Paragraphs to ponder

The pursuit of “risk
sensitivity” led to a re-organisation of bank assets away from lending on the
basis of the banker’s private views about the borrower – regulators considered
this hard to quantify and a little suspect – towards lending on the basis of an
external credit rating. The higher the rating, the lower the capital banks had
to set aside against the loan. Regulators saw this as not only risk-sensitive
but transparent and quantifiable. Banking by numbers was oh so modern.

Here is much more, interesting throughout.


Isn't this an argument against globalization (expansionism in general) and for localism?

I've offered this statement up before - and I'm not sure if most here are coming at this from some tacitly agreed upon party line - but it seems almost universally rejected: perhaps risk-taking in general isn't something to be lauded.

Risk taking in certain endeavors, sure, but in core industries (e.g. banking) it's probably not such a great idea.

And you probably ought not be encouraging it.


Risk taking should be encouraged (by the natural rewards). People are too conservative by nature.

The problem is taking risks when you think you are being conservative because you are relying on ratings, diversification, quants and herd mentality to protect you. When the whole herd goes off the cliff, safety in numbers doesn't save you.

It's no party line, here. I just don't understand why people are so worried about 'saving' a banking system that so obviously sucks. I mean, it's the only system we have, but it's just not that great. The real economy is still stumbling along despite this rattle-trap finance system. Now, what party would endorse THAT line?

"Risk taking should be encouraged (by the natural rewards). People are too conservative by nature."

This is what I refer to: you say that as if it's some natural law. By what metric are people "too conservative?"

I don't know how you can say that when, for example, in this country we have negative savings rates. Or, say, as we stand on the precipice of the most spectacular financial meltdown in the history of this country (if you are to believe Paulson & Co.).

Maybe relative stability is more important than chasing after that extra 5%.


Maybe a better way to phrase my view would be that people are too enamored of the status quo, I'll have to ponder that.

Regarding lending based on privately held view: My family is from a small Southern town which used to have 1 local bank. In the 70's, my parents went to borrow to buy a Honda. The bank president said no because he didn't want to lend for a Honda. My grandparents had gone to the bank in the 60's to borrow to by furniture, and he wanted to know what type and from where first. Another family went to take out a home equity line of credit to build a swimming pool. He said no because he didn't think they needed one. Any system will work great if the people running it are saints.

This seems very borrower-focused. Isn't the relevant hard-to-quantify judgement not "the subjective evaluation of this borrower" but rather "a subjective evaluation of whether we are in a bubble?"

If housing prices decline by 10% a subprime borrower is at risk of default. That seems like a very quantitative problem with subprime mortgages to me, which has nothing to do with the trustworthiness of the borrower.

Now, Andrew Kling has suggested a related point: lenders started "learning" that credit score was the most predictive metric of risk, but they learned it within a very peculiar environment (a bubble). But this isn't too much quantitative reasoning, it's stupid quantitative reasoning. You can't extrapolate from a bubble what will happen in a non-bubble.

Income levels aren't part of credit scores, and they probably shouldn't be. (I know low-income people who take debts very seriously, and some rather rich people who refuse to pay their bills unless they are sued. Probably why they are rich.)

But they should definitely be part of the mortgage information. Although lenders may have been over-relying on credit scores, and incomes may have been falsified, I would be moderately surprised to find out that the income level was just missing.

Just some thought:

Risk of default is variable and dependent on many factors, some of which are "risky" in themselves (e.g., health and divorce). Then, if the loan is for 30 years you need to predict whether a default will occur over the lifetime of the loan.

My gut says that people are generally have positive risk-affinity since the downside, for many, is floored. The CEOs may fail to get money but they rarely lose any no matter how poorly they do. The average Joe, and his 401(k), have the government to protect them. Whether or not we "encourage" risk taking and profit-mongering I would suggest that they will occur anyways since those holding power are pre-disposed to that kind of behavior.

"If housing prices decline by 10% a subprime borrower is at risk of default." - this may be the reality (if you assume 7/1 ARM, 80/20 piggy-back loans for sub-prime borrowers) but I would suggest that both effects are incidental of each other; though sharing similar causes.

Extending the linked piece, it seems that the ratings system has been hijacked. Instead of rating a company and its ability to make good on its bonds the models are being bent to not only rate the first-tier borrower (the mortgagee) but ALSO the second-tier packager (the mortgagor).

Also, from the piece: "One of the implications of this risk sensitivity is that bankers were given incentives to enhance the credit rating of lending to reduce their capital charges and improve their profitability."

I am getting confused as to whom he is referring since in these cases the packager is getting the cash and doesn't need to setup reserves - regardless of the rating those packages get (right?). Maybe its a matter that the reserve requirements information is lost when the repackaging is performed. If I package an "AAA" and a "BB" together I only need to hold reserves as if I had an "A" as opposed to "AAA" + "BB" (average vis-a-vis sum). I guess the comment about covariance falls in here since it they are independent then an average is more reasonable whereas with dependence a sum is necessary.

In all I agree with the article's premises and, to phrase it my own way, the models used did not include any subjective inputs nor were the models used in support of a subjective process of approval/denial. If only single mortgages were done this way, however, there really wouldn't be that large of a problem or the uncertainty in how to go about patching things up. It is because the re-packaging models were designed/used this way that the problem has gotten to where we are today.

Comments for this post are closed