of market downturn start to show up and this change in regime is captured by the fat-tailed short-horizon model risk estimate.

While we can see this as an early warning sign for the April sharp drop of the market, the longer history plot above shows there are occasions when the spread increases and then returns to a lower number when markets are calm. So, the increase in the spread is not always an indication of an approaching market distress. Rather, it’s an early warning for increasing market volatility that may or may not fade away.

Source: FactSet

Economic Sector Perspective

In the following analysis we focus on March - April 2025 and break down the index to economic sectors. We explore how the stand-alone, 1-day 99% VaR (i.e., VaR of the sector, as if only this sector is held and there is no correlation to other sectors within the index) for each group changes through time with an increase in volatility. We see that for certain economic sectors the spread between fat-tailed and normal risk models starts to pick up volatility earlier than it does for other sectors.

The first table below shows the spread between the fat-tailed short horizon model and normal model VaR estimate, which when positive indicates larger volatility captured by the fat-tailed model. (For the largest values we use red coloring while for small or negative values we apply green.)

As of April 3 when markets were already in distress, all spreads are red. However, some of the sectors’ spreads started rising long before that date, including Consumer Services, Distribution Services, Health Services, Technology Services, and Transportation. Nevertheless, other sectors turned in the worst daily returns for the most recent period: Consumer Durables, Electronic Technology, and Non-Energy Minerals.

We track the daily return of each of these sectors in the second data table below, with red indicating large negative returns and green for positive returns.

Spread between risk models (Fat-Tailed Value at Risk – Gaussian 1-Day 99% VaR) for economic sectors as part of S&P 500 composition

Source: FactSet

Daily return of stocks, aggregated on economic sector as part of S&P 500 composition

Source: FactSet

The reasoning behind the widening spread for sectors, where the returns are not the worst in the recent month, is that volatility increases in those sectors, as captured by the fat-tailed model due to the distribution their returns follow.

We compare below the distribution of the fat-tailed short-horizon model Monte Carlo simulations for Consumer Services, where the spread widens early and returns are not as bad as those of Consumer Durables, where the spread reacts much slower. It is obvious the returns of Consumer Services exhibit fat-tailed distribution characteristics, while those for Consumer Durables are relatively normally distributed.

For the first economic sector, applying the fat-tailed model lens adds value as it captures the specific distribution characteristics when compared to a normal model and picks up changes in volatility faster.

For the second one, normal and fat-tailed models will be more or less at par as the returns are pretty normally distributed.

Distribution of Monte Carlo simulations in fat-tailed short-horizon model

Source: FactSet

Thus, for sectors, securities, or portfolios where returns are not normally distributed, we can see additional value added from using a fat-tailed short-horizon model, as it will pick up the changes in market regime faster and will provide a risk estimate that follows closely the actual distribution of returns.

For normally distributed sectors it will be more or less in line with a normal distribution risk estimate. However, in the end the index level distribution as seen on the histogram below follows a distribution that is far from a normally distributed one.

Distribution of Monte Carlo simulations in fat-tailed short-horizon model

Source: FactSet

To get additional insight into what market returns indicate, users within FactSet Portfolio Analysis Risk are at liberty to apply both normal distribution models and fat-tailed models to compare and contrast as well as customize certain assumptions of the models.

This blog post is for informational purposes only. The information contained in this blog post is not legal, tax, or investment advice. FactSet does not endorse or recommend any investments and assumes no liability for any consequence relating directly or indirectly to any action or inaction taken based on the information contained in this article.