Not sure which lesson to do first? Find your level and get a plan →

Statistics · Grade 9-12

Skewness

Skewness is a measure of how asymmetric a probability distribution is around its mean — positive skew has a long tail to the right, negative skew a long tail to the left.

skewness=n(n−1)(n−2)∑(xi−xˉs)3

Quick Answer

Skewness is a measure of how asymmetric a probability distribution is around its mean — positive skew has a long tail to the right, negative skew a long tail to the left. Use it to describe the shape of a histogram or dot plot and to decide which summary to trust. In right-skewed data, such as incomes or house prices, a few very large values stretch the tail and pull the mean above the median; in left-skewed data, such as scores on an easy test, the mean falls below the median. With strong skew, report the median and IQR rather than the mean and standard deviation.

Why This Matters

Skew decides which numbers tell the truth about a data set. If house prices in a town are right-skewed, the mean might be $480,000 while the median is $350,000, and a buyer budgeting for 'the average house' misjudges what most homes cost. Skew also warns you that normal-curve tools such as the 68-95-99.7 rule will give wrong percentages for that data.

Intuitive Explanation

Twenty students take a test. Most finish in 25 to 35 minutes, a few in about 40, and one takes 70. On a dot plot the dots pile up on the left and a thin trail stretches out to the right. That trail is the tail, and the distribution is named after it: right-skewed, or positively skewed. The name points to where the few stragglers are, not to where most of the data sit.

The tail pulls the mean toward itself. The median depends only on the middle position, so the 70-minute student moves it by at most one place. The mean adds every value, so that 70 drags it up. In right-skewed data the usual order is mode < median < mean. Left-skewed data, such as scores on an easy quiz where most students get 9 or 10 and a few get 3 or 4, reverse the order. The skewness formula cubes each standardized distance, (xi−xˉs)3, which keeps the sign and lets far-out values dominate the sum, so a single extreme value can create a lot of skew.

Spread is not skew. A data set can be very wide and perfectly symmetric, like 10, 20, 30, 40, 50, 60, 70, with skewness 0. A data set can also look lopsided because of one outlier rather than a genuine tail; check whether the asymmetry comes from many values trailing off or from a single point before naming the shape.

Think of it like this. A skewed distribution looks like a comet: a dense head where most of the data sit and a long, thin tail streaming off to one side. The skew is named for the direction the tail points, and the mean gets dragged out along the tail while the median stays near the head. A comet has one tail; data with two equal tails are symmetric.

Core idea. A distribution is skewed toward its long tail, and that tail pulls the mean away from the median in the same direction, which tells you which center to report.


When to Use

Describe skewness when a histogram, dot plot or box plot is lopsided, when the mean and median differ noticeably, or when a variable has a floor but no ceiling: wait times, incomes, prices and reaction times all can't go below zero and often trail off to the right. Name it by the tail: a long right tail is positive skew. If the two halves mirror each other, call it symmetric and consider a normal model; if one point causes the lopsidedness, treat it as an outlier question.

Tip. Ask: Is the question about which way and how strongly a distribution's long tail stretches, away from its peak?

How to Recognize It

Before naming a skew, find the peak and the long tail of the display, and decide whether the question wants the direction, the size, or its effect on the summaries.

  1. Does one side of the histogram or dot plot trail off much farther than the other?

    Yes means the data are skewed toward that longer side. If both sides fall away alike, the shape is symmetric and may suit the Normal Distribution.

  2. Is the question only about lopsidedness, rather than peaks, gaps or flatness?

    Yes is skewness. Describing the full outline, including two peaks or a flat shape, belongs to Distribution Shape.

  3. Is the evidence a gap between the mean and the median, such as a mean of 62 and a median of 48?

    A mean well above the median suggests right skew, and well below suggests left skew. Choosing which of the two to report is Mean vs Median.

  4. Is the tail made of many values thinning out gradually, rather than one or two values standing alone?

    Gradual thinning is skew. One or two isolated values, with a gap before them, are better checked with Outlier Detection.

Skewness vs Distribution Shape vs Mean vs Median vs Outlier Detection

A long tail changes the picture, the summaries and the far values all at once, so skew questions get confused with three neighbours. Ask whether you are naming the lopsidedness, the whole outline, the better center, or single unusual values.

Skewness

Meaning
Skewness is the direction and amount of a distribution's lopsidedness, named for the side its long tail points to.
Key test
One question only: which way the long tail runs, and how far.
Formula
skewness=n(n−1)(n−2)∑(xi−xˉs)3
Example
Ages at death in a wealthy country pile up near 80 with a tail toward younger ages: skewed left.

Distribution Shape

Meaning
Distribution shape is the whole outline of the data, including symmetry, skew, flatness and the number of peaks.
Key test
Any feature of the outline, not just which way it leans.
Example
The birth months of a school's students give twelve similar bars: a roughly uniform shape.

Mean vs Median

Meaning
Mean vs Median is the choice of which center to report, a choice that skew usually settles in favour of the median.
Key test
About which center to report, with skew as the reason.
Example
Rainfall with a mean of 3.1 mm and a median of 0.4 mm: the median describes a typical day.

Outlier Detection

Meaning
Outlier detection flags individual values that sit unusually far from the rest, by a stated rule.
Key test
Single isolated values, often after a gap; not a gradual tail.
Formula
x>Q3+1.5 IQR
Example
One reading of 250 bpm among resting heart rates of 60 to 80 bpm is flagged; it is not skew.

Formula & Notation

skewness=n(n−1)(n−2)∑(xi−xˉs)3
The sample skewness is g1=n(n−1)(n−2)∑i=1n(xi−xˉs)3. For a population, γ1=E[(X−μσ)3]. Skewness is zero for symmetric distributions.

How to read it: Skewness is denoted γ1 or g1. Positive values (γ1>0) indicate a right tail; negative values (γ1<0) indicate a left tail; zero means symmetric.

Worked Examples

Example 1 — Name the skew from mean and median

Easy

Books read over the summer by 11 students: 0, 1, 1, 2, 2, 2, 3, 3, 4, 8, 15. Describe the shape of the distribution and say which center to report.
Show the steps and the answer
  1. Find the median: the 6th of 11 sorted values is 2.

    The median marks the middle position and ignores how far the extreme values reach.

  2. Find the mean: 4111≈3.7.

    The mean adds every value, so the 8 and the 15 pull it upward.

  3. Compare: mean 3.7 > median 2, and the values thin out from 4 to 8 to 15 on the high side.

    A mean above the median, with a long trail of high values, signals a right tail.

  4. Name the shape: right-skewed (positive skew). Report the median, 2 books.

    Nine of the 11 students read 4 books or fewer, so 3.7 overstates the typical student.

Answer: Right-skewed; the median, 2 books, is the better typical value (the mean is about 3.7).

Takeaway: When the mean sits above the median and high values trail off, the data are right-skewed and the median is the fairer center.

Example 2 — Wide is not the same as skewed

Standard

Data set A: 10, 20, 30, 40, 50, 60, 70. Data set B: 36, 37, 38, 39, 41, 44, 49. A student says A must be the skewed one because its range is much larger. Which set is skewed?
Show the steps and the answer
  1. Set A: mean = 40, median = 40, and the values are evenly spaced 10 apart.

    Equal gaps on both sides of the middle mean the left half mirrors the right half.

  2. Set B: mean =2847≈40.6, median =39.

    The mean sits above the median, the first sign of a right tail.

  3. Look at B's gaps: 1, 1, 1, 2, 3, 5, growing toward the high end.

    Values spreading further apart on one side is exactly what a tail is.

  4. Compare ranges: A spans 60, B only 13.

    Range measures spread; skew measures lopsidedness, and the two are separate properties.

Answer: Set B is right-skewed; set A is symmetric despite its much larger spread.

Takeaway: Judge skew by whether one side stretches further than the other, not by how wide the data are.

Example 3 — Find the mistake: the tail, not the pile

Application

Scores on an easy 10-point quiz for 15 students: 3, 5, 7, 8, 8, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10. Jay writes: 'Most scores are on the right, so the distribution is right-skewed.' Find the mistake and correct it.
Show the steps and the answer
  1. Locate the pile and the tail: the scores pile up at 9 and 10, and a few low scores (3, 5, 7) trail off to the left.

    Jay named the skew after the pile; skew is named after the tail.

  2. Compute the mean: 12715≈8.47.

    The three low scores drag the mean down toward the tail.

  3. Find the median: the 8th of 15 sorted values is 9.

    The median stays with the pile of high scores.

  4. Compare: mean 8.47 < median 9, which confirms a left tail.

    The mean is pulled toward the tail, so mean below median goes with left skew.

Answer: The distribution is left-skewed (negative skew): the long tail points toward the low scores, and the mean (about 8.47) is below the median (9).

Takeaway: Name skew by where the long, thin tail points; for an easy test, the pile is high and the skew is left.

Where You'll See It

  • House prices in a town

    Most houses cost a normal amount, but a few mansions cost a fortune. That long high-end tail makes the data right-skewed (positive skew), which is why the mean price is higher than the median.

  • Time to finish a race

    Most runners finish near the typical time, but a few very slow finishers stretch a long tail to the right, giving a positively skewed distribution of finish times.

  • Scores on an easy quiz

    If a quiz is easy, most students score high and a few low scores trail off to the left — a left-skewed (negative skew) distribution where the mean is dragged below the median.

Common Mistakes

  1. The slip: Right-skewed means most of the data is on the right side.

    Why it's tempting: The word 'right' makes people picture the bulk of the data on the right. But 'right-skewed' names where the long thin TAIL points, not where the data piles up.

    Instead: Skew is named by the direction of the long tail: right-skewed (positive) has a long tail to the right while most data sits on the LEFT; left-skewed is the mirror image.

  2. The slip: Mean and median are always equal, so they tell you nothing about skew.

    Why it's tempting: In the symmetric examples students see first, mean = median, so the two seem redundant. But asymmetry separates them.

    Instead: For right-skew, the long tail pulls the mean above the median (mean > median); for left-skew, mean < median. The gap and its direction reveal the skew.

  3. The slip: One outlier can't change whether a distribution is skewed.

    Why it's tempting: A single point seems too small to reshape a whole data set. But skewness uses cubed deviations, so far-out points have huge leverage.

    Instead: Because skewness sums (xi−xˉs)3, a single extreme value can dominate the sum and flip a near-symmetric distribution into a clearly skewed one. Always inspect for outliers.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A distribution has mean 52 and median 45. Is it more likely skewed left or skewed right?

    Hint: Which way does a long tail pull the mean?

  2. For the data 2, 3, 3, 4, 4, 4, 5, 5, 20, find the mean and median. What does this say about skewness?

    Hint: Compare the mean with the 5th value of the ordered list.

  3. Seven exam scores are 35, 78, 85, 88, 90, 92, 94. Find the mean and median and state the direction of skew.

    Hint: Is the lone extreme score high or low?

  4. Set A is 1, 2, 3, 4, 5 and Set B is 1, 2, 3, 4, 15. Which set has positive skewness? Compare each mean with its median.

    Hint: Only one value differs between the sets.

  5. A histogram of 1,000 delivery times peaks at 25 minutes and has a long tail out to 90 minutes. Should you use the empirical rule to describe it?

    Hint: What shape does the empirical rule need?

  6. A data set has mean 40 and median 48. A student says: "The mean is lower, so the data is positively skewed." Correct this.

    Hint: Which tail would pull the mean below the median?

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

Is a mean above the median always a sign of right skew?

Usually, but not always. The rule works well for smooth shapes with one peak, but it can fail for small or lumpy data. In 1, 6, 6, 6, 9, 9, 9 the long tail runs left, down to 1, and the skewness is negative, about -1.3. Yet the mean, about 6.57, is above the median of 6. Look at the picture and the skewness itself, not just the gap between the centers.

Why do incomes and house prices skew right rather than left?

Because they have a floor but no ceiling. An income cannot fall much below zero, yet a few people earn hundreds of times the typical amount, so the values can stretch far to the right but not to the left. The same logic runs the other way for scores on an easy test: they cannot go above 100, so the tail stretches left toward the few low scores.

How big does a skewness number have to be to matter?

A common rule of thumb says a skewness between -0.5 and 0.5 is roughly symmetric, between 0.5 and 1 in size is moderately skewed, and beyond 1 is strongly skewed. The sign gives the direction: positive for a right tail, negative for a left one. Treat these cutoffs loosely for small samples, where a single value can swing the skewness a lot.

Can skewed data be made more symmetric?

Often, by taking logarithms. Incomes of 10,000, 100,000 and 1,000,000 are spread very unevenly, but their base-10 logs, 4, 5 and 6, are evenly spaced. Taking logs pulls in the long right tail much more than the bulk of the data, so right-skewed data often become roughly bell-shaped. You can then use mean-based methods on the logs, as long as you translate the results back into the original units.

Learning Path

← Before

Distribution Shape
Skewness

You are here

Next →

Mean vs Median
You need Distribution Shape: reading a histogram or dot plot for center, spread, peaks and gaps. Skewness gives a lopsided shape a name, a direction and a number. Next, Mean vs Median uses it to choose a center: with strong skew the median and IQR describe a typical value better than the mean and standard deviation.

See Also