Applications of state space models in finance
Applications of state space models in finance
Applications of state space models in finance
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
7.4 Empirical results 139<br />
Table 7.4: Summary statistics for the set <strong>of</strong> risk factors (8.1.1992–2.2.2005).<br />
µ a<br />
σ b<br />
ρ1 ρ2 ρ3 ρ4 ρ5 ρ12 Q(12) c<br />
BMR 0.03 (0.68) 13.14 0.14 0.13 0.02 0.02 0.03 −0.08 33.45 (0.01)<br />
V GS 0.03 (0.52) 9.73 0.03 0.10 0.13 0.03 0.02 0.04 81.54 (0.00)<br />
SIZ 0.11 (0.04) 9.80 −0.11 0.11 0.15 0.02 −0.01 0.06 46.70 (0.00)<br />
T S 0.71 (0.14) 90.29 0.00 0.03 0.11 0.00 0.02 0.02 25.09 (0.01)<br />
OIL 0.14 (0.39) 29.51 −0.02 0.04 0.03 −0.03 0.03 0.03 12.98 (0.37)<br />
F X −0.01 (0.82) 10.06 0.01 −0.02 0.00 0.05 0.08 −0.08 14.39 (0.28)<br />
a The mean is expressed <strong>in</strong> percentage terms. Figures <strong>in</strong> parentheses denote p-values for a<br />
simple t-test <strong>of</strong> the null hypothesis <strong>of</strong> a zero mean.<br />
b The standard deviation is expressed <strong>in</strong> annualized percentage terms.<br />
c Q(l) is the test statistic <strong>of</strong> the Ljung-Box portmanteau test for the null hypothesis <strong>of</strong> no<br />
autocorrelation <strong>in</strong> the errors up to order l. The Q-statistic is asymptotically χ 2 distributed<br />
with l degrees <strong>of</strong> freedom. Figures <strong>in</strong> parentheses denote p-values.<br />
the full sample, the highest correlations occur between SIZ and F X (−0.204) and<br />
between SIZ and V GS (−0.227). This <strong>in</strong>dicates that the outperformance <strong>of</strong> small caps is<br />
associated (i) with a fall<strong>in</strong>g dollar, and (ii) with value outperform<strong>in</strong>g growth. The highest<br />
correlation coefficients are measured for Panel E, which is based on fewer observations<br />
than the other subperiods. Dur<strong>in</strong>g this time, the exchange rate factor’s correlation with<br />
the value-growth-spread (−0.370) and the term structure (−0.385) is elevated. Overall,<br />
the observed pairwise correlations <strong>in</strong>dicate that the selected risk factors are far from<br />
be<strong>in</strong>g excessively coll<strong>in</strong>ear, which suggests that none is dispensable.<br />
7.4 Empirical results<br />
Employ<strong>in</strong>g the risk factors def<strong>in</strong>ed above together with the conditional multifactor model<br />
given by (7.6)–(7.10) implies the follow<strong>in</strong>g data-generat<strong>in</strong>g process for the set <strong>of</strong> excess<br />
sector returns, Ri,t. The observation equation can be written as<br />
Ri,t = βi1,tBMRt + βi2,tV GSt + βi3,tSIZt + βi4,tT St + βi5,tOILt<br />
+ βi6,tF Xt + ɛi,t, ɛi,t ∼ N(0, σ2 i ), (7.18)<br />
for i = 1, . . . , 18. The <strong>state</strong> equation is given as<br />
β i,t+1 = β i,t + η i,t, η i,t ∼ N(0, Q), (7.19)<br />
where βi,t+1 and ηi,t are 6 × 1 vectors <strong>of</strong> <strong>state</strong>s and <strong>state</strong> disturbances, respectively;<br />
Q = diag(σ2 ηi1 , . . . , σ2 ηi6 ) is the variance term <strong>of</strong> <strong>state</strong> errors. With reference to (7.11)<br />
we are ultimately <strong>in</strong>terested <strong>in</strong> the factor load<strong>in</strong>gs β i,t−1, which are based only on<br />
<strong>in</strong>formation available at date t − 1. Hence, filtered <strong>state</strong> estimates as def<strong>in</strong>ed <strong>in</strong> (3.13)<br />
serve as time series estimates <strong>of</strong> time-vary<strong>in</strong>g factor load<strong>in</strong>gs.<br />
In order to assess the relative superiority <strong>of</strong> the proposed dynamic specification with<br />
conditional factor load<strong>in</strong>gs modeled as <strong>in</strong>dividual random walks, three alternative <strong>models</strong>