L5 Returned my email… Here is his response. If you can’t tell L5 is a sharp tack and I knew he was thinking correctly, I just wanted to be sure
Thanks for your nice comments, RateLadder! Really good to hear your thoughts.You are absolutely right that interest rate – default (or delinquent) rate doesn’t equal yield ROI.What I was doing (a year’s worth of data then using 1 – current loan %) was a CRUDE approximation. My point was that even a crude approximation using real data from Prosper is, I feel, more relevant than using the Experian data, especially since I feel zeroing in on extended credit stats is very relevant.Now that Prosper has released their ROI estimate tool based on their portfolio data, this is clearly the benchmark to use. As they just released this on Monday, I’ve only played with it briefly, so don’t have full comments for you on it. But it has the advantage of being able to filter based on all the different attributes of Prosper loans, while also giving much better ROI calculations. Basically there are THREE “deducts” from the average rate column — the three negative numbers — that, as I understand it, you want to sum and deduct from the interest rate proposed by a borrower to get the expected yield.I attended a session where they talked about their methodology for determining the ROI, but I didn’t take detailed notes as they said there was a white paper on the site (or maybe in the forums) that goes into this in details. So I plan to get the white paper and think others who are interested in this should get it and read it, too. (Maybe you can post a link to this on your blog.)The key thing they use in ROI is a technique called “roll rates” which estimate on a go-forward basis (based on past data) what percentage of loans in one category will “roll” to the next (worst) category in the next month. For example, their data right now suggests that 3% of C loans will roll to 1-30 days late each month; of those, 25% will roll to 31-60 days late — i.e., 75% “cure” and 25% remain delinquent. Of those that are 61-90 days late, 80% roll (i.e., don’t pay anything), while 20% don’t roll and remain delinquent. Obviously the roll rates vary based on the characteristics of the loan class being studied — credit grade, extended credit scores, etc. — and you can filter on that. They also factor in early payoffs, which is nice, but I think not hugely relevant to me.They did say that their calculations were different than IRR… I didn’t understand the exact differences, but I think the white paper goes into.So, anyway, find the white paper and you (and I) will have answers. Good questions.Feel free to post this to your blog if it’s useful.
Wishing you the very best,