risk
Value-at-Risk (95%)
The loss that should be exceeded only 5% of the time over a given horizon, in dollars or as a share of portfolio value.
Value-at-Risk at 95% confidence is the loss that should be exceeded only 5% of the time over a chosen period. For daily returns, that is about 1 trading day in 20, or 12 to 13 days in a typical year. It is a common level for day-to-day portfolio monitoring.
With historical data, sort the returns from worst to best and read the cutoff at the 5% mark: with 500 daily returns, a simple count puts it at the 25th worst day, while tools that interpolate land between neighboring days. With the normal (parametric) shortcut, the return at the cutoff is the average return minus 1.645 standard deviations, because 5% of a normal distribution lies more than 1.645 standard deviations below its mean. Stated as a positive loss, 95% VaR is therefore 1.645 standard deviations minus the average return; for daily returns, whose average is usually close to zero, that is roughly 1.645 multiplied by the daily volatility.
At 95%, VaR describes a routine bad day rather than a disaster, a loss you should expect on several days each quarter. For a portfolio with 16% annual volatility, about 1% a day, the normal shortcut gives a one-day 95% VaR near 1.65%. When the average return is negligible next to the volatility, a normal distribution puts 99% VaR at about 1.4x the 95% figure; a historical ratio well above that is a possible sign of fat tails, worth weighing against how few observations sit in the tail. The days beyond the cutoff can be much worse than the cutoff itself.
Example with illustrative numbers, using a simple count: you collect 400 days of a portfolio's returns. Since 5% of 400 is 20, the cutoff is the 20th worst day, a loss of 2.1%, so the historical 95% VaR is 2.1%. On a $50,000 account, that means a loss of $1,050 or more on roughly 1 day in 20.
On GNG, the Backtester's tail-risk panel labels this figure VaR 95% and shows it as a negative daily return: the 5th percentile of the run's daily returns, interpolated between neighboring days, with no distribution assumed, marked on a histogram next to VaR 99%. Monte Carlo Forecasting shows Value at Risk (95%) as the dollar result at the 5th percentile of simulated ending values, which covers your whole horizon instead of one day. The Portfolio Optimizer's Stress test panel shows VaR 95 for a scenario when the data allow.
Conditional Value-at-Risk (95%) averages the losses beyond this cutoff, and Value-at-Risk (VaR) explains the general method.
More risk terms
Get full access to GNG Research
Create a free account to access portfolio tracking, advanced tools, and more.