Distribution of closed end fund discounts and implications for portfolio choice

Objective:           Closed End Funds (CEFs) are companies that invest in the securities of other companies for profit.  They have market determined prices that differ from the net asset value (NAV) of the underlying assets and the difference is the discount.  The goal of this study is to determine the distribution that best fits the empirical data of discounts and if that distribution has evolved over time.  The distribution of discounts is important to portfolio managers since the discounts are in fact a major component of returns that can be earned on a portfolio of CEFs. 

Data/Methods:  The sample consists of monthly discounts on US and UK listed CEFs from August 1997 to October 31, 2014. The source of the data is from a propriety database compiled at Gramercy Asset Management. The study will search for the best fit among the Pearson family of distributions using Bayesian methods similar to Markowitz and Usmen (1996). Both the distribution of discount levels and the distribution of discount changes will be estimated.  The estimation will include cross sectional discounts each year for every year in the sample to reveal the parameters that are significant for portfolio choice and how they have changed over time.

Results:  Our sample results so far show that the distributional moments have changed in a number of ways.  With respect to levels the average discount has narrowed over the last 18 years but its volatility has increased recently.  The skewness has steadily declined and has changed from positive to negative.  Kurtosis has remained approximately constant.  As for discount changes, the volatility have been cut in half but kurtosis has approximately doubled.  These changes have significant implications for portfolio managers because they hint that opportunities to invest in CEFs have shifted to those funds with very deep discounts as a result of fattening tails of their distribution.  The study will solidify these important observations in distributional changes by estimating the distributions at various intervals of time.