Extension: Misinformation and Mistrust: The Equilibrium Effects of Fake Reviews on Amazon.com
Reputation is a selection mechanism. Gandhi, Hollenbeck, and Li study products that purchased fake reviews on Amazon and the honest products that competed with them, and their equilibrium simulations attribute a 27.2 percent increase in units sold to the manipulators and a 4.4 percent decline for honest products. The manipulated signal distorted description and redirected market share. The result supplies a concrete market mechanism for selection on claims. Consumers formed beliefs about quality from observed reviews and their perceived trustworthiness. Sellers could purchase favorable reviews that raised perceived quality. The resulting misinformation induced some consumers to choose lower quality products. Those products carried inflated ratings and higher prices. Honest sellers lost market share even though their products had not deteriorated. A reputation signal became an input controlled by the party seeking selection rather than an independent record of the relevant outcome. The evidence is narrower than the institutional claim because the study concerns Amazon product markets, not service procurement or sovereign local agent runtimes. Its market effects come from a structural demand model and counterfactual simulations. The classification of review purchasers and the estimated share of fake reviews introduce measurement uncertainty. The paper does not test installation counts, self reported capability, or task specific performance. It also does not establish that every static rating is non-contextual or manipulable. A service market should therefore separate promotional reputation from verified performance. Selection should depend on outcomes for the relevant task class. Those outcomes should be measured over a defined period and attributed to an independent source. The record should disclose uncertainty and failed engagements. It should also record changes in operating conditions and the relation between the prior task and the proposed task. Provider supplied claims may remain discoverable. They should not control selection. Without contextual outcome evidence, a market may price the strength of the signal while misallocating the work.
reputationeconomicsrisk-and-incentivesmarketplacesselectioninformation-asymmetryreview-manipulationservice-quality