Elsevier UK CH10-H8321 jobcode: FEF 28-6-2007 5:50p.m. Page:224 Trimsize:165
×234MM
Basal Fonts:Sabon Margins:Top:36pt Gutter:15mm Font Size:10/12 Text Width:135mm Depth:47 Lines
224 Forecasting Expected Returns in the Financial Markets
beta, the likely levels of noise, the limited data, and variations in model specification over
time, we suspect that the estimates in Table 10.8 may be hard to better.
10.5 Models optimal horizon, aggression and model
combination
Finding the optimal horizon in the single model case involves maximizing equation
(10.16). That is, the optimal horizon is the one that maximizes the conditional asset return.
Using the model parameters from Table 10.8 (IC and decay beta), for the COMPOS
model, a range of optimal horizons dependent on aggression and transactions costs is as
shown in Table 10.9.
Finding the optimal horizon using the sub-composites constitutes an N model case,
and therefore involves maximizing equation (10.17). As noted previously, this will also
define sub-model ‘weights’ at the optimal horizon.
In order to solve equation (10.17), we require the correlation matrix . Using the
time series of the quintile spreads generated by the sub-composites, we calculated the
fol-
lowing (rounded) correlations: VALUE1/VALUE2
=0.45; VALUE1/MOMENT =−005;
VALUE2/MOMENT
=0.15. These seem reasonably intuitive the value/growth correla-
tion is slightly negative, the value/cash flow price measure is high and positive, and the
growth/cash flow price measure is small but positive and we therefore use them at this
stage to form
.
If we assume transactions costs of 1%, and the parameter values for the decay matrix
from Table 10.8, maximization of equation (10.17) gives estimates of optimal horizons
and weights as shown in Table 10.10.
Table 10.9 COMPOS optimal horizon in months
Transactions costs 0.6 0.8 1.0 1.2
Score =06 12 15 19 26
Score =10 8 10 12 14
Score =14 6 7 9 10
Score =18 5 6 7 8
Table 10.10 Optimal horizon and implied weights for sub-composites
Optimal horizon
(months)
% VALUE1 % VALUE2 % MOMENT
Score =042
Score
=060
Score
=100
Score
=140
Score
=180
Score
=220
24
16
10
8
7
6
56
52
48
45
44
43
10
9
7
7
5
4
34
39
45
48
51
53

Get Forecasting Expected Returns in the Financial Markets now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.