A graph can use only true numbers and still fool you. Here are the eight tricks that show up most, two of them from real news charts, with the misleading version next to the honest one wherever a redraw helps.
A misleading graph shows real data in a way that gives the wrong impression. The most common tricks are a bar-chart axis that doesn’t start at zero, an upside-down axis, a stretched scale, pictures that grow in two directions, uneven steps on the axis, showing only some of the dates, an average that hides the spread, and treating two things that move together as cause and effect.
The 8 tricks, misleading vs. honest
1.
The bar chart doesn’t start at zero
real news example
In 2012 a TV news chart showed the top US income tax rate going from 35% to 39.6% if a tax cut ended. The bars started at 34%, not 0%. The change is 4.6 percentage points, but the second bar was drawn about 5½ times as tall as the first.
MisleadingHonest
Watch the trick: same numbers, the axis moves
Axis starts at 34%: the last bar looks 5.6 times as tall as the first.
Axis starts at 34%: the last bar looks 5.6 times as tall as the first.
How to catch it: Look at the bottom number on the side axis. A bar chart compares lengths, so its axis must start at zero.
Taught in iMasterly Statistics, Grade 2: Bars That Don’t Start at Zero
2.
The axis is upside down
real news example
In 2014 a news graphic about gun deaths in Florida put 0 at the top of the axis. When deaths went up, the line went down, so at a glance it looked as if deaths had fallen. The chart below uses made-up numbers to show the same trick.
MisleadingHonest
How to catch it: Check which way the numbers run on the axis. Bigger numbers should be higher up.
3.
The scale is stretched (zoomed in)
A class’s test average went from 71 to 73 out of 100 over five months. Zoom the axis in to 70.5–73.5 and the line shoots up like a rocket. On the full 0–100 scale it is a small, steady gain.
MisleadingHonest
Watch the trick: same numbers, the axis moves
Axis from 70.5 to 73.5: the line climbs 67% of the chart's height.
Axis from 70.5 to 73.5: the line climbs 67% of the chart's height.
How to catch it: Read the numbers, not the slope. Ask how big the change is compared with the whole scale.
Taught in iMasterly Statistics, Grades 3–4: Stretched and Squished Graphs; The Zoomed-In Line
4.
The picture grows in two directions
Sales doubled from 2,000 to 4,000 bikes. Draw the second bike twice as tall and it also gets twice as wide, so it covers four times the space. Your eye sees 4×, not 2×.
MisleadingHonest
How to catch it: In a picture graph, count the pictures. Every picture should be the same size and stand for the same amount.
Taught in iMasterly Statistics, Grade 2: Giant Pictures
5.
The scale jumps unevenly
The axis reads 0, 10, 20, 50, 100 with equal gaps between. A team with 60 points looks about twice as good as a team with 15, when it really scored four times as many.
MisleadingHonest
How to catch it: Check that the steps on the axis are equal: 0, 10, 20, 30… and not 0, 10, 20, 50.
Taught in iMasterly Statistics, Grade 3: Uneven Scales
6.
Only part of the data is shown (cherry-picking)
Library visits rose over the year from 40 to 58. Show only April to June and it looks like visits are falling, which is a handy story if you want the library’s budget cut.
MisleadingHonest
How to catch it: Ask what happened before and after the dates shown. Why were these dates picked?
Taught in iMasterly Statistics, Grade 4: Cherry-Picking
7.
An average hides the spread
A sign says a river’s average depth is about 4 feet. That sounds safe to wade across, but in the middle it is 9 feet deep. One number can’t tell you how spread out the data are.
What the sign leaves out
How to catch it: When you see an average, ask for the smallest and largest values too, or for a graph of all the data.
Taught in iMasterly Statistics, Grade 6: The Average-Depth River
8.
Two lines move together, so one must cause the other
Ice cream sales and sunburns rise and fall together. Ice cream doesn’t cause sunburn: hot, sunny weather drives both. Statisticians call that hidden third thing a lurking variable.
The data
How to catch it: Before believing “A causes B”, look for something else that could cause both.
Taught in iMasterly Statistics, Grades 5 and 8: Ice Cream and Sunburns; Correlation vs. Causation
Numbers in examples 2 to 8 are made up to show the trick. Examples 1 and 2 describe real news charts; see the sources below.
1Where does the side axis start?Bar charts must start at zero. Line graphs can zoom in, but only if they say so.
2Which way do the numbers run?Bigger numbers go up. An upside-down axis flips the story.
3Are the steps equal?0, 10, 20, 30 is honest. 0, 10, 20, 50 is not.
4What is missing?Check the dates shown, the labels and the units. Ask what happened before and after.
5Is one number hiding the rest?An average needs the spread: the smallest, the largest, or the whole graph.
6Could something else explain it?Two things moving together doesn’t mean one causes the other.
Same numbers, different pictures
In 1973 the statistician Francis Anscombe made four small data sets with the same averages and the same line of best fit. Graph them and they look nothing alike: one is a straight trend, one a curve, two are thrown off by a single odd point. His lesson still holds: always look at the graph, and always ask how it was drawn.
Data Detective, grade by grade
Every grade of the iMasterly Statistics course has Data Detective lessons where kids catch these tricks themselves.
Grade
What kids learn to catch
K–1
Who was counted twice? and who did we really ask?
2
Bars that don’t start at zero, giant pictures, mystery graphs with no labels
3
Stretched and squished graphs, uneven scales, reading the key
4
The zoomed-in line, cherry-picking, when “most” means “a few”
5
Linked isn’t caused, “4 out of 5” small print, percent of what?
6
Averages that hide: the average-depth river, the average paycheck
7
Loaded survey questions and the people who never answered
8
Correlation vs. causation, Anscombe’s quartet, reading “a study found…” headlines
Questions about misleading graphs
What is a misleading graph?
A graph that shows real numbers in a way that gives the wrong impression, for example by starting the axis above zero, stretching the scale, using pictures of different sizes or showing only part of the data. The numbers can all be correct and the picture can still mislead.
What is the most common way graphs mislead?
A bar chart whose axis doesn’t start at zero. It makes small differences look huge. In the 2012 tax-rate example on this page, a 4.6-point change was drawn as a bar about 5½ times as tall.
Is it always wrong for a graph to skip zero?
For bar charts, yes, because the length of the bar is the message. For line graphs, zooming in can be fine when small changes matter (body temperature, for example), as long as the axis is clearly labelled.
What grade learns about misleading graphs?
Many US students meet them in middle-school math, around grades 6 to 8, when they study statistics. In iMasterly’s Statistics course every grade from kindergarten has Data Detective lessons, starting with bars that don’t start at zero and giant pictures in grade 2.
How can students practice spotting misleading graphs?
Use the 6-question checklist on this page on graphs from news sites and ads. A good class activity is the reverse: give students honest data and ask them to make it look as good, or as bad, as possible, then swap and catch each other’s tricks.