This dissertation consists of three papers on the portfolio choices of retail and institutional investors.

The first chapter estimates the effects of product market advertising on household portfolio choice. Product market advertising by raising the awareness of a company's brand is thought to increase the demand for a company's stock as well as its products. I construct a dataset of publicly traded sports sponsors in the US and develop an instrument for investor exposure to advertising via these sponsorships. I show that investors living in a city where local sports teams are sponsored by a given company, local or non-local, are more likely to purchase stocks in that company. The portfolio effects from sports sponsorship are large and suggest that advertising is more important than even local bias.

The second chapter, co-authored with Harrison Hong and Jiangmin Xu, revisits the local bias in household stock investments. Households hold undiversified stock portfolios of firms headquartered near their city of residence. Leading explanations such as the familiarity heuristic assign a causal role to proximity. However, households optimally locate in a city based on unobservables such as optimism about a city's prospects, which can be correlated with latent local-stock demand. Using location choice models, we develop an instrumented Heckman (1977)-style correction for widely-used regressions to show that local bias is also driven by this selection. We then propose a decomposition of local bias into household priors about stocks, using location-choice model residuals, and familiarity, whereby confidence about stock-payoff signals rise with proximity.

The third chapter, also co-authored with Harrison Hong and Jiangmin Xu, estimates the costs of producing information in financial markets. Information in stock markets is often produced by sell-side analysts working at banks, whose research is bought by institutional investors. We estimate the costs and benefits of this information by modeling the effect of analyst coverage on institutional stock portfolios. We quantify two distinct channels: coverage drawing managerial attention to stocks versus improving the precision of managerial signals. We can then calculate the marginal benefit of information as the elasticity of aggregate institutional demand for a stock to the banks allocating one analyst to that stock. Assuming convex cost functions for coverage, we can then measure the marginal cost producing information for each stock.