A primary goal of financial engineering is to confuse the bejeezus out of Them while remaining crystal clear to Us. There’s no point to it if it doesn’t in some way confound the expectations of some Other, whether that Other is the tax authorities, bank capital regulators, rating agencies, customers, or markets generally.¹ But the worst possible outcome is for a product to be unpredictable to whoever built it, mostly because, if it was any good, they built a lot of it, and if it blows up on them it’ll hurt.

There is an obvious tension here: complicated products serve well to confuse Them but are more likely to end up acting up on Us as well.² One fruitful approach is for Us to be just a bit smarter than Them. Another approach, lovely when it works, is to build a product that is so beautifully simple that anyone can understand it, but that has one simple conceptual twist that falls right in the particular blind spot of one particular targeted Them.³

You can bracket the question of whom residential mortgage backed securitizations were designed to confound,⁴ and just take a moment to realize: they kind of screwed the banks that did them, no? I mean, “compared to what” I guess – imagine if Countrywide had done all the lending it actually did, but kept everything on its balance sheet – but the fact that BofA has eighty zillion dollars in putback liability must be discouraging for whoever’s left there on the securitization desk. Like: the whole idea was to put some loans in a box and sell the box to investors; the investors, not you, now own the credit risk on the loans. You own nothing. The loans have nothing to do with you. Sure you signed a piece of paper saying some stuff about the loans, just before you waved goodbye to them, but why would you have read that? Those are just reps and warranties; those are for the junior law firm associates to haggle over. You sold the loans, it’s done, right?

Totally wrong. Doubly wrong: not only are you responsible for loans where your reps were wrong in a way directly related to why the loan defaulted, but you’re also responsible for loans where the reps were wrong and it had nothing to do with the default. If the reps were wrong a lot, and they seem to have been, then you end up on the hook for a surprising amount of the credit risk. You thought you sold the loans and would never see them again. Instead, you’ve got yourself a covered bond.

That’s bad! One way to fix it in future deals would be to make sure the reps and warranties are true but there are important reasons why that’s not a great solution. (These are less “ha ha banks are all fraudy assholes” and more “it is hard to control masses of people and you really want certainty and you never know what a court will find material etc.”). Another way is to revise the terms of securitizations to say basically:

• We sell you loans.
• You’ve gotten used to getting reps and warranties, so we’ll give you reps and warranties.
• But we don’t mean them, and if they turn out not to be true we’ll have various technicalities that prevent us from repurchasing the loans.

Here you can read a Bloomberg article about how Moody’s has decided not to fall for that:

Home-loan securities without government backing probably will be able to get rankings only as high as Moody’s Aa tier if “significant” limits are placed on when and how repurchases can be forced of mortgages that fail to match their stated quality, the New York-based firm said today in a report.

The Moody’s report is wonderful, just a list of “ooh here is a trap we spotted, and here’s another, and here’s another.” I like it, but I’m pretty far from an expert in RMBS documentation; I assume some RMBS structurer is going through it with a checklist and saying “aha, excellent, there are still two traps they missed.” Here are a few they caught:

The risk of loss is higher for RMBS whose R&Ws contain materiality standards that allow loans that are clearly riskier than originally disclosed to remain in the pool because they still fall within the originator’s underwriting or program guidelines or because another factor besides the breach caused or contributed to the loans’ default or loss. For example, a breach reviewer examines a defaulted loan with an initially disclosed LTV of 60% and determines that the appraisal was faulty and the true LTV is 80%. Under an expanded materiality clause, so long as the originator’s program guidelines permitted an 80% LTV, the twenty point difference in LTV would not be material and no breach would have occurred.

So “we can take loans up to 80% LTV, but weighted average LTV is 60%” means “weighted average LTV might be 80%, good luck.”

An enforcement protocol that allows borrower life events, such as divorce, death, or job loss, to render the breach immaterial or prevent the review of a loan for R&W breach will negate the breach review process for many defaulting loans, because defaults often occur in connection with a life event. When a life event occurs and borrower default follows, the RMBS may contain a riskier loan than originally disclosed and leave investors without a put-back remedy. Another provision that weakens the ability of the breach reviewer to put back defective loans is one in which the occurrence of a life event causes certain R&Ws to automatically terminate.

So “this loan is no more than 80% LTV” means “unless the borrower loses her job, in which case, we might have been kidding, LTV was 100%, good luck.” There are a few traps identified by Moody’s that you might find more sympathetic, including the basic “if there’s no breach in the first three years, we’re out.” That could be a legitimate desire for certainty, not a sketchy desire to evade responsibility. You don’t want to be on the hook forever.

It’s easy and fun and feels sort of right to make fun of the banks here a bit. They are after all in a position to just make sure that the reps and warranties are true: they are the ones doing the underwriting, so they are the ones who should be checking to make sure that it’s true.⁵ They would – in theory – be doing that if they were keeping the mortgages on their balance sheet; if they’re not doing it for mortgages they sell, then that seems bad. Moral hazard, don’tcha know.

But I don’t know. To have a functioning securitization market, you really do need two things: you need investors who believe that they’re getting what they’ve been promised, and you need originators who believe that they’re offloading the risks they’re trying to offload. Certainty is a key part of the deal for the securitizer, and I sympathize with the securitizers who want a little more certainty than the previous generation ended up with. On aesthetic grounds if nothing else: you’ve got a beautiful machine to pass credit risk from the lender to the securitization, and it seems a shame to gum it up with a bunch of grubby claims over who was lying about what on whose mortgage application. No wonder the structurers are looking for a way to abstract away from that.

Moody’s Promises Caps on Mortgage-Bond Ratings as Terms Loosen [Bloomberg]
Weak structures can prevent top ratings for new RMBS [Moody’s]

1. This claim is too strong: in a world without rating agencies, capital regulation, etc. (and arguably with weak Modigliani-Miller), you might still invent a CDO as a way to diversify idiosyncratic risk while selling people their preferred amount of systemic risk, or something. But generally you want to:

• confuse tax authorities by deducting things that they weren’t expecting you to deduct,
• confuse capital regulators with magic,
• confuse rating agencies by preying on silly points in their models,
• confuse customers by … I mean, honestly, take your pick, there’s the UK swap mis-selling scandal, which seems pretty awful, just as a recent example, or
• confuse markets generally by “creating value” by paper shuffling in violation of Modigliani-Miller.

2. You can partially address this by writing a 400-page agreement – no one else has read it, so they don’t know what it says, but you wrote it, so you ought to – but this is sometimes fails too, since you can’t anticipate every situation and since nobody actually “writes” a 400-page agreement except maybe some junior law firm associate who doesn’t really know what the product is anyway.

3. There are no perfect examples of this but some good ones; the CPDO is a massively complicated beast but at its core it’s a relentlessly simple exploitation, via Martingale betting, of the fact that S&P rated structured products only for probability of loss and not expected severity of loss. From my own experience I am partial to a product allegedly invented by Harvey Schwartz that I might write about one day when all my friends who sell it are retired. It’s a beaut.

4. I subscribe to the school of thought that answers “capital regulators,” with a side of “ratings agencies.” “Customers” is also a respectable answer and in certain specific cases clearly the right one.

5. This elides the difference between originators who make loans and sponsors who securitize them but, y’know, they can figure it out between themselves. It all back-to-backs, mostly.

Dealbreaker