Thursday, July 27, 2006

What an X% default rate should mean to you, part 5

Excerpted from my full post:

One vital thing that people still seem to fail to consider when it comes to lower credit grade loans (this principle is applicable to other grades, but is less significant):

our default data on HR's, like every other credit grade, comes from new bank card products over a 2 year period. thus, in addition to all the other issues to be considered, such as the mortality curve for loans (that is, where defaults tend to occur)...

the default numbers apply to borrowers of that credit grade who were approved by conventional card issuers for bank card products.

(this is less significant for higher grades, as it is much more likely that the pool of people applying in a given credit grade, is close to the pool of people who are approved for loans in that credit grade. That is, w/the exception of AA's or B's who get far larger loans via prosper than they could via traditional routes, most AA's or B's who get funded on prosper could have, in theory, gotten funded traditionally -- and thus fall within the loans considered for Experian defaults.)

in other words, if every single HR borrower were to be funded, defaults (should)/would be far worse than 19.1%. The question remains -- are we funding more aggressively or less aggressively than traditional issuers, and if more, how much of an offset is there from all the miscellaneous factors that have been brought up thus far?

(I use HR's because the credit grade is open-ended on the lower end -- and this is why i think one of the best improvements would be to further bucket the HR grade, even if defaults for subbuckets are unavailable.)


At 12:58 PM, Anonymous Anonymous said...

I like the idea of the basic letter grades given for each borrower, however I would like to see the addition of the actual numerical score generated for each that placed them in the letter grade.

Or maybe add to the advanced search critera allowing for search
of loans within a specific number value (Ex: Include only 500-540) which would allow a lender to look at higher end HR's.

At 10:38 AM, Anonymous Anonymous said...


Maybe I'm not understanding something, but could you explain why your Excel sheet for a "horribly oversimplified Excel model for modelling defaults" returns a different number than the optimistic IRR at

I thought both of those were attempting to calculate return based on a linear monthly default rate calculated off of Experian yearly default rates?


At 10:56 AM, Blogger cellardoor said...

As I'm not privy to the internal workings of savagenumbers' calculations, I couldn't really say...


Post a Comment

<< Home