I've been playing around with Prosper's Marketplace Performance tool to overcome the limitations present in the exported Prosper data (no detailed credit information). While playing around with some initial queries (if you're trying to figure out how to change things, look for the not-so-easy-to-find "Edit Criteria" link), I stumbled onto some statistical confirmation of something that I've believed for a while: income matters.
This seems obvious. Now matter how well intentioned someone is, they can't repay their lender if they have no cash. Therefore, borrowers disclose their incomes. Prosper is supposed verifying income levels if a loan funds. It's the borrowers who low-ball their official incomes for the listing and then claim in the description that they've got more money available (it's just hard to prove it, trust me) that scare me. As the mortgage industry is learning, stated income loans are risky.
Prosper only has about 6 months of data with income, so this is only a first peek, but the results are interesting. Below are the percentage of distressed loans (late or worse) in the recovered datasets.
|Unstated or <$25k||0.00%||0.75%||2.53%||3.20%||4.64%||11.39%||20.39%|
While this isn't enough information to predict default rates, it's a clear indicator that a verified income is equivalent to elevating the borrower by one credit grade for all but the best borrowers (or, conversely, not having verified income is equivalent to a drop of one credit grade).