Anchoring Bias: Why the First Number You Hear Distorts Every Number After It

A recruitment agency sends you a candidate’s salary expectation: £95,000. You hadn’t formed a number yet — you were planning to benchmark the role against market data, consult with HR, and factor in the team’s existing compensation structure. But now £95,000 is in your head. When you finally arrive at your offer, it’s £88,000. You feel you’ve negotiated well — you came in under their ask. What you don’t notice is that without seeing their number first, your independent assessment would have landed around £78,000. The candidate’s opening figure pulled your final number upward by £10,000, and you experienced the distortion as rigorous analysis.

Anchoring is one of the most reliable and practically consequential effects in cognitive science. The first number influences the estimate and narrows the range of later analysis.

The research

Amos Tversky and Daniel Kahneman introduced the anchoring-and-adjustment rule of thumb in their 1974 paper in Science, one of the most cited papers in the history of psychology. In a simple but devastating demonstration, they had participants spin a wheel of fortune that landed on either 10 or 65 — a completely random number — and then asked them to estimate the percentage of African countries in the United Nations. Participants who saw 65 estimated significantly higher than those who saw 10. A random number, transparently irrelevant to the question, shifted responses by an average of over 20 percentage points.

The finding was counterintuitive because participants weren’t confused or inattentive. They knew the wheel was random. They simply couldn’t prevent the number from influencing their estimate. The anchor operated below conscious awareness, shaping the starting point from which all subsequent reasoning proceeded.

Dan Ariely, George Loewenstein, and Drazen Prelec demonstrated the depth of this effect in a 2003 paper in The Quarterly Journal of Economics. They asked MIT students to write down the last two digits of their Social Security number, then bid on consumer products — wine, chocolate, computer accessories. Students with higher Social Security digits bid 60–120% more than those with lower digits. The anchor was transparently arbitrary. The effect was enormous. Ariely and colleagues called it “coherent arbitrariness” — once an initial anchor is set, subsequent valuations maintain internal consistency relative to that anchor, even though the anchor itself was meaningless.

Birte Englich, Thomas Mussweiler, and Fritz Strack tested whether expertise provides protection against anchoring in a 2006 study published in Personality and Social Psychology Bulletin. They asked experienced German judges to determine sentences for criminal cases. Before deliberating, the judges rolled a pair of dice that had been loaded to land on either 1 or 3 (representing months of sentencing). The dice were explicitly presented as random. Judges who rolled higher numbers delivered significantly longer sentences. Professional training, years of experience, and awareness of the experimental manipulation did not eliminate the effect.

The mechanism

Nicholas Epley and Thomas Gilovich investigated why adjustment from anchors is consistently insufficient in a 2006 paper in Psychological Science. Their finding was that adjustment is an effortful, serial process — you move away from the anchor in small steps, testing each step against your knowledge, and you stop as soon as you reach a value that feels plausible. The problem is that “feels plausible” is a low bar. The first plausible value you reach is usually still closer to the anchor than your true estimate would be if you’d started from scratch.

This means the anchor doesn’t just set a starting point. It determines the direction and range of your search. If the anchor is high, you adjust downward and stop at a high-plausible value. If the anchor is low, you adjust upward and stop at a low-plausible value. The final estimate is anchored not because you accepted the initial number but because you searched for your answer in its neighbourhood.

Gregory Northcraft and Margaret Neale demonstrated the practical stakes in a 1987 study published in Organizational Behavior and Human Decision Processes. They took real estate agents — professionals whose livelihood depends on accurate property valuation — to inspect a house. The agents received information packets that were identical except for the listed price, which was varied across conditions. The listed price significantly influenced the agents’ independent valuations. When debriefed, 81% of the agents denied that the listed price had influenced their assessment. The anchoring effect was invisible to the people it most affected.

This invisibility is the core danger. Unlike other biases, where you might at least suspect you’re being influenced, anchoring feels like independent reasoning. You believe you arrived at your number through analysis. The anchor’s contribution to that number is undetectable from the inside.

You cannot un-hear a number, so the best protection is to form your own estimate before someone else’s number enters the room.

The practical implications

Generate your own estimate before exposure to any external reference point. This is the single most effective deanchoring technique. If you’re evaluating a salary, a timeline, a budget, or a price, form your independent estimate — based on your own data, your own benchmarks, your own reasoning — before you see anyone else’s number. Once you’ve seen an anchor, the damage is done. Epley and Gilovich’s research shows that even motivated, warned participants couldn’t fully adjust. Prevention is more effective than correction.

Name the anchor explicitly. When an external number has already been introduced, write it down and label it: “This is the anchor.” The act of labelling doesn’t eliminate the bias, but it creates a cognitive speed bump — a moment of awareness that the number in your head wasn’t generated by your own analysis. This is particularly important in negotiations, where the first offer is explicitly designed to function as an anchor.

In negotiations and group decisions, be aware that the first number spoken wins disproportionate influence. Whoever sets the opening figure — the first salary expectation, the first time estimate, the first budget number — has shaped the range within which all subsequent discussion will occur. If you’re negotiating, consider whether you want to anchor first (to set the range) or anchor independently (to avoid being pulled). Both strategies have their place, but neither works unless you understand the mechanism.

The bigger picture

Anchoring is not a quirk of laboratory experiments. It operates every time a number enters a decision context: the asking price on a house, the “recommended retail price” crossed out above the sale price, the first estimate a contractor gives, the competitor’s valuation, the friend’s salary. Each of these numbers embeds itself in your evaluative framework and warps everything that follows.

The troubling implication is that the quality of a decision can be profoundly shaped by information that has no legitimate bearing on it. A random number, a competitor’s pricing error, an outdated benchmark — these can move your final judgement by 20%, 40%, or more, while you experience the process as careful, independent analysis.

The professionals who make the best judgements under anchoring conditions aren’t those with the most expertise or the strongest analytical skills. Englich’s judges were experts. Northcraft’s agents were professionals. Both were anchored. The people who resist best are those who have learned to distrust their first number — to ask not just “what do I think?” but “where did that thought come from?” The answer, more often than you’d expect, is someone else’s number wearing the disguise of your own reasoning.