Advances in Investment Analysis and Portfolio Management, Volume 9

Sharpe performance in Table 3 where this strategy has a negative Sharpe

performance.

The abnormal returns of the four strategies in panel A are highly statistically

significant and economically large. The strategy that uses the CAPM model has

the lowest abnormal returns across the four strategies, because this strategy has

a lower mean excess return compared to the other three strategies. All four

strategies have a negative beta on the FTA index when performance is

estimated using the Jensen measure. The strategy that uses the APT model has

the highest performance with the Jensen measure and the FF model has the

highest performance with the FS measure.

When investors face binding investment constraints as in panel B of Table 5,

all four strategies still deliver a higher Sharpe performance than either the FTA

market index or the strategy that uses the characteristics model. The

performance across the four strategies manifests less variation compared to that

in panel A. High mean excess returns and low standard deviation characterize

all four strategies. The smaller means leads to lower abnormal returns in panel

B compared to those in panel A. However the Jensen and FS performance of

the four strategies are still highly significantly positive and economically large.

The strategies that use the EGB and FF models exhibit the highest positive

performance across the four strategies.

The evidence in Table 5 suggests that strategies which use linear asset

pricing models to forecast expected excess returns are able to significantly

outperform alternative passive strategies whenever conditional versions of the

models are used. Furthermore the strategies which rely on linear asset pricing

models are able to outperform the strategy that uses the characteristics model

in terms of higher Sharpe performance and more positive abnormal returns.

These findings support the evidence in Jagannathan and Wang (1996), Hodrick

and Zhang (2001), and Lettau and Ludvigson (2001) among others that

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

stock returns.

CONCLUSIONS

This paper examines the out of sample performance of domestic asset

allocation strategies in U.K. stock returns between February 1981 and April

2000 where linear asset pricing models and a characteristics model of stock

returns are used to forecast expected excess returns. Our findings suggest that

when unconditional versions of asset pricing models are used, the performance

of the strategies that use multifactor asset pricing models is poor. However this

all changes when we use conditional versions of the asset pricing models.

267Characteristics Versus Covariances