How to read a chart

Lesson 1

Read the axes first

Before you look at the shape of a chart, find out what its two axes measure and where they start.

The axis along the bottom usually shows time or the things being compared. The axis up the side shows the amount. Read both before you decide what the chart is saying.

This chart shows the average temperature of the contiguous U.S. each year. Its side axis doesn't start at zero, and a note under the chart says so. That makes changes look bigger than they would on an axis from zero: the warmest and coolest years are 5.4°F apart, yet the line swings across most of the chart. That is fair here, because 0°F isn't a natural starting point for a climate and the chart tells you to read the numbers.

Bars are different. A bar's length stands for its value, so a bar chart that starts above zero misleads: a bar that looks twice as long may be only slightly bigger. When you meet one, read the numbers and ignore the lengths.

Lesson 2

Find the units

A number means little until you know what it counts. The same fact can look very different in another unit.

Look for the unit in the axis label or the title: dollars, percent, people, deaths per 100,000. Charts also shorten big numbers, so $30T on this chart's side axis means 30 trillion dollars.

Here is one fact in three units. At the end of fiscal 2026 the federal debt was $40.2 trillion in total, about $110,000 for each person in the U.S., and 125% of a year's GDP, everything the economy produces in a year. All three are true. They answer different questions.

Check money for one more thing: are these dollars of the day, or adjusted for inflation? This chart uses dollars of the day. A dollar bought far more in 1960 than it does now, so old and new figures aren't like for like.

• How much federal debt is there per American? →
• How big is the federal debt compared with the economy? →

Lesson 3

Rates versus counts

A count says how many. A rate says how many out of how much, which is the fair way to compare different sizes.

A rate divides a count by what could have been counted: per 100,000 people, or per mile driven. Whenever you see a rate, ask "per what?"

These two charts measure the same deaths. The first counts people killed in traffic crashes: 51,091 in 1980 and 39,254 in 2024. The second divides the deaths by the miles Americans drove: 3.35 deaths per 100 million miles in 1980 and 1.19 in 2024. Only the second takes into account how much driving people did.

Rates matter most when you compare places. A state with many people tends to have the most of nearly everything, so a ranking by count often just shows who is biggest. Ranked by rate, Mississippi had the highest traffic death rate in 2024, 1.81 deaths per 100 million miles.

• Which states have the highest and lowest traffic death rates? →

Lesson 4

Margins of error

A survey gives an estimate, not an exact count. Its margin of error shows the range the true figure likely falls in.

Many numbers come from asking some people rather than everyone. This chart's income figures come from a Census Bureau survey, so each state's figure carries a "±", its margin of error.

Massachusetts's median household income is $103,960 ± $527, so the true figure is likely between $103,433 and $104,487. The Census Bureau publishes its margins at 90% confidence, which means likely, not certain.

Rankings can make small gaps look important. When two ranges overlap, treat the two places as too close to call, even if one is ranked above the other.

Smaller places have wider margins, because the survey reaches fewer people there. California's estimate is $99,122 ± $310, while Owsley County, Kentucky's is $22,188 ± $9,297.

• What is the typical household income in your county? →

Lesson 5

Check the source

A chart is only as good as its data. Trace the numbers back to whoever collected them.

1. Who collected the numbers? Look for the agency or group that gathered them, not only the site showing them.
2. Can you open the source? A good citation links to a page where you can see the numbers yourself.
3. How old is the data? Look for the data date, not the date a story was published.
4. What is left out or changed? Look for notes on gaps, estimates, revisions and changed definitions.
5. May it be reused? Look for the license if you plan to republish the chart.

Try those five questions on this chart. Its source and data date (December 31, 2025) are under it, and the source name opens the FBI's own page.

The fourth question matters most here. These are crimes that police recorded and reported, so crimes nobody reported aren't in them. And in 2021 the FBI stopped taking its old summary reports. Many police departments hadn't switched to the newer system, so agencies covering only about 77% of Americans reported, against 88 to 98% in other years.

Chartly isn't the original source of any of these numbers. When you cite one in your own work, name the agency that collected it as well, and check it at the source if you can.