We exploit the impact of hurricane Katrina on insurance companies to study the relationship between bondholder concentration and credit risk. Redemption-driven sales by property and casualty (re)insurance companies exposed to hurricane Katrina are associated with a large drop in bondholder concentration faced by corporate bond issuers. Exploiting this shock to capture exogenous variation in bondholder concentration, we find that greater bondholder concentration is associated with higher bond yield spreads, as well as with firm characteristics associated with credit risk.

Using Internet search volume of dividend-related keywords to measure investor preference for dividends that varies over time and across states, we show that dividend sentiment affects corporate policies and asset prices. Investors search more for dividends when economic conditions are poor, with the peak volume reached during the recent COVID-19 pandemic. Firms initiate or increase dividends when dividend sentiment is stronger, especially in regions with strong dividend sentiment. Shifts in dividend sentiment predict higher investor demand for dividends and higher returns for high dividend stocks. Further, mutual funds that pay high dividends receive more inflows when dividend sentiment is stronger.

Investors' individual arbitrage models introduce idiosyncratic risk into complex asset strategies, driving up average returns and Sharpe ratios. However, despite the attractive risk‐return trade‐off, participation is limited. This is because effective Sharpe ratios in complex asset markets vary with investors' expertise. Investors with higher expertise, better models, and lower resulting idiosyncratic risk exposures realize higher Sharpe ratios. Their demand deters entry by less sophisticated investors. As predicted by our model, market dislocations are characterized by an increase in idiosyncratic risk, investor exit, and persistently elevated alphas and Sharpe ratios. The selection effect from higher expertise agents' more favorable Sharpe ratios is unique to our model and key to our main results.