Luck versus skill over time: time-varying performance in the cross-section of mutual fund returns

ABSTRACT

Using returns histories spanning January 1984 to October 2014 of 5785 actively managed US closed-end equity mutual funds, we address the ‘thorny problems’ highlighted by Fama and French (The Journal of Finance, 2010, vol. 65, p. 1925) that arise due to their resampling procedure. This prevents them from capturing time variation in the parameters of equilibrium asset pricing models. These problems are addressed by combining innovative procedures which allow for testing of multiple break dates on fund-specific parameters along with cross-section bootstraps that remain valid in the presence of time-varying parameters. We find that substantial proportion – 8% – of the estimated versions of the asset pricing model have significant changes in their parameters. The effects of this time variation on the cross-section distribution of the risk-adjusted performance measure are significant and substantially increase centiles of the right tail of this distribution when compared to those produced without time-varying parameters. Our evidence regarding the lack of actively managed US equity mutual funds that generate excess returns is significantly weaker than those of Fama and French but our results do not overturn their pessimistic conclusion regarding the lack of skilled managers. We do find, unlike Fama and French, that managers generating negative returns are just unlucky but have no skill.

KEYWORDS:

• Mutual funds
• capital asset pricing model
• CUSUM test
• linear regression models
• stochastic processes
• U-statistics
• bootstrap

JEL CLASSIFICATION:

• C3
• C4
• G1

Acknowledgements

The authors would like to thank Georgios Chortareas, Keith Pilbeam and Joscha Beckmann for their many valuable comments. The authors acknowledge the helpful comments of attendees at the Fourth International Conference of the Financial Engineering and Banking Society 2014 at the University of Surrey, England, and at the First Conference on Recent Developments in Financial Econometrics and Applications 2014 at Deakin University, Australia and the European Economic and Finance Society 2016 Conference in Amsterdam.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Cuthbertson, Nitzsche and O’Sullivan (2008, 618).

² Adjust-closed price includes adjustments for splits, rights offers and dividend payments.

³ We can allow for non-stationary $Y_t$ and ${\mathbf{X}}_t$ as long as (2) represents a cointegrating relationship among the non-stationary variables.

⁴ $\rho =0$ if and only if the distribution of the errors in our Equation (2) is symmetrically distributed about 0.