What is effect size?

Effect size is how big a difference or relationship actually is: how much the drug lowered blood pressure, how far the new teaching method moved scores. It answers a different question from statistical significance, which only says the difference is probably not luck. A result can be real and tiny at the same time, and effect size is the number that tells you which you have.

Effect size measures how big a finding is. Statistical significance asks whether a difference is likely real rather than chance; effect size asks how much there is, in units you can weigh: minutes of sleep, points on a test, dollars saved. The two come apart often, and sample size is the reason. With enough participants a study can detect almost any real difference, however small: a trial of a million people can find that a sleep app adds forty seconds of sleep and call the result solidly significant. Nothing there is wrong. It is just forty seconds. Researchers therefore report effect size alongside significance, often in standardized form so studies can be compared, with rough labels of small, medium, or large. Context outranks the labels in both directions. A tiny effect can matter enormously at scale: a drug trimming a fraction of a percent off heart attack risk saves thousands of lives when millions take it. A large effect on something trivial matters not at all. The reading habit worth building is one question asked twice: is the difference real, and is it big enough to care about? Significance answers the first. Effect size answers the second, and news stories routinely report only the first.

Effect size is the brightness of a star; significance is whether your telescope can pick it out. Build a big enough telescope and you can detect the faintest speck in the sky. Detection is a fact about your equipment as much as the star, and it says nothing about how bright the star is. That is the trap in "the study found an effect." A study with a million participants is an enormous telescope, and finding a speck through it proves the speck exists, nothing more. Before acting, you want the other number: not "did the telescope see it," but "how bright is it, really." One is significance. The other is effect size.