bdonovan's questions which started this thread are terribly interesting!

The implied question in his post is how to increase the probability that a borrower will pay back a loans.

Another way to put it is, how to reduce defaults.

The better you can predict default rate, the better returns.

I do not know how interest rates are determined for an individual. I suppose that FICO is a primary variable used in that formula. I assume that FICO is a good predcitor of financial reliability-risk.

I suppose if you did a deep interview, a full psychological assessment, and took one of the borrower's children as a hostage, you could increase probability of repayment. I suppose that more in depth data collection would cost more. There has got to be some set of data which are the best predictors of probability of default. I guess those variables could be determined using factor analysis.

Since any statistical prediction only has a particular margin of error, the more loans the less the variance.

This all and bdonovan's question, makes me want to learn more about the science and math of making loans.

I had taken stats in grad school, but I just don't remember enough to solve this problem, now.

I do wonder if LC gathers enough data and provides the data to us for us to be able to improve the prediction of default rates.