← The Casebook

Calibration Certificate No. conservatism-bias

Conservatism Bias

The tendency to revise beliefs too slowly when new evidence arrives, updating in the same direction as Bayesian inference but by insufficient amounts, as if evidence were less diagnostic than it truly is.

calibratedintuitioncalibrated
systematic

documented offset from the normative answer

Status — mixedgrounding ± moderate

Early bookbag-and-poker-chip experiments were widely replicated across multiple labs in the 1960s-1970s, establishing conservatism as a robust finding. However, the phenomenon's interpretation has been contested. DuCharme (1970) showed conservatism could be partly explained by response bias. More fundamentally, the 'source skepticism' account (Fischhoff & Beyth-Marom, 1983) argues conservatism may be a rational response to imperfectly reliable information sources rather than a cognitive error. Ambuehl and Li (2015) found conservatism varies substantially across individuals and correlates with out-of-sample updating behavior, suggesting it may be a stable individual trait rather than a universal bias.

Calibration record

Bookbag and Poker Chip Problemnot reportedMost subjects gave estimates around 70%. The correct Bayesian answer is approximately 97%. Edwards (1968, p. 20-21) reported this as the canonical demonstration that people update from a prior of 0.5 too conservatively. Similar results held across sequential-sample updating paradigms. Phillips & Edwards, 1966 · Multiple experiments; Edwards (1968) summarizes results from several studies
Sequential-Sample Odds Estimationnot reportedSubjects consistently revised their probability estimates more slowly than prescribed by Bayes' theorem. The degree of conservatism increased with larger diagnosticity parameters and larger sample sizes, suggesting difficulty aggregating more diagnostic or more numerous pieces of information. Peterson, Schneider & Miller, 1965 · n varied across experiments; reported in Edwards (1968)
Investor Underreaction to Earnings Announcementsnot reportedThe model successfully accounted for empirical patterns of short-run momentum and long-run reversal in stock returns. Conservatism in the model causes investors to react insufficiently to new information, generating post-earnings-announcement drift where stock prices continue moving in the direction of the earnings surprise. Barberis, Shleifer & Vishny, 1998 · Theoretical model applied to aggregate stock market data

Source of systematic error

Edwards (1968) proposed that conservatism arises from misaggregation of accurately perceived evidence: people can correctly assess individual pieces of evidence but struggle to combine them optimally. Slovic and Lichtenstein (1971) elaborated that aggregating multiple sources of information is cognitively demanding. More recently, Fischer and Maier-Konig (2018) proposed a motivated account in which conservatism reflects loss aversion over changing beliefs. The 'participant skepticism' account (Fischhoff & Beyth-Marom, 1983; Koehler, 1996) suggests subjects may rationally discount experimenter-provided evidence because real-world information sources are imperfectly reliable.

The 'misperception' account (Peterson & Beach, 1967) holds that people have inaccurate subjective sampling distributions. The 'source reliability' account argues conservatism is rational skepticism about experimenter-provided data. Recent work by Ambuehl and Li (2015) finds conservatism is individual-specific and not caused by cognitive limitations like statistical knowledge, supporting a motivational account.

Recalibration procedure

Quantify the diagnostic implications of evidence using Bayes' theorem or structured probability estimation. Explicitly compare your intuitive posterior against the Bayesian calculation. For sequential evidence, use decision aids or algorithms that enforce proper aggregation of multiple signals.

Providing Bayesian feedback improved probability estimation in odds-estimation tasks (Donnell & DuCharme, 1975). Decomposition strategies that separate evidence evaluation from belief updating can reduce conservatism by focusing attention on the logical force of each piece of evidence.

Cross-calibrated against

Calibrated by Ward Edwards, 1968 — Conservatism in human information processing. In B. Kleinmuntz (Ed.), Formal representation of human judgment, Wiley, pp. 17-52.

Uncertainty: The original Edwards (1968) chapter is a book chapter without a DOI. Much of the empirical evidence is summarized rather than presented with full statistical detail. The source skepticism account challenges the interpretation of conservatism as a cognitive bias. Effect sizes for individual experiments are not consistently reported in the primary literature. The degree of conservatism varies substantially across individuals and contexts.

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 →