I was idly looking for patterns in the daily evolution of eurozone government bond spreads (like you do) and thought I would share some findings. The spread of Irish Government bonds over the 10-year German benchmark have of course trended upward during the period since early September 2008 to last week:
If we compute principal components of the spreads of ten euro-currencies we can try to isolate the different factors: separating factors that affect all countries from those that affect Ireland in relative isolation.
Using daily changes in the spreads, the first three principal components explain 80% of the total variation in the ten series.
All ten bonds have roughly equal loadings on the first PC (which alone explains 62%). We can therefore think of PC1 as measuring fluctuations in general aversion to credit risk.
PC2 seems to measure a component which is irrelevant to Ireland — from the loadings this one looks like Club Med vs the North.
But PC3 is an almost Ireland-specific factor, much smaller loadings on the other countries. The big action in PC3 is on just three almost consecutive days in January: the 16th (Anglo nationalization), 19th and 21st.
To me this illustrates just how easily spooked this particular market is. Anglo nationalization was not even demonstrably bad news. When will it settle down to a realistic assessment of Irish risks?
The linear regression equation explaining changes in the Irish spread in terms of three principal components is (t-stats in parentheses):
ΔIreland = 0.020 + 0.015 PC1 + 0.042 PC3 + 0.024 PC4
(17.7) (32.7) (32.0) (15.6)
The constant term reflects the general upward trend in Ireland’s spread (which is not explainable by this method).
(Of course there are many methodological tricks one could explore, but it’s getting late and this seems enough for the present. Probably some readers do this stuff for a living!)