Yesterday a reader posted the following comment…
You rely a lot on DTI - Is DTI a good predictor of default? For the last year, 100 days, or what? Also, what about very high DTI’s (eg >1000%, usually means the person hasn’t been able to verify income) — are they safer than DTIs around 80-100%?
Here is my response…
So far DTI is not correlated to late payments. However, because it weighs so heavily in normal loan applications I think it would be foolish on my part to ignore it. Other lenders seem to agree as rates for higher DTI values are higher.
An unverifiable DTI in my mind is no better than any loan of the same credit and extended credit. It might be argued that it is worse.
The journey continues and some day we may be able to answer your question with confidence and clarity, but for now, this early in the game, I will just wave my hands at it.
I am interested in reader option… What do you guys think?
If you liked this article, vote for it on del.icio.us and stumbleupon.Categories:
Prosper.com, Strategy
Related Articles Related Stores
5 comments ↓
I think that the current lack of correlation between DTI and late payments is due to the strength of the US economy over the past couple of years. Once a recession sets in, I predict that the high DTI loans will tend to be later than low DTI loans.
I think I agree.
I also think that DTI might be a differentiator later in the loan cycle… Still early.
Ok, I pulled the download from couple days ago, and looked at all loans 12 months or older, with a DTI reported. 287 “good” (current, paid, or repurchased) vs 51 “bad” (late or default, I left out the 5 labelled “cancelled” because I don’t know if that means a lender got screwed or what). Assuming variances are equal, t-test shows no significant difference in the DTIs. This means, out of all the “aged” loans, there’s a 95% certainty that the DTI of the perfect payers was not lower than the DTI of the non-perfect payers. more to follow
ok. So far I am not out on a limb
I think you should take out repurchased too. Those loans were bought back by prosper because of fraud. They have a 100% fraud guarantee.
Ok, I had to look up some old statistics lessons; I made a mistake in using a test that assumes the DTIs were normally distributed. Actually log(DTI) are normally distributed, so I re-did the comparisons based on log(DTI), again with loans 12 months or older. 282 perfect loans are paid or current, mean(ie average) DTI 0.1249. 51 bad loans are late or defaulted, mean DTI .1356. 40 very bad loans are 3+ months late or defaulted, mean DTI .1404. 22 defaulted loans have a mean DTI .1393. You can see the averages are different, but the difference is not statistically significant. That is, I can’t say with 95% certainty that defaulters (or any other group of non-perfect borrowers) have a different DTI than perfect borrowers. Among every group, from perfect to defaulted borrowers, there is a huge spread, from almost zero, to over 50% DTI. Also notice that the average in every category is well below .15, a number that seems pretty safe… This is not even addressing the “good” borrowers who have DTI of a zillion because they can’t verify income.
Leave a Comment