Monday, March 12, 2007

A Soap Box Soliloquy on Information Cost and Credit Scoring

As I mentioned previously, I have this new (albeit passing) interest in after a discussion we had in Economics.

While researching Prosper, I found, a blog that Kevin keeps about his investing activities. Though we're pretty much done with Microcredit in class, and I'm not interested in investing through Prosper at this time, I added Rateladder to Google Reader because I find the research that Kevin does interesting.

This post about 3 sigma loans caught my attention:

The data is a little old (2am last night.) But here are listings that at that time where in the 3rd standard deviation. The list of 2nd and 1st standard deviations were much longer, but probably more attainable…

Would you be interested in have this list emailed to you every morning? How about just the auto funding loans were the lender rate is a known quantity? Send me an email (kevin (at) rateladder [dot] com to find out more.

A Credit, $12,800 @ 21%, 23% DTI:

A Credit, $25,000 @ 20%, 26% DTI:

A Credit, $7,100 @ 23%, 22.5% DTI:

I looked through the listings, and was curious why Kevin wouldn't invest in any of these. He has standing orders set in Prosper to only find loans with 0 current delinquencies, 10 or less delinquencies in the last 7 years, and 2 or less public records in the last 10.

I decided to email him and ask:


As you can tell from my blog, I’ve been following yours for a couple weeks. I’m at IT guy by trade (was a DBA for several years) and I’m in school now part time to finish my undergrad in business. I’ve taken some econ and stats classes, and I’m fascinated by your take on prosper listing (where I think most people probably bid based on emotion or gut, you have quite the eye for the statistics).

In short, I’m not really looking to get into investing through prosper, but the whole notion intrigues me. If I did start investing it would just be to say I did it.

I’m curious about your latest post regarding 3 sigma listings. The first one for example (house flip/A paper/21% rate) — why wouldn’t *you* bid on a loan like that.

Your blog is the most interesting feed I have in Google Reader right now — keep up the good work. =)


Look kids, a blog circle-jerk.

In all seriousness, what has kept me reading is his attention to detail. Prosper allows you to download their entire database every night, and people like Kevin are using this to build models around their data. He's even gone so far as to build a suite of tools for power lenders to analyze the data and take the emotion out of lending (which is what the borrowers want by posting pictures of premature babies and puppies with their leg in a cast).

More important than Prosper itself or Rateladder's tools or Microcredit as a concept, what I think is the real story is Prosper's open access to data. Information asymmetry talks about the inequality between borrower and lender: the borrower only knows what the lender tells them (or is required to tell them to make the loan). On a level playing field, this can lead to adverse selection (bad credit risks are most likely to seek out loans) and moral hazard (once the loan is made, the borrower might engage in activities that are not in the best interest of repaying the loan). Lenders figured this out years ago, which is why Equifax, Experian, and Transunion now collect borrowing and repayment activities for all of your trade lines.

To make heads or tails of that information, your credit history is fed into a model that scores your credit-worthiness. Based on this number, lenders decide if to give you credit, how much, and at what rate. If you ask Andrew, who spends his work days lending people money, these guidelines are sometimes more flexible than others. Many lenders will discard outliers or jack up the rate; what makes Andrew so good at his job is validating and challenging the number a model spits back at him, not being a slave to it.

While Prosper runs the credit history of lenders and assigns them a score, their model is subject to interpretation. Kevin from Rateladder is using both the credit information and the data he gets on all Prosper lenders on a nightly basis to build his own scoring model for those borrowers. He is, in essence, trying to reduce the information asymmetry between Prosper borrowers and lenders to relieve moral hazard and adverse selection.

The big three credit bureaus aren't letting just anyone have access to their information, which in this case can be to your detriment. Your beacon score goes up or down based on what zip code you live in, even though you're the same borrower you were before you moved. In the Prosper model, not only does the Invisible Hand decide what you're interest rate is going to be, but lenders have the ability to make better decisions about what risk you are by popping the hood and looking at all the data available, not just a number that comes out of the hopper.

In my book, that's pretty fucking cool.

No comments: