Anchoring
Also known as: Anchoring-and-adjustment, Focalism
The first number you see drags every estimate that follows toward it.
What it means
Anchoring is the tendency, when judging an unknown quantity, to start from an initial value — the anchor — and then adjust outward toward a final answer, but to adjust too little, so the estimate ends up biased toward the starting point. The proposed mechanism is twofold: deliberate but insufficient adjustment away from a self-generated anchor, and a more automatic priming-like process in which the anchor makes anchor-consistent information more accessible. What makes anchoring so striking is that it works even when the anchor is obviously arbitrary, irrelevant, or randomly generated, and even when people are warned about it or paid for accuracy. Its robustness is a boundary condition worth noting: experts are not immune, and forewarning offers little protection, though considering reasons the anchor might be wrong can blunt the effect. In practice, anchoring shapes pricing, salary negotiations, real-estate listings, legal damage awards, and forecasting, which is why setting the first number is often a strategic act rather than a neutral one.
Why it happens
The two mechanisms are less rivals than a division of labour, and which runs depends on where the anchor came from. When you generate it yourself — estimating the boiling point of water on Everest by starting from 100C at sea level and working down — you know the starting value is wrong and deliberately move away from it. Adjustment is effortful, so you stop at the near edge of the plausible range rather than its centre, and the estimate lands short. When someone hands you an anchor, something quieter happens. Testing whether the target is above or below that number sends you searching memory for evidence that it is roughly that large, and the anchor-consistent facts you surface then feed the number you give. The first process is lazy adjustment; the second is biased recall.
What the evidence shows
Anchoring has come through the replication crisis better than almost anything else of its vintage. In the first Many Labs project, which re-ran a set of classic effects across thirty-six samples, the four anchoring items all replicated and ranked among the largest effects in the project. They are four items rather than four studies: four scenarios — the San Francisco to New York distance, the population of Chicago, the height of Everest, babies born per day — drawn from a single 1995 experiment and scored separately. All four came out above the original estimate, though the authors note the comparison is not like-for-like: the original figure pools fifteen questions in a test-retest design, while the replication used random assignment between subjects. Even discounted for that, it is an unusual result for a 1970s social-cognition finding. But the win has a specific shape: the items are factual estimates of unknown quantities, such as the height of Everest or the population of Chicago, where the person holds no firm prior and the anchor sits inside the question itself. Within that design, the effect is about as dependable as psychology gets.
Where it breaks down
The folklore version — any number in the vicinity contaminates any judgment — is wider than the evidence supports, and two limits do most of the cutting. First, the anchor has to be engaged with explicitly. Incidental environmental anchors, where a number is merely present in the setting rather than asked about, produce no reliable effect: across three studies of consumer price estimates, incidental anchors did nothing while conventional anchors in the same experiments worked robustly. Second, the bias thins out where money is at stake. A re-examination of the celebrated social-security-number result found only very weak effects on valuations of ordinary goods and none at all on binary lotteries. Notice the shape of that: the arbitrary-anchor demonstrations that made the effect famous — a digit string with no bearing on anything moving what you would pay — are the ones that have held up worst. The standard summary, that anchoring works even when the anchor is plainly random, is carrying more weight than this literature can bear. Anchoring is powerful for factual estimates you are asked to reason about, and shakier for what you will actually pay.
Using it in practice
Two asymmetries follow. If you are setting the number, move first and be specific: an anchor only bites if the other side engages with it, which argues for a figure precise enough to look reasoned rather than plucked. If you are receiving one, the mechanism points to the antidote. Because the anchor works by making anchor-consistent evidence easy to reach, deliberately generating reasons the number is wrong restores some of what it crowded out — which is why considering the opposite blunts the effect where a bare warning does not. The stronger discipline is procedural: commit to your own estimate in writing before you see anyone else's. An anchor has far less to grip when the estimate already exists.
Examples
Shoppers told a tin of soup is '12 per customer' buy far more than those shown 'no limit' — the number 12 anchors how many feels normal.
Card machines suggesting fifteen, twenty and twenty-five percent collect bigger tips than the same machines offering ten, fifteen and twenty. The printed options quietly define what a normal tip looks like.
Mock jurors asked whether damages should top a low figure award far less than those asked about a high one, even when told the figure was picked arbitrarily.
Estate agents shown a higher listing price valued the same house higher than those shown a lower one, even after touring it in person — and denied the listing had swayed them. This is the favourable case for the effect rather than the shaky one: an appraisal is an estimate of an unknown quantity rather than a price the agent will pay, and the listing is an anchor they are explicitly asked to reason about.
A physician handed a colleague's provisional diagnosis before examining the patient tends to land near it — the same selective-accessibility process running in a clinic, where testing the suggestion is what makes the evidence for it easiest to bring to mind.
First described in Tversky & Kahneman (1974).
Key references
- Shanks, D. R., Barbieri-Hermitte, P., & Vadillo, M. A. (2020). Do incidental environmental anchors bias consumers' price estimations? Collabra: Psychology, 6(1), 19. doi.org/10.1525/collabra.310
- Klein, R. A., et al. (2014). Investigating variation in replicability: A "many labs" replication project. Social Psychology, 45(3), 142-152. doi.org/10.1027/1864-9335/a000178
- Fudenberg, D., Levine, D. K., & Maniadis, Z. (2012). On the robustness of anchoring effects in WTP and WTA experiments. American Economic Journal: Microeconomics, 4(2), 131-145. doi.org/10.1257/mic.4.2.131
- Epley, N., & Gilovich, T. (2006). The anchoring-and-adjustment heuristic: Why the adjustments are insufficient. Psychological Science, 17(4), 311-318. doi.org/10.1111/j.1467-9280.2006.01704.x
- Strack, F., & Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility. Journal of Personality and Social Psychology, 73(3), 437-446. doi.org/10.1037/0022-3514.73.3.437
- Mussweiler, T., Strack, F., & Pfeiffer, T. (2000). Overcoming the inevitable anchoring effect: Considering the opposite compensates for selective accessibility. Personality and Social Psychology Bulletin, 26(9), 1142-1150. doi.org/10.1177/01461672002611010
- Northcraft, G. B., & Neale, M. A. (1987). Experts, amateurs, and real estate: An anchoring-and-adjustment perspective on property pricing decisions. Organizational Behavior and Human Decision Processes, 39(1), 84-97. doi.org/10.1016/0749-5978(87)90046-X
- Jacowitz, K. E., & Kahneman, D. (1995). Measures of anchoring in estimation tasks. Personality and Social Psychology Bulletin, 21(11), 1161-1166. doi.org/10.1177/01461672952111004