Anchoring Bias

In 2019, Softbank valued the office-sharing company WeWork at $47 billion. That number became the anchor. When WeWork's IPO prospectus revealed staggering losses, analysts didn't evaluate the company from scratch -- they adjusted downward from $47 billion. "It's still worth $20 billion," some argued. "Maybe $15 billion." The company eventually went public at roughly $9 billion, and even that figure was generous -- WeWork filed for bankruptcy in November 2023. Each successive estimate orbited the original anchor like a planet that couldn't escape its star's gravity, even as the fundamental business case disintegrated beneath it.

The First Number Wins

Anchoring bias describes our tendency to rely disproportionately on the first piece of information we encounter when making judgments or decisions. That initial number, figure, or estimate becomes a reference point -- an anchor -- that pulls all subsequent thinking toward it, even when it has no logical bearing on the question at hand. This is NOT the same as confirmation bias, though the two often work in concert. Confirmation bias is about selectively seeking evidence that supports what you already believe; anchoring bias is about a specific numerical or quantitative starting point warping how you evaluate all information that follows. You can be anchored by a number you don't believe, have no attachment to, and know to be arbitrary. That's what makes it distinctive and, frankly, alarming.

The concept was first identified by psychologists Daniel Kahneman and Amos Tversky in a landmark 1974 paper, "Judgment Under Uncertainty: Heuristics and Biases." Their core insight was that people estimate quantities not by calculating from first principles but by starting from an available value and adjusting -- and that adjustment is almost always insufficient.

How the Anchor Takes Hold

The machinery of anchoring operates through what researchers call "selective accessibility." When you encounter a number -- even one you consciously dismiss as irrelevant -- your brain begins generating reasons why that number might be correct. This process happens automatically, below the threshold of awareness. Fritz Strack and Thomas Mussweiler demonstrated this in a series of studies at the University of Wurzburg in the late 1990s, showing that anchors work by activating information from memory that is consistent with the anchor value. When participants were given a high anchor for the length of the Mississippi River, they spontaneously recalled facts about long rivers; when given a low anchor, they recalled facts about short ones. The anchor doesn't just sit passively in your mind. It actively recruits supporting evidence, constructing a case for its own accuracy before you've made any deliberate effort to evaluate it. Your brain treats the anchor as a hypothesis and immediately starts confirming it, which is why even experts who recognize the anchor and actively try to resist it still show measurable anchoring effects in controlled experiments.

Kahneman and Tversky's most famous demonstration involved a rigged wheel of fortune that landed on either 10 or 65. Participants were asked to estimate the percentage of African countries in the United Nations. Those who saw 10 guessed around 25 percent. Those who saw 65 guessed around 45 percent. Everyone knew the wheel was random. The anchor moved their estimates by twenty points anyway.

Anchoring at Personal and Systemic Scale

The effects of anchoring are not confined to laboratories. In salary negotiations, the party who names a number first typically controls the final outcome. Linda Babcock and Sara Laschever documented that women who failed to set an initial anchor in salary discussions earned, on average, $500,000 less over their careers than those who did. The anchor doesn't just influence one conversation -- it compounds across a lifetime of negotiations, raises, and promotions calibrated to the original figure.

At the systemic level, anchoring shapes entire legal systems. In a study by Birte Englich, Thomas Mussweiler, and Fritz Strack, German judges with an average of fifteen years of experience determined a prison sentence after rolling dice rigged to produce either a high or low number. Judges who rolled high gave sentences averaging eight months; those who rolled low averaged five months. A completely irrelevant random number shifted sentencing by roughly fifty percent.

Real estate markets offer another systemic example. When licensed agents appraised a home after touring it, those given a higher listing price produced systematically higher appraisals than those given a lower one -- despite viewing the same property with full access to comparable sales data. Asked whether the listing price influenced them, most confidently said no.

Where This Breaks Down

Anchoring bias is a powerful explanatory concept, but applying it carelessly creates its own problems.

First, not every initial number is an anchor in the technical sense. Sometimes the first number you encounter is genuinely informative -- a market price, a benchmark, a well-researched estimate. Dismissing all first offers or initial figures as manipulative anchors can lead to irrational counter-anchoring, where you overcorrect so aggressively that you end up further from the truth than if you'd accepted the original number. A person who refuses to engage with a home's listing price because "it's just an anchor" may waste months pursuing a valuation that has no basis in market reality.

Second, the popular advice to "always make the first offer" oversimplifies the research. Anchoring power depends heavily on information asymmetry. When you know less than the other party about the true value of what's being negotiated, going first can set an anchor that reveals your ignorance and damages your position. The advantage of anchoring belongs to the better-informed party, not simply the faster-talking one.

Third, anchoring research predominantly involves numerical judgments in simplified settings. Real-world decisions involve multiple competing anchors, evolving information, and deliberation with others. The strength of a single anchor diminishes when people have access to rich data, collaborative discussion, and time to reflect -- which means anchoring is most dangerous precisely when these safeguards are absent: under time pressure, in low-information environments, and when you're deciding alone.

Fourth, awareness of anchoring bias can itself be weaponized. Negotiators who know their counterpart has read about anchoring sometimes use this knowledge strategically -- setting an absurdly extreme anchor and then retreating to a position that seems "reasonable" by comparison but is still favorable to them. Knowing about anchoring doesn't immunize you; it changes the game.

How Anchoring Connects to Other Biases

Anchoring rarely operates in isolation. It interacts with several other cognitive patterns in ways that amplify its effects.

Loss aversion makes anchors particularly sticky when they represent a past price or a prior valuation. Once you've anchored on what you paid for a stock, selling it below that price feels like a loss -- even if the current market value is well-established. The anchor and the loss aversion reinforce each other, trapping you in a position long after the evidence has changed.

The sunk cost fallacy creates similar entanglement. When you've invested time or money based on an anchored estimate, the anchor becomes harder to revise because revising it would mean acknowledging that the investment was poorly calibrated. WeWork's investors didn't just anchor on $47 billion because it was the first number -- they anchored because walking it back meant admitting they had overpaid by billions.

Heuristics more broadly provide the context for understanding anchoring. Anchoring is itself a heuristic -- a mental shortcut that substitutes a simpler question ("How far is this number from the anchor?") for a harder one ("What is the actual value?"). Understanding the broader landscape of heuristics helps explain why anchoring persists even among people who know about it: the brain defaults to its shortcuts under cognitive load, stress, or time pressure.

Base rates offer the most direct antidote to anchoring. When you enter a negotiation or evaluation with well-researched base rate data -- the median salary for this role, the average price per square foot in this neighborhood, the typical valuation multiple for this type of business -- you create an internal anchor grounded in evidence rather than in whatever number the other party happens to mention first.

The Anchor Audit

The self-test is straightforward. Before any significant decision involving numbers -- a purchase, a negotiation, a forecast, an evaluation -- ask yourself: "Where did my first number come from, and would I have arrived at the same estimate if I'd never seen it?"

This question is harder to answer honestly than it sounds. The internal experience of being anchored is one of quiet confidence. You don't feel pulled toward the anchor. You feel like you've made an independent judgment that happens to be in the general vicinity of a number you encountered earlier. The hallmark of anchoring is that it feels like reasoning. The moment to apply this self-test is any time you notice a specific number already occupying your mind before you've done your own analysis -- a list price, a first offer, a headline figure, a friend's estimate. That number is your trigger. When you notice it, pause. Do your own research. Generate your estimate independently. Write it down before you compare it to the anchor. The gap between the two reveals how much gravity the anchor was exerting.

What this feels like in practice is a strange doubling of perspective. You hold your "independent" estimate in one hand and the anchor in the other, and you notice a magnetic pull between them -- a quiet urge to adjust your number toward the anchor, to split the difference, to call it close enough. Resisting that pull is the whole exercise.

Back to the Anchor That Sank a Company

WeWork's collapse wasn't caused by anchoring bias alone, but anchoring explains why so many intelligent people held on so long. The $47 billion valuation wasn't just a number -- it was the gravitational center of every subsequent conversation about the company's worth. Each new estimate adjusted down from that anchor rather than building up from fundamentals. Had investors started from base rates -- revenue multiples for comparable real estate companies, occupancy economics, unit profitability -- they would have arrived at a figure a fraction of $47 billion. But the anchor was set, and the adjustment was never sufficient. It almost never is.