bursting (presumably while the alphas worked). Starting in December 1999,
however, just prior to the Dow’s peak in January 2000, the robust strategy
started to outperform. Furthermore, the outperformance was more or less con-
sistent post bubble, starting in the fall of 2001.
Given that the sample period for the FE means and covariances is 30 months,
it is not unreasonable to conclude that once the history contains an extremely
volatile market event, the robust strategy starts to differentiate itself from the
control strategy.
Robust Control
− 40
− 20
0
20
40
60
80
100
Mar-97Jun-97Sep-97Dec-97Mar-98Jun-98Sep-98Dec-98Mar-99Jun-99Sep-99Dec-99Mar-00Jun-00Sep-00Dec-00Mar-01Jun-01Sep-01Dec-01Mar-02Jun-02Sep-02Dec-02Mar-03Jun-03Sep-03Dec-03Mar-04Jun-04Sep-04Dec-04Mar-05Jun-05Sep-05
Figure 3.3 Net of benchmark performance (30 bps to trade).
− 15
− 10
− 5
0
5
10
15
20
25
30
Mar-97Jun-97Sep-97Dec-97Mar-98Jun-98Sep-98Dec-98Mar-99Jun-99Sep-99Dec-99Mar-00Jun-00Sep-00Dec-00Mar-01Jun-01Sep-01Dec-01Mar-02Jun-02Sep-02Dec-02Mar-03Jun-03Sep-03Dec-03Mar-04Jun-04Sep-04Dec-04Mar-05Jun-05Sep-05
Robust
Figure 3.4 Net of control strategy performance (30 bps to trade).