Discussion of the paper: The peer performance ratios of hedge funds P. Masset 27 Oct. 2016
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments Overall impression The paper proposes a novel method to assess the performance of hedge funds: “peer performance ratios”. This measure … is based on a methodology which appears very solid from a technical standpoint is easy to communicate leads to results that are perfectly in-line with the existing literature (e.g., large funds underperforming smaller funds) leads to better results than alternative performance measures when it is used to predict future performance This paper thus appears as highly relevant for both practitioners and academics. I do however have a few comments and suggestions regarding its structure and some additional tests, which could further strengthen the results.
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments Structure The paper is rather long, some sentences do not appear as essential, there are redundancies, and the structure is not always clear/logical. For instance, the questions at the beginning of section 4 could lead to some ambiguity (do you simply paraphrase the hypotheses or do you introduce new research questions?) I would suggest to drop these questions and shorten the introduction of section 4. I would also suggest to present the dataset in a dedicated section. The link between the literature (discussed in section 2.1) and the hypotheses (2.2) should be made clearer. Actually, I am not even sure that it is essential to present these hypotheses explicitly: at present they tend to take the lead over the genuine core of the paper, which is to propose a novel performance measure (while the primary aim of the hypotheses is basically to validate this measure).
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments Hypothesis 2 According to this hypothesis, large but old funds should perform relatively better than large young funds. You justify this hypothesis by the fact that large HFs with a long track record want to avoid damaging their reputation by “failing unconventionally” – they therefore tend to herd more than other HFs. I am not fully convinced by this argument: First, from a theoretical perspective, I would actually expect the opposite to happen (more herding = less originality & talent involved in the investment process lower performance) Second, the interaction term in the regression is very small and seems to be clearly dominated by the age and size effects. Moreover, I wonder “who” are those young but nevertheless big HFs? They are probably launched by managers that have been quite successful in the past. I therefore believe that it would be useful to control for the “manager” in the analysis.
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments Hypotheses 3a and 3b 3a: you consider a rather short horizon to analyze the predicting power of your performance measure is it really possible to exploit these results? I think that it would be interesting to consider a longer horizon (as a robustness test and/or to relate your results with the question of performance persistence). 3b: You do not justify this hypothesis. Which economic arguments make you believe that “the best of two worlds can be achieved by using the outperformance ratio as a second pass filter in selecting hedge funds”? Have you also tried to reverse the sequence (i.e., use first your ratio and then refine the analysis by using the competing performance measures)? D and S, in particular, could potentially lead to interesting results if used as second pass filters (they seem to be capable of identifying “original” but not necessarily excellent funds - they would thus be nice candidates for a second pass filter).
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments General comments / suggestions (1) Talent: you underline the importance of being able to identify those managers that have a particular talent for finding genuine opportunities on the markets. Given the importance of this issue, I think that it would be relevant to explain in more detail why and how your measure can effectively capture the “talent”. Investment style: you analyze the performance by investment style. It could be interesting to refine this approach by taking into account the exposure of the funds to the various risk factors. Tables: Table 3 is not very clear. Table 5: are the results for the Top and Bottom portfolios sorted on the basis of D and S correct or could they have been inverted? (the very strong negative relation between the performance measure and the returns is a bit surprising). Table 6: the R2s of the multivariate regressions seem to be very low what is the economic significance of these results?
Additional tests & comments Introduction Structure Hypotheses Additional tests & comments General comments / suggestions (2) Figure 3 / Performance results for the quintile portfolios: it would be interesting to see the results for each type of strategy (are they comparable/similar?) The middle panel shows that the difference between the performance obtained when using your measure and competing measures is actually rather small and seems to vary depending on the period under consideration: the difference is clearly positive only since 2009 (during the financial crisis, portfolios based on the relative and peer alpha have delivered a better performance) To what extent could your results be sensitive to the specific period considered in this study? (the period 2009-today, in particular, appears as quite peculiar: the golden age for HFs seems to be gone and it is becoming increasingly difficult for managers to distinguish themselves from their competitors). I like your conclusion in the introduction: it is concise and very effective at outlining the contributions of your work!