Arbitrage in Closed-end Funds: New Evidence

Arbitrage pressures that could equalize closed-end fund share prices with fund portfolio values appear to be largely absent in an extensive data set. Observed fund behavior violates the static arbitrage bounds of Gemmill & Thomas (2002) and is inconsistent with the dynamic arbitrage bounds of Pontiff (1996). Furthermore, Fama & French (1992) regressions run on arbitrage portfolios designed to pro t from closed-end fund mispricings generate excess returns that are either signi cantly negative or insigni cantly different from zero, suggesting that arbitrageurs lack a pro t incentive. If arbitrage is absent, observed fund pricing behavior likely re ects changing investor sentiment about fund prospects. ∗Department of Economics, Vassar College, 124 Raymond Ave. #424, Poughkeepsie, NY 12604. [email protected] Mea cuplpa. Using a data set that contains nearly every closed-end fund trading in the United States and Canada over the period 1985-2001, this paper nds strong evidence that arbitrage pressures are weak to nonexistent in closed-end funds. The 462 closed-end funds exhibit behavior that stands strongly in violation of the static arbitrage bounds said to constrain fund discount and premium levels by Gemmill & Thomas (2002). In addition, discounts and premia also violate the dynamic arbitrage bounds that Pontiff (1996) argues vary with interest rates and the opportunity costs of undertaking arbitrage activities. Most importantly, if you estimate Fama & French (1992) three-factor regressions on the excess returns of portfolios that try to pro t from the mispricings of closed-end funds, you nd negative risk-adjusted returns in virtually all cases, suggesting that rational traders lack a pro t incentive. This would explain why, empirically, the mispricings found in closed-end funds often persist seemingly inde nitely and why they appear to move largely without reference to fundamental factors. Simply put, with rational traders absent, the market is dominated by sentimental noise-traders who move prices unpredictably and without reference to fundametals, as suggested by Lee, Shleifer & Thaler (1991) and Chopra, Lee, Shleifer & Thaler (1993). Closed-end funds—mutual funds whose shares trade like stock—have presented a puzzle to nancial theorists for decades because they constantly violate the law of one price. In the United States, closed-end funds are required by law to publish the contents of their portfolios weekly. In addition, many funds now present daily updates about their portfolio values on their websites. Consequently, investors have very precise and timely information about fund portfolio values. Given that closed-end funds issue only a xed number of shares, this should imply that, fund expenses aside, the price of each share should simply be the total portfolio value divided by the number of shares outstanding. Unfortunately, closed-end fund share prices often differ hugely from portfolio values. What is more, these mispricings often last for years at a time, so that there appears to be little pressure for prices to return to par with portfolio values.