The median is the middle value in a dataset when all numbers are arranged from smallest to largest. It tells you the point where exactly half the values fall below and half fall above. This makes it equivalent to the 50th percentile, a fact that unlocks much of its practical meaning: if the median household income in your city is $65,000, half of households earn less than that and half earn more.
Understanding what the median actually communicates, and when it can mislead you, requires looking beyond the basic calculation.
How to Find the Median
Start by sorting your data from smallest to largest. What happens next depends on whether you have an odd or even count of values.
With an odd number of observations, the median is the single middle value. You can locate it by finding the (n+1)/2 position, where n is your total count. In a dataset of 9 values, that’s the 5th number: four values sit below it, four above.
With an even number of observations, there’s no single middle value. Instead, you take the average of the two center numbers. In a set of six values sorted in order, you’d average the 3rd and 4th values. If those two numbers are 3 and 4, your median is 3.5.
What the Median Actually Tells You
The median answers one specific question: what is a typical value in this dataset? It does this by finding the centerpoint of position, not of magnitude. This distinction matters enormously. The median doesn’t care how far away the extreme values are. It only cares about which value sits in the middle when everything is lined up.
Think of five people in a room earning $30,000, $40,000, $50,000, $60,000, and $70,000. The median income is $50,000, and the mean is also $50,000. Now replace the highest earner with someone making $5,000,000. The median stays at $50,000. The mean jumps to over $1,000,000. The median still reflects what a “typical” person in the room earns. The mean no longer does.
This resistance to extreme values is what statisticians call robustness. The median remains stable against the influence of outliers, which is why researchers increasingly recommend using it as the foundation for detecting unusual data points rather than relying on the mean.
Median vs. Mean in Skewed Data
When data is perfectly symmetric, like a bell curve, the median and the mean land in the same spot. In a normal distribution, the mean, median, and mode are all identical. This alignment is one reason the normal distribution is so useful in statistics: it doesn’t matter which measure of center you choose.
Real-world data is rarely that tidy. When data is skewed, the mean and median split apart, and the direction of that split tells you something important. In a right-skewed distribution, where a long tail stretches toward higher values, the mean gets pulled to the right of the median. In a left-skewed distribution, the mean shifts to the left of the median. The mean chases the tail; the median holds its ground.
This is exactly why the Social Security Administration reports both average and median wages. The distribution of workers by wage level is highly skewed, with a small number of very high earners pulling the average up. The median wage is substantially less than the average wage. If you only saw the average, you’d overestimate what a typical worker earns. The median gives a more grounded picture of the middle of the pack.
Why Income and Housing Use the Median
Whenever you see economic statistics reported as medians, the reason is almost always skew. Home prices, household income, net worth, medical costs: these distributions all have long right tails. A handful of multimillion-dollar homes in a neighborhood can inflate the average price far beyond what most houses actually sell for. The median home price tells you what the middle buyer actually paid, which is far more useful if you’re trying to figure out whether you can afford to live there.
The same logic applies to salary data. If a company reports that the average employee earns $120,000, that number might be dragged up by a few executives making seven figures. The median salary strips that distortion away and shows you what the person in the middle of the pay scale takes home.
Reading the Median on a Box Plot
Box plots (also called box-and-whisker plots) are built around the median. The line inside the box represents the median, or the 50th percentile. The left edge of the box marks the 25th percentile, and the right edge marks the 75th percentile. The distance between those two edges is the interquartile range, or IQR, which captures the middle 50% of your data.
This is also why professional reporting guidelines call for presenting the median alongside the IQR rather than on its own. A median of 42 tells you where the center is. A median of 42 with an IQR of 38 to 51 tells you where the center is and how tightly the data clusters around it. The IQR serves as the median’s natural partner the same way the standard deviation pairs with the mean.
When the Median Can Mislead
The median’s greatest strength is also its biggest limitation. Because it only reflects the value at the center position, it ignores the actual magnitude of every other data point. This makes it resistant to outliers, but it also means it can hide important information.
If your dataset changes dramatically at the extremes but the middle value stays put, the median won’t budge. Imagine tracking donations to a charity over several months. If a few major donors triple their contributions but everyone else stays the same, the median donation won’t change at all, even though total revenue increased significantly. The median focuses on what’s common and filters out what’s rare.
This creates a real tradeoff. Extreme values aren’t always errors. They can be a natural feature of the system you’re studying, like disaster impacts, epidemic costs, or investment returns. In those contexts, a single catastrophic event can matter more than a thousand ordinary ones. Using the median on skewed data means you’re deliberately ignoring rare extremes and focusing on the typical case, which may not reflect the full picture your analysis needs.
Another subtle pitfall arises with small datasets or data that clusters around a few values near the center. When many observations share the same value near the midpoint, the median can stay locked on that value even as the rest of the distribution shifts. You won’t notice a change where there is one if that change doesn’t cross the median’s position.
Choosing Between Median and Mean
Neither measure is universally better. The choice depends on the shape of your data and the question you’re asking. Use the median when your data is skewed or contains outliers and you want to describe a typical value. Use the mean when your data is roughly symmetric and you need a measure that accounts for the magnitude of every observation, such as when calculating totals or rates.
In practice, reporting both can be the most informative approach. A large gap between the mean and median immediately signals skew. If the mean is much higher than the median, you know a few large values are pulling the average up. If they’re close together, your data is roughly balanced. That gap itself becomes a diagnostic tool, telling you about the shape of your distribution before you even look at a chart.