Manipulated Statistics

How Statistics Lie Without Technically Lying

Statistics can be technically accurate yet profoundly misleading. A pharmaceutical company might report that its drug reduces heart attack risk by 50% — a relative risk reduction that sounds dramatic. But if the baseline risk was 2 in 1,000, the drug reduces it to 1 in 1,000. The absolute risk reduction is 0.1%, which sounds far less impressive. Both numbers are 'true,' but they tell very different stories.

This is not an edge case. It is the standard way health statistics, economic figures, and policy outcomes are presented to the public. Learning to distinguish relative from absolute numbers, to ask about base rates, and to demand context for any statistic is a core critical thinking skill.

The Most Common Manipulation Techniques

Truncated axes. A bar chart showing approval ratings from 48% to 52% with a Y-axis starting at 47% makes a 4-point difference look enormous. If the axis started at 0%, the bars would appear nearly identical.

Averages hiding distributions. A company reports average salary of $120,000. But if the CEO earns $10 million and most employees earn $50,000, the average is meaningless. The median — the midpoint where half earn more and half earn less — tells the real story.

Cumulative vs. per-period figures. '10 million jobs created in four years' sounds impressive until you learn that 3 million of those were recovering jobs lost in year one. The net creation is quite different from the gross figure.

Denominator tricks. 'Country X had 500 murders last year, versus 100 in Country Y.' But Country X has 50 million people and Country Y has 1 million. Per capita, Country Y's murder rate is five times higher.