Advances in Investment Analysis and Portfolio Management, Volume 9

Active asset allocation strategies that use linear asset pricing models now

outperform the strategy based on the characteristics model in terms of a higher

Sharpe performance and more positive abnormal returns. Furthermore, all

strategies that use linear asset pricing models are able to significantly

outperform alternative passive strategies even when the investor faces binding

investment constraints. The improved performance of linear asset pricing

models when we use conditional versions of the models supports a number of

studies from the empirical asset pricing literature that document that

conditional versions of asset pricing models are more able to explain cross-

sectional patterns in stock returns (see Cochrane (1996), Jagannathan and

Wang (1996), Hodrick and Zhang (1996) and Lettau and Ludvigson (2001)).

Our findings support the usefulness of conditional asset pricing models in

forecasting expected excess returns in domestic asset allocation strategies and

provides some support for the use of these models relative to a characteristic

based model of stock returns.

NOTES

1. The momentum effect stems from Jegadeesh and Titman (1993).

2. We focus on expected returns because Merton (1980) points out that estimates of

expected returns are more unstable than the covariance matrix and Best and Grauer

(1991) document the sensitivity of optimal portfolio weights to even small changes in

expected returns.

3. Jagannathan and Ma (2001) find that the sample covariance matrix performs just

as well as other estimators of the covariance matrix when investors face binding

investment constraints.

4. This is set equal to 0.1. Using alternative values of t has no impact on the analysis

for the performance measures used in this study.

5. See Grinblatt and Titman (1989), Chen and Knez (1996) for a discussion of these

points.

6. We do not include the Utilities sector because data is not available for the whole

period.

7. When the characteristics model is estimated, the ln (market value) is used.

8. The FTA index is a value-weighted index of the largest companies on the London

Stock Exchange.

9. Connor and Korajcyzk (1991) regress the demeaned values of the factors on

statistical factors derived from asymptotic principal components analysis.

10. Keim and Stambaugh (1986) and Fama and French (1988) among others also

show that stock and bond returns are partly predictable through time. Fama (1991) and

Cochrane (1999) provide a review of stock return predictability.

ACKNOWLEDGEMENT

Helpful comments received from an anonymous reviewer.

268 JONATHAN FLETCHER