How to Read Charts and Spot Misleading Ones
Charts are everywhere: the news, ads, work slides, your electric bill. Learn the five-step habit for reading any chart, then the handful of tricks that make honest numbers tell a dishonest story.
Read any chart in five steps
Most people look at the shape first and decide what it means in a second. Slow down for ten seconds and go in this order instead:
- Title. What is this chart claiming to show?
- Axes and units. What's along the bottom (x-axis) and up the side (y-axis)? Dollars, percent, thousands, per person?
- Scale. Where does the y-axis start? Are the steps even? Is anything missing?
- Source and date. Who made the data, and how old is it? No source is a red flag.
- Then the pattern. Now look at the shape: what's going up, down, biggest, smallest? Is the difference actually large?
Pick the right chart for the job
| Chart | Best for | Watch out for |
|---|---|---|
| Bar | Comparing amounts across groups | Bars must start at zero |
| Line | Change over time | Uneven time gaps, zoomed axes |
| Pie | Parts of one whole (adds to 100%) | Too many slices, totals not 100% |
| Scatter | Whether two things move together | Assuming one causes the other |
| Table | Exact values | Hard to see trends |
Axes and scales: the biggest trick
A truncated y-axis starts at some number above zero. That zooms in on the top of the bars, so tiny differences look huge. For bar charts, this is almost always misleading, because we judge bars by their length. A bar twice as tall should mean twice as much.
Line charts are different. They show change, so it's often fine (even helpful) for the axis not to start at zero. Body temperature from 97°F to 104°F would be a flat line if the axis started at zero. Just read the numbers on the axis before you judge how big a change is.
A coffee shop sold 480, 495, 510 and 520 thousand cups over four years. Move the y-axis start and watch the “growth” change.
Other scale tricks
- Stretching or squashing the chart makes the same line look steep or flat. Read the numbers, not the slope.
- Uneven intervals: years like 1990, 2000, 2020, 2021, 2022 spaced evenly make recent change look faster.
- Two y-axes (one on each side) can make any two lines look like they move together. The creator chose both scales.
- Log scales go up by multiples (1, 10, 100, 1,000). They're useful for things that grow fast, but a straight line on a log scale means steady percentage growth, not steady amounts.
- Missing time: a chart that starts at a low point or ends at a high point can hide the bigger picture. Ask what happened just before and after.
Counts vs percentages and rates
A count is a raw number: 900 car crashes. A rate divides by the size of the group: 30 crashes per 100,000 people. Big places almost always have bigger counts, just because more people live there. To compare fairly, use rates.
- City A: 900 thefts, 3,000,000 people → 900 ÷ 3,000,000 = 30 per 100,000
- City B: 300 thefts, 200,000 people → 300 ÷ 200,000 = 150 per 100,000
- City B has fewer thefts but a rate five times higher.
Percent vs percentage points
If a rate goes from 10% to 15%, it rose 5 percentage points. But it also rose 50 percent (5 is half of 10). Both are true. Headlines pick whichever sounds more dramatic, so check which one you're reading.
Relative vs absolute risk
“Doubles your risk!” sounds scary. But if the risk goes from 1 in 10,000 to 2 in 10,000, the extra risk is just 1 in 10,000. Always ask: out of how many?
Small numbers, big percentages. A town that goes from 2 cases to 6 has a “200% increase.” That's true, but it's also only 4 cases. Look for the actual counts behind any percentage.
Percents that don't add up. A pie chart must add to 100%. If a survey let people pick more than one answer, the total can be over 100%, and a pie chart is the wrong tool. Use bars.
Averages can hide the story
“Average” usually means the mean: add everything up and divide by how many. The median is the middle value when you line them all up. When a few values are extreme, the mean gets pulled toward them and the median doesn't.
- $40k, $45k, $50k, $55k, $60k → mean $50k, median $50k
- A billionaire moves in with $10 million a year.
- New mean: about $1.7 million. New median: $52.5k.
- The median still describes a typical neighbor. The mean doesn't describe anyone.
- Home prices and incomes are usually reported as medians for exactly this reason.
- An average without the spread (range, or how much values vary) can hide a lot. A class average of 75 could be everyone at 75, or half at 100 and half at 50.
- Want the full lesson? See mean, median and mode.
Correlation is not causation
Two things are correlated when they tend to rise and fall together. That does not mean one causes the other. When you see a chart with two things moving together, consider these explanations first:
| Explanation | Example |
|---|---|
| A hidden third factor | Ice cream sales and sunburns rise together. Hot, sunny weather causes both. |
| Reverse cause | Towns with more firefighters have more fires? No, towns with more fires hire more firefighters. |
| Pure coincidence | With enough data, unrelated things line up by chance. Tyler Vigen's “Spurious Correlations” site collects many funny examples. |
| Real cause | Sometimes it is! But you need more evidence, like a controlled experiment. |
How do scientists show cause? The strongest way is a randomized experiment: randomly split people into groups, change one thing for one group, and compare. Random assignment spreads hidden factors evenly across the groups.
What's wrong with this chart?
Put it all together. Each chart below has one main problem. Pick it, then read why.
FAQCommon questions
Does every chart have to start at zero?
Bar charts (and filled area charts) should, because you judge them by length. Line charts and scatter plots don't have to, since they show change and position. Just read the axis labels before judging how big a change is.
What's the difference between the x-axis and y-axis?
The x-axis runs left to right along the bottom. It usually shows categories or time. The y-axis runs up the side and usually shows the amount being measured.
Why do people use pie charts if they're criticized?
They're fine for showing a few parts of one whole, like a budget split into 3 or 4 pieces. They get hard to read with many slices or slices of similar size. A bar chart is usually clearer.
What's a “per capita” number?
Per person. Divide the total by the population. It lets you compare places of different sizes fairly.
How can I tell if a chart's data is trustworthy?
Look for a named source, like a government agency or a published study, and a date. Then check that the chart actually matches what the source says.