Abstract: The closed-end fund (CEF) puzzle, which focuses on the difference between the net asset value (NAV) and the CEF price, has been the focus of academic research for decades. This dissertation includes both theoretical and empirical work studying the existence and fluctuations of equity CEF discounts. Chapter Two sets up an equilibrium asset pricing model for the global CEF. Assuming no communication between the NAV country and the CEF country, the equilibrium NAV and CEF prices are characterized separately in closed forms. In equilibrium, the NAV only depends on the current dividend value and the subjective discount rate. However, the equilibrium CEF price also depends on future dividend dynamics until the open-end date. Furthermore, comparative sensitivity analyses and simulations show how the CEF discount fluctuates with different parameter configurations. In conclusion, this chapter finds three factors important to understand the CEF puzzle: (1) The equilibrium interest rate is partially exogenous with respect to the CEF; (2) The NAV portfolio cannot perfectly be hedged by the market assets space; (3) The total number of outstanding CEF shares is fixed in the financial market. Chapter Three investigates the correlation between the NAV return and the CEF discount in global CEFs by estimating the convenience yield using a two-factor dynamic state space model and constructing liquidity risk measures combining the work of Datar, Naik, and Radcliffe (1998) and Amihud (2002). The empirical results show that the CEF discount is negatively correlated with the NAV return but positively correlated with the liquidity risk measure, which is consistent with Chan, Jain and Xia (2008). Chapter Four proposes a theoretical model for the CEF discount change on the open-end announcement using a representative agent portfolio optimization model. Before the announcement, the CEF discount dynamic follows a mean-reverting process. After the announcement, the discount dynamic is modified to converge to zero on the open-end date, which changes the budget constraint of the portfolio optimization. Assuming that the portfolio holding of CEF shares is unaffected by the announcement, the discount change on that date can be solved numerically. Empirical applications also present positive supports for the proposed explanation.