Measures of Dispersion

The full picture

Imagine two vegetable vendors in your local market. Both earn an average daily income of ₹1,500. But Rajan's earnings swing from ₹400 on slow days to ₹3,200 during festivals, while Meena's rarely go below ₹1,200 or above ₹1,800. Their averages are identical, yet their financial lives are completely different. This is exactly the problem that measures of dispersion solve: they tell you how spread out data values are around the average. In this chapter you will learn four key measures — range, quartile deviation, mean deviation, and standard deviation — plus a tool to compare two datasets fairly, called the coefficient of variation.

Range is the simplest measure: Range = Maximum value − Minimum value. If monthly rainfall data for Kerala ranges from 18 mm to 312 mm, the range is 294 mm. Range is quick to calculate and easy to understand, but it has one serious weakness — it only looks at the two extreme values and ignores everything in between. One unusually high or low reading (an outlier) can make the range look enormous even if the rest of the data is tightly packed. Quartile deviation (QD) fixes this by ignoring the top and bottom quarters. First, arrange data in order and locate Q1 (the value that divides the bottom 25% from the rest) and Q3 (the value that separates the top 25%). Then QD = (Q3 − Q1) ÷ 2. Because QD focuses on the central 50% of values, it is not distorted by extreme outliers — making it useful for skewed data like income distributions.

Mean deviation (MD) goes further: it measures the average distance of every data point from the centre. The formula is MD = Σ|X − X̄| ÷ N, where |X − X̄| means you take the absolute value of each deviation (ignoring whether it is above or below the mean). A kirana shop owner recording daily sales of ₹4,200, ₹3,800, ₹4,500, ₹3,600, and ₹4,900 would find the mean is ₹4,200 and the deviations are ₹0, ₹400, ₹300, ₹600, and ₹700, giving MD = ₹400 — meaning on a typical day, actual sales are ₹400 away from average. Mean deviation treats all deviations equally, which makes it intuitive but less mathematically convenient than standard deviation.

Standard deviation (SD) is the most important and widely used measure of dispersion. Instead of ignoring the sign like MD does, SD squares each deviation, averages the squares (giving variance), and then takes the square root: SD = √[Σ(X − X̄)² ÷ N]. Squaring has a deliberate effect — large deviations get amplified much more than small ones. This means SD is sensitive to outliers and penalises high variability. For example, if a student scores 55, 60, 70, 80, 85 in five tests, the mean is 70 and the SD tells you how much scores typically deviate from that mean. SD is expressed in the same units as the original data (marks, rupees, kilograms), so it is directly interpretable.

The coefficient of variation (CV) solves one remaining problem: how do you compare the variability of two datasets with different units or different averages? CV = (SD ÷ Mean) × 100, expressed as a percentage. Consider a savings bank account giving a mean annual return of 4% with SD of 0.3%, and a stock mutual fund giving mean returns of 18% with SD of 9%. The bank account's CV = (0.3 ÷ 4) × 100 = 7.5%, while the mutual fund's CV = (9 ÷ 18) × 100 = 50%. The fund is far more variable relative to its own average. CV lets you compare apples with oranges — whether you are comparing price volatility across commodities, income inequality across states, or risk across investment types. Lower CV means more consistency; higher CV means more uncertainty relative to the mean.

An Indian example

Priya is choosing between two coaching institutes for her CA Foundation entrance preparation. Institute A charges ₹40,000 and its last batch's scores (out of 200) were: 148, 152, 155, 160, 162, 165, 168, 170. Institute B also charges ₹40,000 and its scores were: 115, 120, 140, 160, 175, 185, 190, 195. Both institutes report the same mean score of 160. But when Priya calculates standard deviation, she finds Institute A has SD ≈ 7 marks while Institute B has SD ≈ 30 marks. Institute A's students consistently score around 160; Institute B's results are unpredictable — some students jumped to 195, but others scored only 115. Priya, who needs a reliable qualifying score, chooses Institute A. Without measuring dispersion, both institutes looked identical on paper.

Key concepts covered

• Range, quartile deviation
• Mean deviation, standard deviation
• Coefficient of variation

Common misconceptions to watch for

• Misconception: If two datasets have the same range, they have the same dispersion. Correction: Range only captures the gap between the maximum and minimum — it tells you nothing about how the values in between are distributed. Two datasets with range 100 can have very different standard deviations if one clusters tightly around the mean and the other spreads evenly across the whole range.
• Misconception: Standard deviation is just the average distance from the mean, the same as mean deviation. Correction: Mean deviation uses absolute values to treat all deviations equally, while standard deviation squares each deviation first. Squaring magnifies large gaps more than small ones, so SD is always greater than or equal to MD, and is far more sensitive to outliers — that is intentional, not a flaw.
• Misconception: A higher coefficient of variation always means the data is bad or the investment is poor. Correction: CV measures relative variability, not quality or performance. A startup with CV of 80% on its profits may be far more profitable overall than a fixed deposit with CV of 5%. CV tells you how consistent results are relative to the average; you must always judge it alongside the mean return or mean value to make a fair decision.