Don’t Be Deceived by Statistics
It’s sometimes said that statistics don’t lie, but that’s a lie. Statistics are powerfully effective at deceiving. Don’t let yourself be deceived. Consider these three examples.
Semi-attached numbers are the numbers you see in a claim that some product is a specified percentage better. “Better than what?” you might ask. And you should ask yourself that question, since just knowing that the delectable munchies on the supermarket shelf have 37 percent fewer calories is meaningless unless you know with what the 37 percent is being compared.
The use of unfair poll questions. The wording of a poll or survey can be adjusted to encourage a particular answer. For example, the question, “Do you feel you should be taxed so some people can get paid to stay home and loaf?” is likely to get a different response than “Do you think the government should help people who are unable to find work?” By adjusting the wording, crafty poll designers can rig polls to get the answers they want. So don’t be too quick to accept a report that “87 percent of respondents claim…” unless you’ve seen the questions that were asked.
Absolute vs. relative risks is a common means to deceive, especially in ads. For example, if you’re told that a particular drug reduced the incidence of a certain condition by 50 percent, you might be impressed. But if you knew that without the drug, only two people out of 100 were likely to get the condition, and by using the drug, only one did—i.e., only one less person out of 100—you might react differently.
But these are two ways of saying the exact same thing. The 50 percent (one out of two), which is the relative risk, makes the effect seem much larger than it really is. The one out of 100 (the population being tested), which is the absolute risk, puts the effectiveness of the drug in its proper perspective.
Be especially careful if a claim uses both an absolute and a relative risk. For example, think twice about a claim that a new drug reduced [dreaded condition] by 50 percent, while [product they don’t want you to buy] helps only one out of 100. This is obfuscation at its best.
For more examples of how statistics can mislead, check out this slide presentation by Huff.
In a clever tongue-in-cheek analysis, Sebastian Wernicke uses statistical analysis of TEDTalk presentations to create "the optimum TEDTalk.” It’s a masterpiece!
In general, just be aware of how easy it is to become persuaded (read: misled) by claims that are statistics-based. This 47-second video, for example, persuaded me that the odds of a long life are much better if I cut back on junk food and take up extreme skiing instead. Want to join me?
