**Amit Goyal, ***Emory** University*

**Sunil Wahal, ***Emory University*

We examine the selection and termination of investment managers by plan sponsors. We build a unique dataset that comprises hiring and firing decisions by approximately 3,600 plan sponsors over a 10-year period from 1994 to 2003. We find that plan sponsors hire investment managers after these managers earn large positive excess returns up to three years prior to hiring. Despite general persistence in investment manager returns, this return chasing behavior does not deliver positive excess returns thereafter; post-hiring excess returns are indistinguishable from zero. Plan sponsors terminate investment managers after underperformance but excess returns after being firing are frequently positive.

**Warren Bailey, ***Cornell University*

**Haitao Li, **Cornell University

**Xiaoyan Zhang, ***Cornell University*

We provide a comprehensive empirical analysis of hedge performance using the stochastic discount factor (SDF) approach. We explicitly take into account the no-arbitrage restriction on asset pricing models, which ensures appropriate valuation of derivatives and dynamic trading strategies used by hedge funds. Using the SDFs of a wide variety of asset pricing models, we evaluate the performance of hedge fund portfolios sorted on styles and characteristics. Without the no-arbitrage restriction, a few models are able to explain the hedge fund returns. With such restriction, these models fail to explain the returns on a couple of style portfolios.

Link to PDF File: Option Valuation with a Multifactor Term Structure of Volatility

**Peter Christoffersen, **McGill University

**Steve Heston, ***University of Maryland*

**Kris Jacobs, ***McGill University*

The substantial academic literature on theoretically sound improvements to the seminal Black-Scholes model has left industry practice largely unaffected. Option traders recognize the systematic biases in the Black-Scholes model, but typically rely on non-theoretical curve fitting methods to match observed implied volatility surfaces. Standard stochastic volatility models such as Heston (1993) reduce the biases of the Black-Scholes model, but are unable to remove these biases completely. The objective of our study is to reduce the gap between industry and academia by establishing and implementing a parsimonious two-factor stochastic volatility model. The model is empirically tractable, in that the option valuation formula is available in closed form, and relatively parsimonious.

**Joel M. Vanden, ***Dartmouth College*

Today there is considerable empirical evidence suggesting that at least some options in the economy are non redundant assets. Furthermore, it is well known from the work of Kraus and Litzenberger (1976), Harvey and Siddique (2000), and Dittmar (2002) that higher order moments such as coskewness and cokurtosis are important for explaining risky asset returns. The goal of this research project is to investigate the intersection of these two lines of literature by showing how the market coskewness and market coskurtosis models are altered when a non redundant option is optimally traded. Due to the non redundancy of the option, it is shown that the economy’s stochastic discount factor depends not only on the market return and the square of the market return, but also on the option return, the square of the option return, and the product of the market and option returns. This leads to an asset pricing model in which the expected return on any risky asset depends explicitly on the asset’s coskewness with option returns. The empirical results suggest that option coskewness may capture some of the same risks as the Fama-French factors SMB and HML. Furthermore, the option coskewness model outperforms several competing benchmark models. These results suggest that the factors that drive the pricing of nonredundant options may also important for pricing risky equities.

**Gary Gorton, ***University of Pennsylvania*

Geert Rouwenhorst, *Yale University*

Commodity futures are one of the oldest asset classes, yet little is know about commodity futures. The main reason for this appears to be a lack of data. There are no long time series of commodity futures returns; no benchmark index that extends back very far, and no panel data of significant length. Existing indices are short or cannot be reproduced (because of lost data over the years). We propose to build a comprehensive data set of commodity futures returns in order to study basic properties of commodity futures and to test hypotheses about commodity futures. We have already built a data set of commodity futures monthly returns back to 1959 for commodity futures traded on the CBOT, CME, and LME. See our attached paper “Facts and Fantasies about Commodity Futures.” The plan for the proposed project is fourfold: (1) for the current data set that we have constructed we want to fill in the contracts that were traded in the past, but that are not traded now (these were omitted from the source data that we used); (2) we want to extend the commodity futures return series back in time — if possible to the start of futures trading in the US in the 1850s; (3) we want to recalculate and investigate the stylized facts that we have computed in the current working paper; (4) we want to test hypotheses concerning why futures markets exist.

Relatively little is known about commodity futures as an asset class. We believe that the construction of a data set that has long times series, the documentation of basic stylized facts about commodities, as well as information in the form of tests of hypotheses about the existence of these markets, will be very valuable for the investment community.