Anchoring
Claim assessed: A number seen before an estimate pulls the estimate towards it, even if it is random.
Published , updated · how we work
In short
Anchoring is the tendency to start an estimate from a number you saw beforehand and not to move far enough away from it. Tversky and Kahneman (1974) showed that the effect appears even when the number is clearly random. It is one of the classic effects that came out stronger, not weaker, in a large replication. The more fragile part is another one: how much arbitrary anchors can change the prices people are willing to pay.
What it says
The wheel of fortune. In the experiment described by Tversky and Kahneman (1974), participants had to estimate various percentages, for example the percentage of African countries in the United Nations. Before that, a wheel was spun in front of them, giving a number between 0 and 100. Participants first said whether the real percentage was higher or lower than the number on the wheel, then gave their estimate.
The group that saw 10 on the wheel gave a median estimate of 25%, and the group that saw 65 estimated 45%. The number had nothing to do with the question, and participants had seen it generated at random. Payoffs for accuracy did not reduce the effect.
Tversky and Kahneman also describe a version with no number given. Two groups of high school students had 5 seconds to estimate a product written on the blackboard. Some saw 8×7×6×5×4×3×2×1, others 1×2×3×4×5×6×7×8. The median estimate was 2,250 for the descending sequence and 512 for the ascending one. The correct answer is 40,320. Their explanation: people do the first few multiplications, which give a larger number in the first case, and then adjust too little.
Why it happens. Tversky and Kahneman propose insufficient adjustment: you start from the anchor and move away from it, but stop too early. Strack and Mussweiler (1997) propose a complementary mechanism, selective accessibility: when you compare the anchor with the real value, what comes to mind is mostly information consistent with the anchor, and that information then shapes your estimate.
In experts. Englich, Mussweiler and Strack (2006) asked experienced legal professionals to set a sentence in a case presented in the study, after seeing a sentencing demand. The sentences moved towards that demand even when the professionals knew it had been set at random, or when they had thrown the dice that set it themselves. Professional experience did not reduce the effect.
The replication. The Many Labs project (Klein et al., 2014) repeated 13 classic effects in 36 samples from several countries, with 6,344 participants in total. Anchoring was tested with four questions from the study by Jacowitz and Kahneman (1995): participants received a number that was clearly too high or too low, then estimated. The effect appeared for every question, and the authors note that the original study had underestimated it.
In prices. A widely cited 2003 study reported that arbitrary anchors, for example the last two digits of a US social security number, changed a lot how much participants asked or offered for products and experiences: by roughly 50% to 200%, as summarised by Maniadis, Tufano and List (2014). We describe it only through that summary, because one of its authors is Dan Ariely (see below). Maniadis, Tufano and List repeated one of those experiments, with an annoying sound that participants asked to be paid for listening to. They found an effect about half the size of the original, and close to zero before excluding outliers as the original study did. Reviewing other replications too, they conclude that the 2003 study greatly overestimated the effect: anchoring in economic valuations is real but economically small.
Example
You want to buy a second-hand car. The seller opens with a high price. Even if you negotiate, the conversation starts from their number: a price below the asking price already feels like a win, even if it is still above what the car is worth. The example illustrates the mechanism; it is not a study cited here.
How to apply it
The steps below are a practical way to apply the idea, proposed by us based on the studies cited.
- Estimate before you see other people’s numbers. What something is worth, how long it will take, what it should cost: write down your answer before the first number is on the table.
- Don’t rely on knowing about the effect. In 1974, payoffs for accuracy didn’t reduce it, and for legal professionals experience didn’t either.
- Look for reasons the anchor is wrong. If the effect comes from anchor-consistent information coming to mind, as Strack and Mussweiler (1997) propose, it is worth actively looking for arguments in the other direction.
- Start from a relevant reference. If you are going to start from a number anyway, pick it yourself: the price of similar cars, the duration of similar projects.
Limits and nuances
- The effect depends on what you know. Klein et al. (2014) note that the differences between samples may come from how much each group knew about the height of Everest or the distance to New York. Studies often use quantities people don’t know well, where the anchor has a lot of room to act.
- The replication used explicitly given anchors. In Many Labs, the number was presented as information before the question. That is not the same as a wheel of fortune or the digits of a personal number, which are visibly unrelated.
- We excluded Dan Ariely’s work. He co-authored a 2012 study retracted in 2021, after the data of a field experiment he supervised were shown to be fabricated; who fabricated them can’t be told from the data. His 2003 study on prices is not among the sources; we describe it only through the replication by Maniadis et al. (2014), which shows much smaller effects.
- What we read. We read Tversky and Kahneman (1974) and Klein et al. (2014) in full. Of Maniadis et al. (2014) we read the October 2013 working-paper version, from the University of Nottingham: the introduction and the empirical part, with the tables. For Strack and Mussweiler (1997) and Englich et al. (2006) we read the abstracts.
- Revised on 7 October 2026. We removed the study by Ariely et al. (2003) from the sources, under our exclusion rule, and rewrote the paragraph on prices based on the replication by Maniadis et al. (2014), re-read for this revision.
Sources
- Amos Tversky, Daniel Kahneman (1974). Judgment under uncertainty: Heuristics and biases
- Fritz Strack, Thomas Mussweiler (1997). Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility
- Birte Englich, Thomas Mussweiler, Fritz Strack (2006). Playing dice with criminal sentences: The influence of irrelevant anchors on experts' judicial decision making
- Richard A. Klein, Kate A. Ratliff, Michelangelo Vianello et al. (2014). Investigating variation in replicability: A "Many Labs" replication project
- Zacharias Maniadis, Fabio Tufano, John A. List (2014). One swallow doesn't make a summer: New evidence on anchoring effects
See also: Availability heuristic, Confirmation bias, Consider the opposite, Decoy effect, Planning fallacy, Priming