What is the difference between Value-at-Risk (VaR) and Expected Shortfall (CVaR)?

Value-at-Risk (VaR) identifies the maximum dollar or percentage loss expected over a given holding period at a specific confidence level (e.g. 99%). However, VaR is blind to the magnitude of losses beyond that cutoff. Expected Shortfall (CVaR, or Conditional VaR) calculates the expected average loss given that the VaR threshold has been breached. Unlike VaR, Expected Shortfall is a mathematically 'coherent' risk measure that satisfies the property of subadditivity, which is why the Basel Committee mandated CVaR under the Fundamental Review of the Trading Book (FRTB).

How does the square-root of time rule scale VaR across holding periods?

Under the assumption that asset returns are independent and identically distributed (i.i.d.), portfolio variance scales linearly with time (T), and standard deviation scales with the square root of time (sqrt(T)). Therefore, a 10-day regulatory VaR is computed as 1-day VaR multiplied by sqrt(10) (approximately 3.162). While standard in regulatory reporting, this rule assumes zero return autocorrelation and can understate risk during liquidity crises when volatility clusters.

How does Basel III traffic-light backtesting evaluate a VaR model?

Basel III requires banks to backtest their 99% 1-day VaR models against daily P&L over 250 trading days. In a 99% model, the expected number of exceedances (days where loss exceeds VaR) is 2.5 days per year. The Basel framework categorizes results into three zones: Green Zone (0 to 4 exceedances, model accepted with no capital penalty), Yellow Zone (5 to 9 exceedances, supervisory review and scaling multiplier penalty), and Red Zone (10 or more exceedances, model rejected and maximum capital surcharge applied).

Value at Risk (VaR) & Expected Shortfall (CVaR) Workbench | Portfolio Stress-Testing

Methodology	Horizon Loss (%)	Dollar Loss ($)	Underlying Modeling Assumptions
Parametric (Normal)	1.95%	$194,500	Closed-form Gaussian Z-score × Portfolio Volatility
Parametric (Student-t)	2.28%	$228,000	Fat-tailed Student-t distribution (df=5 degrees of freedom)
Historical Simulation	2.12%	$212,300	Empirical non-parametric percentile of historical returns
Monte Carlo Simulation	2.05%	$204,800	5,000 Cholesky-correlated geometric Brownian motion paths
Expected Shortfall (CVaR)	2.58%	$258,400	Coherent average loss beyond VaR threshold (Basel III FRTB)

Historical Shock Event	Peak Crisis Drawdown	Portfolio Impact (%)	Simulated Dollar Loss	Survival Evaluation
2008 Lehman Bankruptcy (GFC)	-24.8% SPY / +8.2% IEF	-11.6%	-$1,160,000	Significant Liquidity Strain
2020 Covid Liquidity Shock	-19.5% SPY / +5.1% IEF	-9.7%	-$966,000	Manageable Cushion
2022 Inflation & Rate Shock	-18.1% SPY / -14.9% IEF	-16.8%	-$1,682,000	CORRELATION FAILURE
1987 Black Monday Crash	-22.6% SPY / +3.8% IEF	-12.0%	-$1,204,000	Acute Margin Call Alert

Institutional Downside Risk & Coherent Measures Methodology

Value-at-Risk (VaR) was pioneered by J.P. Morgan in the early 1990s as a universal risk metric to answer a daily CEO question: "What is our maximum potential loss over the next 24 hours?" While revolutionary, standard VaR suffers from a critical mathematical defect: it violates the principle of subadditivity, meaning merging two portfolios can mathematically produce a higher combined VaR than the sum of their individual VaRs.

1. The Subadditivity Failure of VaR vs. Coherent Expected Shortfall

A risk measure $\rho(\cdot)$ is mathematically coherent if and only if it satisfies four axioms: monotonicity, subadditivity, positive homogeneity, and translation invariance. Value-at-Risk fails subadditivity:

VaR(X + Y) > VaR(X) + VaR(Y) (VaR Penalizes True Diversification in Fat-Tailed Regimes)

CVaR(X + Y) ≤ CVaR(X) + CVaR(Y) (Expected Shortfall is Always Coherent)

2. Basel Committee Fundamental Review of the Trading Book (FRTB)

Recognizing that VaR completely ignores tail risk severity beyond the quantile threshold (the "cliff effect"), the Basel Committee on Banking Supervision transitioned the regulatory market risk capital framework from 99% VaR to 97.5% Expected Shortfall under the FRTB framework. Expected Shortfall measures the conditional expectation of loss in the tail:

ES_α = E[ L | L ≥ VaR_α ] = (1 / (1 - α)) × ∫_α^1 VaR_u du

Frequently Asked Questions

Why does Parametric VaR understate risk during financial crises?

Parametric (Variance-Covariance) VaR assumes that financial asset returns follow a Gaussian normal distribution. In empirical financial markets, actual return distributions exhibit severe excess kurtosis (fat tails) and negative skewness (sharp drops happen far faster than rallies). A 4-sigma market crash is virtually impossible under a normal distribution (occurring once every 126 years), yet major financial markets experience 4-to-6 sigma shocks every 3 to 7 years.

How does the Cholesky Decomposition work in Monte Carlo simulation?

Assets in a portfolio do not move independently; equities, bonds, and commodities exhibit dynamic cross-correlations. To generate realistic correlated random variables, the algorithm performs a Cholesky decomposition on the asset covariance matrix Σ to find a lower-triangular matrix L such that L × L^T = Σ. Multiplying a vector of independent standard normal random variables ε by L produces a vector of correlated shocks that faithfully preserve empirical asset relationships.

Why did stocks and bonds both fall during the 2022 market shock?

The classic 60/40 portfolio relies on an empirical negative correlation between equities and Treasuries (the "stock-bond hedge"). However, this negative correlation only holds when inflation is low and growth is the dominant macro variable. When inflation spikes unexpectedly, central banks are forced to hike policy rates aggressively, which simultaneously crushes bond prices (via duration discounting) and equity multiples (via rising discount rates). In 2022, the stock-bond correlation flipped from -0.30 to +0.65, rendering traditional 60/40 diversification ineffective.

Value-at-Risk (VaR) & Expected Shortfall (CVaR) Workbench

Portfolio & Risk Parameters

Asset Allocation Weights

Tripartite VaR & Expected Shortfall Comparison

Return Distribution & Tail Cutoff Visualizer

Historical Crisis Tail-Risk Shock Replay

Institutional Downside Risk & Coherent Measures Methodology

1. The Subadditivity Failure of VaR vs. Coherent Expected Shortfall

2. Basel Committee Fundamental Review of the Trading Book (FRTB)

Frequently Asked Questions