At the outset of the current crisis in the financial and credit markets, we had no serious economic problems. Inflation was under control, economic growth was good, unemployment was low, and there were no major credit problems in the banking system.
The dark cloud on the horizon was about $1.2 trillion of subprime mortgages (most had been securitized), about $200 billion to $300 billion of which were believed to be held by FDIC-insured banks and thrifts. The rest were spread throughout the world.
The likely losses on these assets were estimated by regulators to be roughly 20%. Losses of this magnitude would have caused pain for banks that held the assets, but would have been quite manageable, particularly for an industry that had after-tax earnings of roughly $150 billion in 2006 and had capital of $1.4 trillion.
He makes one good point I haven't seen elsewhere, which is that MTM focuses only on the asset side of the balance sheet, ignoring potential offsets on the liability side. He also includes an actual example of an anonymous bank that makes the point quite clearly:
Below is an example of the distortion that occurs when using mark to market (MTM) accounting. This example involves an actual case study from an anonymous bank that made loans and securitized them as a mortgage backed security (MBS). Expected losses and expected cash flows for the MBS differ dramatically from MTM write downs. The Bank is required to record MTM losses of $913 million as opposed to the maximum expected lifetime losses of $100 million, resulting in a significant overstatement of losses and having a negative impact on tangible common equity.