In the first chapter of this dissertation, I describe a method for representing institutional investors’ portfolio holdings as a graph, in which funds connect to stocks through patterns of common ownership. I then demonstrate that changes to a firm’s position within this network are closely related to future stock market performance. Specifically, stocks moving toward the center of the holdings network outperform those drifting toward the periphery by approximately 4.1%, annually, adjusting for standard risk factors, consistent with a model in which short-sale constraints combined with increasing dispersion in investors’ beliefs signal potential overvaluation. After controlling for a number of additional variables, including the “breadth of ownership” measure proposed by Chen, Hong, and Stein (2002)—a local indicator of a firm&rsquos network importance—stocks with the largest decrease in holdings network centrality still underperform by 2.2% per year.

In the second chapter, using a novel data set consisting of Schedule 13D filings and amendments over a seven-year period, from 2003 to 2010, I present evidence that managers of large investment portfolios exploit periods of perceived investor distraction to minimize the adverse impact of the disclosure of large sales on future transactions. Specifically, managers reporting substantial decreases in holdings favor Friday disclosure over disclosure on other weekdays, and prefer to release the news in the hours after markets close. Moreover, investors who go on to make future sales are significantly more likely to pursue an opportunistic filing strategy. Employing event study methodology, I test for underreaction to Friday filings, but find no support for investor inattention to Friday 13D disclosures. Investors seem to rapidly incorporate available information from regulatory disclosures into stock prices, correctly attributing heavy selling to liquidations and not informed trading.