Arnold Kling is defending Paul Krugman:
The Fed is treating the problems in financial markets as a liquidity crisis. What that means is that the assets that the institutions are holding, such as mortgage-backed securities, are really worth $X, but their current market value is, say, $X times 0.95, and no one will lend the institutions the money to enable them to carry those assets to maturity. If the Fed acts as lender of last resort, then eventually the Fed will get paid back out of the cash flows from those securities--more likely sooner, as investors regain their confidence.
That's if the mortgage-backed securities are really worth $X. But if those securities are actually worth $X times 0.95, then the Fed will take a huge loss on behalf of its owners, the taxpayers.
Krugman thinks that housing market defaults are going to spread from the subprime segment of the mortgage market to the prime segment in a big way, so that mortgage-backed securities really are worth less than $X. I am less pessimistic than Paul in my outlook for home prices and mortgage defaults, but the probability that his forecast turns out to be approximately true is certainly greater than zero. So I don't see how you can say he is irresponsible for speaking out, even if it turns out that his forecast is wrong.
My argument is that the securities are almost certainly worth 0.95X, or perhaps 0.8X. But that isn't enough to produce a general insolvency crisis. Few of these firms hold the majority of their assets in risky mortgage backed securities, or permutations thereof. They will take losses, but banks take losses all the time. They won't go bust unless the crisis of confidence dries up their credit, turning a duration mismatch between their assets and liabilities into a fatal error.
But I certainly agree with Arnold Kling about this:
I am seeing a lot of people argue that the investment banks should not have to mark down the value of their mortgage-backed securities to the market value of less than $X. But the alternative of historical-value accounting (what the assets were worth before the market turned sour on them) is worse.
To make a long story short, the reason that the U.S. taxpayers took such a big hit in the S&L crisis of the early 1980's is that S&L's claimed to be solvent using historical-value accounting, and so they kept borrowing more money even though they were actually insolvent. Market-value accounting provides better protection against insolvency.
There is no such thing as a perfect accounting system, but mark-to-market is a much more conservative accounting standard than the historical cost rule it replaced. In general, accountants like to reduce a CFO's discretion as much as possible; that's what mark-to-market does.
On the other hand, we could reconsider the proposal of my old financial accounting professor, Roman Weil: rely less on rules than on broad principles, and rotate the auditors, by law, every five to seven years. But just changing the rule back to "use historical cost" would, IMHO, be a big mistake.