Anchoring Bias: The First Number Wins

A wheel of fortune is spun in front of you. It lands on 65. Now: what percentage of African countries are members of the United Nations? Your answer, statistically, will be higher than if the wheel had landed on 10 — even though you know perfectly well that a random number on a spinning wheel has no bearing on African geopolitics. This is anchoring bias, one of the most durable and counterintuitive findings in cognitive psychology. It means that the first number you encounter — regardless of its relevance — colours every numerical judgment that follows.

The Founding Experiment

The wheel-of-fortune study was conducted by Amos Tversky and Daniel Kahneman in 1974 as part of their foundational work on heuristics and biases. Participants watched a rigged wheel land on either 10 or 65, then estimated the percentage of UN member states from Africa. Those who saw 65 gave a median answer of 45%. Those who saw 10 gave a median answer of 25%. The wheel was obviously random. Participants knew it was random. It didn't matter. The anchor had done its work.

Tversky and Kahneman described the underlying process as anchoring and adjustment: when estimating an unknown quantity, we start from an initial value (the anchor) and adjust from there. The problem is that adjustment is typically insufficient. We stop adjusting before we reach the true answer because any value in the neighbourhood of the anchor feels intuitively "in the right ballpark," even when it isn't. We anchor, we nudge, we stop too soon.

Arbitrary Anchors Have Real Power

What makes anchoring particularly striking is that anchors do not need to be credible or even related to the question at hand. In a study by Dan Ariely, Drazen Prelec, and George Loewenstein (2003), participants were asked to write down the last two digits of their Social Security number and then bid on various items (wine, chocolates, computer accessories). Those with higher Social Security endings consistently bid significantly more — a 10-20% effect for items across the board. The last two digits of a government identification number determined the willingness to pay for unrelated commercial products.

In another demonstration, researchers asked people whether Mahatma Gandhi died before or after age 9, or before or after age 140 (obviously impossible anchors). Subsequent estimates of Gandhi's actual age at death were significantly pulled in the direction of the absurd anchor. People "knew" the anchor was nonsense — and were still influenced by it.

Anchoring in the Real World

Retail and Pricing

Retail pricing is built on anchoring. The "was £120, now £79" sale tag works because £120 functions as an anchor that makes £79 feel like a bargain — even if the product was never actually sold at £120, or if £79 is above the fair market price. The "suggested retail price" on packaging serves the same function when the actual selling price is lower. Restaurants place an extremely expensive item at the top of the menu not necessarily to sell it but to anchor the price expectations of diners, making the second-most-expensive option feel reasonable by comparison. The first number on the page reshapes all subsequent judgments.

Subscription services use anchoring when they present tiers: a very expensive "premium" tier makes the mid-tier look affordable, even if the mid-tier is what they primarily want to sell. The premium tier is an anchor, not a product.

Salary Negotiation

Research consistently shows that whoever makes the first offer in a salary negotiation creates an anchor that significantly influences the final settlement. Studies by Adam Galinsky and Thomas Mussweiler found that the first offer was the strongest predictor of final agreement, with deviations from it typically being small. Conventional wisdom once held that you should never make the first offer — but the research suggests the opposite: making an ambitious first offer anchors negotiations in your favour. The adjustment process tends to work against whoever concedes first.

This has practical implications both for job seekers (stating a salary expectation slightly above your target tends to shift the final number upward) and for employers (low initial offers create anchors that candidates may never fully escape).

Legal Sentencing

A disturbing body of research suggests anchoring operates even in criminal sentencing. In a 2006 study by Birte Englich, Thomas Mussweiler, and Fritz Strack, experienced legal professionals were given a case file and asked to roll a die (rigged to show either 3 or 9) before deciding on a sentence. Those who rolled 9 recommended sentences averaging 8 months; those who rolled 3 recommended sentences averaging 5 months. A meaningless random number — one that participants knew was meaningless — significantly affected their sentencing recommendations.

The same mechanism may operate when prosecutors make sentencing recommendations, when a defence lawyer suggests a lenient alternative, or when a prior sentence in a similar case becomes salient. The starting point contaminates the judgment, even in trained professionals making high-stakes decisions.

Medical Diagnosis

Clinical medicine is not immune. When doctors form an early diagnostic hypothesis, it serves as an anchor. Subsequent information tends to be interpreted in light of it, and adjustment away from the initial diagnosis — even in response to contradicting evidence — is typically insufficient. This "anchoring on diagnosis" is considered one of the main cognitive mechanisms behind diagnostic error in medicine, where early impressions fail to be revised when the evidence demands it. It connects closely to confirmation bias, which drives selective attention to evidence that supports the anchor.

Why Adjustment Is Always Insufficient

The anchoring-and-adjustment account explains the bias, but it also raises the question: why do we systematically under-adjust? Research by Nicholas Epley and Thomas Gilovich (2006) suggests that adjustment stops when people reach the edge of their "plausible range" — the first value that doesn't feel obviously wrong. Because we start from the anchor and move toward plausibility, we stop as soon as we reach a value that seems defensible, rather than continuing to the most accurate estimate. The direction of movement is correct; the distance is wrong.

A complementary account is confirmatory hypothesis testing: once an anchor is set, we selectively retrieve information consistent with it, which makes it feel more validated and reduces the perceived need to adjust further. The anchor recruits its own supporting evidence, making itself feel more accurate than it is.

The Resistance of Experts

A natural question is whether experts — people who know what numbers should look like in their domain — resist anchoring better than novices. The research answer is: somewhat, but not as much as you would hope. Northcraft and Neale (1987) found that real estate agents' property valuations were significantly influenced by the stated listing price — an anchor — even while insisting they were using objective comparable sales data. The anchoring effect was smaller than for inexperienced amateurs, but it was still substantial and still unconscious. Domain expertise reduces anchoring bias but does not eliminate it.

Countering Anchoring

The difficulty with debiasing anchoring is that knowing about the bias does not neutralise it. Merely warning participants that an anchor is irrelevant does not prevent it from influencing their estimates. Some strategies that do help:

• Generate a counter-anchor: Before making a numerical judgment, explicitly generate a value in the opposite direction and consider the evidence for it. This forces engagement with information that would otherwise be screened out.
• Prepare your own anchor first: In negotiation contexts, forming an independent reference before encountering the other party's number reduces (but does not eliminate) the anchoring effect of their opening offer.
• Consider the extreme: Research by Mussweiler and Strack shows that considering reasons why the anchor might be wrong — deliberately generating disconfirmatory evidence — reduces its effect more than simply trying to ignore it.
• Base rates and reference classes: Grounding estimates in systematic data (what does this type of item typically cost? what is the average sentence for this category of offence?) provides a competing anchor derived from evidence rather than arbitrary salience.

In a world built on numbers — prices, deadlines, statistics, offers — anchoring is everywhere. The first number is rarely neutral. It is a rhetorical act, a cognitive trap, and a negotiation strategy all at once. Knowing this doesn't make you immune; it does, at least, make you less surprised when you notice the pull.

Sources & Further Reading

• Tversky, A., & Kahneman, D. "Judgment Under Uncertainty: Heuristics and Biases." Science 185, no. 4157 (1974): 1124–1131.
• Ariely, D., Prelec, D., & Loewenstein, G. "Coherent Arbitrariness." Quarterly Journal of Economics 118, no. 1 (2003): 73–106.
• Englich, B., Mussweiler, T., & Strack, F. "Playing Dice with Criminal Sentences." Personality and Social Psychology Bulletin 32, no. 2 (2006): 188–200.
• Galinsky, A. D., & Mussweiler, T. "First Offers as Anchors." Journal of Personality and Social Psychology 81, no. 4 (2001): 657–669.
• Epley, N., & Gilovich, T. "The Anchoring-and-Adjustment Heuristic." Psychological Science 17, no. 4 (2006): 311–318.
• Wikipedia: Anchoring effect