A chart has three categories labelled 33 percent, 33 percent and 33 percent. The total is 99. Before concluding that someone lost one percent of the data, check whether the displayed values were rounded.
Imagine three equally sized groups. Each represents one third of the whole, or 33.333… percent. Rounding each value to a whole percent produces three labels of 33. The arithmetic of the labels differs from the arithmetic of the original fractions.
The same process can go above 100
Here is a second invented example. Four categories contain 33.5, 33.5, 16.5 and 16.5 percent of the observations. They total exactly 100. Using a conventional half-up whole-number rounding method produces 34, 34, 17 and 17, which total 102.
This is not proof that every total above 100 is harmless. It shows why the rounding rule and underlying values matter. Different rounding conventions can produce different labels at a halfway value.
Check whether the categories should add up at all
A multiple-choice survey may allow each respondent to select several answers. Its percentages can legitimately total far more than 100 because the categories overlap. That is a different explanation from rounding a set of mutually exclusive shares.
Read the question wording and chart note. “Percentage of respondents selecting each option” does not mean the same thing as “share of all responses”. If the denominator is unclear, the total alone cannot resolve it.
- Are the categories mutually exclusive?
- Does the display include every category?
- Are values rounded, and to what precision?
- Are counts or a downloadable table available?
When writing your own summary, do not silently adjust a category to force the displayed total to 100. That changes a reported value without explaining the method. Retain the source’s figures and note the rounding where relevant, or calculate consistently from the original counts and explain what you did.
The useful habit is to distinguish the data from its presentation. A one-point discrepancy deserves a check, but a neat total of 100 is not by itself evidence that the chart’s categories, denominator or interpretation are sound.