Anchoring Bias

Understanding Anchoring Bias

A Precise Definition

Anchoring bias is the phenomenon by which exposure to an initial value — the anchor — pulls subsequent numerical or evaluative judgments toward that value. The anchor can be relevant (a comparable house's price) or irrelevant (the last two digits of one's phone number). It can be supplied by the situation, by another person, or by oneself. Whatever its source, once it is in mind, it exerts gravitational pull on whatever judgment comes next.

The bias is not about being persuaded by the anchor. People often know the anchor is suspect and try to correct for it. The problem is that the correction is typically insufficient — they move away from the anchor, but not far enough to reach the answer they would have given without it. The final judgment lies somewhere between the anchor and the unbiased estimate, with the anchor doing more of the work than the judge appreciates.

Two Big Findings

Two findings from the research literature are particularly worth holding onto. First, the anchoring effect is large. In the original studies and many replications, judgments under different anchors differ by a substantial fraction of the gap between the anchors — often roughly half or more. Second, the effect persists when participants are told that the anchor was randomly generated and is unrelated to the question being judged. Knowledge of the anchor's irrelevance reduces but does not eliminate its influence.

Different Kinds of Anchors

Researchers distinguish several types. Externally provided anchors are numbers given by another person or by the situation — an opening offer, a quoted price, a published statistic. Self-generated anchors arise when a judge thinks of an initial estimate themselves — for example, starting from a known reference point and adjusting. Implicit anchors are numbers that appear in the background, perhaps as part of an unrelated question, and still affect a subsequent judgment. Different mechanisms may dominate in different cases, but the directional effect — pull toward the anchor — is broadly shared.

Not the Same as Suggestion

It is tempting to read anchoring as a form of social suggestion or persuasion, but the experimental evidence shows it is something deeper. People anchor on numbers they generated themselves, on numbers they were told are irrelevant, and on numbers that come from a transparently random source like a spinning wheel. Whatever the bias is, it is not simply a willingness to be swayed by the speaker.

The Research Foundation

The Wheel of Fortune Study

The most famous demonstration of anchoring comes from Tversky and Kahneman's 1974 paper. They asked participants to estimate the percentage of African nations in the United Nations. Before answering, each participant watched a wheel of fortune spin and stop on a number — but the wheel was secretly rigged to stop on either 10 or 65. Participants first wrote down whether they thought the true percentage was above or below the wheel's number, then estimated the percentage itself. Those who saw a 10 estimated, on average, around 25%. Those who saw a 65 estimated, on average, around 45%. The wheel's number — explicitly random, transparently irrelevant — had moved estimates by roughly 20 percentage points.

Multiplication Problems

In another study from the same line of research, participants were given five seconds to estimate a multiplication expression: 1 × 2 × 3 × 4 × 5 × 6 × 7 × 8 or 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1. Although the products are identical, the ascending sequence produced lower estimates (median around 512) and the descending sequence produced higher ones (median around 2,250). The actual answer is 40,320. Two key features of anchoring appear here: an initial computation acts as an anchor, and the final judgment falls far short of the truth because adjustment from the anchor is insufficient.

Insufficient Adjustment

Tversky and Kahneman's original interpretation was that judgment under uncertainty often proceeds by anchoring on an initial value and then adjusting. The adjustment, however, terminates prematurely — typically as soon as a "plausible" range is reached — so the final judgment remains too close to the anchor. This account, sometimes called the anchor-and-adjust heuristic, applies particularly well to cases where the judge consciously starts from a known reference and adjusts.

The Selective Accessibility Model

Later work, particularly by Strack and Mussweiler in the 1990s, proposed a different mechanism for many anchoring effects. On their account, evaluating the anchor — asking "Is this quantity higher or lower than the anchor?" — activates information consistent with the anchor being correct. That information remains active in memory and is then used when generating the final estimate. The bias arises from selective retrieval rather than from explicit anchoring-and-adjustment. Both mechanisms likely operate in different situations.

Replications and Boundary Conditions

Anchoring effects have been replicated in dozens of countries, in many languages, and across decades. Replications have shown the effect with anchors that are irrelevant social-security numbers, room temperature numbers in unrelated questionnaires, and even the year a participant was born when used in front of a numerical estimate task. The effect tends to be smaller — but not absent — when the anchor is implausibly extreme, when judges have strong domain knowledge, and when they are given explicit incentives and time to deliberate.

Anchoring in Experts

One of the more striking findings is that experienced professionals are not immune. Studies of real-estate agents have shown that listing-price anchors significantly affect their appraisals of properties, even though the agents deny being influenced by the listing. Similar findings have been reported for judges receiving sentencing recommendations and for physicians considering diagnostic possibilities. Expertise reduces but does not eliminate anchoring.

How It Works

Anchoring as a Shortcut

From a cognitive-economy perspective, anchoring is not a glitch but a reasonable strategy. When asked to estimate an unfamiliar quantity, starting from any available reference point and adjusting from there saves the mental work of building an estimate from scratch. The trouble is that the strategy systematically over-weights the starting point, especially when the starting point is salient and the question is hard.

Two Mechanisms, Often Both Active

The two main mechanistic accounts — insufficient adjustment and selective accessibility — are not mutually exclusive. For self-generated anchors, where the judge knows they are starting from a reference and explicitly trying to adjust, the adjustment account fits well. For experimenter-provided or environmentally supplied anchors, where the judge merely had to consider the anchor first, the selective accessibility account often fits better. In real-world cases like negotiations, both mechanisms can be at work in the same exchange.

Why Adjustment Stops Too Early

Adjustment from an anchor proceeds until the judge feels they have reached a plausible range. But "plausible" is a wide zone, and the brain stops at the near edge of it rather than searching for the center. The result is a final estimate that is closer to the anchor than to the truth. Faster, less deliberate adjustment produces larger anchoring effects; slower, more incentivized adjustment reduces them somewhat.

The Role of Memory and Priming

Considering whether a quantity is above or below an anchor — the question typically used in anchoring studies — activates examples and associations consistent with that anchor. If asked whether Gandhi was older or younger than 144 when he died, you are likely to recall facts that suggest he lived a long life. When you then estimate his actual age at death, those facts disproportionately shape the answer. The anchor reshapes the contents of memory available for the next judgment.

Numerical and Magnitude Effects

Anchoring is partly a numerical phenomenon — the brain treats numbers on a mental number line, and exposure to a large number activates the upper region of that line. This explains why even seemingly nonsensical anchors (like the last digits of a phone number) can have small but measurable effects on subsequent numerical estimates. The general principle is that whatever is in the cognitive workspace when a judgment is made shapes that judgment.

Everyday Examples

Sticker Prices and Discounts

A jacket marked "Was $300, now $120" is rarely a worse deal than the same jacket listed plainly at $120, but it usually feels like one. The $300 is doing anchoring work: the value of the jacket is judged relative to the higher number, making the $120 feel like a bargain. Retail pricing has been studied extensively in this context, and the anchoring effect on willingness to pay is consistent and substantial.

Restaurant Menus

An expensive item near the top of a menu — sometimes called a decoy entree — shifts the perceived value of the dishes that follow. Diners do not necessarily order the expensive item; they order the next-most-expensive one, which now looks reasonable. Menu engineering as a discipline relies heavily on anchoring of this kind.

Salary Negotiations

When asked early in a recruitment process, "What are your salary expectations?", a job candidate who quotes a number sets an anchor for the entire subsequent negotiation. Studies of negotiations consistently show that the first offer — even an aggressive one — strongly predicts the final settlement. Candidates who refuse to be first, or who anchor with a careful but ambitious figure, tend to fare better than those who let the employer set the anchor.

Real Estate

A house's listing price functions as an anchor for the offers it receives, even from buyers who have done their own comparable analyses. Sellers who price aggressively often receive higher offers in absolute terms, although they also risk longer time on market and may eventually accept less if the listing seems stuck. The anchoring effect helps explain why a listing's first asking price is so consequential.

Estimating Time and Money

Project managers asked how long a task will take often start from a previous similar task — an anchor — and adjust for differences. The result is the well-known phenomenon of underestimating project timelines, since the original anchor rarely accounted for the complications that the new project will encounter. The anchor-and-adjust pattern is part of the explanation for the planning fallacy.

Tip Suggestions

Many payment screens now offer suggested tip percentages — 18%, 20%, 25%. Studies have shown that the displayed defaults change the average tip, with higher suggestions producing higher tips. The customer is free to enter any amount, but the suggested numbers anchor the decision.

Charity Asks

Fundraising appeals that present donation options — "$25, $50, $100, $250" — anchor donors' choices in ways that affect the median donation. Including a higher option pulls average donations up, even among donors who do not choose the high option, because the high option recalibrates the sense of what a "normal" gift looks like.

Where It Shows Up

Negotiation

The first offer effect — that the party who makes the first credible offer in a negotiation tends to end up with a better outcome — is one of the most reliable findings in negotiation research. Anchoring is the central mechanism. The first offer defines a numerical landscape against which all subsequent offers are evaluated. Skilled negotiators take this seriously, opening with carefully calibrated numbers and, when on the receiving end, working actively to neutralize the anchor before responding.

Pricing and Marketing

The anchoring effect underlies many standard pricing tactics: list-price-then-discount, decoy products that make the target option look like a deal, and tiered pricing structures where a high-end option makes the middle option attractive. These tactics are robust enough to have become foundational in retail and software pricing.

Judicial Sentencing

Studies of judges and legal decision-makers have shown that proposed sentencing recommendations — from prosecutors, probation reports, or even seemingly irrelevant numerical primes — influence final sentences. Experiments with experienced judges presented with cases involving randomly generated sentencing demands have shown measurable anchoring effects on the sentences they then proposed.

Medical Diagnosis and Treatment

A patient's chart that already contains a tentative diagnosis biases subsequent clinicians toward that diagnosis. A reference range on a lab result anchors interpretation of the result. The first treatment proposed by a colleague anchors what alternatives are considered. Medical training increasingly emphasizes structured approaches that resist these effects.

Real Estate Appraisal

As noted earlier, listing prices serve as anchors even for trained appraisers. The implication is that the most experienced eyes in the market are not exempt from the bias, although their estimates may anchor less than non-experts'. Blind appraisal protocols — where the appraiser does not see the listing or seller's price — have been proposed to address this.

Finance and Forecasting

Analysts' earnings estimates tend to cluster around previous estimates and around what others are publishing — a pattern partly explained by anchoring. The same is true for many macroeconomic forecasts. Forecasts that "anchor" too heavily on the recent past struggle to predict regime changes.

Public Policy

When a budget conversation starts from "last year's number," the anchor is the previous budget. Proposals are evaluated as deviations from it. This makes large reallocations difficult even when underlying conditions have changed and a different starting point would be more appropriate. Zero-based budgeting was developed in part as a structural counter to the anchoring effect in this context.

Related and Often-Confused Biases

Framing Effects

Framing effects — judgments shifting depending on whether outcomes are described as gains or losses — share some psychological territory with anchoring. Both involve sensitivity to how information is presented. They are distinct, however: anchoring concerns the influence of a specific value, while framing concerns the structure of the description.

Status Quo Bias

The status quo bias — a preference for things to stay as they are — interacts with anchoring when the status quo also serves as the anchor. A budget that has been a certain size for years is both familiar (status quo) and a numerical anchor; both effects push against change.

Confirmation Bias

Confirmation bias can deepen an anchor's effect: once you commit to an initial estimate, you look for evidence that supports it. The two biases reinforce one another, with anchoring providing the initial commitment that confirmation bias then defends.

Availability Heuristic

The availability heuristic — judging frequency or probability by the ease of bringing examples to mind — overlaps with the selective accessibility account of anchoring. Both involve memory shaping judgment, but they are activated by different cues and applied to different kinds of questions.

Halo Effect

The halo effect is a different kind of anchor: an overall impression of a person or thing pulls subsequent judgments about specific attributes. A candidate's first impression anchors how their later answers are read, much as a number anchors a later estimate.

Dunning-Kruger Effect

The Dunning-Kruger effect interacts with anchoring in self-assessment. An early estimate of one's competence can anchor later self-judgments, and the miscalibration the effect describes can be reinforced by anchoring on past, possibly inflated, self-ratings.

Real-World Consequences

Money Left on the Table

In negotiations of every kind — salary, real estate, vendor contracts, divorce settlements — the anchor set early in the conversation predicts a large share of the final outcome. Parties who do not appreciate this can leave significant value on the table by deferring to the other side's opening number. Conversely, parties who use anchoring strategically can extract concessions disproportionate to the underlying merits.

Distorted Valuations

Markets are not immune. Initial public offering pricing, analyst estimates, and broker quotes all serve as anchors for participants whose own analysis is then biased by them. In thinly traded or hard-to-value markets, the gap between true value and anchored value can persist for long periods.

Unfair Outcomes in Justice

When judicial sentences are pulled by recommendations or by superficial numerical primes, defendants in similar cases receive systematically different treatment. The legal system has responded with sentencing guidelines, presumptive ranges, and other structural counters — although these counters themselves can act as anchors that may be over- or under-calibrated.

Misallocated Healthcare

In medicine, an early diagnostic anchor can lead to missed alternative diagnoses, sometimes with serious consequences. Reflective practices that explicitly consider "What else could this be?" before settling on a working diagnosis are designed to disrupt this pattern.

Bad Planning

Anchoring on prior project timelines, optimistic feasibility studies, or politically chosen budget figures contributes to systematic underestimation in large projects. Reference-class forecasting — comparing the current project to a distribution of similar past projects, rather than starting from a single anchor — was developed in part as a counter.

Manipulation by Sophisticated Actors

Once anchoring is understood, it can be used. Negotiators, marketers, fundraisers, and political operators routinely deploy anchors to steer the responses of others. Some of this is straightforwardly persuasive communication; some shades into manipulation. Either way, awareness of the bias on the other side is the first line of defense.

How to Recognize It in Yourself

Notice the First Number

Whenever a numerical decision is about to be made — a counter-offer, a budget proposal, a project estimate — pause to ask what number is already in your head, where it came from, and how much of your reasoning is implicitly starting from there. The number that arrived first is the one most likely to anchor the conversation. Naming it, even silently, weakens its grip.

Ask Whether You Adjusted Far Enough

If your final estimate is uncomfortably close to a salient anchor, ask whether you would have arrived at the same number if you had started from a different anchor. If the answer is plausibly no, your adjustment was probably insufficient. A useful exercise is to imagine three different anchors and see how far apart your estimates would be.

Look for Asymmetric Effort

Adjustment is typically faster and easier in one direction than another — usually away from the anchor by a comfortable distance and no further. If your reasoning has done most of its work pulling the answer slightly away from the anchor, rather than building the answer from independent considerations, the answer is probably anchored.

Watch for Suspect Sources

Anchors arrive from a wide variety of sources, not all of them credible. The listing price set by a motivated seller, the salary band quoted by a recruiter with an incentive to keep it low, the time estimate offered by a hopeful project sponsor — each of these is a number with a story behind it. Asking who set the anchor and what their incentive was helps de-anchor.

Track Your Initial Versus Final Estimates

If you make many estimates in a domain, it is worth occasionally logging both your initial impression and your final, considered estimate. Over time, you can see how often the gap between them is large versus small. A consistently small gap suggests you are anchoring on your first impression more than is healthy.

How to Counter It

Generate Your Own Reference Points

Before being exposed to someone else's anchor, generate your own reasoned estimate. In salary negotiations, this means doing the research and writing down a target range before the recruiter asks. In project planning, it means estimating effort from first principles before reading the previous team's estimates. In purchasing decisions, it means knowing what you would pay before seeing the sticker. A well-prepared internal anchor competes with the external one.

Use Multiple Anchors

If a single anchor exerts gravity, multiple anchors dilute it. Looking up several comparable properties before evaluating a real estate listing, or considering several base-rate cases before estimating a probability, generates a distribution rather than a point and reduces the influence of any one number. The technique is related to reference-class forecasting and to formal Bayesian reasoning.

Commit to Ranges, Not Points

Where possible, express estimates as ranges, with the bounds set by considering plausible high and low scenarios rather than by adjusting outward from a central anchor. Forecasting research has shown that people forced to articulate confidence intervals are typically less anchored — and better calibrated — than those who quote a single number.

Consider the Opposite

As with confirmation bias, the consider-the-opposite prompt has shown real effects in anchoring research. Before settling on an estimate, ask: "What reasons can I generate for thinking the answer is much higher than the anchor? Much lower?" This exercise activates information that the anchor's selective-accessibility effect was suppressing.

Make the Anchor Explicit

Strange as it sounds, naming the anchor out loud reduces its effect somewhat. In a negotiation, saying "Your opening of X is interesting; let me set it aside and walk through how I came to my own number" creates psychological distance from the anchor. The technique is used by experienced negotiators precisely because it disrupts the anchor's quiet pull.

Use Structured Approaches

Decision processes that require explicit reasoning — pre-mortems, structured analytic techniques, formal decision matrices, blinded appraisal — make anchoring harder by routing judgments through procedures that do not depend on a single starting number. The effects can be substantial in domains where the procedure can be enforced.

Be Strategic About First Offers

If you are negotiating, do not surrender the anchor lightly. When circumstances permit, make a thoughtful first offer based on a defensible analysis. When you cannot, prepare to neutralize the other side's anchor — by acknowledging it, articulating your own reasoning explicitly, and moving the conversation to the underlying issues before counter-offering.

Limits of Debiasing

Awareness Is Not Enough

Studies that have warned participants about anchoring before exposing them to anchors have found that the warning reduces but does not eliminate the effect. Knowing about the bias is necessary but not sufficient; the anchor still does work even when its presence is named. This is part of why structural counters tend to be more reliable than personal vigilance.

Strong Anchors Are Hard to Neutralize

Some anchors are simply too central to a problem to be set aside. The previous year's budget, the listing price of the house, the prosecutor's recommended sentence — these are not numbers you can pretend you have not seen. The best one can do is generate strong independent reasoning, use multiple reference points, and route the decision through processes that make the anchor's influence visible and contestable.

Experts Anchor Too

The finding that real estate agents, judges, and physicians show anchoring effects is sobering for any institution that relies on expert judgment. Expertise reduces noise and increases overall accuracy, but it does not abolish anchoring. Institutions that depend on expert judgment can supplement it with decision aids, blinded protocols, and structured processes — but they should not assume that expertise alone is enough.

Limits of Range-Setting

Asking people to commit to ranges helps, but ranges themselves can be anchored. A range built by adjusting outward from an anchor will often be too narrow and too centered on the anchor. The best ranges are constructed from independent considerations of upper and lower bounds — what is the highest reasonable case, and what is the lowest? — and then combined.

Structural Solutions

As with other robust biases, the most durable counters are structural. Blinded appraisal, multiple-bid auctions, zero-based budgeting, sentencing guidelines, reference-class forecasting, and standardized clinical protocols all work by limiting the influence that any single anchor can have. They embed disagreement and external reference points into the decision process so that no individual judge has to resist the anchor alone.

Living With Anchoring

The realistic stance is that you will be anchored. The questions are by what, by whom, and with what consequences. The best defense is preparation — knowing your own anchor before someone else gives you theirs — and a habit of generating independent estimates whenever the stakes warrant it. Awareness, paired with practice and structure, will not abolish the effect, but it can keep it small enough to matter less than it otherwise would.