Last Wednesday I wrote a highly critical post about the agreement between Bank of America(BAC) and the government (Treasury, the Fed, and the FDIC) to terminate BAC’s asset guarantee agreement in exchange for a payment of $425 million. I’ve learned some more about this and I think I can reconstruct the government’s perspective on this issue, with the help of someone knowledgeable about the transaction.
A good place to start is Schedule A of the Termination Agreement. A few relevant facts:
- BAC requested the termination on May 6. Note that this is the day after the stress test results were released, although I’m not sure how much of a factor that was. Apparently at this point BAC felt comfortable going without the guarantee. It took an additional four months to work out the terms, but Treasury and BAC decided to use January 16-May 6 as the period that the guarantee was in effect. You could argue that the guarantee was really in effect until the Termination Agreement was signed, because BAC could have changed its mind, but I don’t think using May 6 as an end date is too unreasonable.
- The $118 billion pool of assets was identified in advance of the announcement, but not down to the level of individual securities. Because the assets were mainly or entirely from Merrill Lynch, the vast majority of them were already marked to market. One of the things the government had to do was verify BAC’s marks. Another thing they had to do was verify that the assets did not violate any of the conditions of EESA, the bill that governed the usage of TARP money. This is one reason that negotiations on the initial deal took time.
- By May 6, the parties had already agreed to exclude $14 billion of assets, and disagreed about another $42 billion. For purposes of calculation, then, they assumed that they would have excluded half of that $42 billion, or another $21 billion. This brought the “covered pool” down to $83 billion. (This is a bit of a fiction since the final pool was never identified, but the only purpose of the fiction was to calculate the termination fee.)
If you accept those assumptions, the calculations on Schedule A for the Fed’s loan commitment fee ($57 million) and the foregone dividends ($69 million) more or less make sense. The calculation for the warrants also seems right. They used a strike price of $13.30 and the market price on May 6, pro-rated for the reduction in the asset pool from $118 billion to $83 billion, and came up with a warrant value of $140 million.
I have two quibbles with this so far. The first is this practice of using a preceding 20-day average stock price for the exercise price. Given that most banks are coming to the government when their stocks have just fallen precipitously, this seems like a surefire way to set your exercise price too high. But given that this has been standard practice since early in the bailouts, that has nothing to do with this deal in particular.
The second is pro-rating the amount of preferred stock and warrants by the size of the asset pool. On January 15, when they agreed on $118 billion as the size of the asset pool, both sides knew that the pool had to be verified, and they probably knew that the pool would likely get smaller. In that case, since they both knew that the pool would be adjusted when they agreed on $4 billion as the premium, $4 billion was the appropriate premium for the post-adjustment pool, not the pre-adjustment pool; the adjustment was already priced in.* But that’s a relatively small issue.
The big issue is that BAC and the government pro-rated the $4 billion in preferred stock by the effective term of the guarantee – 4 months, while the original term was 5-10 years, depending on the type of security. Because they came up with a weighted average term of 5.4 years (which I don’t dispute) this reduced the premium for the insurance from $4 billion to about $230 million (pro-rating by the size of the post-adjustment pool brings it down to $159 million).
The government’s argument for this is, roughly, that if you prepay the annual premium on your homeowner’s insurance at the beginning of the year, but then you sell it after four months, you get a rebate for the last eight months. Put another way, if $4 billion is the right price for 5.4 years’ worth of insurance, then $230 million is the right price for four months’ worth. (BAC’s argument, presumably, is that since no definitive agreement was signed they shouldn’t have to pay anything.)
My argument, on the other hand, is that it’s more like buying a homeowner’s policy when there’s a 50% chance that your house has asbestos (which would increase your liability premiums). Four months into your policy period, you get your house inspected and find out there’s no asbestos. Now you can’t get a rebate from your insurer, because in this conceptual example the insurer already priced in a 50% chance of asbestos. A similar example would be someone buying health insurance (in the individual market) at the moment that he has a 50% likelihood of having cancer. In either case, there are two possible states of the world, and you are buying insurance against the bad state of the world. When the good state occurs, you can’t get your money back.
Another way to think of this is as an option. If BAC had bought an asset guarantee for four months, I agree that the premium would have been a lot less than $4 billion because these assets could take many years to deteriorate. (However, you could also argue that since these assets are marked to market, their values deteriorate as soon as expectations of default increase; you don’t need to wait for the actual defaults). But even though BAC only used the guarantee for four months, they got more than that: they got four months of insurance, plus an option to buy another five years of insurance if they found themselves in the bad state of the world.** If the first four months were worth $230 million, then the option was worth a lot more than $230 million.
The other thing that can be said in defense of the Termination Agreement is that to get more, the government would have had to go to court to enforce a term sheet that was largely performed but never finalized as a definitive agreement, and that’s not what the government should be doing with its time these days. I’d say that’s a matter of opinion.
* Alternatively, perhaps the working assumption was that as assets got disqualified from the pool, other assets would be added. In that case, had things gone badly, BAC would have been pushing to replace the assets that got disqualified with new assets, and the effective coverage would have been $118 billion. That is, the fact that the pool only got smaller already reflected the fact that things got better for BAC after January 16, not worse, and therefore that fact should not be used when estimating the value of the guarantee.
** Actually, it’s more complicated and slightly better for BAC, because for some reasonable amount of time (six months?) they had short-term insurance and they had the right to exercise the option on long-term insurance by signing a definitive agreement. Furthermore, if you accept the underlying logic of the Termination Agreement and the metaphor of selling your house and getting an insurance rebate, even after signing the definitive agreement BAC still had the option to terminate the agreement and get a rebate. Refundable level premiums make sense when the risk of loss is uniformly distributed and unchanging over time; when the risk of loss changes over time and the buyer of insurance can see how it changes, then they are an invitation to moral hazard. Come to think of it, if you live in the fire zone of the Oakland Hills and you buy your insurance policy on July 1, should you get a 50% rebate if you cancel it on January 1 after fire season is over?
By James Kwak