This study approaches the issue of hedge fund return persistence from two interesting angles: the selected funds invest a significant portion of their assets in Asian countries, and the Kolmogorov-Smirnov multi-period test of persistence is used (the only previous paper to use this method to test hedge fund return persistence is by Agarwal and Naik (2000)).

The databases used, provided by EurekaHedge Advisors and HedgeFund Intelligence AsiaHedge, cover the period from January 1999 to March 2003, and list post-fee returns. Using management fees and performance fees, pre-fee returns are calculated to test the persistence of pre-fee and post-fee returns. Because this dataset focuses on Asian hedge funds, the degree of overlap with the most used databases is low.

Single-period tests of persistence are conducted via a contingency table of winners (funds that beat the median) and losers. These tests are the Cross-Product Ratio (CPR) and the Chi-square test. According to Carpenter and Lynch (1999), the Chi-square test is well-specified, powerful and more robust to the presence of survivorship bias compared to the CPR method. Formation and holding periods are successively 1, 2, 3, 6, 9 and 12 months. The results produced by the two methods are similar: winners and losers exhibit significant and positive return persistence at the 1% level for horizons up to and including 9 months. The pre-fee return persistence is stronger. Nevertheless, at a 12 month horizon, no evidence of persistence is found. This is consistent with several previous studies on this topic where the persistence level diminishes as the time-horizon lengthens.

The Kolmogorov-Smirnov multi-period test of persistence is also used. It is particularly convenient in the context of non-normal distribution of returns, which is one of the main characteristics of hedge funds. Pre-fee and post-fee returns of winners and losers exhibit significant and positive return persistence for the 1, 2 and 3 month horizons. It remains significant for winners at the 6 and 9 month horizons. This corroborates the two-period test results: at a 12 month horizon, no evidence of persistence is found, and the persistence level decreases as the time-horizon lengthens. While winners persist as much as losers with pre-fee returns, winners are more persistent than losers with post-fee returns. This can be explained by the fact that the performance fee is only paid if the hurdle rate is reached.

To adjust their measures for style, the authors separately examine the persistence of funds with long/short equity, macro, relative value and multi-strategy styles. All the test methods indicate similar results to those obtained with the whole sample.

Return persistence can be generated by the systematic risk or by the stock picking skills of the manager. In order to examine whether the systematic risk can explain all of the persistence, 7 factors are selected: two Asian equity factors, an Asian bond factor, a US equity factor, and Fama and French ‘s factors (size, book-to-market and momentum). The funds are given a monthly rank in deciles based on their return in the last sub-period. The ranking is made with pre-fee and post-fee returns. The risk-adjustment uses the US Treasury rate. The regression of the return series on the 7 factors displays an R-squared of about 45%. The Jensen alpha (in post-fee returns) between the highest return decile and the lowest return decile is statistically positive at 2.65 % per month. Since the mean excess return of the spread is 2.01% per month, this suggests that the systematic risk cannot explain all of the persistence, and also that stock picking skills are a source of return persistence.

In the second part of this study, the authors conduct principal component analysis and construct a multi-factor style model. The aim is to identify the diversification benefits of investing in specific classes of hedge funds, and also to avoid the potential self-misclassification problem. The first component explains about 33% of the cross-sectional variation in returns, while the second component explains 8% of the cross-sectional variation. An attempt to identify these components is made, with a differentiation by style (macro, multi-strategy and CTA) and investment region (Asia, Emerging Asia, Japan and Global). For example, the first component is identified as the common Asian component. Among the first 9 components, 7 are identified. They explain 64% of the cross-sectional variation in returns.

In the third part, the relationship between the characteristics of Asian hedge funds and their returns is investigated, in the cross-sectional Fama and MacBeth (1973) framework. The following results are reported: the management fees and the performance fees do not generate higher returns, the fund size is weakly related to returns, the redemption period is strongly and positively correlated to returns, the size of the holding company is positively related to returns and the fund age and the size of the minimum investment are not related to returns. Consulting the matrix of correlation coefficients, larger funds have higher management fees and higher minimum investment amounts and the redemption period is positively correlated with both management fees and performance fees.