more than just soft information is lost when lenders rely on generic credit scores. Practical obstacles — and in some cases political considerations — exclude from the scores factors, such as income and education, that self-evidently affect creditworthiness. Moreover, score-based lenders, like Friedrich Hayek’s central planners, rely on “statistical information” that ignores “crucial circumstances of time and place.” From their far-away perch, they cannot recognize substance abusers, nor can they distinguish workers in plants scheduled to close from judges with lifetime tenure.
He argues that we need more traditional banking, involving credit judgment and originate-to-hold, and less securitization using credit scores. My thoughts.
1. Although in theory underwriting judgment could lead to wise decisions to over-ride credit scores, in practice I do not think this happens. Human underwriters are not geniuses. And they do not necessarily know when the information they have is really news to the scoring system. Maybe the scoring system does not explicitly know that someone is a substance abuser, but it may nonetheless observe behavior that reflects that substance abuse. I believe that judgmental over-rides tend to lead decisions that are worse, not better.
2. As Bhide points out, some of the shift toward originate-to-distribute was influenced by capital regulations.
3. In mortgages, the private securitization market remains pretty dormant. That is, without a guarantee from Freddie and Fannie, investors are reluctant to buy mortgage securities just based on loan-to-value ratios and credit scores.
4. I think that for bank regulation, stress testing is the least bad way to promote safety and soundness. Stress tests for mortgage portfolios should include scenarios of falling house prices. Stress tests for consumer portfolios should include rising unemployment.
Between the early 1990s and the peak of the housing bubble, there was no significant increase in the total market share of securitized mortgages. This doesn’t address the comments above directly. But, as with so many conversations about the housing bubble, it seems clear that it would be a wholly different conversation if it wasn’t perched atop a foundation of shared, faulty presumptions.
Even at the height of the private securitization boom, it was GSEs and FHA losing market share. Much of that, clearly went to private securitization. But banks and other private lenders were also taking some of that market share.
Facebook, Google, and the others know a lot, lot more about individuals than what a mere credit score could reveal, and supplemented by some financial history could probably build a profoundly superior statistical product. I’m guessing the investment market would pick up a lot with access to those more predictive Info-Scores (depending on legal / political limits).
Then again, also depending on those limits, maybe Facebook, etc. would keep the info in-house and invest their own cash hoards. Maybe Cowen is right that in the end, Google becomes a hedge fund, though not just with their super-AI’s trading, but investing on the basis of in-house personalized info too.
It seems to me that stress-testing would have one of two predictable consequences:
-(>90%) The banks see the stress test results, hate the numbers, fiddle with the assumptions to change the numbers, and go on as usual.
-(<10%) The banks see the results, realize they're much more vulnerable than they had believed, tighten credit while fleeing to quality, and cause a recession.
If each bank is allowed to choose its own path, the ones that choose option 1 will offer higher returns than the ones that choose option 2 (unless the stressful situation comes to pass within a year or two). Option 2 management teams will be replaced by their boards or in buyouts.
Am I missing something?
The banks don’t have nearly as much autonomy in developing the stress tests as it may appear. The FRB and OCC dictate the underlying macroeconomic scenarios (which do include massive cuts to HPI and sharp increases in unemployment, among many other stresses). They also have very strict criteria in place to judge the quality and review of a bank’s internal models and projection methodologies.
In my opinion, the risk is not that the stress test is not stressful enough, or that banks will somehow opt out of the stressful components. The risk is that the simulated stress is very severe, but one that is implemented in an identical fashion in each bank. Regulators are providing a strong incentive that each bank align their risk exposures. That’s great news if the next crisis looks anything like the last one; Banks will be well-prepared. It is horrible news if the next crisis happens in a way is not caused by underestimated credit risk. Every bank will have the same portfolio so to speak.