What is anchoring bias?

Three components are worth separating, because they fail in different ways:

• The anchor. A salient value arrives first — a price paid, a quoted figure, a headline number — and becomes the starting point for the judgment, regardless of how relevant it actually is.
• Insufficient adjustment. You do move away from the anchor, but you stop as soon as the new figure feels defensible rather than when the evidence runs out, so estimates land systematically close to it.
• Range compression. The anchor narrows the outcomes you seriously consider. Scenarios far from it never get modelled, which is why anchored forecasts are both wrong and confident.

Anchoring is not the claim that reference points are useless. A well-chosen reference — a base rate, a valuation multiple derived from comparables — is exactly what good analysis uses. The error is accepting whichever number happened to appear first as though it were that reference.

The effect was documented by Tversky and Kahneman in the 1970s under the heading "anchoring and adjustment". In their best-known demonstration, participants spun a wheel of fortune before estimating the percentage of African countries in the United Nations. Those who landed on a high number gave markedly higher estimates than those who landed on a low one — even though everyone could see the number was produced at random.

Later work found the effect survives almost every attempt to switch it off. It persists when the anchor is obviously irrelevant, when participants are warned about it, when they are paid for accuracy, and among experts working inside their own domain — real estate agents valuing houses, judges setting sentences, engineers estimating costs. Anchoring is not a novice error; it is how numerical judgment works.

Two mechanisms explain it. Adjustment from a starting value is effortful and terminates early, at the near edge of the plausible range rather than its centre. And an anchor primes memory selectively: once a number is present, information consistent with it becomes easier to retrieve, so the anchored estimate arrives with reasons attached and feels independently derived.

Further reading: the anchoring effect in judgment and decision research.

The pattern is easiest to see in someone else's decision. These are the shapes it takes most often:

• Trading

  A position bought at 100 and now trading at 70 is held because 'it just needs to get back to where I bought it'. The entry price is a fact about your history, not about the asset, but it has become the yardstick for fair value.

• Investing

  A stock that fell from 400 to 250 looks cheap purely because of the old high, while a stock that rose from 20 to 60 looks expensive. Both judgments are anchored on a past price rather than on current earnings, growth, or risk.

• Valuation and fundraising

  A startup's last round valuation frames the whole negotiation, so both sides argue in a band around that number even after the underlying business has materially changed in either direction.

• Analyst estimates

  The first forecast you read — a price target, a growth rate, a market size — becomes the centre of your own range, and your final estimate lands suspiciously close to it despite doing independent work.

• Salary and negotiation

  Whoever names a figure first sets the zone in which the rest of the conversation happens, and the eventual agreement usually sits nearer that opening number than either side expects.

• Round numbers and price levels

  Targets, stops, and 'cheap' thresholds cluster at round numbers and 52-week highs and lows because those values are salient, not because they carry information about future returns.

Three forces keep it in place. Adjustment is expensive: moving away from a starting value requires generating reasons, and the search stops as soon as a number feels acceptable. Anchors prime evidence: the presence of a figure makes consistent information easier to recall, so the anchor supplies its own justification. Salience beats relevance: round numbers, highs, lows, and prices you personally paid are vivid and easy to hold in mind, and vividness is repeatedly mistaken for informativeness.

This is why "ignore the anchor" fails as advice. By the time you notice the number, it has already shaped the range you are working inside. The remedy is procedural — decide the order in which information reaches you, and produce your own estimate before anyone else's number is on the table.

In a trading account, anchoring most often appears as the entry price acting as a target. A position is held not because the current evidence supports it but because selling below the purchase price would make the loss real — which is anchoring feeding directly into loss aversion. The same mechanism caps winners early, when a round-number target set months ago overrides everything that has happened since.

For investors it distorts valuation. "Cheap" and "expensive" get defined against past prices rather than against cash flows, so a stock down 40% from an unjustified high looks like value and a compounding business at a new high looks unaffordable. In negotiation and fundraising it decides outcomes before analysis begins: the first figure named frames the entire zone of agreement.

It compounds with other biases too. An anchor sets the belief that confirmation bias then defends, and the narrowed range it produces is one of the cleanest routes to overconfidence.

You cannot unsee a number, but you can control when it reaches you and how much of the decision it is allowed to carry:

1. 1

   Estimate before you look

   Write your own number — value, target, probability, size — before reading anyone else's. An anchor you never saw cannot pull your estimate, and comparing afterwards tells you something useful about both figures.

2. 2

   Work in ranges, not points

   State a plausible low and high before settling on a central case. Ranges force you to think about the distribution rather than adjusting timidly away from one salient value.

3. 3

   Argue the far side on purpose

   Deliberately build the case for an outcome well beyond your estimate in each direction. Considering an extreme is the most reliable way to loosen an anchor, because it makes insufficient adjustment obvious.

4. 4

   Treat your entry price as sunk

   The only question a position asks is whether you would buy it today at today's price with today's evidence. Your purchase price belongs to the past, and holding it as a target is anchoring wearing the clothes of discipline.

5. 5

   Define levels by rule, not by roundness

   Derive stops, targets, and sizing from volatility, risk per trade, or valuation — written in advance — so that salient numbers on a chart cannot quietly become your decision criteria.

6. 6

   Watch your own language

   Phrases like "back to where I bought it," "it used to trade at," "cheap compared to its high," or "they opened at X so" are markers that a reference number, not the current evidence, is carrying the argument.

• Why day traders lose money

  The hub piece: how an hours-old price quietly distorts value judgments, alongside six other biases behind the failure rate.

• Loss aversion

  The entry-price anchor is what turns an ordinary decline into a 'loss' you refuse to realise, which is where anchoring does most of its financial damage.

• Sunk cost fallacy

  Both keep a decision tied to a past number rather than to the outcomes still available from here.

• Confirmation bias

  Once an anchor has set your estimate, evidence supporting that number gets favoured over evidence against it.

• Recency bias

  The most recent price or result is the most available anchor, which is how a short stretch quietly becomes your reference point.

• Overconfidence

  Anchoring narrows the range of outcomes you consider, and a narrow range feels like precision rather than a missing tail.

The Bias Checker screens for all of them at once, and the articles go deeper on each.