Value at Risk
A statistical risk-management tool that estimates the probability that a portfolio could suffer a large loss over a given period, focusing on the tail of the loss distribution.
As defined in Lars Peter Hansen on Risk, Ambiguity, and Measurement
Other definitions
A statistical measure that estimates the maximum likely loss of an investment portfolio over a given period, derived from historical data.
John Kay and Mervyn King on Radical UncertaintyRisk measurement method that estimates potential losses over a time horizon at a given confidence level; Roberts argues it misses real tail risk.
Tim Harford on Adapt and the Virtues of Failure 05/23/2011
Where it comes up · 11
- Nassim Taleb on the Financial Crisis 03/23/20098 claims
- Cathy O'Neil on Weapons of Math Destruction2 claims
- Gerd Gigerenzer on Gut Feelings2 claims
- A conversation between Nassim Nicholas Taleb and Stephen Wolfram at the Wolfram Summer School 20211 claim
- Tim Harford on Adapt and the Virtues of Failure 05/23/20111 claim
- Lars Peter Hansen on Risk, Ambiguity, and Measurement1 claim
- Paul Pfleiderer on the Misuse of Economic Models1 claim
- Nassim Nicholas Taleb on the Precautionary Principle and Genetically Modified Organisms1 claim
and 3 more
Most important claims · 10 of 20Search all claims for “Value at Risk” →
- Value-at-risk models are fundamentally flawed because they derive risk estimates from historical data sets that did not and could not contain the extreme events that cause crises — which is especially damning since protection against such extreme events is exactly what the models are built to provide.John Kay and Mervyn King on Radical Uncertainty
- A risk measure like Value at Risk may be fine when used passively to observe someone's risk, but becomes far more problematic once it is set as the benchmark for someone who has an incentive to take risk, because very smart agents in industries like finance will game the measure.Paul Pfleiderer on the Misuse of Economic Models
- Value at Risk models examine the tails of distributions where empirical evidence is thin and sparse, so they can work very well through normal times and then completely miss exactly when needed; excessive confidence built during normal times is precisely what gets you burned in low-frequency tail events.Lars Peter Hansen on Risk, Ambiguity, and Measurement
- The Turkey Illusion describes models—like Value at Risk or rating-agency models doing Bayesian updating on five years of data—that, in a rising market before 2008, could only predict continued rises; like a turkey fed daily that grows ever more confident it won't be killed until the day before Thanksgiving, such models fail catastrophically when the environment is one of uncertainty rather than calculable risk.Gerd Gigerenzer on Gut Feelings
- Risk models like Value at Risk failed catastrophically in the 2008 crisis because they assumed a normal distribution that ignores rare, non-normally-distributed events; this exemplifies Bertrand Russell's and Taleb's turkey problem, in which accumulating daily evidence of safety misleads you about an underlying process right up until the day the farmer chops off the turkey's head, showing that doing 'the best you can' with available data can leave humans systematically unprepared.299 - Steps In The Right Direction
- Calibrating risk models to recent extreme history, such as the 10 percent 1987 crash, guarantees failure because large deviations do not have predecessors, and the next large deviation will be different and worse than the last.Nassim Taleb on the Financial Crisis 03/23/2009
- Putting risk into a mathematical formulation can make trained, rational people insensitive to it: even when they know the model is flawed and rely on bad assumptions about error distributions or black swans, daily reassurance from the model that 'everything's fine' lulls them into a false sense of security—making the model a weapon of math destruction.Cathy O'Neil on Weapons of Math Destruction
- The industry standard of worrying about 95% (or 99%) Value at Risk meant institutions never looked beyond that threshold, which let people 'stuff risk into the tail'—concentrating extreme losses beyond the measured cutoff—to game the VaR measure while appearing safe.Cathy O'Neil on Weapons of Math Destruction
- Value at Risk would be a great tool if finance were predictable (a world of calculable risk), but real finance is high uncertainty; calculating a large bank's VaR requires estimating thousands of risk parameters and millions of correlations from 5-10 years of data, which borders on astrology rather than science, and VaR calculations have not prevented any crisis while fostering an illusion of certainty.Gerd Gigerenzer on Gut Feelings
- Financial institutions use value-at-risk (VaR) as their primary risk metric, but VaR only measures the loss at a specific quantile (e.g., 95th percentile) and tells you nothing about losses beyond that quantile—in fat-tailed domains, most catastrophic losses occur in the tail that VaR ignores, making it a dangerously flawed risk metric that persists in use because it is simple and fits institutional incentives.A conversation between Nassim Nicholas Taleb and Stephen Wolfram at the Wolfram Summer School 2021