The paper version of this article has an inset called How Repo 105 Worked. Here is what it says.
Banks use repurchase agreements, known as repos, all the time for short-term financing. One borrows cash and gives the other securities, such as government bonds, as collateral. Both agree to unwind the arrangement on a set date. The deals, which run only for days or weeks, are accounted for as financings, and remain on the books with banks recording and asset--the cash--and a matching liability in the promise to buy back the collateral.
Lehman's 105 was different -- insteach of handing over securities equivalent to the cash it received, the bank gave more than was necessary. The point was to exploit a loophole allowing such over-collateralized deals to be accounted for as true sales. Lehman then reported its obligation to repurchase the securities at a fraction of the full cost, and used the cash it had received to pay off its liabilities, thereby "shrinking" its balance sheet.
Now, if you understand this, you are smarter than me!
The first paragraph seems OK, except that perhaps "matching liability" should be replaced with "corresponding liability."
The second paragraph seems just plain wrong to me. Lehman's 105 was not "different" in the sense the cash loan was over-collateralized. Over-collateralization (the "haircut") is normal in repo transactions.
And the point was not to exploit the "loophole" of treating the repo transaction as a "true" sale. After all, repo is short for "sale and repurchase agreement." There is a legitimate sense in which there was a sale.
And if one does treat the transaction as a sale, then I think there is a legitimate reason for why one should not have to record the corresponding liability to repurchase the asset (pay back the cash loan).
An example. Imagine that I have asset (capital) worth $100 and no debt. I am worth $100. Imagine that I repo my asset for $90 in cash: What does my balance sheet look like?
I am an economist (not an accountant); so I would say that your B/S should now look like this:
Assets: $100 (capital) + $90 (cash)
Liabilities: $90 (cash loan -- promise to repurchase asset)
Wealth: $100 (unchanged)
But as this is a repo, let's take the "sale" part of repo seriously. That is, imagine that I have "sold" my asset. In fact, there is a sense in which I have sold it: I have diverted control of the asset to my creditor, who holds it as collateral. It's not like I (or other shareholders) can access this asset in the event of default. If this is true (and it is), then why should I record the asset on my books? OK, suppose I don't. Then my B/S looks like this:
Assets: $90 (cash)
Liabilities: ...what...you want me to record the $90 loan? Shouldn't I leave this loan off my books? After all, I am not counting the corresponding asset as something I own. And I do not have to pay back my loan -- I have the (legal) option of not doing so; in which case the creditor gets to keep my asset. Moreover, if I do include the $90 liability, then my reported net worth drops to zero! Let's be reasonable here and not include the liability; in this case...
Net worth: $90 ($10 less than prior to the repo, but still higher than zero, which would grossly underestimate my net worth)
If there was a problem with what actually transpired, it probably resides in the fact that the transaction was not disclosed (off balance sheet items were not reported). While this sounds like funny business (and it no doubt was), I do not think that any law was broken (and keep in mind that everything was disclosed to Ernst & Young, the accountants in this case).
As the Financial Times article explains, there are a lot of grey areas in the theory of accounting. If a firm exploits these grey areas to make their books look (temporarily) better, who is to blame? The firm, the "independent" accountants, or the accounting principles themselves?
If your answer is the firm, then be prepared to condemn most firms (and individuals too) of the practice of keeping some assets and liabilities off balance sheet. Be prepared to blame governments as well, since many important government liabilities are not recorded in "official" government debt measures. In short, this is not a problem that resides with just a few Lehman executives.