Before you make almost any estimate, somebody has usually already handed you a number: a price tag, an opening offer, a figure in the question itself. That number is supposed to be irrelevant. It isn't. Your answer drifts toward it, even when you know it means nothing. Psychologists call this anchoring, and it is one of the best-known results in the study of human judgement.

Below you can test it on yourself. First, how it looks in the lab.

A wheel that tells you nothing

In a classic 1974 paper, Amos Tversky and Daniel Kahneman had people spin a wheel of fortune that was rigged to stop on either 10 or 65. The people were then asked a factual question: what percentage of the countries in the United Nations are African? Those who saw 10 gave typical guesses near 25%. Those who saw 65 guessed near 45%. A spin of a wheel, which obviously carries no information about geography, moved answers by about twenty points.

It is not only wheels. In a later study, people were asked whether Gandhi died before or after age 9, or before or after age 140. Both anchors are absurd, yet guesses for his actual age at death averaged around 50 after the low anchor and 67 after the high one. In another set of experiments, by Dan Ariely, George Loewenstein and Drazen Prelec, 55 business students wrote down the last two digits of their Social Security number, then said whether they would pay that many dollars for items such as a wine, a keyboard or a box of chocolates, and finally stated their top price. Students whose number fell in the top fifth stated prices typically about three times as high as those in the bottom fifth.

Try it: five estimates

You will see five quick questions. Each one starts with a random number and asks whether the true value is more or less than that. Then you slide to your best guess. Don’t look anything up. The anchors are random, high for some questions and low for others, and it changes each time you play.

Five answers from one person prove very little: you may simply have known some of these. The studies above work because they average over many people. But if your high-anchor guesses came out higher than your low-anchor guesses, you just felt the pull yourself.

Why a useless number works

One explanation is anchor and adjust. You start at the number you were given and move toward what feels right, and you stop as soon as the value seems plausible. Psychologists Nicholas Epley and Thomas Gilovich found that adjustments from numbers people generated themselves tend to be insufficient because they stop once a plausible value is reached, unless people are able and willing to search for a more accurate estimate. Start high and you stop at the high end of what seems plausible. Start low and you stop at the low end.

The toy model below shows that idea in its simplest form. The truth is 100. The shaded band is the range of values that would feel plausible to you. Drag the anchor and watch where the walk from the anchor stops. Then narrow the band, as if you knew the subject better.

Toy model: walk from the anchor and stop when it feels plausible
Anchor: 180Truth: 100
Plausible range: 100 ± 40
▼ anchor● where you stop| truth (100)

Notice two things. A high anchor and a low anchor stop at opposite edges of the band, so they pull in opposite directions. And a narrower band, which is what knowing more gives you, shrinks the damage. That matches a pattern in the research: anchoring is weaker when people know a lot about the thing being judged, though it rarely disappears. A second explanation adds that the anchor also makes facts that fit the anchor come to mind more easily, so even a careful search is tilted.

Why you can’t just decide to ignore it

You can’t see anchoring happen from the inside. In a study of house pricing, students and real-estate agents were shown the same house with different listing prices. The agents said they ignored the listing price, but both groups were equally influenced by it. Knowing about the effect isn’t a shield either.

What does help is working on the answer itself. Epley and Gilovich’s result points at the same fix: a more careful search for the real answer shrinks the pull. In practice that means making your own estimate before you see anyone’s number, asking what the opposite anchor would have made you say, and being most suspicious of the first number in any negotiation, including the one you offer yourself.