Skewness and Kurtosis in Statistics: A CAIIB ABM Study Guide

Every CAIIB candidate learns to calculate the mean and standard deviation of a data set, but banking data rarely behaves symmetrically — a handful of huge corporate loans, a spike in delinquent accounts, or a cluster of near-identical retail deposits can all distort a distribution's shape. That's where skewness and kurtosis in statistics come in: two measures from the Business Mathematics & Statistics module of Advanced Bank Management that tell you whether a data set leans to one side and whether it has unusually heavy or light tails. This article works through both measures with formulas, interpretation rules and a solved numerical, exactly as they appear in the CAIIB ABM paper.

📊 What Skewness and Kurtosis Actually Measure

A frequency distribution is described by four properties: central tendency (mean, median, mode), dispersion (range, standard deviation), skewness, and kurtosis. Skewness tells you whether the distribution is symmetrical or lopsided — whether values are bunched on one side with a long tail stretching the other way. Kurtosis tells you how peaked or flat the distribution is compared with a normal curve, and specifically how much probability mass sits in the tails.

In banking, few real distributions are perfectly symmetrical. Loan sanction amounts are typically positively skewed — most loans are small or medium-sized, but a few very large corporate exposures pull the tail to the right. NPA recovery percentages, branch profitability figures, and transaction values behave the same way. Recognising the shape of a distribution before applying a statistical technique — averaging, forecasting, or setting a threshold — prevents a bank from treating an atypical outlier as the "normal" case.

The chapter on Estimation builds directly on this idea: an estimate built from a skewed sample using only the mean can mislead, because the mean is pulled toward the long tail while the median stays closer to where most of the data actually sits.

📐 Measuring Skewness: Formulas and Interpretation

The most commonly tested formula in CAIIB ABM is Karl Pearson's coefficient of skewness:

Skp = 3(Mean − Median) ÷ Standard Deviation

The factor of 3 approximates the empirical relationship between mean, median and mode in a moderately skewed distribution. A simpler, mode-based version — Skp = (Mean − Mode) ÷ Standard Deviation — is used when the mode is reliably known; the "3×" version is preferred when the mode is difficult to pin down, which is usually true of banking data.

When ranked data or quartiles are given instead of raw values, examiners switch to Bowley's coefficient of skewness: Skb = (Q3 + Q1 − 2×Median) ÷ (Q3 − Q1). This version only needs the three quartiles, making it useful for grouped or open-ended frequency tables common in branch-wise banking data.

💡 Exam Tip: A positive skewness coefficient means Mean > Median > Mode (long tail to the right); a negative value means Mean < Median < Mode (long tail to the left). A value of zero indicates a symmetrical distribution.

The more rigorous, moment-based measure is the coefficient of skewness γ1 = μ3 ÷ σ³, where μ3 is the third central moment and σ is the standard deviation. This form is what most statistical software reports and is the one that generalises cleanly into the kurtosis formula below.

📏 Measuring Kurtosis: Peakedness and Tail Risk

Kurtosis is calculated from the fourth central moment: β2 = μ4 ÷ σ⁴. For the normal distribution this value equals exactly 3, so statisticians usually quote excess kurtosis = β2 − 3, which is zero for a perfectly normal curve. This single subtraction is the detail examiners most often test.

Distributions are classified into three shapes based on this value:

• Mesokurtic — kurtosis = 3 (excess = 0): peakedness matches the normal curve.
• Leptokurtic — kurtosis > 3 (excess > 0): a sharper peak and fatter tails than normal, meaning extreme values occur more often than a normal curve would predict.
• Platykurtic — kurtosis < 3 (excess < 0): a flatter peak and thinner tails, meaning values are more evenly spread and extremes are rarer.

⚠️ Common Mistake: Students often assume high kurtosis means "more spread out." It actually means the opposite of dispersion in the everyday sense — a leptokurtic distribution concentrates most values tightly near the mean, but the small share that escapes the peak lands unusually far out in the tails.

Practically: a leptokurtic distribution of loan losses means small losses are common, but the rare large loss can be more extreme than a normal-curve assumption suggests — a key reason banks avoid assuming losses follow a neat bell curve.

🧮 Worked Numerical: Calculating Both Measures

Consider recovery percentages from six stressed accounts in a quarter: 40, 45, 48, 50, 52, 95. The mean is 55, the median (average of the 3rd and 4th ranked values) is 49, and the standard deviation works out to approximately 19.1.

Applying Karl Pearson's formula: Skp = 3 × (55 − 49) ÷ 19.1 = 18 ÷ 19.1 ≈ 0.94. The positive value confirms the distribution is skewed right — exactly what you'd expect, since five accounts cluster near 40–52% recovery while one outlier at 95% pulls the mean upward.

A second branch with recovery percentages of 48, 49, 50, 51, 52, 50 — tightly bunched with no outlier — would show a much smaller standard deviation and a near-zero skewness, testing closer to mesokurtic rather than the first branch's leptokurtic profile.

This is the pattern CAIIB numericals test repeatedly: compute mean, median and standard deviation first, then plug them into the skewness formula, and separately reason about kurtosis qualitatively from how tightly (or loosely) the bulk of the data sits around the centre.

📌 Remember: Always compute standard deviation before skewness — every skewness formula in the CAIIB syllabus uses it as the denominator, so an arithmetic slip there cascades into the final answer.

🏦 Why Bankers Use These Measures

Skewness and kurtosis inform how a bank reads its own portfolio data. A credit team checks whether exposure sizes are positively skewed before treating the average exposure as a meaningful summary, feeding directly into risk management and credit rating exercises where concentration matters as much as the average.

Kurtosis has a parallel role in tail risk: a leptokurtic distribution of default frequencies warns analysts that clusters of defaults are more likely than a normal-distribution model predicts, one reason banks stress-test portfolios rather than relying purely on averages. This complements the broader toolkit covered under classification and tabulation of banking data, where figures are organised before shape-based measures are applied.

The same logic connects to hypothesis-testing techniques such as the chi-square test in bank statistics — both are diagnostic tools bankers use before trusting an average or a stated relationship. Just as ABM builds these quantitative skills, the CAIIB elective on employee engagement in banks builds the people-side skills needed to run a branch well.

🧠 Practice MCQs: Skewness and Kurtosis

Want chapter-wise mock tests with 100+ MCQs? Start practising free →

❓ Frequently Asked Questions

Master the Full Statistics Module for CAIIB ABM

Skewness and kurtosis round out the shape-analysis toolkit that sits alongside central tendency and dispersion in the CAIIB ABM syllabus. For the complete topic list and the official exam structure, refer to the IIBF CAIIB curriculum, and browse more Advanced Bank Management guides or the full CAIIB course on iibf.store to keep building your statistics fundamentals chapter by chapter.