Anchoring effect

The anchoring effect (also anchoring bias) is the tendency for numerical estimates to be pulled toward an initial reference value the person was exposed to immediately beforehand, even when that reference is uninformative or explicitly random. Classic demonstrations include estimates of the percentage of African countries in the UN shifting with a spun roulette wheel, judgments of property value shifting with arbitrary listing prices, and sentencing recommendations shifting with prosecutor demands. The effect occurs across numerical-judgment domains and is robust to many warnings, payment incentives, and expertise levels.

The effect was introduced by Tversky and Kahneman (1974) in their canonical Science paper on judgment under uncertainty as one of three foundational heuristics-and-biases findings alongside availability and representativeness. The anchoring-and-adjustment mechanism they proposed treats the anchor as a starting point from which the person adjusts insufficiently toward the target value. A subsequent account — the selective accessibility model (Strack & Mussweiler 1997; Mussweiler & Strack 1999) — proposed instead that anchors prime anchor-consistent information from memory, making the comparison process biased rather than the adjustment process. Contemporary reviews (Furnham & Boo 2011; Lieder et al. 2018 on rational-analysis accounts) suggest both mechanisms operate in different paradigms.

Three points are routinely missed in popular treatments. First, the effect is robust in numerical-estimation paradigms but more contested in valuation tasks, where some studies show smaller effects after pre-registration and incentive controls. Second, anchoring is difficult to eliminate even with explicit forewarning, expertise, or financial stakes — making it an unusually stable empirical phenomenon. Third, the underlying mechanism is genuinely contested; popular treatments often present anchoring-and-adjustment as if settled, but the field has moved toward a more nuanced multi-mechanism view.

Anchoring matters because numerical estimates show up everywhere — salary negotiations, real estate prices, jury awards, sentencing decisions, medical risk estimates, charitable giving, performance ratings — and the first number on the table tends to drag every subsequent estimate toward itself. The effect compounds in institutional settings where templates, defaults, and previous decisions become anchors for future ones. The first salary offer in a negotiation, the listing price of a house, the prosecution's sentencing recommendation, the previous year's budget — each of these tends to anchor the eventual outcome more than its objective merits warrant.

Even experts are not immune. Englich, Mussweiler, and Strack (2006) found that experienced judges' sentencing decisions were measurably influenced by transparently irrelevant numerical primes, including a recommendation written by a journalism student or a number generated by rolling dice. Real-estate agents, physicians estimating prognoses, and salary negotiators all show the effect on professionally relevant tasks. The persistence across expertise, motivation, and stakes is what makes anchoring practically consequential.

For individuals, recognizing anchoring's pull is one of the most useful additions to a personal decision toolkit. Practical mitigations exist — generating an independent estimate before exposure to comparable cases, structurally separating the anchor from the decision, deliberately re-anchoring with multiple reference points — but they require setup, not just awareness.

The anchoring effect was identified in Tversky and Kahneman's foundational 1974 Science paper as the third of three core heuristics they documented, alongside representativeness and availability. In their canonical demonstration, participants spun a wheel of fortune that landed on either 10 or 65, then were asked to estimate the percentage of African countries in the United Nations. Those who saw "10" estimated 25% on average; those who saw "65" estimated 45%. The wheel-of-fortune number was transparently irrelevant, yet estimates were pulled toward it.

Subsequent decades of research developed several complementary mechanistic accounts. The classic anchor-and-adjust account proposed that people start at the anchor and adjust toward what they believe is correct, but stop adjusting before they should — producing estimates pulled toward the anchor. Mussweiler's selective accessibility theory provides a different mechanism: encountering an anchor activates anchor-consistent information in memory, which then disproportionately informs the subsequent judgment. The two accounts apply to different situations; both have empirical support.

Modern research distinguishes carefully between experimenter-provided anchors (where someone presents a number and asks a related question) and self-generated anchors (where the person produces a number from prior knowledge and adjusts from it). The mechanisms appear to be partly distinct, and recent replication work has emphasized that moderators of one type may not apply to the other (Röseler & colleagues, 2024).

Distinguishing the two main types matters because mitigation strategies differ.

• Experimenter-provided (incidental) anchoring. The anchor is provided by an external source, often appearing irrelevant to the judgment at hand. The wheel-of-fortune in Tversky and Kahneman's experiment is the classic case. The list price of a house or the opening offer in a negotiation are real-world examples. The mechanism is largely automatic and absorbed without conscious deliberation. Mitigation requires structural separation: generating an independent estimate before exposure to the anchor, or considering multiple reference points to dilute the pull of any one.
• Self-generated anchoring. The person produces an initial estimate from prior knowledge and then adjusts toward a more considered answer. The classic example is being asked the boiling point of water on Mount Everest: most people start from 100°C (the sea-level boiling point they know) and adjust downward, but typically not by enough. Mitigation requires more aggressive adjustment than feels comfortable, and ideally a deliberate question about the magnitude and direction of adjustment.

Anchoring effects have been replicated across knowledge questions, probability estimates, price estimates, sentencing decisions, and judgments about one's own behavior (Englich et al., 2006; Mussweiler, multiple). The effect is generally stable across cultures, age groups, and education levels, though effect sizes vary substantially with task structure. A 2024 replication of Epley and Gilovich's seminal moderator studies confirmed the basic anchoring effect but found that proposed moderators — need for cognition, cognitive load, forewarning — did not reliably reduce it (Röseler et al., 2024).

What is well-established. Classic anchoring effects in numerical-estimation tasks are robust and well-replicated. The Many Labs replication projects and subsequent meta-analyses have confirmed substantial effect sizes across knowledge questions, sentencing, and estimation domains. Anchoring resists most simple debiasing interventions — explicit warnings, financial incentives, and instructions to consider opposite values typically do not eliminate the effect, though structural interventions can attenuate it.

Where the evidence is more contested. Recent well-powered replication work has found that some anchoring effects on willingness-to-pay and product valuations are smaller than originally reported. Li and colleagues (2025), in Economic Inquiry, replicated a willingness-to-pay anchoring study with much higher statistical power and estimated the effect at 3.4% rather than the 31% originally reported. The careful conclusion is that classic numerical anchoring is robust, but some specific applications — particularly in valuation tasks — may have smaller effect sizes than the early literature suggested. Scientific consensus is that anchoring is real and substantial in many contexts, with effect magnitudes that depend meaningfully on task structure.

"Anchoring is just being persuaded by salesmanship." No. The effect operates even when the anchor is transparently random — a number from a wheel of fortune, a date, a digit drawn from a hat. It also operates when participants are warned about it, when they are paid for accuracy, and when they are explicitly told the anchor is irrelevant. The bias is automatic and largely outside deliberate control.

"Smart people are immune to anchoring." They are not. Experts on professionally relevant tasks — judges sentencing, real-estate agents pricing, physicians estimating prognosis — show the effect on tasks they know well. Higher cognitive ability does not reliably reduce susceptibility.

"Just being aware of anchoring is enough to avoid it." Largely false. Awareness reduces but does not eliminate the effect. The reliable mitigations are structural: generating independent estimates before exposure to comparable cases, considering multiple reference points, deliberately considering both higher and lower anchors. Personal vigilance alone is insufficient.

"Anchoring effects are huge in every context." The effect is robust in classic numerical-estimation tasks but more variable in some applied domains. Recent well-powered replication work has found smaller effects in willingness-to-pay tasks than the early literature suggested. The careful conclusion is that anchoring is real and substantial in many contexts, but specific magnitudes depend on task structure.

Consider a salary negotiation. The candidate has researched comparable roles and believes the right range is $110,000–$130,000. The hiring manager opens with $95,000 — anchored, perhaps unconsciously, on the candidate's previous salary or a budget constraint. Without active resistance, the anchoring effect will pull the candidate's counter-offer toward $95,000. Even if the candidate argues for $115,000, the eventual settlement is likely to land closer to the manager's anchor than to the candidate's research-based range.

The structural protection is to commit to the research-based range before hearing the manager's number. The candidate can also re-anchor explicitly: rather than counter-offering against the manager's $95,000, propose a range starting at $130,000 (the top of the research range) and let the negotiation re-center higher. Negotiation researchers consistently recommend making the first offer when possible, precisely because of anchoring — the first number sets the gravitational center of the subsequent discussion. Awareness alone would not protect the candidate; the structural decision (commit to the range, anchor first) does.

• Cognitive bias — the broader category of systematic deviations from rational judgment. Anchoring is one of the three foundational biases.
• Heuristic — the mental shortcuts from which biases including anchoring emerge. Anchor-and-adjust is one of the named heuristics; selective accessibility is the cognitive mechanism.
• Confirmation bias — a related systematic error in which encountered information disproportionately shapes subsequent reasoning, also through selective-accessibility-like mechanisms.
• Decision hygiene — the structural practices that mitigate anchoring and other systematic errors before they occur.
• Priming — the broader phenomenon in which exposure to a stimulus influences subsequent processing of related information. Anchoring is a numerical priming effect.
• Selective accessibility — Mussweiler's mechanistic theory of anchoring, in which encountering an anchor activates anchor-consistent information in memory.
• Negotiation theory — the broader field studying strategic interaction over outcomes. The first-offer advantage is a direct application of anchoring research.
• Negativity bias — anchoring effects can interact with negativity bias in social and self-evaluation contexts — negative anchors produce stronger adjustments than positive ones in some paradigms.
• Survivorship bias — another well-documented bias, though structurally distinct: anchoring is a cognitive bias while survivorship is a statistical bias in the data.

The LifeByLogic Cognitive Bias Susceptibility tool measures anchoring using a validated task structure from the Heuristics-and-Biases Inventory (Berthet, 2023): the user is exposed to a deliberately irrelevant number before estimating an unrelated quantity. The bias score reflects how much the user's estimate was pulled toward the anchor, calibrated against published population shifts. The full methodology is documented on the tool methodology page.

Take the Cognitive Bias Susceptibility assessment →

This entry is intended as a citable scholarly reference. Choose the format that matches your context. The retrieval date should reflect when you accessed the page, which may differ from the entry's last-reviewed date shown above.

LifeByLogic. (2026). Anchoring Effect: Numerical Estimates and Bias. https://lifebylogic.com/glossary/anchoring-effect/

LifeByLogic. "Anchoring Effect: Numerical Estimates and Bias." LifeByLogic, 2 May 2026, https://lifebylogic.com/glossary/anchoring-effect/.

LifeByLogic. 2026. "Anchoring Effect: Numerical Estimates and Bias." May 2. https://lifebylogic.com/glossary/anchoring-effect/.

@misc{lblanchoringeffect2026,
  author = {{LifeByLogic}},
  title = {Anchoring Effect: Numerical Estimates and Bias},
  year = {2026},
  month = {may},
  publisher = {LifeByLogic},
  url = {https://lifebylogic.com/glossary/anchoring-effect/},
  note = {Accessed: 2026-05-15}
}