← The Casebook

Calibration Certificate No. anchoring

Anchoring

Anchoring is the tendency for an initial reference value, even an arbitrary or irrelevant one, to pull subsequent numerical estimates and judgments toward it.

calibratedintuitioncalibrated
≈ 20 pts

the pull of a visibly random anchor on the median estimate (wheel-of-fortune, Tversky & Kahneman 1974)

Status — mixedgrounding ± strong

Anchoring is among the more reliably reproduced effects in social/cognitive psychology: it was successfully replicated in the Many Labs project (Klein et al., 2014; total N=6,344 across ~36 labs), where anchoring tasks were among the largest and most robust effects, and multiple meta-analyses report large effects with effect sizes similar in published and unpublished studies (Röseler and colleagues' Open Anchoring Quest work). However, the magnitude is contested: high-powered re-evaluations find much smaller effects than the underpowered originals. Li, Weigel, Ferraro & Messer (2025, Economic Inquiry) raised power from 46% to 96% and found a 3.4% effect (95% CI [-3.4%, 10%]) where the original reported ~31% — about seven times smaller, with a CI including zero. So the existence of anchoring replicates well, but classic effect sizes appear inflated.

Calibration record

Wheel of fortune (original demonstration)Median estimates ~25% (anchor 10) vs ~45% (anchor 65); inferential statistics not reported in the original paperEstimates assimilated to the random anchor; the visibly arbitrary number shifted the median judgment substantially. Amos Tversky, Daniel Kahneman, 1974 · University of Oregon students (exact n not reported in paper)
Real-estate appraisal by professionalsAppraisals shifted roughly in step with the listing manipulation (an ~$11,000 higher listing yielded ~$14,000 higher appraisals, per reports of the study)Appraisals tracked the manipulated listing price for both experts and amateurs; most agents denied being influenced (only ~19% cited listing price). Gregory B. Northcraft, Margaret A. Neale, 1987 · Real-estate agents and students (Organizational Behavior and Human Decision Processes, 39, 84-97)
Playing dice with criminal sentencesMean recommended sentence higher under high-anchor vs low-anchor dice (large effect reported); exact d as publishedSentencing recommendations of experienced judges/prosecutors were pulled toward the blatantly random dice anchor. Birte Englich, Thomas Mussweiler, Fritz Strack, 2006 · Experienced German legal professionals (Personality and Social Psychology Bulletin, 32(2), 188-200)
Selective accessibility mechanism studiesnot reported hereSupported the selective accessibility model: solving the comparative anchor question makes anchor-consistent information more accessible for the later absolute judgment. Fritz Strack, Thomas Mussweiler, 1997 · Student samples (Journal of Personality and Social Psychology, 73, 437-446)
Putting adjustment back in (self-generated anchors)not reported hereFor self-generated anchors, people genuinely adjust but insufficiently; head movements affirming/denying values changed answers only for self-generated anchors, supporting a dual-process view. Nicholas Epley, Thomas Gilovich, 2001 · Student samples (Psychological Science, 12(5), 391-396)

Source of systematic error

Two mechanisms are now distinguished by anchor type. For experimenter-provided (external) anchors, the dominant account is the selective accessibility model (Strack & Mussweiler, 1997): testing the hypothesis that the anchor is the answer selectively activates anchor-consistent information, which then biases the absolute judgment, much like semantic priming. For self-generated anchors (e.g., starting from a known value and adjusting), Epley and Gilovich (2001, 2006) revived the original anchoring-and-adjustment account: people adjust away from the anchor but stop at the near edge of a plausible range, yielding insufficient adjustment.

The original Tversky-Kahneman 'insufficient adjustment' explanation was challenged by selective-accessibility/priming accounts in the 1990s. Epley and Gilovich reconciled them as a dual-process picture: adjustment for self-generated anchors, accessibility/priming for provided anchors. Numeric-priming and conversational-inference accounts have also been proposed.

Recalibration procedure

Consider-the-opposite: deliberately generate reasons the true value could be far from the anchor (especially much lower/higher), and where possible source your estimate from independent data before any reference number is shown. Mussweiler, Strack & Pfeiffer (2000) found that explicitly considering anchor-inconsistent arguments reduced the anchoring bias.

Mussweiler, Strack & Pfeiffer (2000), 'Overcoming the Inevitable Anchoring Effect: Considering the Opposite Compensates for Selective Accessibility,' Personality and Social Psychology Bulletin, 26(9), 1142-1150, showed the consider-the-opposite strategy attenuates anchoring; forewarning alone is generally weak.

Cross-calibrated against

Calibrated by Amos Tversky, Daniel Kahneman, 1974 — Judgment under Uncertainty: Heuristics and Biases. Science, 185(4157), 1124-1131.

Uncertainty: The 1974 paper reported the wheel-of-fortune medians (~25% vs ~45%) without inferential statistics or exact sample sizes, so those specifics are as the paper presents them. Effect-size figures for Northcraft & Neale (1987) and Englich et al. (2006) are summarized from secondary reports and the abstracts; I did not retrieve the full original PDFs to verify every coefficient, so treat the dollar/sentence magnitudes as approximate. For the Röseler 'Open Anchoring Quest' / '50 Years of Anchoring' meta-analytic work I confirmed the dataset and the claim that published and unpublished effect sizes are similar, but I could not retrieve a single verified pooled Cohen's d, so no specific meta-analytic d is asserted. The Jacowitz & Kahneman (1995) URL is the plausible PSPB record but was not independently fetched. The honest headline: existence of anchoring replicates well (Many Labs), but classic effect magnitudes are likely inflated by low statistical power (Li et al., 2025), which is why replication status is 'mixed' rather than 'robust.'

File your own case

Open the same case on your own draft.

Paste a memo, a research draft, or a strategy argument. It is scored against all 175 cards, and the strongest two or three risks come back with the evidence quoted and one practical next check.

Open a case on your draft →