A Generalized Approach to Measure Market Timing Skills of Fund Managers: Theory and Evidence

Introduction

Among the theoretical models on fund manager performance, Merton’s (1981) dual beta model is widely used for testing the market timing ability of fund managers. It should be noted that market timing in this paper does not refer to the late trading controversy. The Merton model is popularized through the Henrikkson and Merton (1981) empirical implementation. This model is transformed into different variations and used even in the recent times. For an example, see Ferson and Schadt (1996) who use a conditional version of Henrikkson and Merton (1981). Merton’s model is a response to the single beta or constant risk model of Jensen (1968) which has been recognized to be overly simplistic (see Jensen (1972), Lee and Rahman (1990), Treynor and Mazuy (1966) and Grinblatt and Titman (1989)) since informed money managers are apt to change their portfolio risks in response to market conditions, for instance see Ferson and Schadt (1996) who control for publicly available predetermined information variables in their models. In Merton’s model, money managers change the risks of their portfolios depending on their information about the future return on the market relative to the risk-free security. The market timing skills of money managers is assessed by measuring whether the managers anticipate the market return and change the portfolio risk accordingly.

Lee and Rahman (1990) describe the information structure in Merton’s model as being too coarse and the dichotomy of market returns as either outperforming or underperforming the risk-free rate as unsophisticated. They develop a model using a continuously varying noisy information signal about the future market return and arrive at a quadratic regression similar to that used by Treynor and Mazuy (1966).

Recent applications of Merton’s model can be found in Ferson and Schadt (1996), Bollen and Busse (2001)Chance and Hemler (2001), and Farnsworth et al. (2002). Ferson and Schadt (1996) extend the conditional CAPM framework to performance evaluation. In this framework, the risk of a portfolio may change over time due to changes in the public information structure as well as management skill in anticipating market conditions. The conditional CAPM framework for performance evaluation has been criticized by Kryzanowski et al. (1997) and Chance and Hemler (2001). Kryzanowski et al. (1997) argue that if fund managers make their investment decisions based strictly on the information set, the Jensen’s alpha of the portfolio will be zero. Thus the model cannot capture the abnormal performance of fund managers that can be attributed to their real-time market timing skills.