The empirical rule, also called the 68-95-99.7 rule, is a quick way to describe data that follow a bell curve. Knowing only the mean and the standard deviation, you can say where most of the values lie, how unusual a particular value is, and how many observations you can expect to fall far from the center. Test scores, heights, measurement errors and many industrial processes behave this way.
How the calculation works
- 68% of values lie within mean ± 1 standard deviation
- 95% lie within mean ± 2 standard deviations
- 99.7% lie within mean ± 3 standard deviations
- Z-score = (value − mean) ÷ standard deviation
- Share within the same distance = erf(|z| ÷ √2), and the share above a value = 50% × (1 − erf(z ÷ √2))
Example
IQ scores have a mean of 100 and a standard deviation of 15. About 68% of people score between 85 and 115, 95% between 70 and 130, and 99.7% between 55 and 145. A score of 130 has a z-score of 2: it is as close to the mean as 95.45% of people, and only 2.28% score higher. In a group of 1,000 people, you would expect about 2.7 scores beyond three standard deviations.
Where it is used
Quality control uses the rule to set control limits, so that a value beyond three standard deviations triggers a check. Teachers use it to grade on a curve, doctors to interpret reference ranges, and analysts to judge whether a return or a measurement is ordinary. The same logic is behind the six sigma methods in manufacturing.
Good to know
The rule is an approximation for bell-shaped data. Skewed data, such as incomes, or data with heavy tails, such as stock returns, put far more values in the extremes than the rule predicts. Check a histogram first. With a small sample the observed shares will vary around the theoretical ones, and the standard deviation itself is only an estimate.
Related tools
To work with exact probabilities for any range, use the normal distribution calculator, and for the position of a single value, the z-score calculator. To find the standard deviation of your data, use the standard deviation calculator.