Banks argue that loans should not be marked down if they’re still “performing.” As long as borrowers are meeting their contractual obligations, there’s no reason to take a writedown…This is how we become Japan. Emergency bailout facilities allow banks that otherwise would have failed under the weight of bad loans to hold those loans to maturity — pretending the bad ones will be paid off in full over time….
One problem is that it’s much more difficult to determine the fair value of a loan than it is the fair value of a security, where more liquid markets with more frequent price quotes make measurement relatively easier. With loans, banks must rely on internal models…
For instance, what estimates are banks using in their models?…Determining fair value is largely subjective…
Felix adds (my underline):
The calculation which needs to be done is pretty complex, and involves the future path of three uncertain variables. First there’s the lender’s own cost of funds: at the moment it’s low, but there is a chance it could rise substantially by the time the extended loan matures. Secondly there’s the income stream from the loan: while most of these loans are performing right now, and making their interest payments on time, there’s a significant chance of future default on many of them. And thirdly there’s the future course of property prices, or other assets securing the loans.
The last two, of course, are highly (but not perfectly) correlated: if the value of collateral declines, then the chances of the borrower defaulting on the loan increase….
In the real world, however, bankers are human, and they’re liable to fudge the figures so that extending the loan always makes sense: tweak a volatility assumption here, put in a favorable interest-rate assumption there, and it’s not hard to get the answer out that you wanted in the first place. They have a very strong incentive to do this, because much of the time if they take their losses now, they’ll become insolvent: everybody wants to survive, first and foremost.
Let’s say you are going to give out a mortgage, and you want to model how much money you’ll make off of it. One of the big variables is “recovery”, which for this conversation is identical to collateral – if the loan goes bad, how much can you repossess and what is it worth? This is why collateral is such a big deal, and why banks are so on the ball keeping an eye on it. When you model this, there’s a large percent chance you’ll get your payment, and a small percent chance the other side will default and you’ll only get the recovery collateral. Now let’s move some variables around: another identical way to say this is that there’s a 100% chance you’ll get the collateral, and a very high chance you’ll also get the difference between the collateral and the full payment. The recovery is a floor of value for the loan.
Now in normal times, we can do this. If you have a mortgage, and you go into foreclosure because of an unexpected health care cost, unemployment spell, poor moral hygiene, whatever, we can do some statistical magic to get a good sense of how much we can resell your house for, how much we can ‘recover’ from the loan. Indeed we do it when we give you the mortgage, to figure out what kind of interest to charge.
What about not normal times? What if you go into foreclosure because the economy is collapsing, and during times when the economy is collapsing is also the same exact time when the collateral isn’t worth much? The quants call that recovery risk or ‘downturn lgd’. The moment you most need collateral, the moment where most of the value of the loans in the models will be determined by collateral, is the moment when the collateral is worth the least.
So Felix is right about the correlation part I underlined, but there’s another correlation (they sneak up on the financial engineers!) hanging above this – call it correlation to macroeconomic state, or just plain-old unemployment. If the economy takes a nose dive, and unemployment rises, you are more likely to default but what you recover, that important floor in value, also takes a dive, since nobody wants to buy your collateral.
The book at the beginning of this entry calls it “The Next Challenge in Credit Risk Management” (it’s from 2005). Here’s a paper/article from Risk Magazine’s “The Cutting Edge” feature in 2007. These are ironic since it is in fact the current challenge we need to deal with right now, not something for the future’s sake. How’s it doing? Now asking some risk managers how to calculate this recovery risk is like asking your magic teacher how to make a horcrux. To avoid speaking too much, I’m going to kick it to Wikipedia to discuss how this works in practice (my underline):
Under Basel II, banks and other financial institutions are recommended to calculate ‘Downturn LGD’ (Downturn Loss Given Default), which reflects the losses occurring during a ‘Downturn’ in a business cycle for regulatory purposes…
The calculation of LGD (or Downturn LGD) poses significant challenges to modelers and practitioners. ….Many institutions are scrambling to produce estimates of Downturn LGD, but often resort to ‘mapping’ since Downturn data is often lacking. Mapping is the process of guesstimating losses under a downturn by taking existing LGD and adding a supplement or buffer, which is supposed to represent a potential increase in LGD when Downturn occurs.
Trying to use statistics to figure out what recoveries on bad housing loans while there is an increasing foreclosure rate for the next 18 months, the housing market remains weak and U3 is approaching 10% is like using statistics to figure out what it’ll be like when aliens land – there’s no real history of data for this. So they ‘guesstimate’ it.
If you think that much of the poor underwriting associated with mortgages in the past 5 years was predicated on house price increases, then the recovery here drove most of the value of the mortgage. If the recovery is very high because the house went up in value, you might be able to top the contract off with enough fees and penalties to make you somewhat indifferent on whether or not the loan gets paid as long as you get the house to resell. The turnaround is that if the value of the collateral becomes a disaster, something unknowable, the state that you are in is more confusing than ever.
Some additional thoughts:
1) This is one of those situations where if we had some simple dumb rules as to what to do to assess this value of recovery in housing markets, rather than leave it to statistical fortune-telling on a data set of size zero, it would clean up a lot of uncertainity. It may be inefficient; but note that when we need this information the most – whether or not there’s enough capital to survive the next two years of a major crisis – is when we are the most confused.
2) Guesstimating on the local level also requires regulators to guesstimate their guesstimates as well. It’s very arbitrary, and custom designed to increase ill-will between smaller non-connected banks and their federal regulators, an area I want to dedicate more thought to in the upcoming weeks.
3) Having another stress test for the largest banks, this time with the models and guestimates in place already, might give us significantly more information. There will be less picking-and-choosing models and data sets, and we’ll see how the actual largest banks are performing with new data over time but the framework consistent.
4) This is also how real economies get wrecked. The cycle between unemployment and underwater and foreclosed mortgages and insolvent banks will continue to spiral even though the largest 20 banks are backed by our government. One way to cut this knot would be to allow underwater/foreclosed mortgages to become renters, with a note to allow the books to keep the property on at full price. Now the collateral can be on the book at a marked-down price tailored to the end of the renting agreement, while the stream of income rent payments all goes to profit instead of principle reduction for the foreseeable future. People more familiar with this part of the Basel – that would be a win, right?