In this dissertation, David Holtz investigates the impact of personalized content recommendations and proposes methods to estimate unbiased treatment effects in online marketplaces. Chapter 1 reveals that increasing the personalization of Spotify's recommendations boosts podcast streaming but reduces individual consumption diversity and highlights the "engagement-diversity trade-off." In Chapters 2 and 3, Holtz presents advanced methods like graph cluster randomization to reduce bias in estimating treatment effects within online marketplaces, using Airbnb as a case study, and finds that interference bias impacts estimates significantly, with its magnitude influenced by the supply or demand constraints of the market.