Goldman Scandal?

Felix Salmon has a good discussion on NYT’s 3000-word story on Goldman and synthetic CDOs.

How Toxic Update

In my initial post on solving the toxic assets problem I focused on the foreclosure rate to estimate losses. But that may not be quite correct. The key statistic is the foreclosure rate on mortgages in the population of mortgages financed by CDO's. This rate may be higher, I need to check on that. The issue is what are the maximum losses in all the mortgages that underlay CDO's.

How Toxic are Toxic Assets?

Brad Setzer has a good post on the Treasury's new plan to deal with toxic assets. Clearly dealing with the credit crisis is the key factor in getting any recovery going. And it is the most difficult. Primarily, I think, for political economy reasons. The problem is who should bear the brunt of the losses. But there is also a coordination problem involved.

To see this, first note that so-called "toxic assets" are complex collections of mortgages. These CDO's are derivatives whose value depends on the payoffs on the underlying assets. So the losses on the whole universe of these assets cannot be bigger than the losses on the underlying assets. The problem is that nobody knows what individual collections -- CDO's are worth. Hence, they trade for little value in the market, and banks are reluctant to sell them at low prices; choosing instead to keep them on their balance sheets. And then they are frozen from lending and instead hoard bailout funds until the value of the CDO's pay off (they hope).

Now let us make a back of the envelope calculation. A really high foreclosure rate is 8%, so let us take 10%. If foreclosure takes place not all value is lost for the mortgage holder, but let us suppose it is. Then if one held the entire portfolio of mortgage backed securities the maximum losses on the total portfolio would be 10%. The big problem today is that banks hold CDO's of uncertain value -- nobody wants to pay a lot for a tranche that could have lots of bad mortgages. But other tranches must have good mortgages. Even if we consider CDO-squared's -- collections of collections -- the underlying risks ought to be no more than 10% figured conservatively.

Suppose then that the government took all CDO's and held them to maturity. It would, at worst, earn 90 cents on the dollar. So if it exchanged for these assets a security that pays 90 cents, and gave each bank a pro rata share (this would have to be based on nominal value not market value) then banks would receive 90 cents for sure instead of some uncertain value. Wouldn't this plan break even for the government? And wouldn't the banks -- as a whole -- be better off? Couldn't lending re-start? If the answer is yes, then the fundamental problem is information and coordination. It is a problem of who to allocate the losses to.

What am I missing?

Dedicated reader Vijay Krishna reminds me that this solution is related to an old puzzler on Car Talk:

The problem is that you have to take two identically looking pills each day. You are isolated from your doctor for 30 days and cannot get resupplied. If you don't take them both you die (you could change the proportions and volumes). But you spill the pills on the floor. What are you to do?

The solution offered by Click and Clack is to take all the pills from the floor, crush them up, mix them up in water, and divide the solution by the number of days. Then each day take a share of the solution equal to one divided by the number of days left.

The similarity is clear. Taken in total we have a collection of pills or mortgages that is not all that harmful. But arbitrary portions of the collections can be extremely hazardous to your health.

Citigroup's Descent

The failure to adequately manage risk is the theme of this article in the NYTimes. One of the highlights concerns Citigroup's analysis of the risk of its mortgage holdings:

...when examiners from the Securities and Exchange Commission began scrutinizing Citigroup’s subprime mortgage holdings after Bear Stearns’s problems surfaced, the bank told them that the probability of those mortgages defaulting was so tiny that they excluded them from their risk analysis, according to a person briefed on the discussion who would speak only without being named.

Another amazing item is that Citigroup mainly relied on ratings agencies rather than its own risk managers to assess the risks from assets like CDO's. One might understand this if Citi was just selling the assets to gullible investors (which it did), but it held huge slices itself, and it is the losses on the assets it retains that have caused such large losses.

It is amazing to read how a company could lose $220 billion in market value in just two years. As Brad DeLong emphasizes, however, the losses in market value dwarf the announced losses on CDO's and other real estate loans. DeLong's argument is that the risk premium on holding Citi equity has risen as the crisis worsened. After all, the probability that the government will dilute the shares in a takeover has increased dramatically. He takes the rise in the risk premium as the major factor causing the value of Citi's shares to decline so dramatically.

DeLong's analysis is based on the observation that banks are institutions that trade on reputation and trust. The borrow short and lend long. Hence, they are always on the edge in a financial crisis. The risks bank take that are sensible in normal times appear insane in rough times.

Rescue package soon to follow.