A housing table can show three percentages that total 99.9 rather than 100. Another chart may total 150. These results do not necessarily have the same explanation. The first could reflect rounding, while the second could reflect overlapping categories or a different denominator. The right response is to read the definitions before correcting the numbers yourself.
Begin by asking whether the categories are supposed to form one complete, nonoverlapping whole. Then check the rounding, population and missing information. This guide uses invented examples to explain those checks. It does not validate a particular dataset or claim that every unusual total is harmless.
Check whether the categories form a complete whole
For percentages to sum to 100 in the ordinary way, each member of the defined population must belong to exactly one of the displayed categories, and all categories must be included. If the table is only a selection, a total below 100 may be expected.
Imagine an invented inventory of 100 homes divided into 40 studios, 35 homes with one bedroom and 25 homes with two bedrooms. Those categories cover the whole hypothetical inventory without overlap, so the percentages total 100.
If a chart displays only studios and homes with one bedroom, its visible total is 75 percent. The missing 25 percent is not necessarily an error. It is outside the displayed selection and should be explained by the title or note.
Do not invent an other category until you understand what was omitted. The remaining share might contain several defined categories, missing responses or records outside the display. Use the source's explanation rather than assigning your own meaning.
Recognize a small rounding difference
The Census Bureau's Current Population Survey definitions explain that rounded percentages in a distribution may not add to exactly 100.0 percent. Rounding changes the displayed values even when the underlying calculation covers a complete whole. [1]
For an invented example with three equally sized groups, each exact share is one third, or 33.333 percent continuing indefinitely. Displaying each to one decimal place gives 33.3 percent. The displayed sum is 99.9 percent.
Nothing is missing from that hypothetical population. The difference comes from displaying each share with limited precision. A note about rounding is more accurate than changing one category to 33.4 without explaining an adjustment method.
Do not assume every discrepancy is rounding. A total of 80 percent would require a different explanation in this simple example. Investigate the categories, denominator and missing data before deciding what caused the difference.
Do not silently alter published percentages
If you reproduce a source table, preserve its published values and relevant notes. Do not adjust a category solely to force the total to 100. An unexplained change makes your version disagree with the source and can imply precision the source did not provide.
If your own presentation uses a documented rounding method, state it and retain the underlying values. Keep a record of the method so another person can reproduce the display. Do not mix adjusted and unadjusted percentages without labeling them.
When an exact total is important for a calculation, use the appropriate underlying data rather than rounded display values where available. Check whether the source permits that calculation and whether the categories remain compatible.
For a simple chart intended only to explain proportions, it may be enough to show the published rounded percentages with a concise note. The choice should serve the reader's understanding, not make the table look artificially perfect.
Look for overlapping categories
A question that allows multiple answers can produce percentages above 100. Each percentage may use the same respondent count while a person appears in more than one category. The sum then counts selections rather than distinct people.
In an invented survey of 100 households, 70 report using a bus and 60 report using a train. If 40 report both, the shares total 130 percent. The number reporting at least one of those modes is 70 plus 60 minus 40, or 90 households.
The 130 percent total does not mean there are 130 households. It reflects overlap between the two answers. This example is arithmetic only and does not describe travel in any actual city.
Read the questionnaire or category definitions when overlap is possible. Labels such as amenities used, reasons for moving or services considered can describe multiple selections. Do not turn them into a pie chart unless the data actually represent nonoverlapping parts of one whole.
Keep the denominator visible
A percentage is a count divided by a defined total. If the total changes from one row to another, adding the percentages may have no useful meaning. Copy the population description along with the value.
For an invented example, 20 of 50 renter households is 40 percent, while 30 of 150 owner households is 20 percent. Adding 40 and 20 does not describe the combined share. Across both groups, the combined count is 50 out of 200, or 25 percent.
A simple average of the two percentages would be 30 percent, which is also wrong for that combined population because the groups have different sizes. Use the relevant counts and compatible definitions when calculating a pooled percentage.
Do not combine percentages from different years or geographic units merely because the labels sound similar. Even when the arithmetic is possible, the resulting population needs a coherent definition. State what the combined figure represents.
Distinguish a row percentage from a column percentage
In a table with several categories on two dimensions, a percentage may describe a row total or a column total. The same underlying count can therefore produce different percentages depending on the question.
Imagine an invented table with 30 renter households and 20 owner households in one category. If there are 100 renters overall, the 30 represents 30 percent of renters. If the category contains 50 households overall, those same 30 represent 60 percent of that category.
Both statements can be correct. They answer different questions: the share of renters in the category versus the share of the category that rents. Confusing them changes the meaning even though the underlying count is unchanged.
Read the table heading, footnote and denominator label. If they are unclear, do not guess from the arrangement of the cells. Locate the documentation or ask the publisher before using the percentages in a housing comparison.
Treat missing responses separately from zero
A blank or missing response is not automatically a no answer. A table may exclude missing responses from its denominator, include an unknown category or handle them through another documented method. Check which approach the source uses.
For an invented question answered by 80 of 100 households, 40 yes responses are 50 percent of respondents but 40 percent of all households approached. Neither label can be omitted without making the figure ambiguous.
Do not fill the 20 missing responses with no to make a complete chart. That would add information you do not have. Keep the missing group visible or use the publisher's stated denominator and explain it.
If two sources handle missing responses differently, their percentages may not be directly comparable. A difference in the displayed shares could partly reflect the denominator rather than a change in the behavior you are investigating.
Read nested categories carefully
Some tables show a broad category followed by its subcategories. Adding the parent and all children counts the same observations more than once. Indentation, labels and table notes can help identify the hierarchy.
For an invented inventory, 60 apartments may include 25 studios and 35 larger apartments. Adding 60, 25 and 35 produces 120, but there are still only 60 apartments in that parent category. The smaller groups explain the total rather than add to it.
When making a summary, choose a consistent level of detail. You might show the parent categories or their compatible subcategories. Do not mix levels in one total unless the relationship is explicitly accounted for.
Keep the original table available when simplifying it. A clean chart can hide the indentation that originally signaled a nested category. Your new labels should preserve the meaning rather than only the visual appearance of the numbers.
Use a short diagnostic sequence before sharing
First read the title and identify the population. Next determine whether the displayed categories are complete, exclusive and at the same level. Then check rounding and missing response notes. Finally, reproduce any calculation from compatible underlying values where appropriate.
Write a plain explanation of the result. “Shares total 99.9 percent because each is rounded to one decimal place” is useful when that is actually the cause. “Respondents could select more than one answer” explains a different kind of total.
If the source does not resolve the discrepancy, describe it as unresolved rather than repairing it silently. You can still use other verified information from the table while withholding the uncertain conclusion.
The purpose is not to make every housing percentage sum to 100. It is to understand what is being counted and how. Once the population, categories and precision are clear, the total becomes a check on meaning rather than a cosmetic target.
Sources and scope
[1] U.S. Census Bureau. Current Population Survey Subject Definitions, Rounding
Source checked October 6, 2026. All counts, categories and percentages in examples are invented. Calculations illustrate arithmetic and do not validate a specific survey or housing dataset.