02.12.2012 Views

Applications of state space models in finance

Applications of state space models in finance

Applications of state space models in finance

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

118 6 Time-vary<strong>in</strong>g market beta risk <strong>of</strong> pan-European sectors<br />

probability density<br />

probability density<br />

probability density<br />

1.2<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0.0<br />

(a) OLS<br />

Median = 0.249<br />

F(0) = 30%<br />

−1.0 −0.5 0.0 0.5 1.0<br />

1.2<br />

(d) RW<br />

Median = 0.298<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 25%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

2.0<br />

1.5<br />

1.0<br />

0.5<br />

(g) GRW<br />

Median = 0.254<br />

F(0) = 14%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

probability density<br />

probability density<br />

probability density<br />

1.2<br />

(b) tG<br />

Median = 0.278<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 24%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

1.2<br />

(e) MR<br />

Median = 0.260<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 25%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

(h) MSM<br />

1.2 Median = 0.244<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 27%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

probability density<br />

probability density<br />

probability density<br />

1.2<br />

(c) SV<br />

Median = 0.289<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 28%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

1.2<br />

1.0<br />

0.8<br />

(f) MMR<br />

Median = 0.295<br />

F(0) = 23%<br />

0.6<br />

0.4<br />

0.2<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

(i) MS<br />

1.4<br />

1.2 Median = 0.250<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

F(0) = 29%<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0<br />

Figure 6.13: Histograms <strong>of</strong> Spearman’s out-<strong>of</strong>-sample rank correlations (100 samples).<br />

correlations are smaller than zero. This means that the risk <strong>of</strong> generat<strong>in</strong>g a mislead<strong>in</strong>g<br />

signal is reduced.<br />

When evaluat<strong>in</strong>g a model, a common way to take the risk related to a forecast explicitly<br />

<strong>in</strong>to account is to calculate an <strong>in</strong>formation ratio (IR). Alexander (2001, p. 445)<br />

def<strong>in</strong>es an <strong>in</strong>formation ratio as “the mean prediction error divided by the standard deviation<br />

<strong>of</strong> the prediction error”. In the context <strong>of</strong> the cross-sectional analysis us<strong>in</strong>g rank<br />

correlations, an <strong>in</strong>formation ratio for a given model<strong>in</strong>g techniques n can be def<strong>in</strong>ed as<br />

IRn = E(ρS � n)<br />

, (6.44)<br />

V ar(ρS n)<br />

where ρS n = {ρS n,1, ρS n,2, . . . , ρS n,T }. The computed <strong>in</strong>formation ratios are reported <strong>in</strong><br />

Table 6.8. For the chosen out-<strong>of</strong>-sample period, the GRW model is confirmed to <strong>of</strong>fer<br />

the best risk-adjusted forecast<strong>in</strong>g performance, followed by the MMR and the RW model.<br />

To check which <strong>of</strong> these three <strong>models</strong> yields the best forecast<strong>in</strong>g performance over a more<br />

representative period <strong>of</strong> time, an out-<strong>of</strong>-sample period <strong>of</strong> ten years based on 520 samples<br />

Table 6.8: Information criteria <strong>of</strong> out-<strong>of</strong>-sample rank correlations.<br />

OLS tG SV RW MR MMR GRW MSM MS<br />

0.63 0.65 0.64 0.70 0.64 0.72 1.42 0.62 0.65

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!