This dissertation uses novel data to provide micro-level perspectives on the behavior of fund managers and investors. Chapter 1 examines the effect of a trader's personal co-investment or skin-in-the-game on fund risk-taking. Using a unique dataset from an online social trading platform, I uncover a source of exogenous variation in trader's skin-in-the-game to investigate the causal effect of skin-in-the-game on fund risk-taking. I find that having no skin-in-the-game significantly increases trader's incentive for risk-taking. The findings provide evidence in support of skin-in-the-game as an important mechanism to align incentives of traders and investors. Chapter 2 studies individual fund investor extrapolation. In particular, I examine fund investment and withdrawal events at individual investor level. The sample is constructed from brokerage account data of global retail forex investors from the social trading platform. The findings provide evidence of fund investor extrapolation, in which the past performance consistency of fund investment has a significant effect on investor's withdrawal decision. This effect is more pronounced for investors from more developed countries. The results also highlight that investor's withdrawal decision depends not only on past performance and volatility but also on the consistency of past performance. However, none of these factors positively predicts future performance. These results support the view that fund investors over-extrapolate as they tend to extrapolate based on past performance measures and do not profit from doing so. As U.S. adults increasingly obtain news through mobile devices rather than desktop computers, Chapter 3 compares "mobile sentiment" with "desktop sentiment" in predicting future stock returns and liquidity. I construct unique data scraped from Google, which sometimes produces very different results on mobile vs. desktop search due to different ranking practices (e.g. a link with text consisting of negative words about a stock is shown on mobile but not on desktop). Thus, I collect daily Google search results separately on mobile and desktop platforms for tickers of stocks in the S&P 500 index. I conduct textual analysis on the search results. I find that negative mobile or desktop sentiment predicts abnormal return reversal in the following week with mobile serving as a more significant predictor than desktop. I show that this reversal is mainly driven by stock over-pricing. That is as investors become more optimistic due to recent good news, the stock is over-priced and will later revert back to its fundamental value (i.e. lower future returns), whereas as investors become more pessimistic due to recent bad news, the stock is unlikely to be under-priced and have reversal. In addition, the effect of mobile sentiment on returns becomes more pronounced than desktop sentiment in stocks of high retail interests. This supports the idea that going mobile is a preferred way to obtain trading information among less sophisticated investors. I find weak evidence that mobile sentiment relates more to liquidity measured by effective spreads and volume than desktop sentiment. The results also suggest that sentiment is mutually Granger-causal with either return or liquidity. In the end, my results highlight the growing relevance of mobile media in disseminating financial news, and provide suggestive evidence that compared with desktop computer users, mobile users are less informed and more akin to sentiment investors.