subprime

so anyway, this is quite typical of the ill-informed rubbish people are writing about the subprime crisis.

the bare bones of it are easy to understand. this is how an RMBS works:

Lender A has 1000 mortgages. You can expect x amount of money over the period of those mortgages. you can also expect some number of the mortgages not to be paid. people lose their jobs, get sick, get caught by rapidly increasing interest rates. this is the credit risk of the mortgages.

Lender A is only allowed to keep so much risk on its balance sheet. it's the deal between banks and the rest of us: they do not have to have enough money to pay everything they owe, but can only hold the potential to lose this much and no more. the rules were fixed by the Basel Accord.

but every loan Lender A can make increases the benefit flowing to it. so the question is, how can it make the lovely money from loans without taking on more risk than it is permitted? the answer is securitisation.

this is what Lender A does. it sells the loans to a special trust (they have various names, including structured investment vehicles and special-purpose entities). Lender A therefore gains benefit for having made the loans--and unless it provides credit support, its interest in the loans is now over. the money the borrowers pay goes to the SIV (it owns the loans). but how does the SIV pay for them? it has no money of its own, even though it will get all the cashflow from the loans. well, the SIV sells the cashflows from the loans to investors. in effect, it borrows the capital to pay Lender A from the investors, and gives them promissory notes in exchange.

these notes are a very good deal for investors, generally, and are not especially risky for them. this is because they are carefully structured, with well-chosen loans that have carefully figured default scenarios based on historical records (basically, if you have had 10,000 similar loans and only a couple of hundred went bust, you can draw a normal distribution and say that at worst, y loans will go bust with 95% confidence) and have "credit enhancements", which are mechanisms to prop up the notes if loans go bad (such things as having more than enough loans to pay the notes down -- overcollateralising, and having insurance for the top tranche). the most notable structural enhancement is usually that the notes have more than one level, or tranche. the A tranche will generally be rated AAA; the B tranche generally A. the B tranche carries much more of the risk, and goes bust first, giving support to the A tranche. the notes pay interest from the cashflows from the loans, and pay the principal back on the expiry of the note, if there's enough money.

so what went wrong in the subprime crisis? basically, when RMBS are composed of "prime" loans (loans to people with excellent credit histories and good jobs), they are very sound investments, but do not make you much money (because the borrowers are not charged high interest rates). the A tranche of a prime-loan RMBS is a very safe investment, probably suitable for Granny's pension fund (which would generally only invest in government bonds). but some lenders made loans to people who did not have good credit histories and did not have good jobs. these loans are called "subprime". from a risk-analysis point of view, the most significant difference between these types of loan is that if you make default scenarios for prime loans, you can be fairly sure they will be accurate in most financial conditions. sure, lots of people will default if there's a recession, but you can model that and allow enough room in your model for significant defaults (overcollateralising by sufficient loans that even if the worst happens, you still have enough paying loans to pay the interest on the notes). however, subprime loans are more unpredictable. so the risk is harder to quantify.

but lender A wants those loans off its balance sheet. they are very risky (although very lucrative). so it makes the RMBS out of them very attractive to borrowers. it can do this because the interest rate a note pays is a function of the interest rate the borrower pays on the loan. if you're paying 7 per cent and the note is paying 6.5 per cent, that works. (in case you're interested, the interest on the loans will be the central bank rate plus a margin, which I think is usually 2 per cent; and the interest rate on the notes is generally the interbank rate, which is similar to the central bank rate, plus a margin that is usually a percentage point or similar--and usually the SIV has a deal with a third party to take on the risk that the interest rates become very different). subprime borrowers pay very much higher interest rates than do prime borrowers, so subprime notes have much higher interest rates too.

the crisis has occurred because a/ banks bought tons of the subprime RMBS because their risk/return ratio seemed very acceptable (they pay luscious interest rates for what was sold as reasonable risk--remember, they are only allowed so much risk, and if they can make more from it, all the better) and b/ many more loans defaulted than the models suggested would. it's key to understanding this that you grasp that everyone involved relied on Lender A telling the truth about its borrowers. no one else gets to see the raw figures. i doubt anyone did anything downright fraudulent, but certainly massaging the figures would happen.

so back in the summer (or winter, here), people started realising that they were holding bad paper. the default rates for subprime lenders had risen, and investors realised that they were holding much more risky investments than they had thought. so they wanted to sell, fast. the price of subprime RMBS on the market fell.

so why is there a "credit crunch"? because no one knows who is holding how much of this shit! but everyone knows that banks are holding a lot of it. they treated the AAA tranches of RMBS as though they were Treasuries, which they are not (Treasuries are effectively risk free--if you cannot trust the US to repay its debts, who can you trust?). so banks are refusing to lend each other money, because they don't know how risky it is. this is quite astonishing. we are saying that, say, Citibank will not lend HSBC money because it is uncertain that it will be paid back. this notion was exacerbated by the collapse of Northern Rock, a previously sound bank.

***

so what does Keegan have wrong? i'll give just a flavour of how bad his article is:

banking operations, via such wonderfully opaque financial 'instruments' as collateralised debt obligations (CDOs) and structured investment vehicles (SIVs), became increasingly untransparent.

opaque must be a synonym for "i can't understand them" because CDOs and SIVs are perfectly transparent if you know what you're looking at. (in a CDO, you might have a pool of, say, car loans, which pay the interest on notes in the same way as the mortgages in an RMBS.) the problem is not that they are not transparent but that they are difficult to "look through". you cannot be sure that the risk you are supposed to be taking on fairly reflects the risk that the underlying obligations carry.

SIVs sound like real vehicles. Unfortunately, they don't seem to work very well.

this simply isn't true. they work extremely well for the purpose they were invented. they very effectively transfer risk from the lender to the investor. this is not the problem with them at all.

Similarly, it is not much good if bank balance sheets look wonderful but there is insufficient liquidity in the system.

this is just plain meaningless. he is confusing two things: a/ improving your balance sheet by transferring loans off it and b/ improving it by investing in bonds that seem to offer good risk/return ratios. the first thing is not affected by the subprime crisis: it remains good business for banks to create RMBS. the second is affected: people are now unsure whether banks really do have the financial positions they claim to.

Things were not helped by what people saw when they were forced to put those 'off balance sheet' items back on the balance sheet.

this is plain ridiculous. no one has to put the loans back on their balance sheet. nothing can make them. they no longer own them. the investors do, in effect. the subprime crisis does not affect lenders in this sense unless they rely on RMBSs to provide them with capital for their operations (as Northern Rock does).

Banks and building societies like to know with whom they are dealing. I say 'like', but of course should say 'liked'. In the world of securitisation and the 'parcelling out' of risk, trust and intimate knowledge of what things were worth and who was holding them became old hat.

this is total nonsense. banks have to be able to quantify risk. that is all. it goes no deeper than that. you can call it "trust" if you like, but the truth is that they trust the money, not the person. if you pay, and keep paying, they "trust" you. if they think you'll keep paying, they "trust" you. if they think you might not, they don't. that's all.

So what about the impact of all this on the real economy? Nobody really knows.

no, they don't, but mostly the range of possibilities goes from bad to worse, particularly given the bursting of the housing bubble in the States.

posted by Dr Zen @ 10:20 am