Narayana made a speech yesterday, posted here. Most of this was on target. I particularly liked his comments on the issue of the implicit subsidization of housing by the Fed, and the conclusion that recovery in housing construction is not a necessary condition for aggregate recovery (and increases in housing construction should not happen for a long time given the huge stock of housing we have built on the back of incentive problems in mortgage markets). As well, I was interested in his analysis of the current pressing problem for the Fed: what to do with the large stock of mortgage-backed securities (MBS) on the Fed's balance sheet. As Narayana points out, based on an in-house study at the Minneapolis Fed, if the Fed holds the MBS until maturity, there will still be $250 billion of the stuff on the Fed balance sheet in 2030. He advises the MBS be sold off, but slowly. However, he loses me in his analysis of what he thinks is the key problem in MBS sales by the Fed, i.e. a commitment problem. Maybe Narayana has been spending too much time talking Chari, who appears to have taken an overdose of commitment, but his argument didn't make any sense to me, and I don't think commitment is at the crux of the problem.
Narayana gives the following example. Sarah has 2 houses on the same block, wants to sell both, and presumably wants to maximize the present value of the payoffs. Consider the following strategies: (i) sell both houses today; (ii) sell one house today, and commit to selling the other one year from now; (iii) sell one house today, with no commitment about when the other will be sold. Narayana appears to be telling us that (ii) is optimal. Why? Apparently, if both houses are sold at once Sarah gets a low price for both. If she follows (ii), then Sarah gets a high price for both. However, if Sarah cannot commit, as in (iii), then for some reason she can't help herself and is compelled to sell the second house as soon as she disposes of the first. This doesn't make any sense in terms of how housing markets work, let alone being a useful analogy to the market for mortgage-backed securities. Presumably we want to think of Sarah's block as a closed economy. Then we would like to ask what Sarah is going to do with the assets she acquires when she sells her house, what assets the buyer sells to buy the house, and how portfolios are going to be arranged in equilibrium once the sale occurs. One could certainly think of conditions where the timing of the sales would be irrelevant for prices.
It seems to me a given that the Fed will have to suffer a loss in any sales of MBS - these were bought at high prices with the goal of subsidizing the housing market, and willing buyers are going to be had only at lower prices. However, the key question is whether sales of MBS by the Fed will affect other asset prices. The answer depends on whether the Fed is somehow doing a better job of intermediating the MBS than the private sector would do. Essentially the Fed is transforming the MBS into reserves, which are held by banks (and exchanged in interbank transactions), and further transformed into bank liabilities. What difference would it make if the banks were holding the MBS directly rather than the reserves? Maybe not much. Why not sell the MBS as soon as possible, take the losses, and get this over with?
Commitment is always an issue in central banking, but the relevant commitment problem for the Fed is its implicit commitment to an inflation target. Whether the Fed can meet this commitment given how it has stuck its neck out is the burning question.