Here's the 01/15/08 update to my late loan statistics charts.
These charts show statistics for the performance of ALL prosper.com loans. Each curve represents the set of loans that were created in one calendar month. The vertical axis is the fraction of those loans that have "gone bad", in other words are 1 month late or worse (up to and including default). The horizontal axis is the observation date. All data comes from Prosper.com's performance web page.
A larger, more readable version of that chart can be found here
If we slide the curves so that the horizontal axis represents days since origination month, all the curves fall on top of one another, and we can see how remarkably similar they are. (I've left the oldest three months off of the "slid" chart, because those months do seem to be a bit different.) It will be interesting to learn how much these curves "bend" as we go past 1 year loan age. There is a hint in the data that the default rate may decrease as the loan ages, but it isn't yet clear. I described a mathematical model and showed some curve fits which attempted to estimate how the default rate might be changing over time several months ago on the old Prosper.com forum. I'll update that eventually and post it here as more data becomes available.
If you look out one year from origination (ie 360 days) you will see that about 20% of Prosper's loans have gone bad. You can also see that this is remarkably consistent from month to month (ie the different curves). One can only conclude that the default rate of Prosper loans is in the neighborhood of 20% per year. Loans originating after Feb'07 are going bad at a slightly lower rate, probably because Prosper increased the minimum credit score required for a Prosper loan at that time.
Shamefully, Prosper continues to report much lower default rate numbers to the press. See a recent example in the Chicago Tribune article of 01/13/07, which talks of a 4.7% default rate. http://www.chicagotribune.com/business/yourmoney/chi-ym-started-0113jan13,1,842983.story?ctrack=1&cset=true
Posts
