Noise-trading, Costly Arbitrage, and Asset Prices: Evidence from US Closed-end Funds

The behavior of US closed-end funds is very different from that of the UK funds studied by Gemmill and Thomas (2002). There is no evidence that their discounts are constrained by arbitrage barriers, no evidence that higher expenses increase discounts and no evidence that replication risk increases discounts—but strong evidence that noise-trader risk is priced. The differences between US and UK funds may be due to the fact that small investors dominate US funds while institutional investors dominate UK funds, or because the sample selection method for the UK funds chooses only funds that are relatively easy to arbitrage. ∗Department of Economics, Vassar College, 124 Raymond Ave. #424, Poughkeepsie, NY 12604. E-mail: [email protected]. Telephone: (845) 437-5209. I would like to thank Osaka University’s Institute for Social and Economic Research for support while I completed the final drafts of this paper. I also gratefully acknowledge the diligent and tireless research assistance of Rebecca Forster. All errors are my own. Vassar College Economics Working Paper # 71 Responding to the invitation of Gemmill and Thomas (2002) in the final paragraph of their paper, I compare their results on UK closed-end funds with those of US closed-end funds. I find that the behavior of US funds is extremely different from the behavior shown by their sample of UK funds. In both countries, closed-end funds operate as actively managed investment companies that do not redeem their own shares at par with portfolio values in the way that mutual funds do. Instead, the shares of closed-end funds trade on stock exchanges where supply and demand determine their prices. The upshot is that their per-share stock prices typically differ from their per-share portfolio values—quite often by large amounts. The resulting discounts (underpricings) and premia (overpricings) have been intensely studied because they are cases where arbitrage pricing and the Law of One Price at least appear to be violated. Numerous papers reviewed by Dimson and Minio-Kozerski (1999) debate whether behavioral explanations or rational factors like fund expenses, mispriced portfolios, or outstanding tax liabilities can do a better job of explaining why fund market values deviate from fund portfolio values. The purpose of Gemmill and Thomas (2002) is to add to the literature by testing competing rational and behavioral explanations using UK data. They focus on the ability of arbitrage to constrain discount and premium levels to be consistent with rational factors as well as whether the discounts and premia of the funds in their sample appear to be affected by the leading behavioral explanation for closed-end fund mispricings, noise-trader risk. As explained by DeLong, Shleifer, Summers, and Waldmann (1990), noise-trader risk is the risk faced by rational traders that irrational “noise traders” may not only cause an asset to be mispriced, but may also, on an ongoing basis, cause any given mispricing to widen rather than narrow. Noise traders can do this because their trading actions are completely unpredictable and noisy. As such, the random trading of the noise traders confronts rational traders with a unique, non-diversifiable risk that deters them from fully and immediately rectifying the mispricings caused by the noise traders. The result