234 CORRELATION BREAKDOWNS IN ASSET MANAGEMENTTable 12.5Regression equation of correlation USD–JPY changes explained
by volatility differences
Variable Coefficient Std. Error t-statistic Prob.
DUSDV 14.29453 8.641957 1.654085 0.0987
DJPYV 2.155165 7.527247 0.286315 0.7747
R-squared 0.004810 Mean dependent var 0.135836
AdjustedR-squared 0.003020 S.D. dependent var 8.821442
S.E. of regression 8.808111 Akaike info criterion 7.192801
Sum squared residuals 43136.04 Schwarz criterion 7.208300
Log likelihood −2004.791 F-statistic 2.687343
Durbin–Watson stat. 1.988807 Prob(F-statistic) 0.101714
Notes: The dependent variable; differences of exponential correlations between the equity US market
and the equity Japanese market. Explanatory variables are DUSDV: differences of exponential volatility
of the equity US market; and DJPYV: differences of exponential volatility of the equity Japanese market.
Number of observations: 558 daily changes.
Table 12.6Regression equation of correlation USD–JPY changes explained
by volatility differences
Variable Coefficient Std. error t-statistic Prob.
DUSDV 14.29993 8.634812 1.656078 0.0983
R-squared 0.004663 Mean dependent var 0.135836
AdjustedR-squared 0.004663 S.D. dependent var 8.821442
S.E. of regression 8.800849 Akaike info criterion 7.189364
Sum squared resid 43142.40 Schwarz criterion 7.197114
Log likelihood −2004.833 Durbin–Watson stat 1.989381
Notes: The dependent variable; differences of exponential correlations between the equity US market
and the equity Japanese market. Explanatory variables are DUSDV: differences of exponential volatility
of the equity US market. Number of observations=558 daily changes.
the plot of residuals of correlation between the USD–JPY higher changes
explained by volatility differences.
Our estimates demonstrate that correlation jumps can be modeled rather
than the complete time series in all the reported cases. Table 12.8 shows the
measures of the improvement obtained for all the correlations. As for out of
sample predictability in our daily intermarket correlation series, the results
reported in Table 12.9 are coherent with those obtained in the sample period.
In particular, correlation jumps mean absolute errors are everywhere lower