Searching for mutual fund winners? the strategy is to outbid both, the benchmark and the peer group

ABSTRACT

Standard Fama-French-Carhart models define ‘winners’ as funds that generate the highest excess returns given the factor risks involved; however, they do not provide information on whether such winners are outperforming their prospectus benchmark or their peer group. In addition, existing literature relying on these models, by and large, does not find evidence of persistence in performance. In this paper, we propose a two-stage procedure that allows investors to select ‘true’ winners(losers) which generate the highest factor-risk-adjusted performance relative to the benchmark and the peer group simultaneously. Utilizing both adjustments at the same time results in a strong predictive ability, leading to a selection of funds that persist in performance. Our true winner funds have statistically significant superior benchmark-adjusted alphas, peer group adjusted alphas and Sharpe ratios one year ahead, which are significantly different from those generated by the true loser funds. The results are robust to extended investment horizon, and alpha estimation method, and they are not driven by outliers, size of fund-sorts, or any particular period within our sample.

KEYWORDS:

• US equity mutual funds
• benchmark-adjusted alphas
• peer-group-adjusted alphas
• performance ranking

JEL CLASSIFICATION:

• G11
• G12
• G2

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ As per Mateus et al. (2019c), both methodologies in 1) and 2) provide qualitatively and quantitatively similar results, with method 1) by Angelidis et al., (2013) approach being more commonly tested in the literature, hence adopted in this paper.

² Morningstar categorizes US funds according to their portfolio holdings into five ‘global categories’ (Large Cap Value, Large Cap Growth, Large Cap Blend, Mid Cap and Small cap).

³ The Small Cap fund category is not included as they do not satisfy our minimum funds requirement per performance quartile, as outlined in the methodology.

⁴ In this paper we do not aim to search for any variables omitted from the standard factor models..

⁵ The benchmarks for each Morningstar global category are presented in the Table 1.

Table 1. Final sample.

⁶ All factor data is obtained from Kenneth French’s website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html..

⁷ https://www.cfauk.org/-/media/files/pdf/pdf/5-professionalism/3-research-and-position-papers/benchmarks-and-indices.pdf (Accessed: 22/10/2022).

⁸ We require minimum 30 months of returns in each of the rolling windows.

⁹ Detailed results of significance of these differences are available on request from authors..

¹⁰ Significance of the differences in Sharpe ratios, AGT or APB alphas is available on request..

¹¹ Evidence from the industry supports this notion, see for instance : https://www.westernsouthern.com/touchstone/insights/the-unique-positioning-of-mid-cap-investing (Accessed: 22/10/2022).

¹² Where outperformance is found in both top and bottom quartile. Horst et al. (2001) claim that funds that take risks unaccounted for by Jensen’s model are the ones that perform better if they survive 36 months post-ranking, creating a U-shaped pattern or performance across portfolio octiles. However, when applying Carhart four-factor model, the look-ahead bias in their study is found to be negligible.

¹³ Where outperformance is found in the bottom quartile..

¹⁴ The list of false winners/losers is available upon request.

¹⁵ The significance of the differences is less pronounced in the bottom quartile.

¹⁶ We also include year dummies and results are qualitatively similar. Tables are available upon request.

¹⁷ Please note that if we have simply selected top (bottom) 10% of all funds each year to form our top (bottom) deciles, some of our winner (loser) deciles could be populated by a single style.

¹⁸ In addition, results are not sensitive to the impact of outliers: we winsorize the data in each quartile, each year, and each style by trimming the outliers using a 5% and 95% cut-off point and find the same results from the repeat analysis..