Stock vs. bond closed-end funds: The investor sentiment theory revisited

Closed-end investment funds (CEFs) hold portfolios of various financial assets. If the efficient market hypothesis holds, share prices of a CEF should equal the net asset value (NAV) of the underlying portfolio.  However, in reality most CEFs are traded at a discount to NAVs. This anomalous pricing of CEFs has been puzzling economists for decades. Our objective is to make a contribution to this ongoing debate, and specifically, take a new look at the investor sentiment theory and its role in explaining CEFs’ discounts. We are especially interested in investigating whether CEFs discounts behave differently during tranquil and turbulent periods.

One of our contributions to the literature is that we separate CEFs by their type. We decided to look at stock and bond CEFs and their discounts separately after we noticed some differences in funds’ behavior depending on their type and since we had daily returns data available for 16 bond and 16 stock CEFs for the period of May 7, 2004, through February 17, 2011. Instead of investigating the dynamics of each fund, we focus on the movement of the common factor, f_t, which might explain discounts, d_i,t, of all funds of the same type (i=1 or 2, depending whether it was a stock or a bond CEF): 

                                                                                   d_i,t =λ_if_t + ς_i,t

We allow factor loadings, λ_i , to vary across funds and also include an idiosyncratic component, ς_i,t, for each fund i. Once the common factor, f_t , is identified, we estimate its dynamic conditional correlation (DCC) with an investor sentiment variable (using Chicago Board Options Exchange Market Volatility Index (VIX) as a proxy). Employing Engle’s DCC estimator for MGARCH models is another contribution that we make to the existing literature on CEFs’ discounts. All of the existing papers assume that the relationship between CEFs’ discounts and investor sentiment is constant over time and, therefore, can be estimated with the constant conditional correlation (CCC) estimators.

The empirical results of our research confirm our intuition that discounts on stock CEFs are formed in a different way than bond funds’ discounts. First, we find strong evidence of CCC between VIX and stock CEFs’ discounts, while the correlation between VIX and bond funds’ discounts varies over time, increasing especially during the crisis of 2008. Second, we uncover a new puzzle about stock funds’ discounts. We estimate the correlation between stock funds’ discounts and our investor sentiment proxy (VIX) to be negative, contrary to our intuition. Further investigation is needed.