What does rounding hide when I compare values in a chart?

It can hide a real difference between values by giving them the same displayed label. Imagine a chart of two groups with values 4.2 and 4.4, labeled to the nearest whole number. Both labels read 4, even though the underlying values differ by 0.2. This is an invented example. [S1]

Treat a rounded label as an approximation, not necessarily as the exact value. The more your comparison depends on a small difference, the more important it is to check the values behind the labels. OpenStax advises rounding only the final answer when possible and notes that rounding can produce errors. [S1]

Not from those labels alone. You can say that the displayed labels are equal; you cannot conclude that the underlying values are equal unless the chart or its data confirms it. If the numbers are rounded to the nearest whole number, different values can share the same label. The labels also do not tell you exactly how far apart the original values were.

For example, 4.2 and 4.4 both display as 4 under that rounding rule, but they are not equal. If the distinction matters, look for a data table, a note explaining the rounding, or values displayed with more precision. [S1]

First, find out how the labels were rounded: to the nearest whole number, tenth, or another place. Then seek the unrounded values if the chart does not show enough detail. Compare those values directly rather than trying to recover exact numbers from rounded labels; a rounded label may not identify one unique original value.

Worked example: suppose a chart shows A = 7 and B = 7, rounded to the nearest whole number. The labels alone do not settle whether A and B are equal or which is larger. If the underlying values are A = 7.2 and B = 7.4, B is larger by 0.2, despite the matching labels. These numbers are invented to illustrate the point. [S1]

Keep your conclusion at the level the chart supports. You might say, “Both values are shown as 7; the chart does not show whether they are exactly equal.” If the bars look different in height, describe that as a visual difference, but do not claim an exact gap based only on rounded labels.

A useful question to ask is: “What are the unrounded values, and to what place were these labels rounded?” If you need a precise comparison and cannot get the underlying values, report that limitation instead of guessing. A chart’s labels alone may not reveal its rounding method or hidden values.