A headline lands in your feed: something has doubled the risk of something else, or a chart shows a line climbing off the top of the page. The number is doing a lot of work in that moment. It looks like evidence, and it arrives without context — no baseline, no time frame, no note about who was counted or what was actually measured. Most of us absorb the impression and scroll on. The impression sticks anyway.

This is not about becoming a statistician. It is about learning the handful of questions that separate a real finding from a number dressed up to look like one. Most misleading statistics in news coverage are not lies. They are accurate figures placed in a frame that flatters a point, and the frame is usually easy to check once you know where to look.

Ask What the Number Is Compared To

The most common trap is a statistic with no baseline. A story reports that a habit, food or behaviour is "linked to a 40 percent higher risk" of some condition. Forty percent sounds enormous. But a percentage increase tells you nothing until you know how common the thing was to begin with.

Suppose a condition affects two people in every thousand. A forty percent increase takes that to roughly three in a thousand. The relative change is real and the arithmetic is honest, but the practical difference is small. Reporters lean on relative risk because it produces bigger, more clickable numbers; absolute risk — the actual counts — is duller and far more useful to a reader.

Relative risk without a baseline is a mood, not a measurement.

So when you meet a percentage change, ask two things: compared with whom, and how common is this in the first place? If the article never answers either, treat the figure as a prompt to look further rather than a conclusion.

The Time Frame Is a Choice, Not a Fact

Any line chart can be made to tell a story by choosing where it starts. Begin a graph at a trough and the trend looks triumphant; begin it at a peak and the same data looks like decline. Neither chart has to be inaccurate. Both are edited.

Two habits help here. First, look for the dates on the horizontal axis. If a chart has no labelled years or months, that omission is itself the finding — someone chose not to show you the window. Second, when a claim rests on a short period, ask what the longer series looks like. A five-year rise in rents, temperatures, crime or sales can look dramatic and still sit comfortably inside a twenty-year pattern of noise.

Watch for comparisons across mismatched periods too: this quarter against the same quarter last year is sensible; this month against a single unusual month three years ago is not. Seasonality trips up a surprising amount of coverage, especially anything involving weather, retail, travel or employment.

Correlation Is Not Causation, and Headlines Know It

Two things rising together do not prove that one causes the other. Ice cream sales and drowning deaths climb in the same months, because hot weather drives both. The third factor — the confounder — is doing the real work, and it is often missing from the story.

There are other directions the arrow can point. If people who sleep badly report more anxiety, does poor sleep cause anxiety, does anxiety cause poor sleep, or does something else cause both? Observational studies frequently cannot tell, which is why careful reporting uses cautious language: "linked to", "associated with", "may increase".

Compare the study's own wording with the headline. A paper that says a factor is "associated with" an outcome has not shown cause and effect. If the headline has quietly dropped that hedge, you have found the gap between the research and the story.

Percentages Hide Their Denominators

A percentage is a fraction in disguise, and the bottom half of the fraction matters more than the top. "One in five teenagers" means nothing until you know which teenagers, how many were asked, and how they were selected.

Also watch the difference between percent and percentage points. A rate falling from four percent to two percent is a halving — a fifty percent drop — but only two percentage points. Both descriptions are technically correct, and one of them will be doing the persuading.

Small bases distort large numbers in the other direction. A town with three reported burglaries last year and six this year has seen cases double. That is accurate and almost meaningless as a trend, which is why comparisons of rare events need several years of data before they say anything at all.

Check Who Was Asked, and Who Answered

Surveys are where shaky numbers most often enter the news. Two questions cut through most of the noise:

  • Who was in the sample? A survey of readers of one website, members of one organisation, or people who happened to be online at a particular hour is not a picture of the public.
  • How many responded? A survey sent to thousands that a few hundred people completed can be informative, but it needs to be described that way. Self-selected respondents are usually not representative.
  • What was actually asked? Wording shifts answers. "Do you support the policy?" and "Do you support the policy if it raises your taxes?" can produce very different results from the same people.
  • How big are the subgroups? A dramatic finding about a slice of the sample — one age band, one region — may rest on a few dozen people.

None of this makes surveys worthless. It makes them worth reading carefully, especially when a single poll is presented as a settled fact. One poll is a snapshot; a series of well-run polls is a signal.

A Short Checklist Before You Share

  1. Find the baseline: how common is this normally?
  2. Find the time frame: is the period stated, and does a longer view change the picture?
  3. Check the verb: does the source say "causes" or "is associated with"?
  4. Convert the percentage: what does it mean in absolute numbers?
  5. Ask who was counted, and who was left out.
  6. Trace the number to its origin: a study, a press release, an industry body, a single anonymous source.

That last point is worth dwelling on. Many stories recycle a figure from a press release that summarised a study that measured something slightly different from what the headline claims. Following the trail takes a minute and often explains the oddity you noticed.

When to Slow Down

You do not need to interrogate every number you see. Save the effort for the stories that will change what you do — a health claim that might affect a decision about treatment, a financial figure that might shape where you put savings, a legal change that might affect your rights. In those cases, the questions above are a starting filter, not a final answer. For anything touching your health, money or legal position, speak to a qualified professional who can look at your circumstances rather than a population average.

For everything else, a simpler habit will carry you a long way. When a number surprises you, pause before you repeat it. Ask what it is compared to, over what period, and measured in whom. Often the article answers all three and the finding holds up. Sometimes it does not, and you have saved yourself from passing along a figure that was never quite saying what it seemed to say.

Photo: AS_Photography / Pixabay