Hacker News Reader

Comment by kristjansson | Hacker News Reader

The problem with MBS was always the zero-sum nature of the alchemy. They took 100 low-quality loans in, and returned 10 high-quality loans, 30 ok-ish loans, and 60 dog-shit loans. No harm no foul, until the dog-shit tranches were marketed as ok-ish, and alchemists believed they we're really creating gold.

View on HN

I talked to a banker at the time (from BoA I think) and he said these things came about by the banks re-arranging the structure again and again, running it through the risk and credit rating algorithms and then tweaking and retrying until they got something with a good rating. So essentially that's like finding exploits in the other parties risk calculation algorithms or credit rating algorithms until you find some hack that labels crap as gold. So I think of it as hacking basically. The banks figured out how to hack Moody's algorithms, as it were.

Don't forget through all of that madness Moody's was rating utter crap as AAA

https://en.m.wikipedia.org/wiki/Credit_rating_agencies_and_t...

If you have a metric, eventually it'll be gamed.

CDOs weren't the problem. You see, mixing together dog-shit __ONCE__ isn't that big of a deal.

CDO-squared (CDOs of CDOs) were a problem. If you took those 60 dog-shit loans, mixed them together, and created 10 high quality loans out of them, problems began to occur.

Even then, it wasn't the CDO-squared that collapsed the whole thing. It was the CDS: the "insurance" (so to speak) on the CDO-squared that made things go haywire.

I am fascinated by the number of layers of abstractions built on top of each other in these financial products... Bankers seem to rival programmers in their ability to stack abstractions on top of each other until it is impossible to understand how the underlying system relates to the top layer.

You could say that MBS/CDOs were 'leaky abstractions', and that there were equivalents to 'code smells' to suggest they weren't the golden debt they masqueraded as. I can't think of other professions that develop the same level of abstractions that programmers and bankers create. Layers and layers of inputs and outputs.

You could also say they were deliberate fraud.

The number of big-name banks whose actions are questionable [1] - if not actually criminal - is astonishing.

[1] As reported by insider papers like the Financial Times.

A structured products desk is basically a microservices team

Nested leveraging is a common theme. Each layer is leveraged and but supposedly protected by another layer's risk management.

Go through enough layers and you eventually hit an insurer that has zero chance of being able to cover all of the losses of everyone. It could be CDS of CDOs of CDOs, or coins being backed by other coins.

Thankfully the average person is less affected by crypto losses than mortgages, or in the case of the depression, stocks.

Actually AIG and other bond insurers really got the bailouts too.

I remember a guy bragging about helping another firm backdate its books to get bail out money. He was bragging. Still remember his name and how he got his pill head daughter a Cush company paid job in London.

Were the risk models assuming no correlation between the risk of the underlying securities?

Mortgage-backed securities weren't exactly the problem, it was merely the vehicle in which a trans-Atlantic banking regulation arbitrage trade was carried out, and the trade accidentally became fully self-financing. US mortgage originators and banks were allowed to take risk if they sold it off onto the capital markets, and EU banks were allowed to take risk as long as it was deemed "safe" by ratings agencies.

The truly pernicious part is the self-financing spiral. Banks issued mortgages, which deposit cash in the home seller's account. Seller puts the cash into a money market fund. MMF gets short-term interest by financing AAA-rated securities. EU banks use this financing to buy senior tranches of MBSes. These MBS purchases allow banks to issue more mortgages by freeing up their capitalization.

Like, zero-sum alchemy will only generate small frauds and relatively limited losses. You need large amounts of leverage and some kind of self-financing spiral to generate catastrophic de-leveraging events.

The central MBS problem seemed to be ratings-based-regulation + free-market-unregulated-ratings-agencies.

In hindsight, it should have been structured like government insurance servicers: subject to random and regular audits, with $$$ fines if they turn up anything fishy.

If you don't want corruption, engineer a system where not being corrupt is more profitable...

Reply on news.ycombinator.com