|Ph.D Student||Sharon Garyn-Tal|
|Subject||Two Essays on Performance Evaluation|
|Department||Department of Industrial Engineering and Management||Supervisors||Professor Reisman Haim|
|Full Professor Simoen Beninga|
|Full Thesis text|
In this dissertation I estimate the fund manager's contribution to the return of her investors, alpha, examining several existing approaches and elaborating from them. The dissertation is divided into two chapters.
The first chapter is co-authored with Beni Lauterbach. In this chapter we document comprehensive and updated evidence on the performance of and correlation between various regression frameworks for estimating "alpha" - the excess return offered by the fund to its investors. Examining over 1000 U.S non-specialized mutual funds in 2001-2009, our main findings are: 1) alpha's magnitude, the fund's classification as good or poor, and the fund's ranking relative to other funds, all strongly depend on the regression framework. The differences are substantial even among somewhat-similar-specification regression models; 2) funds' performance persistence can be traced. However, persistence is strongly dependent on the evaluation period, and sometimes evaporates once more complex (and probably more correct) pricing-models are used as benchmarks; 3) when we compute the net economic alpha (an alpha that takes into account the fact that the alternative of investing in ETFs is also costly), the average net economic alpha across the 2001-2009 period is below 1% per year in absolute value, regardless of the regression framework used. Thus, the net economic alpha correction, employed for the first time in this study, portrays the mean return of the mutual fund industry as only slightly inferior to ETFs. In addition, given that the net economic alpha measure provides a more accurate estimate of the excess return offered to fund investors, it should receive more attention.
In the second chapter I suggest a new methodology for evaluating short-term excess return (short-term alpha). The intuition behind the new methodology is derived from the forward rate calculation and, as opposed to the widely used out-of-sample alpha method, it does not require that the betas remain constant over a short period of time. I compare the new methodology with the out-of-sample alpha method, as well as with the one-year regression method, and _nd substantial score and ranking differences between the short-term estimates. I analyze the short-term alpha estimation methods based on aspects of expected value and variance and conclude that the new methodology is the better one. Simulation tests also support this result. Finally, the new methodology also yields performance scores and rankings that are the most consistent with their long-term counterparts.