Hedge funds, fund attributes and risk adjusted returns

See for example the U.S. Securities and Exchange Commission (www.​sec.​gov).

Edwards and Caglayan (2001) find some types of funds having significant and positive relationships between fund age and performance.

There were a total of 49 observations that fell outside the three standard deviations interval. When these outliers were removed, the Kolmogorov-Smirnov Z test still pointed to non-normally. The test statistic did, however, improve from 7.80 to 5.23. We thank the anonymous referee for this comment.

Agarwal et al. (2009) test how managerial incentives such as a hurdle rate and high-watermark affect hedge funds performance. They find positive coefficients of hurdle rate and high-watermark without directly controlling for performance fee in their models.

http://​www.​ft.​com/​cms/​s/​0/​cf7f91e2-f3f0-11dd-9c4b-0000779fd2ac.​html. Financial Times, “Hedge fund investors have a great chance to cut fees”, James Mackintosh, 6 February 2009.

Following the suggestions of the referee, the Durbin-Wu-Hausman test was conducted to investigate possible endogeneity. The endogeneity tests were run using “msratio” and “mreturn” as instrumental variables. The test results indicated that there is no endogeneity in the postulated relation between the hurdle rate and high watermark variables. Upon further examination of our data, the number of funds in the sample offering a hurdle rate and/or high watermark points us to the same conclusion, i.e. that there is no endogeneity. As we explain in the first paragraph on page seven, 855 funds (24%) offer a hurdle rate and 2974 funds (83.3%) offer a high watermark rate. 759 funds (21.3%) offer both hurdle rate and high watermark and 501 funds (14%) offer none of the two. If a fund manager happens to offer both hurdle and high watermark rates, then endogeneity could be a specific issue in this category. However, in our sample only 21.3% offer both thus partially mitigating the issue of endogeneity. Therefore, it would be difficult to conclude that hurdle rates can explain high watermark rates. Even if a problem of endogeneity existed, the proportion of funds offering both compared to the rest of the sample is such that the overall results of the study would not qualitatively differ. We also conducted the same Durbin-Wu-Hausman test using a different instrumental variable (performance fee) to attempt to verify the presence of endogeneity. However, the use of “performance fee” as an instrumental variable appeared to be inappropriate because the coefficient belonging to the variable is statistically significant in the model specified in Table 4. If the variable was to be excluded from the model, the error term would be correlated with “performance fee.” Thus, it would violate the conditions to be satisfied by instrumental variables. Lastly, correlation tests were conducted between ‘high watermark’ and the residuals from the model in Table 4, and between the hurdle rate (and high watermark rate) and the residual from the model expressed in Table 4. There appeared to be no statistically significant finding pointing to endogeneity. The coefficients themselves were also very close zero, supporting our conclusion with respect to endogeneity.

One may question whether due to some variation in the hurdle rates, some type of censored regression may be more appropriate than a logit model. Our re-examination of the data revealed that the hurdle rates are fairly homogenous such that approximately half of the funds that use a hurdle rates use a standardized benchmark across the funds (e.g. libor and eurobor) and the other half offer hurdle rates within a relatively narrow interval such that it would be acceptable to treat all hurdle rates the same (dummy = 1).

Before removing outliers, hurdle rate was statistically significant and negatively associated with fund performance. However, in examining the new results it is no longer significant.