lhcole77,
You make an interesting point. I was aware that there were certainly differences in the FICO 08 alogorithms as they pertain to the three CRAs, but I had always assumed that to the extent that those were being marketed as the FICO 08 Classic scores that the nature of those changes was just to try to apply a sort of "leveling" of the data available from each CRA to the end goal of having as consistent a score as possible generated by the three CRAs.
I never really contemplated that the agencies may have influenced the development of the algorithm to effect other policy ends. My understanding was always that the algorithms are built to recognize specific predictive factors, combine these together, and generate a score which translates to a direct probability-of-default over a two year period. It seems so incredible to me that the AAoA of accounts at one agency would yield a different level of predictiveness versus the AAoA at a different agency. AAoA is a pretty objective measure. The raw data is the same in both places, yet the algorithm as applied treats the data as two different ends of the spectrum. I would think that FICO research would say "AAoA of these parameters tends toward XYZ probability of default." It now makes me wonder if they build all elements of the FICO 08 algorithms on the same general principles but alter the version for each CRA based only on input data from that same corresponding CRA.
What you say about UTIL% definitely makes sense to me. If the algorithm across different CRAs treats a same AAoA differently, why would it not possibly treat a same UTIL% differently across CRAs?
Any FICO insiders or other industry insiders have any insight on this?
