Expected Shortfall (ES): Tail Risk and Extreme Loss Scenarios

Expected Shortfall (ES), also known as Conditional Value at Risk (CVaR), is a risk assessment measure that quantifies the potential loss in value of an investment portfolio, given that a loss greater than the Value at Risk (VaR) has occurred. While VaR provides a threshold value such that the probability of a loss exceeding this value is at a specified confidence level, ES goes a step further by considering the average loss that occurs beyond this VaR threshold.

In other words, ES gives us a deeper insight into the tail risk of a distribution, particularly focusing on the worst-case scenarios.

How to Calculate Expected Shortfall (ES)?

Mathematically, Expected Shortfall (ES) is calculated as:

Where:

• $\text{ES}_{\alpha}$ is the expected loss given that losses exceed the Value at Risk (VaR) threshold.
• $E[X \mid X \geq \text{VaR}_{\alpha}]$ represents the conditional expectation of losses beyond the VaR level.
• $\text{VaR}_{\alpha}$ is the Value at Risk (VaR) at a confidence level $\alpha$ (e.g., 95% or 99%).

In essence, ES provides the average of the worst-case losses beyond the VaR.

Expected Shortfall Calculation Example (Historical Method)

The historical method needs no assumption about the shape of returns: sort past returns and average the worst ones.

Take 100 daily returns of a $100,000 account and a 95% confidence level. The worst 5% of days is the worst 5 days:

• VaR (95%) is the edge of that tail: about 3.5%, or $3,500. On 95 days out of 100 the loss was smaller.
• ES (95%) is the average of the five tail losses: (6.1 + 4.8 + 4.2 + 3.9 + 3.5) / 5 = 4.5%, or $4,500.

So VaR says “a bad day costs at least $3,500”, and ES says “when a bad day comes, expect to lose about $4,500 on average”. The gap between the two is the size of the tail VaR doesn’t see.

Expected Shortfall Formula Under a Normal Distribution

If returns are assumed to be normal with mean $\mu$ and standard deviation $\sigma$, ES has a closed form:

Where $z_{\alpha}$ is the standard normal quantile (1.645 at 95%, 2.326 at 99%) and $\varphi$ is the standard normal density. In practice:

With a daily volatility of 1.2% (and $\mu \approx 0$), the 95% VaR is about 1.97% and the 95% ES about 2.48% of the account. Real returns have fatter tails than the normal curve, so this method usually understates ES; the historical method above doesn’t have that problem.

Calculating Expected Shortfall in Python

import numpy as np
 
def expected_shortfall(returns, level=0.95):
    """Historical ES: the average of the returns at or below the VaR cut-off, as a positive loss."""
    returns = np.sort(np.asarray(returns))
    cutoff = max(1, round(len(returns) * (1 - level)))   # 100 returns at 95% -> the worst 5
    var = -returns[cutoff - 1]
    es = -returns[:cutoff].mean()
    return var, es
 
var, es = expected_shortfall(daily_returns, 0.95)

Importance of ES in Trading

Expected Shortfall (ES) plays an important role in trading for several reasons. It provides a more comprehensive risk assessment than Value at Risk (VaR) by focusing on the tail end of the loss distribution, making it an essential tool for enhanced risk management. Traders often use ES to conduct stress testing, helping them understand potential losses during extreme market events. Additionally, financial regulators frequently require firms to report ES as part of their risk management practices, ensuring regulatory compliance. By incorporating ES into their strategies, traders can also optimize their portfolios to minimize the impact of potential extreme losses.

Expected Shortfall vs VaR

Use both: VaR for the everyday limit, ES for how deep a breach can go. ES is always at least as large as VaR at the same confidence level.

Limitations of Expected Shortfall

• Few data points in the tail: with one year of daily data (about 250 days) at 97.5%, ES is the average of only 6 or 7 days. One extra crash moves it a lot.
• Harder to backtest: a VaR model can be checked by counting breaches. ES has no such simple test, so a wrong model goes unnoticed longer.
• Model risk: the normal formula understates fat tails; the historical method assumes the past window contains the kind of crash that comes next.
• Window and horizon choices: a calm 12-month window gives a low ES right before a volatile period. Compare several windows, and include a stressed period.
• It is still an average: ES says nothing about the single worst outcome, so pair it with maximum drawdown and stress tests.

ES in Action

Imagine a portfolio worth $1,000,000. The trader uses a ES model at a 95% confidence level to estimate potential extreme losses.

Application to Portfolio:

• Input Parameters: 95% confidence level, portfolio ($1,000,000).
• VaR Calculation: At 95%, $\text{VaR} = \$40,000$ (threshold loss).
• ES Estimate: The average loss beyond VaR is $\text{ES} = \$50,000$.

Risk mitigation approach:

1. Risk Benchmark: The trader uses $\text{ES} = \$50,000$ to gauge extreme downside risks.
2. Hedging: To reduce ES, protective options are purchased, decreasing exposure by $\$10,000$.
3. Portfolio Adjustment: Shifting 10% of high-risk assets to low-volatility instruments lowers the overall ES.

Analysis: ES provides actionable insights to manage extreme losses. Incorporating it into your strategy ensures your portfolio aligns with your risk tolerance while adapting to market dynamics.

Combining ES with Other Tools

To gain more insights, ES can be combined with other risk management tools such as:

• Stress Testing: Simulating extreme market conditions to understand potential impacts on the portfolio.
• Scenario Analysis: Evaluating how different hypothetical scenarios affect the portfolio.
• Monte Carlo Simulations: Running numerous simulations to model the behavior of asset prices and estimate risk.
• Beta and Correlation Analysis: Understanding how different assets within the portfolio interact and affect overall risk.

Key Points

• Beyond VaR: Expected Shortfall (ES) measures the average loss beyond the Value at Risk (VaR) threshold, offering a deeper understanding of tail risk.
• Focus on Extreme Losses: ES provides insight into potential losses during rare but severe market events, making it essential for robust risk management.
• Enhanced Risk Assessment: By considering losses in the tail of the distribution, ES offers a more comprehensive view of downside risk than VaR alone.
• Portfolio Optimization: ES is widely used in portfolio construction to minimize exposure to extreme losses while maintaining performance goals.
• Sensitivity to Assumptions: The accuracy of ES depends on the chosen confidence level, time horizon, and method of calculation (e.g., historical or Monte Carlo).
• Risk-Averse Strategies: Investors and institutions use ES to design conservative strategies, particularly in volatile or uncertain market conditions.
• Regulatory and Institutional Relevance: ES is frequently used in compliance frameworks and capital adequacy assessments due to its focus on worst-case scenarios.
• Complementary Metric: Pair ES with VaR and other risk measures for a holistic understanding of both typical and extreme risks.
• Stress Testing Integration: ES can be integrated into stress-testing models to evaluate resilience under adverse market scenarios.
• Applicability Across Assets: ES is used for individual assets, portfolios, and complex derivatives, making it versatile across financial products.

Conclusion

ES is a powerful tool for understanding and managing tail risk in trading portfolios. By focusing on the worst-case scenarios, it provides a more comprehensive view of potential losses than VaR alone. By leveraging ES, traders can better prepare for and navigate the uncertainties of financial markets, leading to more robust risk management and decision-making processes.