strategies that use linear asset pricing models tend to underperform passive
combinations of the risk-free asset and the domestic market index. There is an
improvement in the out of sample performance using the FS measure. The
underperformance is substantial for the strategies which use the FF and EGB
models. These latter two strategies also have negative Sharpe performance.
When investors face binding investment constraints as in Table 4, the Sharpe
performance of all ﬁve asset allocation strategies increases. The strategies now
have considerably lower volatility. The strategy that uses the characteristics
model has the highest Sharpe performance across the ﬁve strategies. However
it still underperforms the market Sharpe measure. This strategy also generates
Table 4. Performance of Asset Allocation Strategies: Restricted.
Sharpe Minimum Maximum
Char 0.335 3.26 0.103 –25.01 14.73
CAPM 0.143 2.67 0.054 –28.34 7.58
FF 0.204 2.82 0.072 –30.13 6.43
EGB 0.194 2.85 0.068 –30.12 6.70
APT 0.241 2.82 0.085 –28.16 6.88
Char CAPM FF EGB APT
* Signiﬁcant at 5%.
The out of sample performance of monthly asset allocation strategies is evaluated between
February 1981 and April 2000. Expected returns are estimated from a characteristics model of
stock returns (Char) and unconditional versions of linear asset pricing models of the CAPM, Fama
and French (1993) (FF), Elton, Gruber and Blake (1996) (EGB) and APT models. Panel A includes
summary statistics of performance of the ﬁve strategies that also includes the Sharpe (1966)
measure. Panel B reports the performance measures of Jensen (1968) measure and the conditional
measure of Ferson and Schadt (1996) (FS). The t-statistics (in parentheses) are corrected for
heteroskedasticity using White (1980). The asset allocation strategies are estimated with no short
selling allowed in the risky assets and an upper bound constraint of 20% in each risky asset. All
performance numbers are monthly %.
264 JONATHAN FLETCHER