Kurtosis Calculator

What Is Kurtosis?

Kurtosis measures the "tailedness" of a distribution—how much probability resides in the extreme tails versus near the center. A distribution with high kurtosis has heavy tails and a sharp peak. Low kurtosis indicates light tails and a flat peak.

While skewness measures asymmetry (left-right imbalance), kurtosis measures tail behavior (extreme-value frequency).

Types of Kurtosis

Mesokurtic (kurtosis ≈ 3): Normal distribution. The baseline for comparison.

Leptokurtic (kurtosis > 3):

• Heavy tails, sharp peak
• More outliers than normal
• Examples: Stock returns, income distributions

Platykurtic (kurtosis < 3):

• Light tails, flat peak
• Fewer outliers than normal
• Examples: Uniform distribution, grades on an easy test

Kurtosis vs. Excess Kurtosis

Two conventions exist:

Kurtosis: Raw fourth moment; normal distribution has kurtosis = 3

Excess Kurtosis: Kurtosis minus 3; normal distribution has excess kurtosis = 0

This calculator reports both. Excess kurtosis is convenient because zero represents normality, but some software reports raw kurtosis.

The Formula

Sample kurtosis (Fisher's formula):

$\text{Excess Kurtosis} = \frac{n(n+1)}{(n-1)(n-2)(n-3)} \sum_{i=1}^{n}\left(\frac{x_i - \bar{x}}{s}\right)^4 - \frac{3(n-1)^2}{(n-2)(n-3)}$

This formula (Fisher's) provides an unbiased estimate of population excess kurtosis.

The fourth power makes extreme values dominate—a single outlier dramatically increases kurtosis.

Interpreting Kurtosis Values

These guidelines are approximate. Statistical significance depends on sample size.

Why Kurtosis Matters

Risk Assessment

In finance, kurtosis reveals "tail risk":

• High kurtosis → More frequent extreme events
• Stock markets show excess kurtosis of 3-10
• Models assuming normality underestimate crash probability

Outlier Expectation

High kurtosis means more outliers are natural:

• Outlier detection should use appropriate thresholds
• IQR-based methods may flag too many "outliers"

Statistical Method Selection

Many methods assume normality:

• t-tests require approximately normal distributions
• Regression assumes normal residuals
• High kurtosis indicates these assumptions may fail

Data Quality Checks

Unexpected kurtosis can reveal data problems:

• Extreme kurtosis from a few outliers
• Mixture of different populations
• Measurement errors

Kurtosis in Different Fields

Finance and Economics

Stock returns consistently show leptokurtosis—"fat tails." The 2008 financial crisis was a 25-sigma event under normal assumptions, but such events occur more often than the normal distribution predicts.

Example: Daily S&P 500 returns have excess kurtosis around 5-10, meaning extreme days are far more common than a normal model suggests.

Insurance

Claim amounts often show high kurtosis. Reinsurance exists specifically because large claims (the tail) can be catastrophic.

Manufacturing

Quality measurements should be mesokurtic (normal). High kurtosis might indicate:

• Inconsistent processes
• Mixed batches
• Measurement problems

Psychology

Test scores designed to be normally distributed should show kurtosis near 3. Departures suggest ceiling/floor effects or other problems.

Historical Background

Karl Pearson introduced kurtosis around 1905 as part of his system for classifying distributions. He used the term "excess" for departures from the normal distribution.

The word "kurtosis" comes from the Greek "kyrtos" meaning curved or arched. Pearson was describing the peak's curvature, though modern understanding emphasizes tail behavior.

The misconception: Many textbooks incorrectly describe kurtosis as measuring "peakedness." Research has shown kurtosis is primarily about tail weight, not peak shape. A distribution can have a flat peak yet high kurtosis if it has heavy tails.

Comparing Distributions by Kurtosis

The t-distribution shows how degrees of freedom affect kurtosis—fewer degrees means heavier tails.

Testing for Significant Kurtosis

Standard error of kurtosis:

$SE_{kurt} \approx 2\sqrt{\frac{6}{n}}$

Note: SE of kurtosis is approximately twice SE of skewness.

Rough test:

$z = \frac{\text{Excess Kurtosis}}{SE_{kurt}}$

If |z| > 2, kurtosis significantly differs from normal.

Example: n = 100, excess kurtosis = 2

$SE_{kurt} = 2\sqrt{6/100} = 0.49$

$z = 2 / 0.49 = 4.1$

Highly significant—distribution has heavier tails than normal.

Kurtosis and Sample Size

Kurtosis estimates are notoriously unstable for small samples:

• A single extreme value dramatically increases kurtosis
• Reliable estimates require n > 100
• For n < 50, interpret cautiously

As samples grow, kurtosis estimates stabilize and confidence intervals narrow.

Kurtosis and Outliers

There's a circular relationship:

• Outliers increase kurtosis
• High kurtosis suggests more outliers are expected
• But a few extreme outliers might inflate kurtosis artificially

Always visualize your data. Histograms reveal whether high kurtosis reflects genuinely heavy tails or a few unusual points.

Robust Alternatives

When standard kurtosis is too sensitive to outliers:

Moors' kurtosis: Based on quantiles rather than moments

$\text{Moors} = \frac{(E_7 - E_5) + (E_3 - E_1)}{E_6 - E_2}$

Where $E_1$ through $E_7$ are octiles. Less sensitive to extreme values.

L-kurtosis: Part of the L-moment system, providing stable estimates.

Kurtosis and Transformations

Transformations affect kurtosis:

• Log transforms often reduce kurtosis of right-skewed data
• Power transforms can adjust tail behavior
• Box-Cox transformations optimize toward normality

After transformation, verify kurtosis is closer to 3 (or excess kurtosis closer to 0).

Related Calculations

Kurtosis pairs with:

• Skewness Calculator — Complementary shape measure
• Standard Deviation Calculator — Second moment
• Mean Calculator — First moment
• Outlier Calculator — Extreme value detection
• Five-Number Summary Calculator — Distribution overview

Summary

Kurtosis measures the weight of a distribution's tails, indicating how frequently extreme values occur relative to a normal distribution. High kurtosis (leptokurtic) signals heavy tails and more outliers; low kurtosis (platykurtic) indicates light tails and fewer extremes. In finance, kurtosis reveals tail risk that normal-based models miss. In quality control, unexpected kurtosis signals process problems. Combined with skewness, kurtosis provides the complete picture of distribution shape beyond the basic measures of center and spread.