On understanding mutual fund terminations
Journal of Economics and Finance, Jan 2009 by Bu, Qiang, Lacey, Nelson
Published online: 25 January 2008
© Springer Science Business Media, LLC 2007
Abstract
We examine the determinants of US mutual fund terminations and provide estimates of mutual fund hazard functions. We find that mutual fund termination correlates with a variety of fund specific variables as well as with market variables such as the S&P; 500 index and the short-term interest rate. We also test the underlying assumptions of the semi-parametric Cox model and reject proportionality, thus calling to question the use of this model in forming estimates of mutual fund hazard functions. We find that different fund categories exhibit distinct hazard functions depending on the fund’s investment objectives.
Keywords Mutual fund termination * Market condition * Proportionality * Cox model * Hazard function
JEL Classification G11
1 Introduction
While previous research tells us much about mutual fund performance, we still do not know very much about fund terminations. This paper examines terminations of mutual funds through the hazard function so as to better understand the mutual fund industry. For example, if poorly performing funds tend to be terminated after only a short time period, then knowledge of termination determinants will allow investors to better time their mutual fund investments. On the other hand if terminations tend to occur only after longer protracted periods of time, then short-term performance signals are weak and not as useful.
Brown and Goetzmann (1995), the forerunners of mutual fund termination studies, found that US mutual fund disappearance is a function of lagged relative returns, relative fund size, fund expenses, and fund age. According to their results, higher lagged relative returns and larger relative fund size reduce a funds’ termination probability, while higher expense ratios increase the chance of fund closure. Moreover, their test results show that younger funds have a higher probability of disappearance. Brown and Goetzmann (1995) also examined the interaction effects between variables but failed to find significance. They concluded that fund managers are mainly concerned about growth and therefore that poor track records are strong predictors of fund termination. They also found that poor performance in the fund’s early years is not as powerful a factor in fund disappearance compared with more recent performance. In the Brown and Goetzmann study lagged predictors are defined as 1 year or longer and thus cannot capture termination in reaction to shorter term (less than 1 year) movements in macroeconomic variables.
Lunde et al. (1999), studied UK unit trust terminations using the semi-parametric Cox model and found that termination probabilities are negatively related to fund returns and that higher average returns in a fund category reduce the termination probabilities of funds in that category. LTB found an inverse U-shaped hazard function such that young and very old funds were least likely to close, and attributed closures to a learning process where investors extract information concerning poor fund performance. Funds that survived the so-called learning period were more likely to be perceived as possessing a good track record and thus have a low hazard rate. Two implications of the LTB study are worthy of note. First, the Cox model assumes proportional hazard which could be violated if the independent variables, or covariates, are interactive with time. The Cox model assumes a multiplicative relationship between the underlying hazard function and the log-linear function of the covariates, which is also called the proportionality assumption. In practical terms, it is assumed that given two observations with different values for the covariates, the ratio of the hazard functions for those two observations does not depend on time. The Cox model is also known as proportional hazard model due to the underlying proportionality assumption