This thesis focuses on the role of information and beliefs in finance. In particular, it explores how to interpret data generated as an outcome of market equilibrium. The era of digitization has made data and analytical tools increasingly available. Consequently, it is essential to critically consider how data is generated in specific contexts to avoid misinterpreting results obtained from generalized statistical or algorithmic methods. Data sources and information technology are key concepts of the first chapter. Investors face the challenge of approximating the high-dimensional prediction function for the pay-off of a risky asset. Optimal decision-making involves choosing a biased estimator to lower variance, as well as excluding data sources to manage complexity. Historical returns appear predictable due to improved technology not available at the time. The second chapter formalizes the problem of regulating dividend payments in the banking sector, where asymmetric information leads to strategic deviations from the optimal pay-out policy. Specifically, the information concerns an imminent shock that threatens only a subset of banks, and the model explores the link between the size and scope of this shock and the opportunities and pitfalls associated with regulatory measures. The third chapter investigates sentiment analysis of social media posts as a novel source of data for understanding investor beliefs. However, it is important to recognize that communication on social media is itself an equilibrium outcome. Interactions on social media between investors with differing opinions allows for an equilibrium with information sharing. This equilibrium aligns better with empirical patterns observed in the literature than an alternative scenario where sentiment solely reflects misguided beliefs.