Wednesday, March 29, 2006

YAFEP, Yet another financial economics perspective

This Prosper.com post rips off by my own, earlier post, and only about 20 million financial economics 101 classes…and applies

1.I suspect that sustainable return rates, after adjusting for default, from E borrowers +- one grade, will be far better than stock returns. My real job, my own research, my academic experience, all lead me to believe that over the long term, stock market returns will be in the 4-10% range. Or, you could just assume what most mainstream economists believe -- 4-10%. It is certainly possible to be investing during an unparalleled era of prosperity (most of the 90's), but as a result, people's expectations of long term stock market returns are ridiculously high -- 20%+ long term comes to mind. In contrast, the risk factor of lending to individuals w/poor credit seems unlikely to dry up anytime soon. That's where the edge is, I bet -- you can make money from disintermediating banks when lending to A borrowers w/minimal risk, but also minimal returns -- its going into the subprime markets where banks fear to tread, where the inefficiencies are really juicy (and potentially, so too is the risk of losing your shirt).

2.I suspect that large, well diversified baskets of loans are safer than a similar basket of stocks. Let me explain what I mean by "safer": To rehash stuff I've posted before:

a)insofar as you believe default correlations among individual prosper loans are close to 0/can be chosen to get close to 0 (very bad economic states notwithstanding -- I recognize the 0 assumption is totally unrealistic)...
b)insofar as you can diversify across _lots_ of prosper loans, you can get much closer to 0 risk than with stocks(again, exaggerating to some degree here as well).

In one paradigmatic financial economic framework, individual stocks have systematic risk (risk that anything in the stock market possesses) and idiosyncratic risk (risk particular to that company only), uncorrelated to other companies' idiosyncratic risks. By diversifying over enough companies, you push the risk of your portfolio down to just systematic risk, and that's what economists tell you to do.

Similarly, for prosper, I would say that the proportion of idiosyncratic risk for a specific loan to the systematic, personal loan market risk, is far greater than the analogous ratio for individual stocks, so w/sufficient diversification, your rock-bottom market risk, in absolute terms, is much lower.

Realistically, the 0 correlation assumption is fairly unlikely, because even if you look at subprime lenders like Providian, you see that they got knocked around when the economy went downhill and subprime borrowers defaulted a lot more...However, insofar as we aren't just doing credit card loans (and hopefully, not getting greedy, like Providian management did)-- instead, we pick from everything from breast implants to cat breederies to student loans to travel cards to medical loans, we do get better diversification protection...though, as always, this makes Experian historic default probabilities even less likely to be appropriate for our loan predictions.

Note: An imperfect, but useful example of correlation values as ranging
between -1 and 1:
-1: if stock a goes up 1%, stock b goes down 1%
0: no relationship between price moves
+1:if a goes up 1%, stock b goes up 1%
(You could also replace all of Stock B moves with some other constant factor -- that would still keep the correlation measures constant)

Note: I'm to some degree conflating levels of risk w/correlation between assets, which isn't the ideal way to describe these phenomena...

Note: Another example of what I mean by risk...say you have 100 million HR loans of equal size and riskiness, and every single one is completely unrelated to the others (again, a hugely extremely wrong assumption I'm making for the sake of argument). If so, if the true default rate of HR loans is 19%, it would be infinitesimally unlikely for the number of defaults to be very far from 19 million. If you know for sure that this will happen, it's not a particularly risky asset anymore. Or, think about betting on a fair coin -- heads, I give you a million dollars, tails, you give me 900,000. Clearly, even though you expect to make $50,000 from this bet, you probably don't want to take it on because it is too risky (unless you're a lot wealthier than I am :) ). But what if I told you that for each flip, I give you 10 cents/you give me 9 instead, and that we could play 10 million times? You have the same expected return, but you're almost assured to make close to $50,000 every time.

Note: I suspect the 0 correlation assumption rapidly weakens w/credit grade.

Note: I'm really arguing for better Sharpe ratios for consumer loans.


Read more!
/body>