Ask three risk managers how risky a portfolio is and you may get three different numbers — all of them correct. Value at Risk, Conditional Value at Risk and stress testing are complementary tools, and most risk platforms report all three. The challenge for traders is knowing which question each one answers.

Value at Risk: the threshold

Value at Risk (VaR) estimates a loss level that should not be exceeded over a given horizon at a given confidence. A one-day 99% VaR of 50,000 means that, under the model's assumptions, losses should exceed 50,000 on roughly one day in a hundred.

VaR is intuitive and widely used in regulation and reporting. Its weakness is equally well known: it says nothing about how large losses are on the days the threshold is breached. Two portfolios with identical VaR can have very different tail behaviour.

CVaR: what happens beyond the threshold

Conditional VaR, also called Expected Shortfall, averages the losses in the tail beyond the VaR level. It directly addresses VaR's blind spot and is mathematically better behaved — it rewards diversification consistently, which VaR does not always do. Banking regulators adopted Expected Shortfall for market-risk capital in the Basel framework's Fundamental Review of the Trading Book for these reasons.

VaR tells you where the cliff edge is. CVaR tells you how far the fall might be.

Stress testing: the scenarios models miss

Both VaR and CVaR are typically estimated from historical or simulated return distributions. If the past sample did not contain a particular crisis, the model will not anticipate it. Stress testing fills that gap by applying explicit scenarios — a repeat of 2008, a sudden 300-basis-point rate move, a liquidity freeze in a single market — and measuring the result.

Good stress tests are specific and uncomfortable. They include correlated moves, widening spreads and reduced liquidity at the same time, because that is how real crises arrive.

How the three fit together

  • Use VaR for day-to-day limit monitoring and communication.
  • Use CVaR when sizing positions or comparing strategies with different tail shapes.
  • Use stress tests to ask what could break the portfolio, regardless of probability.

What to look for in a risk platform

When evaluating risk software, check whether it supports multiple estimation methods (historical, parametric, Monte Carlo), lets you define custom stress scenarios, handles options and other non-linear instruments correctly, and documents its assumptions. Transparency of method matters more than the number of metrics on the dashboard.

A worked illustration

Consider two hypothetical portfolios, each with a one-day 99% VaR of 1% of capital. The first holds a diversified basket of liquid equities; on its worst days, losses cluster just beyond the threshold, averaging perhaps 1.3%. The second sells short-dated out-of-the-money options. On most days it earns small premiums, and its VaR looks identical — but on the rare day a large move occurs, losses can be many multiples of the threshold.

VaR treats the two as equally risky. CVaR immediately separates them, because it averages the tail. A stress test that shocks the underlying by 10% and doubles implied volatility would reveal the second portfolio's exposure even more starkly. The lesson is not that one measure is right and the others wrong, but that each sees a different part of the distribution.

Backtesting the risk model itself

Risk models should be tested like trading strategies. Count how often actual losses exceeded VaR over a long period: at 99% confidence, roughly 1% of days should breach. Far more breaches suggest the model underestimates risk; far fewer may mean it is too conservative to be useful. Breaches that cluster together indicate that the model adapts too slowly to changing volatility.

Quorient Editorial Desk

Articles credited to the Editorial Desk are researched and written for Quorient and reviewed by the editor, Callum Reid, before publication, in line with our Editorial Policy. Spotted an error? Request a correction.