| Closed-end funds often trade at prices that are substantially less than the values of their portfolios. These discounts are positively correlated, and the average discount exhibits significant time-series variation. Previous research has failed to support a rational basis for this anomaly. De Long, Shleifer, Summers, and Waldmann (1990) develop a model of noise traders in financial markets that offers an explanation for the pricing of closed-end funds. In this model, noise traders make systematic forecasting errors related to investor sentiment, and rational investors face impediments to profiting from these errors. The risk from noise traders causes closed-end funds to trade at a discount, which varies with investor sentiment.; The first essay in this dissertation directly tests the noise trader model using transactions data on a sample of 23 domestic equity closed-end funds. In contrast to the predictions of this model, the relative order-flow imbalance caused by individual investors does not have a statistically significant effect on large daily changes in discounts. Trading activity associated with institutional investors is strongly associated with these fluctuations. This is the first study to directly attribute discount fluctuations to the trading by a particular clientele. In addition, this research shows substantial discount changes have occurred on a daily basis, and large changes typically occur when trade size, trade frequency, and share volume increase.; Unlike closed-end funds, exchange-traded funds can be arbitraged by institutional investors through an in-kind process of share creation and redemption. The second essay in this dissertation examines the relation between discounts and arbitrage for World Equity Benchmark Shares (WEBS), a series of 17 exchange-traded funds listed on the American Stock Exchange. The analysis of WEBS leads to several important observations. The average premium on WEBS is quite small, but arbitrage fails to eliminate occasional periods of premium persistence. In addition, significant cross-sectional differences in premium levels exist. The transactions data do not support the hypothesis that noise traders influence the premiums on WEBS. |
