Anchoring

What it is

The founding experiment is deliberately absurd. Tversky and Kahneman spun a wheel of fortune in front of students at the University of Oregon, rigged to stop at ten or at sixty-five, then asked what percentage of the countries in the United Nations were African. Those who saw ten answered twenty-five per cent on average; those who saw sixty-five answered forty-five. Everyone had watched the wheel produce the number.

The effect appears wherever somebody has to produce a figure they cannot look up. Birte Englich, Thomas Mussweiler and Fritz Strack took the design into a courtroom, where experienced legal professionals read a case file and, in one study, threw dice themselves to generate a prosecution demand before passing sentence. Higher rolls produced longer sentences, and the paper reports that expertise and experience did not reduce the effect. Gregory Northcraft and Margaret Neale ran the property version, giving estate agents a house to value alongside a listing price, and the valuations followed the listing.

Two accounts of the mechanism have competed since the beginning. Tversky's was adjustment: you start at the anchor, move away from it, and stop too early. Kahneman's was selective accessibility, a priming account in which the anchor makes anchor-consistent knowledge easier to bring to mind, so the judgment gets assembled from a skewed sample of what a person already knows. Neither has won. Adam Harris, Fi Blower and Sophie Rodgers ran four experiments on a test the literature treats as a signature of selective accessibility and did not obtain the evidence, then ran a fifth by a different method and again found none.

In effect

Commercially, anchoring is the most used idea in pricing. The bottle at the top of a wine list, the recommended retail price struck through on a tag, the enterprise tier almost nobody buys, the expensive handset displayed beside the mid-range one: each presents a number that could plausibly be a price, at the moment the decision is being made.

The condition matters more than the tactic. Robert Sugden, Jiwei Zheng and Daniel Zizzo tested real buying and selling decisions on familiar products while varying the kind of number people met first, and found anchoring only where the anchor was framed as a plausible price for the good, with stronger effects among sellers than buyers. Their paper is titled Not all anchors are created equal. Ben Newell and David Shanks reached the same boundary from the theory side, reviewing the literature to test whether stray numbers act on behaviour broadly and automatically and concluding that anchoring supports neither claim.

For a buyer, the practical reading is that the first number is a claim rather than a fact. A struck-through price asserts that an item was once offered at that figure and did not sell. It carries no information about what the thing is worth, and it is admissible enough to move you anyway.

What it does not say

It does not say that any number will do. The most famous demonstration outside psychology, in which people bid more for wine and chocolates after writing down the last two digits of their social security number, has not held up: Drew Fudenberg, David Levine and Zacharias Maniadis found very weak effects when they re-ran that manipulation on market goods and none at all on lotteries, and a later paper in the American Economic Review reported the same fragility.

It does not say that knowing about the bias protects you. Knowing about the subject does. Andrew Smith, Paul Windschitl and Kathryn Bruchmann found across four studies that people with more knowledge of the thing being judged anchored less. The only tested defence that works in the moment is deliberately generating reasons why the anchor is wrong, which reduced anchoring in both studies by Mussweiler, Strack and Pfeiffer.

It does not say you can find out whether you are personally resistant. Lukas Röseler and colleagues assembled the largest open collection of anchoring data available and reported that the measures used to score individual susceptibility are unreliable, which is why a decade of work relating anchoring to personality has produced contradictory results.

The individual anchoring indices that circulate, including the forty-one per cent for estate agents, come from single studies and are mostly reported without attribution in the popular literature. Treat them as illustrations of scale rather than as measurements.

---