Fixed Weight and Rolling Window Clones

AIM 77.5: Distinguish between "fixed-weight" and "rolling-window clones," explain their construction and compare the differences between these strategies.

Two versions of the multifactor model are used to create linear clones that replicate the factor exposures of hedge funds. The first version is the fixed-weight methodology, which uses the whole sample and has the portfolio weights constant through time. The second version is the rolling-window methodology, which is used to reduce the look-ahead bias that is inherent in the fixed-weighted version. The look-ahead bias exists in the fixed-weighted method because the entire data set of returns is used to construct the clones (i.e., all available data points over the 1986 through 2005 time period). The risk factors are the same as before, for both methods, except that the change in volatility factor is excluded because no liquid asset is readily available to replicate this volatility.

Fixed-weight clones: The fixed-weight clone model does not include an intercept. The sum of the weights of the beta coefficients from the regression is constrained to equal 1 (i.e., 100%) so it can calculate the weighted averages of the five factors that remain in the model. Negative weights are allowed and indicate short selling. Passive investors prefer this approach.

Rolling-window clones: The linear clones constructed with a 24-month rolling window also use the five factor model and renormalize the factors across time and by fund. The rolling window clone's volatility will not be the same as that of the entire sample history (as used by the fixed-weight methodology); however, if there are no dramatic shifts in volatility, they will be similar. Although the rolling-window reduces the look-ahead bias in the fixed-weight clones, there are disadvantages, which include greater estimation error and more frequent rebalancing. Therefore, more active investors engaging in dynamic asset-allocation will prefer the rolling-window method.

Figure 5 summaries the results of the linear clone portfolios relative to the hedge fund category that each clone is replicating. The results of analyzing the performance of the linear clones indicate that the Managed Futures and Global Macro linear clones outperform the fund for both the fixed-weight and the rolling-window constructions. Also, the Event Driven and the Emerging Markets style clones are not able to perform as well as the actual hedge funds.

Figure 5: Comparing Fixed-Weight and Rolling-Window Clones

Fixed-Weight Linear Clones

24-Month Rolling-Window Linear Cbnes

TASS Style Category

Expected Return %

Clone Return %

Difference between Fund and

ization Factor for Leverage

Expected Return %

Clone Return %

Difference between Fund and

Clone's Returns %

Managed Futures

13.64

27.97

14.33

2.76

11.84

19.24

7.40

Global Macro

11.38

15.54

4.16

2.60

9.01

12.97

3.96

Equity Market Neutral

8.09

10.00

1.91

2.18

5.71

4.43

-1.28

Fund of Funds

8.25

9.29

1.04

1.69

7.34

5.67

-1.67

Dedicated Short Bias

5.98

6.70

0.72

2.26

2.58

6.83

4.25

Multi-Strategy

10.79

10.32

-0.47

2.04

8.97

5.33

-3.64

Convertible Arbitrage

8.41

7.4

-1.01

1.71

4.04

2.78

Income

Arbitrage

9.5

8.35

-1.15

2.07

7.80

4.47

Equity

Hedge

14.59

13.12

-1.47

2.18

11.90

9.08

-2.82

Event Driven

13.03

9.84

-3.19

1.78

11.65

6.96

-4.69

Emerging Markets

20.41

14.77

-5.64

2.13

21.12

5.17

-15.95

0 0

Post a comment