Although I read it two months ago, I haven’t made much in the way of comments on Andrew Ross Sorkin’s Too Big Too Fail, figuring the ground was well plowed by others. However, the bit that I found the most shocking has not gotten the notice it deserves, so I am writing it up now.

The book presents a number of truly remarkable incidents from the crisis, so it might seem surprising that one stands out among the revelations and juicy tidbits. For instance, in what amounts to a bit of Treasury defending its record, the book recounts how Neel Kashkari developed a “break the glass” memo, so the powers that be allegedly did plan for an emergency. But anyone who has done real business planning or project management would recognize this document was useless, that it was at far too high a level of abstraction to be helpful when the shit hit the fan.

Similarly, it was an eye-opener to read that absolutely no one in the officialdom had the foggiest idea of the legal and practical implications of putting Lehman into bankruptcy. Lehman had engaged Harvey Miller, the preeminent BK lawyer, and no one bothered to talk to him. In fact, Miller’s highly developed deal sense told him Lehman was going to be rescued precisely because he was so removed from the discussions.

But there is one particularly striking tidbit that I though someone would have picked up on by now (and hopefully add their own commentary). But since (to my knowledge) no one has done so, let me point it out.

Readers of TBTF may recall that the book gives the travails of Lehman center stage, with the AIG unravelling a secondary but parallel story. Sorkin drops in the proximate cause for the AIG rescue: the magnitude of its credit default swaps portfolio: $2.7 trillion notional, with $1 trillion of that concentrated among 12 financial institution (p. 236). We get a flavor of how badly run the place is: management admits to having “antiquated systems” (p. 364) and is not able to get a estimate its cash needs (p. 365). Amazingly bankers poring over its financials in the course of trying to raise funds discover a $20 billion sinkhole that was somehow overlooked by the AIG top dogs, the result of losses in its securities lending business. That increased the amount of money needed at that juncture from $40 billion to $60 billion (p. 337). Ouch.

So the story thus far is that AIG is a great big mess that will bring everyone down if it goes. Got that. Geithner accepts that picture, persuades Bernanke. AIG is on the verge of bankruptcy, according to Sorkin, mere “minutes away” (p. 399). The Fed agrees to extend a $14 billion loan to get it through the trading day but it wants collateral. Collateral? From a broke company? How is that going to happen?

Then we get this bit:

Wilmustad understandably wondered how they were supposed to come up with $14 billion in the next several minutes. Then it dawned on them: the unofficial vaults. The bankers ran downstairs and found a room with a lock and a cluster of cabinets containing bonds – tens of billions of dollars’ worth, dating mostly from the Greenberg era. They began rifling through the cabinets, picking through fistfuls of securities that they guessed had gone untouched for years. In an electronic age, the idea of keeping bonds on hand was a disconcerting but welcome throwback. (p. 400)

WTF? This is a company about to go out of business, then it suddenly remembers it has a secret stash….worth at least 1/6 of the initial government rescue commitment? $14 billion was only what they coughed up to satisfy the Fed. How much more was left in those cabinets?

And more important, WHO SUPPOSEDLY OWNED THIS PAPER? This wasn’t held by the subsidiaries; otherwise, AIG would not have been able to pledge it to the Fed. And if it was a parent company holding, why wasn’t it repoed or sold earlier? What entity took the semi-annual interest payments? Take the $14 billion we know about, and assume a 5% interest rate. That’s $700 million. Where did it go? Was it reinvested? Disbursed?

The language further suggests that bonds in this secret trove, while mainly accumulated under Greenberg, had more recent additions, presumably under Martin Sullivan, perhaps Wilmustad. This “unofficial vaults” designation strongly implies this was a secret, off balance sheet cache that threw off a hefty amount of annual income by virtue of its staggering size. That would mean it could be used by the CEO at his sole discretion, for anything from bribes to unreported executive payments that might then be used to open foreign bank accounts or pay for personal or business expenses.

And these “unofficial reserves” continued AFTER Greenberg was ousted over accounting improprieties. Business Week gave a short summary:

Investigators believe that AIG may have goosed its financial performance with dubious transactions and improper accounting. Last fall, the insurer paid $126 million in fines to the Securities & Exchange Commission and Justice Dept. for deals it structured for outside clients that allegedly violated insurance accounting rules, although AIG admitted no wrongdoing. The company also came under the glare of New York Attorney General Eliot Spitzer for its role in bid-rigging with broker Marsh & McLennan Cos. (MMC ), which led to the ouster of Hank’s son Jeffrey as CEO there. AIG admitted no wrongdoing, but two of its executives plead guilty and left the company.

This time, investigators initially focused on two transactions involving Berkshire Hathaway’s (BRK ) General Re Corp. unit. The deals essentially amounted to a $500 million loan that was dressed up on the books as premium revenue. That allowed AIG to boost its sagging reserves at a time when investors thought they were too low. The problem: AIG never assumed any of the risk associated with insurance underwriting. On Mar. 30, the company acknowledged that “the transaction documentation was improper” and should never have been classified as insurance premiums.

Since then, AIG ’s problems have escalated. In its Mar. 30 release, the company itself identified several problem areas. They include transactions with supposedly independent companies that were in fact controlled by AIG; bond transactions that may have allowed it to claim gains without actually selling the bonds; misclassified losses; and questionable estimates on deferred acquisition costs. Investigators and state regulators are looking into some 60 transactions involving these and other possible accounting shenanigans. “Greenberg strived for a steadily rising stock price,” says a source in Spitzer’s office. “He used mechanisms now being revealed as deceptive and improper.”

Perhaps there is an innocent explanation for this huge stash. However, but in all my years in financial services (and having had billionaire clients who would be completely within their rights to run their enterprises as personal cookie jars to the extent the law allows), I have never heard of anything remotely this suspect. Given AIG’s history, there is every reason to believe this is what is appears to be: a slush fund created for Greenberg’s personal use that was never accounted for properly.

The simple test of whether this arrangement is proper or not is whether the cache was fully accounted for on AIG’s balance sheet. If not, no wonder Greenberg had to go to such lengths to boost reported earnings. He would have needed to hide the truly remarkable amount of skimming needed to put these reserves aside.