Endorser Controversy Spillovers and Firm Reputation Loss Using Sponsorship Exposure and Multimodal Stakeholder Signals
Main Article Content
Abstract
Celebrity endorsers and digital creators provide firms with attention, identity transfer, and access to highly segmented audiences. These benefits also create exposure to reputational shocks that originate outside the firm. When an endorser becomes publicly controversial, sponsoring firms face a difficult empirical problem: the incident may be irrelevant to product quality, yet consumers, investors, and media audiences may still revise beliefs about brand judgment, governance, and cultural fit. This study examines endorser controversy spillovers using a panel of 1,138 public controversies involving 642 endorsers and 391 publicly traded sponsoring firms from 2013 to 2024. The dataset combines sponsorship-contract evidence, advertising exposure, social-media text, search behavior, product-review sentiment, consumer purchase proxies, equity returns, option prices, analyst language, and brand-owner responses. Endorser controversy severity is measured with a multimodal classifier that separates norm violation, legal allegation, discriminatory speech, deceptive promotion, performance scandal, and political polarization. Sponsorship exposure is measured by contract visibility, campaign recency, audience overlap, endorser-brand fit, and creative dependence. The results show that controversy spillovers are concentrated in high-fit, high-recency relationships where the endorser is visually or narratively central to the campaign. A one-standard-deviation increase in sponsorship exposure predicts a 0.48% lower three-day abnormal return and a 5.6% decline in brand-specific consumer sentiment. Firms that pause campaigns without clarifying replacement and governance steps experience slower sentiment recovery than firms that give precise corrective disclosures. Machine-learning tests show that exposure architecture predicts market reaction more accurately than controversy severity alone.