In the 1954 classic text, How to Lie With Statistics, author Darrell Huff asserts that there are hundreds of ways that numbers can lie to you. Statistical interpretation is a subjective business - trends and correlation can be employed by the statistician to support pretty much any hypothesis. Graphs and charts can distort reality. Ratios and proportions can be meaningless comparisons. What we might think as “fact” can actually be a wildly incorrect (or intentionally deceptive) conclusion. Don’t be fooled by the claims of diets, products, or even automotive vendors: here are a few ways statistics can lie to you.

Correlation doesn’t equal causation.

If cereal commercials have taught us anything, it’s that kids who start the day off with breakfast do better in school. Must be something in Fruit Loops that boosts our brainpower, right? Not really. This study may be correct in stating that kids who eat breakfast are more likely to do well in school, but it makes a common statistical error in concluding that breakfast causes better grades. When the phenomenon was actually investigated, researchers discovered that food early in the day had no effect on the intelligence of children. Instead, the new research concluded that kids who had circumstances keeping them from eating breakfast are also the type to have trouble in school.

“Average” isn’t average.

If you’re like me, you’ll be a little discouraged when you find that the average American household income is $70,000. And I thought I was doing pretty well for myself. But this figure doesn’t really tell us what the average American is making. The problem with this number is that America’s most wealthy individuals are skewing this number heavily. In reality, almost 70% of Americans earn less than the average income - so don’t be too hard on yourself.

Are samples representative?

Suppose you flip a coin 5 times, and “heads” comes up four times. Would it be accurate to say that the coin lands on heads 80% of the time? It did for your sample, but you really haven’t flipped the coin enough times to determine actual probability. Statisticians can (intentionally or not) use this sampling trick to draw inaccurate conclusions. This very thing happened in the 1936 election, when pollsters predicted the certain defeat of FDR against Alfred London. But the surveys only reached those people with telephones, at the time a luxury afforded only to a portion of the population. As we now know, this statistical “certainty” ended up being completely meaningless, because researchers did not collect an appropriate sample size.

So what does this all mean for you?

It’s healthy to remain skeptical even of claims that sound like they’re based in data and statistics. It’s easy for someone to abuse figures and cloud the truth about any issue - whether it’s for the sake of selling a product, making a  public service announcement, or attempting to identify the root of a problem. Now you know how to spot a few shady statistical practices - so don’t let numbers sucker you into making decisions you otherwise wouldn’t.