What is the Empirical Rule?
The empirical rule (also called the "three-sigma rule" or "68-95-99.7 rule") is a fundamental statistical principle that applies to normally distributed data. It states that:
- 68.27% of data falls within 1 standard deviation (σ) of the mean (μ)
- 95.45% of data falls within 2 standard deviations (2σ) of the mean
- 99.73% of data falls within 3 standard deviations (3σ) of the mean
The Empirical Rule Formula
Given a dataset with mean μ and standard deviation σ:
- 1σ range: [μ − σ, μ + σ] — contains ~68.27% of observations
- 2σ range: [μ − 2σ, μ + 2σ] — contains ~95.45% of observations
- 3σ range: [μ − 3σ, μ + 3σ] — contains ~99.73% of observations
An Example of How to Use the Empirical Rule
Suppose the average height of adult men in a population is 175 cm with a standard deviation of 7 cm:
- 1σ range: 175 ± 7 = [168 cm, 182 cm] — about 68.27% of men fall in this range
- 2σ range: 175 ± 14 = [161 cm, 189 cm] — about 95.45% of men fall in this range
- 3σ range: 175 ± 21 = [154 cm, 196 cm] — about 99.73% of men fall in this range
This tells us that if you randomly select a man from this population, there is a 95.45% chance his height will be between 161 cm and 189 cm.
Where is the Empirical Rule Used?
- Finance & Economics: Analyzing stock returns, portfolio risk assessment, and Value at Risk (VaR) calculations. A stock with an average daily return of 0.1% and a standard deviation of 2% can be expected to return between −3.9% and +4.1% on 95.45% of trading days.
- Manufacturing & Quality Control: The "Six Sigma" methodology is directly based on the empirical rule. Products within 3σ of the target have a 99.73% conformance rate; the goal is to push this to 6σ for near-perfect quality.
- Science & Medicine: Determining normal reference ranges for clinical tests (e.g., blood pressure, cholesterol levels) and deciding when a result is statistically "abnormal."
- Education: Interpreting standardized test scores. If the mean SAT score is 1050 with a standard deviation of 200, about 68% of students score between 850 and 1250.
- Engineering: Predicting component lifetimes, load tolerances, and safety margins.
- Meteorology: Establishing normal temperature and precipitation ranges.
FAQs
When does the empirical rule apply?
The empirical rule applies when your data follows a normal (Gaussian) distribution — the familiar symmetric bell-shaped curve. If your data is heavily skewed or has extreme outliers, the rule may not hold. Always verify approximate normality before applying it.
What is a standard deviation?
Standard deviation (σ) measures how spread out data is around the mean. A small standard deviation indicates data is clustered tightly around the mean; a large one indicates data is more spread out. It is calculated as the square root of the variance.
What is the difference between 1σ, 2σ, and 3σ?
Each sigma level captures more of the data distribution. 1σ covers the central 68.27%, 2σ extends to 95.45%, and 3σ covers 99.73%. Points beyond 3σ are considered statistical outliers (only 0.27% of data).
How is this different from Chebyshev's theorem?
Chebyshev's theorem applies to any distribution and guarantees that at least 1−1/k² of data lies within k standard deviations. For k=2, at least 75% of data is within 2σ. The empirical rule is stronger but requires a normal distribution, giving exact percentages of 68%, 95%, and 99.7%.
What does "outside 3σ" mean in practice?
Values beyond 3 standard deviations from the mean are extremely rare in a normal distribution (only about 0.27%, or 1 in 370 observations). In quality control, this is called a "defect." In finance, it is an "outlier event" or "tail risk."