Lenders Properly Assessing Risk?

Prosper provides a reasonable amount of information on the borrowers to allow lenders to estimate risk. I've been quite curious, however, on how well the lenders are interpreting this information and generating risk-adjusted bids on loans. To that end, it's time for some more data crunching.

My handy-dandy tools have calculated the distress rate (loans in all the bad status categories from Late to Defaulted) for various interest rate bands.

Start	End	Distress Rate	Spread
0.00%	5.00%	12.82%	-10.32%
5.00%	7.50%	1.92%	4.33%
7.50%	10.00%	1.44%	7.31%
10.00%	12.50%	2.89%	8.36%
12.50%	15.00%	4.01%	9.74%
15.00%	17.50%	5.72%	10.53%
17.50%	20.00%	5.91%	12.84%
20.00%	22.50%	9.99%	11.26%
22.50%	25.00%	14.29%	9.46%
25.00%	27.50%	15.06%	11.19%
27.50%	30.00%	21.54%	7.21%
30.00%	40.00%	62.07%	-27.07%

Looking at the table broadly, risk is priced reasonably well except for loans below 7.5% and above 30%. The table's middle section has a slowly increasing distress rate that implies that, roughly speaking, lenders are placing risk reasonably. I have to assume that the loans below 7.5% are sweet-heart deals brokered between people who know each other, because it doesn't make financial sense. The CD rates (risk free) have been in the 4.5% - 5% range for a while, and a collateralized home loan has been around 6%. Well, for whatever reason it was a bad risk. Equally, anyone who wants a loan at over 30% appears to be a bad risk as the interest rates are obviously not covering the risk.

My rule of thumb has been that a 1% increase in distress rate should be compensated for by a 1% increase in the interest rate. This was captured in the Spread column. An increasing spread corresponds to improving lender return. Looking at the table, lenders are getting a good deal. Many ranges approach 2% increases in interest rates for a 1% increase in risk. The lender sweet spot is coming in for loans in the 20% area.

Auto Funding Risks

Automatically funded loans are another topic that gets thrown around in forums as a risk items for loans. There is some logic to the idea that auto-funded loans are risk items: borrowers who are willing to accept a higher interest rate to get the money now are more desperate, desperation results in people taking risks, higher risks mean higher distress and default rates. Thanks to our handy-dandy statistics processing, we have the distress rates (i.e. late or worse) for auto funded loans (Close When Funded) versus competitive bid loans (Open For Duration):

Credit Grade	Open For Duration	Prosper Average	Close When Funded
AA	1.46%	1.43%	1.28%
A	2.30%	3.31%	6.86%
B	2.38%	3.53%	6.45%
C	3.02%	4.02%	5.74%
D	3.31%	6.15%	9.91%
E	6.26%	11.96%	16.50%
HR	11.68%	19.45%	25.80%

From the table, it's obvious that auto-funded loans are a risk item. For all but AA grade listings, it results in a 2x - 3x distress rate. This is a large and statistically significant spread. It's dangerous if you're using the Prosper average for bidding criteria as you'll be under-pricing risk for auto-funded loans and over-pricing open market loans.

Unlike unverified accounts, however, it's possible to price the loans appropriately to handle the risk for all but the HR loans. Remember, risk is not bad, as long as you're compensated appropriately for taking the risk. The rule of thumb I've been using is that every 1% increase in the distress rate should be compensated for by a 1% increase in interest rate.

More On Verified Bank Accounts

As I had observed a few days ago, unverified bank accounts are statistically dangerous for loans. Over 83% of listings that closed with unverified accounts are showing signs of distress (late payments or in default). For mitigating this risk, it's important to understand what this represents.

All loans require a verified bank account, but Prosper does not require it be verified during the listing phase. If a fully funded listing indicates that it does not have a verified bank account, yet the listing was fully funded, the borrower had to get the account verified during the review period. So, what this statistic is indicating is that borrowers who couldn't or wouldn't verify their account in the listing phase are very likely to fall behind on payments.

For this reason, I only look at listings with verified bank accounts. This strategy only works as long as there are other lenders that don't mind unverified bank accounts, and, so far, that's not a problem.

Fixing Those States

For some reason, some of the Listings in the Prosper XML export have odd state codes. I've traced down three of these. AA and AP are synonyms for LA (Lousiana), and AE appears to be NY (New York) in disguise. I have no clue how this one came about.

Update: RateLadder has pointed out that these are US Military Base codes. A little Googling found these definitions:

• AE: Armed Forces Africa
  
• AA: Armed Forces Americas
  
• AE: Armed Forces Canada
• AE: Armed Forces Europe
  
• AE: Armed Forces Middle East
  
• AP: Armed Forces Pacific
  

Getting Loan/Listing Pairings From ProProsper

ProProsper doesn't make it trivial to extract the loan to listing pairing information. It is doable, however. Click on the "Query Analyzer" link in the Tools section of the menu on the left side (after you're logged in). Enter the query info below into the "SQL Statement" box.

SELECT
loan.listingkey,
loan.[key]
FROM
loan, listing
WHERE
loan.listingkey = listing.[key]

Press 'Run' to verify that it works properly, and then press 'Export CSV' to get all 9100+ loan/listing key pairs.

Verified Bank Accounts

I view lenders as bankers. They're attempting to pick an interest rate that appropriately covers the risk of their loan while provided a rate of return. The problem with Prosper is that the statistics reporting has been too granular - it's difficult to spot trends because there's little information to differentiate different loans. All the good bits are in the listing information. Now that ProProsper has provided the link between listings and loans, it's time to analyze to see if there are any unnecessary risks that bankers should avoid.

I'm going to start by looking at the distress rate on verified versus unverified bank accounts. I consider a loan distressed if it's in any of these Prosper status categories: 1 month late, 3+ months late, Late, 2 months late, Defaulted (Delinquency), Repurchased, Defaulted (Bankruptcy). I'm approaching it this way because I'm thinking like a banker, and each of these represents risk, delay, and uncertainty - and that's bad.

Here's the distress rates based on the data from a few days ago looking at whether verified bank accounts matter:

Loans	Distress Rate
All	8.65% (789/9188)
Verified Bank	7.43% (667/8972)
Unverified Bank	83.56% (122/146)

Obviously, unverified bank accounts are a large risk. Huge. Phenomenal. I'll delve more into this later.

Prosper Around The Web

A few goodies from around the web

RateLadder discusses the Prosper site revamp

Another positive is that the site seems WAY WAY faster. They have removed a lot of images and replaced with simple text. I think the new look is cleaner as well, but that is subjective.

EBay entering the micro-loan market?

It appears that Prosper and Zopa are going to be having some competition in the micro financing world. Ebay has purchased a company called MicroPlace

And some self-help: Six Secrets of obtaining Prosper loan funding

A week or two ago I introduced a new concept in the finance world: online peer-to-peer lending through Prosper.com. I have spent some time analyzing what makes loans work on Prosper and come up with the following “Six Secrets” of obtaining funding quickly and at the lowest possible rate.

Updated: Prosper also bought back some loans because of possible identity fraud

Prosper also repurchased 9 defaulted loans, repaying the lenders their bids on those loans. Some lenders believed these 9 loans were part of an identity theft scheme because of suspicious similarities among the listings.

ProPosper Now Free

I saw over on the Prosper developer forums that ProProsper is now free to subscribers. The most valuable item I've seen so far is the data export with listing keys added to the loans. Let the really interesting data analysis begin!