Why Expected Shortfall is More Coherent Than Value at Risk
This article explores why regulatory financial mathematics favors Expected Shortfall (ES) over Value at Risk (VaR) as the primary metric for market risk. It provides a straightforward comparison of both metrics, outlines the mathematical criteria that define a "coherent" risk measure, and highlights the specific flaws of VaR—namely its inability to capture extreme tail risk and its failure to consistently reward portfolio diversification. Finally, it explains how these mathematical realities prompted global banking regulators to adopt ES as the modern standard for calculating capital requirements.
Defining VaR and Expected Shortfall
To understand why one metric is superior, it is necessary to define what each measures:
- Value at Risk (VaR): VaR measures the maximum potential loss of a portfolio over a specific time horizon at a given confidence level. However, it only acts as a threshold. For example, a 99% VaR tells a risk manager the minimum loss that will occur in the worst 1% of cases, but it provides zero information about how severe the losses could get beyond that threshold.
- Expected Shortfall (ES): Also known as Conditional Value at Risk (CVaR), ES solves the threshold problem by calculating the average of all losses that exceed the VaR cutoff. It answers the question: "If things get bad, how much do we expect to lose on average?"
The Four Axioms of a Coherent Risk Measure
In 1999, mathematicians Philippe Artzner, Freddy Delbaen, Jean-Marc Eber, and David Heath published a foundational paper establishing what makes a risk metric mathematically sound, or "coherent." According to their framework, a risk measure must satisfy four axioms:
- Monotonicity: If Portfolio A consistently produces worse returns than Portfolio B under all scenarios, the risk of Portfolio A must be higher.
- Translation Invariance: Adding a guaranteed amount of cash to a portfolio reduces its risk by that exact amount of cash.
- Positive Homogeneity: If you double the size of every position in a portfolio, you double the risk of the portfolio.
- Subadditivity: The risk of two merged portfolios must be less than or equal to the sum of their individual risks.
While both VaR and ES satisfy the first three axioms, VaR fails the fourth axiom—subadditivity. Because Expected Shortfall satisfies all four, it is classified as a coherent risk measure, whereas VaR is not.
The Subadditivity Failure of VaR
Subadditivity is the mathematical representation of the most fundamental rule in finance: diversification reduces risk. Mathematically, for two portfolios \(X\) and \(Y\), a risk measure \(\rho\) is subadditive if:
\[\rho(X + Y) \le \rho(X) + \rho(Y)\]
VaR frequently violates this rule when applied to portfolios with asymmetric return distributions, such as those containing options or credit-default swaps. Under VaR, combining two separate portfolios can paradoxically result in a combined VaR that is higher than the sum of the two individual VaRs.
This creates two major problems for financial regulators:
- Discouraging Diversification: A bank using VaR might be penalized with higher capital requirements for diversifying its assets.
- Regulatory Arbitrage: A financial institution could artificially lower its reported VaR (and thus its required regulatory capital) by breaking a single large company into smaller subsidiary portfolios, hiding the true aggregate risk.
Expected Shortfall is strictly subadditive across all possible return distributions. Under ES, diversification is always mathematically recognized as either reducing risk or keeping it neutral, preventing institutions from gaming their risk models.
Blindness to Tail Risk
Beyond failing the subadditivity axiom, VaR is entirely blind to the shape of the distribution's "tail" (the extreme end where catastrophic losses occur).
If a trader utilizes a strategy that yields small, steady profits 99.5% of the time, but results in complete bankruptcy 0.5% of the time, a 99% VaR model will completely ignore the bankruptcy risk. Because the catastrophic event sits outside the 99% threshold, VaR treats the strategy as completely safe. This creates a perverse incentive for traders to take on massive, low-probability "black swan" risks that boost short-term returns while masking the potential for systemic collapse.
Expected Shortfall eliminates this blind spot. Because ES averages all outcomes in the tail beyond the confidence level, a catastrophic loss in the 99.9th percentile will heavily skew the ES calculation upward. Under ES, it is mathematically impossible to hide extreme risks in the furthest reaches of a probability distribution.
The Regulatory Shift to Expected Shortfall
The 2008 global financial crisis exposed the practical consequences of VaR's mathematical flaws. Banks that appeared well-capitalized under 99% VaR models suffered devastating losses because their models ignored the severity of tail events and failed to properly aggregate risk across diverse trading desks.
In response, the Basel Committee on Banking Supervision (BCBS) overhauled market risk standards in the Fundamental Review of the Trading Book (FRTB), a key component of the Basel III framework. Regulators officially replaced the 99% VaR standard with a 97.5% Expected Shortfall standard. By mandating a coherent risk measure, regulators ensured that modern banking capital requirements accurately reflect both the benefits of diversification and the true cost of extreme market stress.