Saturday, August 9, 2014

Why Death Matters for Central Bank Policy

Noah Smith raises the question: can the Fed influence the interest rate? Although the answer may seem obvious, the question itself reflects a conundrum for neoclassical theory. It is representative of a related but more comprehensive question: does the asset composition of the central bank balance sheet matter?

Let me set aside, for now, the deep question: what is money? I will take for granted the fact that the liabilities of the central bank are special. Perhaps this is due to legal restrictions, as Neil Wallace has suggested, or perhaps it is a matter of social convention. My focus here is not on central bank liabilities; but on their assets.

Figure 1
Figure 1 is a stylized representation of the balance sheet  of the Fed. Like King Midas who turned everything he touched to gold, so the Fed turns everything it purchases into money. Commercial banks hold accounts at the Fed, and when the Fed purchases an asset, any asset, those accounts are credited with the creation of new money.

Historically, the asset composition of the Fed has consisted almost exclusively of short term Federal government bonds, the item in red on Figure 1. In September of 2008 two things happened. First, the size of the balance sheet increased. The RHS of the table in Figure 1 went from $800b to $2,000b overnight. Second, the composition of the asset portfolio changed dramatically.
Figure 2
Figure 2 is a highly stylized representation of what happened following the collapse of Lehman Brothers in the fall of 2008.  The Fed purchased a whole boatload of long-term government debt: And for the first time in its history, it bought mortgage backed securities (MBS). 

The fact that the asset side of the balance sheet went from $800b to $2,000b is referred to as quantitative easing. The fact that the fraction of liabilities held as short term treasury securities went from 94% (750/800) to 38% (750/2000) is referred to as qualitative easing.

Here's the puzzle for neoclassical theory. According to received wisdom (Michael Woodford's Jackson Hole paper is an excellent exposition of this idea) the asset composition of the Fed's balance sheet is irrelevant. If the Fed had bought more short-term government debt instead of intervening in the riskier MBS market; it would not have made one whit of difference to the economy.

Why is that? According to standard neoclassical models, all transactions are carried out by infinitely lived families who take into account the welfare of their descendants. The far sighted paternalistic patriarchs of these families trade assets with each other that are contingent on every possible future event.  

Because the price of long bonds reflects all known facts about the probabilities of future outcomes, central bank asset positions do not influence the market price of risk. When the government takes a new position in the asset markets, the private sector unwinds that position through its own open-market trades.

But that is not what happened. A wealth of evidence shows not just that quantitative easing matters, but also that qualitative easing matters. (see for example Krishnamurthy and Vissing-Jorgensen, Hamilton and Wu, Gagnon et al). In other words, QE works in practice but not in theory. Perhaps its time to jettison the theory.

Replacing all of neoclassical theory with an operational alternative is a daunting task. There is no lack of contenders. Perhaps people are irrational as the behaviorists have claimed. Perhaps the market is segmented and institutional constraints cause pension funds to favor safe assets. Perhaps there are borrowing constraints that prevent some trades from taking place. These are all possibilities and I do not want to suggest that they do not have merit. But there is a much simpler explanation for the failure of the irrelevance result. Human beings do not live forever.

The fact that our lives are finite has consequences for the efficiency of asset markets. Davis Cass and Karl Shell called this idea sunspots. Asset markets are volatile because we all, eventually, meet the grim reaper. And although governments are sometimes overturned, they have much longer horizons than individuals. That simple fact explains why the asset composition of the central bank matters.


  1. I've tried to explain why QE doesn't work in Wallace 1981, but does in the real world. Here is the tack I take (from my current post):

    "Well, first off, people are far from perfectly expert (especially in the super complex modern world), with perfect public information that they can gather, digest, and analyze at zero time, effort, or money cost.

    So, when the government "firm" starts to borrow a lot more, almost no one thinks, MM style, or Wallace style, I'm going to start selling some of my bonds to compensate in equal measure as I see them doing that. And so total borrowing in the market does, in fact, go up, and so do market interest rates. And vice-versa, when the government starts to lend more. People just don't react that way. And it won't be nearly enough if a savvy minority do. They won't control enough money to drive us to Wallace neutrality.

    It's like in Miller and Modigliani's model, if the firms start borrowing a lot more, but the shareholders are mostly not really paying attention, and/or don't know well the implications, so for the most part they don't borrow any less to compensate. In that case, aggregate demand for borrowing would not remain unchanged. The aggregate demand curve for borrowing would, in fact, shift out, and the interest rate would rise."

    1. Richard: yes, I include explanations like that under 'irrational behavior' or imperfect information. Wallace neutrality is actually a very deep critique of monetary theory.

      There was a time, a long while ago, when I believed that we needed a deep theory to go with the deep critique. That's the view that Randy Wright and Steve Williamson still hold. I've moved on from that position.

      I was persuaded by conversations with Don Patinkin that real balances in the utility function is an acceptable short-cut. For me that is no worse a sin than assuming the existence of an aggregate production function. They are both short-cuts to tractable models.

    2. I do see your points. With regard to the point I was making, just predominantly not expert and well informed agents, who are just not seeing and doing the MM thing, I thought perhaps a good paper idea would be to duplicate Wallace's model, but just add a group of uninformed, unexpert investors who ignore the government's actions, and see what happens to the results. You can look also at various percentages of agents being uninformed, and see how the results move as this percentage increases.

  2. Hello Roger, I followed the link from FT Alphaville.

    I am afraid that this is wrong: "Historically, the asset composition of the Fed has consisted almost exclusively of short term Federal government bonds". Moreover, I think it was a common misunderstanding among academics before the financial crisis drove them to take more interest the operational details of monetary policy that led commentators like Krugman to exaggerate the possibility of a liquidity trap in the early days of the financial crisis.

    In fact, to the degree that the stock of base money was considered likely to be permanent, the Fed used to buy debt deliberately evenly distributed across the treasury curve (to avoid disturbing the Treasury's debt management strategy). In the pre-crisis period, I thought that this was a mistake, because it meant that the Fed was probably the largest central bank holder of treasury debt at a time when they were attributing low long-term dollar interest rates to a "saving glut" of Chinese reserves purchases:

    Tim Young

    1. Thanks Tim
      I'm not sure we have a deep disagreement here.

      The Hamilton Wu data show that the Fed held 70% of its SOMA Treasuries in maturities of two years or less up until 2007. After 2007, the composition at the short end drops from 67% to 25%. Two to five year maturities go from 16% to 30%, five to ten year maturities from 6% to 27% and ten to thirty years go from 11% to 18%.

      The important shift however, was not the composition of the Treasury holdings, it was intervention in the MBS market.

    2. The key point that I was trying to make though is that before the financial crisis, the Fed aimed to be neutral with its SOMA holdings, so they should not be described as "short" (unless you think that treasury issuance was in general, "short"). In fact, by comparison with other central banks except the BoJ, the Fed's pre-crisis monetary policy assets were long. The Fed could have managed the duration of its asset portfolio to lean against the pre-crisis boom just as it has done so to try to stimulate the economy post-crisis.

    3. I don't disagree. Thank you for pointing that out.

    4. Thanks, Roger - in that case I suggest you change Figure 1 to something like "bonds (market mix)" to avoid confusion. It took Paul Krugman months to stop writing about the Fed buying treasury "bills"!

      It will be interesting to see whether, after QE (if we ever get there!), the Fed goes back to that policy, or whether it follows other central banks such as the ECB, BoE etc in making repo loans to banks the mainstay of its asset portfolio in normal times. Or perhaps the BoE chooses to follow the Fed and retain and roll over some of its QE bonds.

    5. Tim
      I have a new post on this topic that you may find interesting.

    6. I did read that post, Roger; please see my comment there.

  3. In my opinion, a weakness of the standard theory is that it assumes that participants in financial markets select assets based on their first and second moments (average return and risk). It seems to me that financial-market participants also care about value at risk. For example, for a financial institution that would be the loss beyond which the institution becomes insolvent, and therefore cannot continue to operate and look forward to better days.

    So suppose that the expected return of an asset goes up but so does its kurtosis (a fourth moment). In this case buyers who care about value at risk may decide to not take advantage of the higher return because the probability of a catastrophic loss is now higher even as modestly-sized deviations from the mean are reduced hence keeping the variance the same. Someone looking at only the return and the variance may therefore get the impression that arbitrage is imperfect, when in fact this is not true.

    1. Thanks Constantine
      I agree completely with your assessment. What we are missing is a way of integrating those facts with neoclassical theory. The representative agent paradigm is broken but that does not mean we need jettison every other part of neoclassical theory. The race is on for the right way of integrating asset pricing theory with business cycle theory.

  4. This comment has been removed by a blog administrator.


Note: Only a member of this blog may post a comment.