Sunday, April 02, 2006

Finding Expected Rate of Return on, Part 1 - Gathering Info

The goal of these calculations is to determine the expected annual rate of return (APR) one can expect from making loans on given a 'Credit Grade'.

First, it is important to understand the risk model that employs. When you are a lender you are given the following information for each loan applicant: 1) "Credit Grade", 2) Debt/Income Ratio, 3) Borrower's group, 4) A story of how the applicant will use funds (at the applicants descretion).

When calculating the expected return, what we really care about is the future return of our investment. In order to forecast the future with probably we are primarily concerned with statistically significant trends that relate to the returns (or repayment in this case of the investment). Currently, criteria 3 & 4 above may provide lendors with the 'warm and fuzzies' needed to finalize a loan, but only criteria 1 & 2 have historical data.

According to the 'Credit Grade' is a score from Experian called Experian ScoreX.

This score seems to be slighlty different for most when compared with their FICO score - usually 30-80 points lower. Nonetheless, the most relevant historical data that is provided at is from Experian:

Default rates refer to the percentage of individuals that default on loans (many different kinds of loans) during the course of a year. These default rates only apply to the Experian ScoreX rating system, so they will not accurately relate to other credit scoring systems. Also, note that the default rates only apply to borrowers with debt to income ratios

"Accounts in "default" include those which are charged-off, the borrower has filed bankruptcy, or the borrower is 90 or more days past due on payment. Consumer loan products include auto, mortgage, credit card, and many other types of loans to individuals."

For now, this is the best data we have to work with, although both and its users will readily admit that even these numbers do not apply directly because is a different lending environment. In the future Prosper will have historical data of its own to provide potential lenders. I'm sure this will include real default rates from the site, correlation to the 'Group' function, more detailed debt/income data, and other lending criteria.