When the median is higher than the mean, the data is negatively skewed, meaning a few unusually low values are dragging the average down while most values cluster toward the higher end. This pattern shows up in several real-world datasets, from human lifespan in developed countries to exam scores on an easy test. Understanding why it happens helps you choose the right number to describe what’s “typical” in a dataset.
What Negative Skew Looks Like
Picture a histogram where most of the bars are bunched up on the right side, with a long tail stretching to the left. That’s a negatively skewed (or left-skewed) distribution. The peak of the data sits toward the higher values, and the tail is made up of a smaller number of unusually low observations. In this shape, the three main measures of center line up in a specific order along the number line: the mean sits lowest, the median falls in the middle, and the mode (the most common value) sits highest.
The key principle is simple: the mean follows the tail. In a negatively skewed distribution, the tail points left, so the mean gets pulled in that direction. The median, which only cares about the middle position in a sorted list, barely moves. That gap between the two is your signal that the data isn’t symmetric.
Why Low Outliers Pull the Mean Down
The mean is calculated by adding every value together and dividing by the count. That means a single extreme number has real influence over the result. If you add one very low value to a dataset, the mean drops significantly because that outlier gets folded directly into the sum.
The median works differently. It’s simply the middle value when you sort the data from lowest to highest. It depends on position, not magnitude. Whether the lowest number in your dataset is 10 or 10,000 below the next value, the median stays in the same spot. This is why a handful of extremely low observations can create a noticeable gap where the median sits above the mean, even though most of the data hasn’t changed at all.
Real-World Examples
Human Lifespan in Developed Countries
One of the clearest examples of a negatively skewed distribution is age at death in wealthy nations. Most people live into their 70s and 80s, creating a large cluster of values near the upper end. But deaths in infancy, childhood, or middle age pull the left tail outward. A study examining 74 life tables from around the world found that in 95% of cases, adult lifespan was negatively skewed, with skew values below negative one. For American white males using 2006 data, the median age at death was approximately 79.3 years. The mean (life expectancy) was lower, pulled down by those who died young. This is why researchers working with longevity data generally prefer the median over the mean.
Historically, this distinction mattered even more. During the 20th century, life expectancy (a mean) rose rapidly as infant mortality declined, but the mode of adult lifespan rose much less. The mean was being heavily influenced by the shrinking left tail rather than by people actually living longer at the top end.
Easy Exam Scores
When a test is relatively easy, most students score high and cluster near the top of the scale. A small number of students who were unprepared or guessed poorly end up with very low scores, forming a left tail. The result: the median score is higher than the mean, because those few low scores pull the average down while the median sits comfortably in the high-scoring majority.
Student Height Within an Age Group
The Australian Bureau of Statistics identifies student height as another example of negatively skewed data. Within a specific age group, most students fall near a common height range, but a few individuals on the shorter end extend the left tail. The median height ends up slightly above the mean.
How This Differs From Income Data
You may have heard that the median is better than the mean for describing income. That’s true, but for the opposite reason. Income and wealth distributions are positively skewed (right-skewed), with a long tail of very high earners pulling the mean upward. In that case, the mean is higher than the median. Wealth distributions are particularly extreme: the Gini coefficient for U.S. wealth is about 0.8, compared to roughly 0.48 for earnings, reflecting just how far the right tail stretches.
So while both negatively and positively skewed data create a gap between the mean and median, the direction flips. With income, a few billionaires inflate the average. With lifespan in a developed country, a few early deaths deflate it. In both cases, the median gives you a better sense of what’s typical for the average person.
When to Use the Median Instead
Any time your data is skewed or contains outliers, the median is generally the more useful number. It reflects the experience of the person in the middle of the pack rather than being distorted by extremes at either end. Practical guidelines are straightforward:
- Use the median when you want to describe the “typical” value and your data has a visible tail in either direction.
- Use the median when outliers are present and you don’t want a handful of extreme cases to misrepresent the group.
- Use the mean when the data is roughly symmetric and you need a mathematically precise center, or when every value (including extremes) should genuinely contribute to the summary.
Household income reporting is the classic case: agencies typically report median household income specifically because a few ultra-high earners would make the mean misleadingly high. The same logic applies in reverse for negatively skewed data. If you’re summarizing lifespan, test scores with a ceiling effect, or any dataset where most values bunch near the top, the median tells the more honest story.
How to Spot Negative Skew in Your Data
The quickest check is to compare the mean and median directly. If the median is noticeably higher, your data is likely left-skewed. You can also look at a histogram or box plot. In a histogram, negative skew shows up as a longer left tail with the bulk of bars shifted right. In a box plot, the whisker on the lower end will be longer than the one on the upper end, and the median line inside the box will sit closer to the top.
Most statistical software will also calculate a skewness value. A negative number confirms left skew, and the further below zero it falls, the more pronounced the asymmetry. Values between zero and negative 0.5 represent mild skew, while anything beyond negative one (like the lifespan data, which consistently fell below negative one across dozens of countries) indicates a strongly skewed distribution where the mean and median will diverge substantially.