We interpret the large positive values for ARMRESID over the 2003-06 period as indicative of generally eased underwriting standards.
Other series are likely to convey additional relevant information about underwriting. But, we deliberately chose to exclude many of them.
We have argued that we have five variables that serve as indicators of various aspects of bank and nonbank underwriting standards.
Because they each are related to overall underwriting, they tend to be somewhat correlated; the average simple correlation coefficient between them was 0.55; the multicollinearity of this group of five variables was naturally considerably higher than that.
In the case at hand, however, using input variables that are reasonably connected to underwriting increases our confidence that the first PC is a satisfactory candidate as an indicator of aggregate underwriting.
The first PC as an indicator of underwriting standards
As we might expect from an indicator of underwriting tightness, UWPC rose both with the Federal Reserve and with the OCC measures of underwriting tightness.
They both, however, pointed toward underwriting tightening starting with the 2007 financial crisis, an episode that everyone recognized.
In that respect, UWPC suggests that underwriting eased significantly from 2002 through 2006.
The onset of the financial crisis in 2007 then saw UWPC rise very sharply, by more than double the prior decline, indicating extreme underwriting tightness.
In the case at hand, the resulting impulse responses (IRs) functions can be used to help assess not only the dynamic structure of housing and mortgage markets, but also the caliber of the constructed underwriting variable, UWPC.
(Further descriptions and sources are given in Appendix II.) We chose these variables because we judged that they were important, aggregate variables that were likely to affect or be affected by underwriting, or both.