📚 Trusted by students, educators & researchers since 2017.
If you just finished learning about Variance and Standard Deviation, you might have a lingering, frustrating question in the back of your mind.
When we calculate how far our data spreads out from the average, we run into a problem: some data points are above the average (positive distance) and some are below the average (negative distance). If you add them all up, they cancel each other out to exactly zero.
To fix this, statistics teachers tell you to square the differences. Squaring a negative number makes it positive, which stops them from canceling out.
But wait. If the only goal is to get rid of the negative signs, why don’t we just use Absolute Value? Why don’t we just strip the minus signs off the numbers and take the simple, intuitive average of the distances?
It is the most common question in statistics, and it is almost never answered in the classroom.
The Alternative: Mean Absolute Deviation (MAD)
What your intuition is screaming for actually exists. It is called the Mean Absolute Deviation (MAD).
To calculate it, you simply find the distance of each data point from the mean, drop any negative signs, and find the average of those distances.
$$\text{MAD} = \frac{\sum \vert{}x_i – \bar{x}\vert{}}{n}$$
In many ways, MAD is vastly superior to Standard Deviation:
- It makes perfect logical sense. If you calculate a MAD of ₹5, it literally means: “On average, our data points are ₹5 away from the centre.” (Standard deviation does not actually mean this, even though people assume it does).
- It doesn’t blow outliers out of proportion. If one data point is 10 units away, MAD counts it as 10. Variance squares it to 100, heavily distorting your spread.
So, if Mean Absolute Deviation is more intuitive, easier to calculate, and less susceptible to being ruined by outliers… why does Standard Deviation rule the world? Why did Variance win?
The Three Hidden Reasons We Square
The reason we square variance has nothing to do with making things simple for humans. It has everything to do with making things elegant for mathematics.
Here are the three reasons squaring became the undisputed king of statistics.
1. The Calculus Problem (The Sharp Corner)
Statistics wasn’t built on modern computers; it was built by mathematicians using calculus in the 1800s and 1900s.
Calculus is all about finding the optimal point of a curve (the absolute peak or the absolute valley) by calculating slopes (derivatives). But calculus has a fatal weakness: it cannot handle sharp corners.
If you graph absolute value ($y = \vert{}x\vert{}$), it forms a harsh “V” shape with a sharp, jagged point at the bottom. You cannot calculate a derivative at that sharp point; the math completely breaks down.
If you graph a squared number ($y = x^2$), it forms a beautiful, smooth, U-shaped curve (a parabola). Calculus loves parabolas. It is incredibly easy to find the exact minimum of a smooth curve. Because of this, squared errors became the foundation of modern predictive modeling, most famously in Least Squares Regression.
2. The Alarm System (Punishing Outliers)
Sometimes, we want to be highly sensitive to extremes.
Imagine you are manufacturing airplane engine parts. If a part is 1 millimeter too large, it is a minor issue. If a part is 10 millimeters too large, it is a catastrophic failure that could crash a plane.
When you square the deviations, you are building an alarm system into your math.
- A deviation of $1$ becomes $1^2 = 1$.
- A deviation of $2$ becomes $2^2 = 4$.
- A deviation of $10$ becomes $10^2 = 100$.
Squaring forces the statistical model to pay massive, severe attention to extreme outliers. It mathematically guarantees that a model will try its hardest to avoid huge errors, even if it means accepting a few smaller ones.
3. The Pythagorean Secret (Variances are Additive)
This is the most mind-bending property of Variance, and the real reason statisticians refuse to abandon it.
If you flip a coin, or roll a die, or measure the height of two independent people, their combined chaos behaves exactly like the Pythagorean Theorem ($a^2 + b^2 = c^2$).
If you have two independent variables (let’s call them $X$ and $Y$), you cannot add their Standard Deviations together. You cannot add their Mean Absolute Deviations together. The math will fail.
But you can add their Variances.
Because Variance is a squared metric, the Variance of $X$ plus the Variance of $Y$ equals the combined Variance of ($X + Y$). This magical, additive property is the bedrock of probability theory. Without it, the Central Limit Theorem—the very foundation of statistical inference—would fall apart.
The Verdict
We don’t square Variance just to get rid of negative numbers. We square it because it unlocks calculus, penalises massive errors, and holds the laws of probability together.
We sacrificed human intuition for mathematical power.