Red flags that a figure needs a closer look
Some figures deserve a second look before you repeat them. Learn the common warning signs, such as a missing denominator, a vague source and an unusually neat round number.
Three warning signs matter most when a figure appears without context: a missing denominator, a vague source and an unusually neat round number. A count with no total beside it cannot be judged, because nobody can tell whether it is large. A figure that names no organization, report or date cannot be checked. A tidy round number, such as exactly 50 percent or 1,000 people, may be a rounded estimate rather than a measured count. Two more checks are worth making: whether the parts add up to the stated total, and whether the figure is larger than the population it claims to describe.
None of these flags proves that a figure is wrong. Each one is a reason to slow down and ask a question before repeating it to anyone else. The aim is not to distrust everything. It is to notice when a figure asks for more trust than its presentation has earned, and to know what to ask for next. Most of the checks below take a minute or two and need nothing beyond the article or post in front of you and, sometimes, a short search for the original report.
Signs in how the figure is presented
Figures that arrive without context are harder to test. Watch for "up" or "down" with no starting value, a percentage with no count behind it, or a comparison with "the average" that never says which average or which group. Vague quantities such as "thousands," "a large share" or "nearly double" are not wrong by themselves, but they cannot be checked until someone supplies the actual values.
Emotional or absolute wording is another cue. Phrases such as "shocking surge," "always" or "never" often describe a single case or a small change in a larger picture. A figure presented as proof of a dramatic story, but with no detail on how it was produced, deserves more scrutiny than one presented alongside its method.
Precision can mislead as well. A figure given to a decimal place or two can look more reliable than the method behind it justifies, particularly when the underlying measure is an estimate. The right amount of precision depends on how the number was produced. A report that explains its method is usually easier to trust than one that simply lists exact-looking values without saying where they came from.
Common flags and the question that settles them
The table below pairs frequent warning signs with a question that usually resolves them.
Scroll sideways to see the full table →
| Warning sign | Question to ask |
|---|---|
| No denominator, only a count | Out of how many? |
| Very round numbers, such as exactly 50 or 1,000 | Was this counted, or estimated? |
| Vague or missing source | Who measured this, and where is the report? |
| Large percentage from a tiny base | How many cases sit behind the percentage? |
| No dates, or an unusually short window | What period does this cover? |
| Comparison with an unnamed "average" | Which average, over which group? |
Round numbers are common in estimates and in rounded reporting, so a round figure is a reason to ask, not a sign of fraud. A single flag is usually ordinary. Two or three together, especially in a figure that supports a strong claim, make a much stronger case for checking the original before you rely on it.
Numbers that do not add up
Recalculate what you can. If a report says a figure rose from 40 to 50, the change is 10 units, which is 25 percent of the starting value. If a stated percentage does not match the figures printed beside it, that mismatch is itself worth a question, and it sometimes reveals that two different bases were mixed together.
Breakdowns should add up. When a report splits a total into categories, the parts should sum to the total, and shares of one whole should sum to about 100 percent, allowing for rounding. If five categories of a single question add to 115 percent, either a base has been mixed up or the question allowed several answers at once. Multiple-answer questions can total more than 100 percent legitimately, so check the question wording before calling it an error. A report that does not say which applies deserves the question.
A count of people cannot exceed the population it is drawn from. If a report says 12,000 residents of a village of 3,000 took part in a survey, something is wrong with the figure, the population or the description. Compare any count with a known total, such as an official population estimate for the same place and date. A share of a population above 100 percent is impossible for a single share, so it signals a mistake in the base or the wording.
Deciding what to do with a flagged figure
Tracing a figure back to the first report that published it is the most useful single step, and a trail that ends at an unnamed source or a claim that a figure is "widely reported" is a red flag in its own right.
A flagged figure does not have to be discarded. You can use it with an explicit caveat, look for an independent source that points the same way, or tell the people you are speaking with that the figure is uncertain. A useful rule is to match your confidence to the checks you have actually done. A figure you have traced to its origin and understood is a different thing from one you have seen once in passing. Keep a short note of what you checked, such as "traced to the original release, definitions match, sample size stated." That note lets you reuse the check later and tells anyone you share the figure with how much weight it can bear.
When a figure feeds a decision about money or health, write down the figure, its source, its date and the one check you completed before acting on it. If you cannot name the original report, leave the figure out of that decision, and say so if you mention it to someone else.