Lemons and toxic assets
Following up Geithner versus Darling by Robert Peston:
There are three reasons why the assets held by banks (loans to weak borrowers, packaged in whatever way) might not be worth very much:
- The borrowers really can't (or won't) pay them back
- There is not enough liquidity so the banks can't sell them
- There is a "lemon" discount based on asymmetric information
- [Updated] Buyers or sellers of the assets are somehow irrational. I don't analyse this option in the article below but will try to work out for a future article whether it is a major factor. I am trying not to suggest a bounded rationality explanation for everything in the world!
If the answer is number 1, then the banks have screwed up and we may not want to throw more good money after bad by buying the loans. If the answer is number 2, we should be happy to buy the loans as this will both offer liquidity to the banking system and a profit to the taxpayer. If it's number 3, things are a little more complicated.
In reality of course, all three reasons are contributing. We don't know the proportions but we can do some exploratory work to make a guess.
I want to go into the third point in particular, as it gives us a lot of insight into whether and how the taxpayer should contribute. I'll start with a link to Mark Thoma's very good explanation - though it outlines a simple version of the problem without going into the solution yet. Come back once you've read it and I'll continue.
Back? He outlines some of the problem quite nicely. I will put it in more concrete terms, because it illustrates a more serious problem too - the future market for credit.
Imagine you have 100 people - 50 of them are architects aged 45, employed by prestigious architecture practices, and 50 of them are lawyers aged 35, employed by prestigious law firms. In 2006 all 100 people would have had no trouble getting personal loans for, say, £15,000 each. The banks would have extended the loans on the assumption that even if a few lost their jobs, the majority would repay the loans with interest. A bit of employment insurance, some reasonable assumptions about the assets that these people have to back up the loans, and you'd make some fairly safe assumptions about getting your money back.
Now imagine that we are in a recession (not too difficult I hope), and that all 100 people have just been made redundant. How much do you expect to get back?
Here's the thing: as a society, we know with a high degree of certainty that nearly all of these people will get good jobs within the next 18 months. The recession will almost certainly have ended by then, and we know that these people have the skills to earn a decent income. Sure, the odds are a bit lower than before - perhaps 10 out of 100 or even 20 will be unemployed for a long time. Perhaps their future salaries will be a bit lower than those from their previous jobs. On average, the population's income will be about static next year and go up by 3% after that. But this is unevenly distributed - maybe 20 people lose all their income and the others get a 15% increase.
We still expect to get a high proportion of the money back. But the return is asymmetrical. On the people who have lost their income, we probably lose all our principal. However, the people whose income is up by 10% don't have to pay 10% more - they just repay what they were going to pay anyway. So if you buy this asset, you're definitely going to lose 20% of its value. This is an essential property of debt finance and is one of its flaws.
Now, a new question: are you willing to lend the same group of people more money?
If you could lend to the whole group, you would be confident of getting a good 90% of your money back. You would set the interest rate accordingly - let's say three percentage points higher than before the recession - but overall, you will still make a profit.
The problem is this: the private sector can't lend to the whole group - you can only lend to individuals. You don't know which individuals will default, but the individuals do - or at least they have a better idea than you do. So if you charge three percentage points extra, the creditworthy individuals won't borrow (or won't borrow as much). You end up with only the bad risks, and you lose most of your money.
But again, as a society we know that the group as a whole is going to have an increasing income next year and so can afford to repay.
The conclusion: the state, because it can average across a larger group and because - in this scenario - it is the only lender willing to step up, can get a more stable and more reliable return than any private sector lender.
This applies both to the purchase of old assets and the issuance of new lending. And so there are strong economic arguments for the state to buy, or guarantee, these assets.
Now let's go back to the other two reasons for asset discounts.
1. To whatever extent the debts are really not going to be repaid, the state should give the creditors a haircut. Mitigating this is the fact that the state is in a (relatively) good position both to know the answer to this question and to influence it through economic stimulus. Therefore we should worry a bit less about this than the worst-case scenario would imply. But we should still worry. This component of the asset discount is genuine.
2. The liquidity argument supports state purchases - because the state can never run out of liquidity, as it can print new currency as required. So it should use its comparative advantage in selling liquidity to the private sector in return for undervalued assets. This part of the asset discount disappears as soon as the state acquires the assets.
And the third discount, as discussed above, is much less for the state than for other actors.
So if the second and third elements of the discount are at all significant, the state can make a profit from acquiring the assets. My feeling is that they are, and so it makes perfect sense for the Treasuries in both the UK and US to pursue their respective policies.
But it is tricky to work out the balance between the three factors. Perhaps we can make a model to evaluate it. More on this tomorrow.