Essays on Digital Content Strategies: Creation, Diffusion, and Monetization

Abstract

Content—whether in the form of social media posts, online videos, product reviews, or digital media coverage—lies at the heart of modern business and marketing. It shapes how firms and creators communicate, foster engagement, and build trust with consumers. My dissertation examines how content is created, diffused, and monetized in today’s digital world. The first chapter studies content monetization by examining how sponsored content affects influencer reputation. Using a large-scale dataset of YouTube creator videos, I extract rich, theory-driven video features and apply DiNardo-Fortin-Lemieux reweighting to construct comparable treatment and control groups matched at the influencer-video level. The matched-sample analysis reveals a reputation-burning effect: compared with posting an equivalent organic video, posting a sponsored video costs influencers 0.19% of their reputation, measured by subscriber count. This effect is stronger for influencers with larger audiences. It is mitigated when there is a strong fit between the sponsored content and the influencer’s usual content, and when the promoted brand is less well-known. This chapter empirically tests a key assumption in theoretical work, contributes to the literature on influencer marketing and celebrity endorsement, and offers implications for influencers, brands, and platforms. The second chapter examines the diffusion and monetization of content in the music industry, focusing on how TikTok affects music discovery, consumption, and revenue. I leverage the dispute between TikTok and Universal Music Group (UMG), during which UMG removed its music from TikTok from February to May 2024. UMG argued that TikTok undercompensated artists and labels because consumption on the platform could reduce revenue elsewhere, while TikTok emphasized its promotional and discovery benefits. Using tracks from Sony and Warner as a control group, I conduct a difference-in-differences analysis. Overall, removing UMG music from TikTok did not significantly affect demand for UMG tracks on Spotify and YouTube. However, I find substantial heterogeneity. Tracks previously available on TikTok experienced a 2–3% increase in consumption after removal, suggesting substitution, especially among popular tracks by well-known artists. In contrast, UMG tracks not previously available on TikTok saw a 1–3% decrease in streams, suggesting complementarity, especially for less popular tracks by lesser-known artists. Further analysis indicates that this complementary effect is driven by TikTok’s role in promoting and enabling discovery for artists with a partial presence on the platform. An economic impact analysis shows that TikTok substantially undercompensated UMG, consistent with the terms of their new licensing agreement. The third chapter turns to content creation and explores how media firms can use large language models to generate news content that balances multiple objectives: increasing engagement while maintaining a preferred editorial stance. Using articles from The New York Times, I first show that more engaging human-written content tends to be more polarizing, and that naïvely using LLMs to increase engagement can further amplify polarization. I then propose a constructive solution based on \textit{Multi-Objective Direct Preference Optimization (MODPO)}, which combines Direct Preference Optimization with multi-objective optimization. Using an open-source LLM, I develop a model that increases engagement while maintaining editorial stance. The model achieves this balance by leveraging theory-driven content strategies, such as reducing provocative language while allowing more balanced perspectives. These findings apply broadly to settings where firms use LLMs to create content under multiple objectives, including advertising and social media.