A Metastudy on Anchoring Effect Moderators

Numeric anchoring is a classic and robust judgment phenomenon in which, numeric judgments depend systematically on previously-presented numbers. Competing theoriespropose many moderators, but conclusions from prior meta-analyses are constrained by publication bias and correlational designs. We ran a preregistered metastudy (N = 1,968) to more causally test theory- and practice-relevant moderators of experimenter-provided anchors. Cognitive load and more extreme anchors increased effect sizes, whereas debiasing and less diagnostic anchors reduced them. Precision, directional information, incentives, and a page break between the anchor and estimate showed small or null effects. Greater subjective knowledge predicted smaller anchoring. Overall, the patterns are most consistent with Anchoring-and-Adjustment and Noise accounts and less consistent with the Selective Accessibility Hypothesis and the Attitudinal perspective. Additional exploratory analyses point to diagnosticity and extremity as the most consequential levers for future research.