S.M.M.

Stop Manipulating Me · A Field Guide to Psychological Influence

ENTRY No. T12.19
CATEGORY Deception
CLEARANCE Public / Essential
EDITION 01
Dossier · Manipulation Tactic

Statistical Manipulation

Falsification · Valid-Data Misuse · Widespread
High Alert
How It WorksSEC 01

Statistical manipulation misuses genuine data — through presentation, selection, or framing rather than fabrication — to make it support a conclusion the honest numbers don’t. Because the underlying figures can be real, it is more insidious than making numbers up: the deception hides in the packaging. Darrell Huff catalogued the core repertoire in his 1954 classic How to Lie with Statistics — the truncated y-axis, the missing base rate, the cherry-picked timeframe, the confusion of correlation with causation — and every one of them still runs daily in news graphics, ad copy, and political messaging.

The tactic exploits what might be called the authority of numbers: a chart or percentage reads as objective and rigorous, so audiences apply less scrutiny to it than to prose making the same claim. A y-axis that starts at 90 instead of 0 turns a rounding error into a cliff; a raw count with no denominator inflates a rare event; a “risk doubled” headline hides that the risk went from negligible to slightly-less-negligible. Each move survives a glance precisely because the viewer assumes the graphic is neutral.

Defense is to reconstruct what the presentation stripped away. Reset the axis to its full range, demand the denominator (“out of how many?”), widen the cherry-picked timeframe, and translate relative changes back into absolute numbers. Above all, insist on the source, sample size, and method — a figure with no checkable methodology is decoration, not evidence. An honest statistic holds its shape when you restore the full scale and base rate; a manipulated one only impresses while those are missing.

Warning SignsSEC 02
  • Truncated axis. A y-axis that doesn't start at zero, magnifying a trivial difference into a dramatic cliff.
  • Missing denominator. Raw counts with no base rate — "10,000 cases" out of how many? A percentage without its total.
  • Cherry-picked timeframe. A start and end date chosen to show the trend the author wants, ignoring the fuller series.
  • Relative dressed as absolute. "Risk doubled!" when it went from 1-in-a-million to 2-in-a-million.
  • Correlation sold as cause. Two lines moving together presented as one driving the other.
  • No source or method. A striking figure with no study, sample size, or methodology you can check.
Frequently Paired WithSEC 03
  • Fabricated Statistics · T12.11
    Invented rather than distorted numbers
  • Misleading Comparisons · T12.15
    Baseline distortion in quantitative form
  • Cherry-Picking · T13.1
    Selective-evidence engine behind the timeframe trick
  • Appeal to Data · T2.9
    The authority-of-numbers lever exploited
How the Numbers MisleadSEC 04
  • Stage 01 · Select
    The framer chooses the slice, timeframe, or metric that best fits the desired conclusion and drops the rest.
  • Stage 02 · Present
    Axes, scales, and framing are tuned — truncated, rebased, or relative-only — so the visual overstates the effect.
  • Stage 03 · Assert
    The polished figure is offered as objective proof, and the authority of numbers discourages readers from checking the method.
Counter-ProtocolSEC 05
Defense: Numbers persuade by feeling objective — restore the axis, the denominator, and the method before you believe them.
  • Rebuild the axis. Mentally reset the scale to zero and full range. If the dramatic gap flattens, the drama was in the chart, not the data.
  • Demand the denominator. Ask "out of how many, over what base rate?" A count without its total can be made to say almost anything.
  • Widen the timeframe. Look at the fuller series. A trend that only appears within a hand-picked window is an artifact of the window.
  • Separate relative from absolute. Translate "doubled" into actual numbers. A big relative change on a tiny base is often trivial.
  • Check method and source. Insist on sample size, methodology, and a primary source. No method, no credibility.