SQRT
Returns the square root of a number.
SQRT(<number>)Returns: A number (a scalar). An error when the number is negative.
What it does
SQRT returns the number that, multiplied by itself, gives the input. It is used in finance for scaling risk. Standard deviations and volatilities are built from square roots, and a common rule is that monthly volatility scales to annual volatility by the square root of 12. SQRT is also the way to undo a squared figure, such as going from a variance to a standard deviation.
Syntax and arguments
- number: a number that is zero or positive. A negative number returns an error.
Example
A portfolio's monthly volatility has been measured as 2.1 percent. The board wants the annual figure, using the standard square root of time rule.
Annual Volatility = [Monthly Volatility] * SQRT(12)
SQRT(12) is about 3.46, so 2.1 percent a month becomes about 7.3 percent a year. This is an approximation that assumes returns are independent from month to month, which the report should state.
Example: a second, different use
Turn a variance of monthly errors into a standard deviation, in the same units as the data:
Error Std Dev = SQRT([Error Variance])
The variance is in squared units. The square root returns the typical size of an error in the original units, which an analyst can read directly.
Things to watch
- SQRT of a negative number is an error. A variance calculated from noisy data can be negative only through an error, so check the input.
- SQRT(12) is a common scaling. It assumes the periods are independent, which may not hold for trending or correlated data.
- A square root is not a substitute for a true risk model. It converts one scale to another.
- Wrap the input in a check or IFERROR when the source is unreliable, but prefer fixing the source.
- SQRT and POWER(x, 0.5) give the same answer for non-negative numbers.
In business terms
SQRT appears in risk and forecasting work: scaling monthly volatility to an annual figure, turning a variance into a standard deviation for a control limit, and computing the root mean square error of a forecast. Controllers and analysts use it when they explain how uncertain a number is.