Since Basel II requires banks to utilize pooled PDs per rating grade / segment, I wonder why exactly. Via a logistic regression you can directly estimate an obligors PD, why the extra step to pool them back together?
E.g. why take a small corporate exposure with a PD of, say 0.31% stemming from a logit model, and calibrate it to the long-run average default rate of, say 0.4% for the segment of small corporate exposures.