Skewness and Kurtosis: Shape of Data Distributions

Skewness measures the asymmetry of a distribution — whether its tail stretches further to the right or the left. Kurtosis measures how heavy the tails are and how sharp the peak is, i.e. whether extreme values occur more often than in a normal distribution.

They matter because the mean, median and many statistical tests assume particular shapes. A skewed income distribution explains why "average salary" misleads, and heavy-tailed (high-kurtosis) data warns that rare extreme events are more likely than they look.

Key Concepts

Positive (right) skew: long right tail; Mean > Median > Mode (e.g. incomes). Negative (left) skew: long left tail; Mean < Median < Mode (e.g. exam scores with a ceiling). Symmetric: skewness ≈ 0; mean = median = mode. Kurtosis types: mesokurtic (normal-like, kurtosis = 3), leptokurtic (sharp peak, heavy tails, > 3), platykurtic (flat, light tails, < 3). Excess kurtosis = kurtosis − 3, so normal = 0.

Formulas

Pearson's first skewness: (Mean − Mode)/σ. Pearson's second skewness: 3(Mean − Median)/σ. Moment skewness: γ₁ = Σ(x − x̄)³/(nσ³). Kurtosis: γ₂ = Σ(x − x̄)⁴/(nσ⁴). Excess kurtosis: γ₂ − 3. All are unit-free (units cancel in the ratios). Sign of skewness gives tail direction; magnitude beyond ±1 is considered highly skewed.

Worked Example

Salaries (₹ thousand): 25, 28, 30, 32, 35, 38, 120 (n = 7). Mean: 308/7 = 44. Median: 32. SD: squared deviations sum = (25−44)² + (28−44)² + (30−44)² + (32−44)² + (35−44)² + (38−44)² + (120−44)² = 361+256+196+144+81+36+5776 = 6854; σ² = 6854/7 ≈ 979.1; σ ≈ 31.29. Pearson's second skewness: 3(44 − 32)/31.29 = 36/31.29 ≈ 1.15 — strongly positive, as the single high earner (120) drags the mean far above the median. Kurtosis signal: that one extreme value also fattens the tail, pushing kurtosis above 3 (leptokurtic tendency). The median (32) is the honest "typical" salary here.

Common Misconceptions

1. "Skewness is about which side has more data." It is about the tail's direction — in right skew the bulk sits left and the tail reaches right. 2. "High kurtosis means a sharper peak only." Kurtosis mainly reflects tail weight (outlier-proneness); the peak shape is secondary. 3. "Skewed data is wrong data." Skew is natural — incomes, waiting times and city populations are inherently skewed; use the median, don't "fix" the data.

Real-World Applications

Skewness guides salary reporting (median for skewed pay), insurance pricing for skewed claim sizes, and exam analysis. Kurtosis is critical in finance — heavy-tailed returns mean crashes are likelier than normal models say — and in quality control, where heavy tails signal process instability.

Skewness names the direction of a distribution's tail (and which average to trust), while kurtosis gauges tail heaviness (how often extremes strike). Together they complete the picture that centre and spread begin: the full shape of the data.