Paul Krugman's column today says that if the Aggregator Bank buys up toxic assets, it will unavoidably overpay, resulting in a windfall for shareholders at the taxpayers' expense. He says that we should nationalize banks instead. But nationalizing just means buying the banks, and the government can't buy a bank unless the shareholders are willing to sell. If the government will unavoidably overpay for toxic assets, why won't it unavoidably overpay for the shares of the institutions that own these toxic assets?
Perhaps, the theory is that if the government announces that it will buy the banks and hence will not buy the toxic assets, the market price of MBS's will fall drastically—because banks are only unwilling to sell them in the first place because they hope that the government will overpay for them—and thus the banks', or some banks', market price will fall to zero. With zero market price, the government doesn't need to pay anything for the banks; it can just take them over.
But even then, the government has to do something with the MBS's in its possession. Indeed, it has to do something with the banks in its possession. If it sells the MBS's and/or the banks, it might end up selling them for less than their market price, with the result that buyers will enjoy windfalls at the taxpayers' expense. If it holds the MBS's to maturity, it may lose out on selling opportunities, again at the taxpayers' expense. Is there any reason to think that a government inclined to pay too high a price for MBS's when it buys them will know enough to avoid accepting too low a price when it sells them?
Krugman says, "should the government be in the business of declaring that it knows better than the market what assets are worth?" The implicit criticism here—that government officials evaluate assets less accurately than the market does—overreaches, as it undermines Krugman's preferred approach of nationalization as well as the more limited approach of purchasing the toxic assets. The case for nationalization must rest on other considerations.