Our first step is to identify a global factor common to all yields for each maturity. Technically, this is the first principal component and is illustrated in blue in Chart 1 for the case of French sovereign yields. By construction, it is common to the time series of yields of all nine countries. We then construct a series for each country of the part of their yields that cannot be explained by the global factor. Countries with similar patterns of yield dynamics are then grouped together (France, Germany and Spain) and (Australia, Canada, United Kingdom, United States) in order to identify a “euro area” factor and an “Anglo-Saxon” factor, respectively. Italy is not part of the group used to identify the euro area factor because of its low correlation with the rest of the group. Japan is a factor in itself. For 10-year yields we find that we need to split the “Anglo-Saxon” countries into North America and Australia/UK.
To estimate the euro area factor, we again search for a common factor which explains the co-movement of the residual components of those countries' yields after accounting for the contribution of the global factor. We do the same for the Anglo-Saxon factor. Finally, we regress the yields of each country on all these factors to establish the importance of each factor for each yield. The residual that we cannot explain is termed the idiosyncratic factor. We obtain a decomposition of each yield as illustrated in Chart 1 for the French case.
A global, not a US factor
It is sometimes argued that US markets drive the global financial cycle (Miranda-Agrippino and Rey, 2021). In this case, we would expect to find the highest correlation between US yields and the global factor. We do not find this. The results in Table 1 show that the US correlation is even below the median for both maturities.