Suspicious Perfection: Rating Rewards, Credibility Inversion, and Information Loss

Platforms use rating bonuses and badges to reward good sellers. When ratings can be inflated, stronger rewards can instead weaken the information available to buyers. We study an initially informative rating system in which sellers can pay to raise their displayed score and buyers rationally account for manipulation. Manipulation costs and returns are common across qualities conditional on the honest rating. Under this restriction, credibility can fail to increase between adjacent scores only where the higher score earns an additional reward. Our main result concerns large top-rating bonuses. Once sellers choose only between keeping their honest rating and reaching the top, a larger bonus makes the equilibrium rating system less informative in Blackwell's order along either extremal equilibrium selection. Lower scores retain their honest quality composition but become rarer. As the bonus grows, almost everyone displays the top score, its credibility approaches the population average, and the value of rating information vanishes. Nevertheless, a nearly perfect score can remain more credible than a perfect one. We derive the rate of information loss for exponential manipulation costs and establish a budget-feasible exposure rule that avoids credibility reversals. The results suggest that platform evaluations should examine rating frequencies together with category credibility. Rational buyer skepticism can correct the interpretation of ratings without recovering the distinctions that manipulation has erased.