In this thesis, we study multiple aspects of digital markets, including developing methods that extract customer and product insights from text data, investigating the implications of iBuyers on the racial price differentials in housing market, and designing fair recommendation system based on behavioral characteristics in user interaction with an online Business-to-Business marketplace. Online reviews are integral parts of digital platforms and marketplaces. We benchmark and develop new methods to extract customer and product insights from text data. We comprehensively benchmark several topic modeling methods, including probabilistic and matrix factorization models, for interpretability and prediction accuracy. We identify the relative strengths and weaknesses of these methods within different document scenarios, providing a basis for model selection for researchers. Beyond this, our work also contributes by developing a new model that generates more interpretable topics with better predictions. iBuyers are a new type of housing market intermediary that rely heavily on digital technologies and algorithms, also offer online platforms for buyers to search and purchase homes directly without real estate agents. We investigate the implications of iBuyers on the housing market. We quantify robust estimates of iBuyers’ mitigating effect on racial price differentials, separate market-level effect and transaction-level effect, compare the iBuyers’ role with traditional housing market intermediary, flippers, and investigate the heterogeneity of neighborhood racial compositions and buyer income levels. Our work bridges studies on the impacts of innovative technologies and those on racial price differentials in the residential real estate market. Our work contributes to the literature on the impacts of digital platforms in the U.S. housing market, the discussion of the intermediary roles of different types of flippers, and the emerging body of literature in Fintech show evidence that the loan automationreduces racial bias in mortgage outcomes through the channel of removing face-to-face interactions and human judgment. Recommendation system is the key operation that facilitates the transactions between the supply and demand sides in many online marketplaces. We design recommendation system based on behavioral characteristics in user interaction with an online marketplace and investigate the implications for fair recommendation. We leverage the intrinsic differences based on user behaviors to identify the advantaged groups and disadvantaged, and develop a fair recommendation framework that leverage the decision path from the advantaged groups to help with the disadvantaged groups. By providing high quality recommendations for different types of sellers, we better allocate the resources to the sellers. Our new model that combines the logit and decision tree to better describe the user decision path, increases the interpretability of the results and serves as alternative method for fair recommendation. Our work bridges the gap between the algorithmic fairness and recommender systems. The hypotheses development and testing provide insights on how to improve the online marketplace, and themarket design based on users’ behavioral characteristics.