What is the theory behind QE2?
The underlying theory, in which asset prices are directly linked to the outstanding quantity of assets, dates back to the early 1950s.4 For example, in preferred-habitat models, short- and long-term assets are imperfect substitutes in investors' portfolios, and the effect of arbitrageurs is limited by their risk aversion or by market frictions such as capital constraints. Consequently, the term structure of interest rates can be influenced by exogenous shocks in supply and demand at specific maturities. Purchases of longer-term securities by the central bank can be viewed as a shift in supply that tends to push up the prices and drive down the yields on those securities.So, as we have been hearing for a long time, the basic idea is based on 1950s "preferred-habitat" ideas. Financial markets are segmented by maturity, and so if monetary policy alters the relative supplies of securities of different maturities, then this can move the yield curve. In this instance, the short end of the nominal yield curve is pegged by the interest rate on reserves, so if the Fed swaps reserves (overnight maturity) for Treasury securities with an average duration of about 7 years (which is what the Fed is doing), then nominal bond yields on long-maturity assets should fall. One might hope that we could do better than 1950s-era modeling and, indeed, there are some modern preferred-habitat models. For example, Yellen cites a paper I have seen which is this one, by Vayanos and Vila. This is much more sophisticated than the 1950s stuff, but it's basically a theory where we explain the term structure of interest rates by putting assets in the utility function. Thus, we can safely say that we are still waiting for a good theory of preferred habitat.
Is there empirical work suggesting that QE2 could be quantitatively significant, or that would tell us how large a program is needed to give us the effects we want - inflation rate about 1% higher, and less unemployment? Yellen cites two pieces of evidence. The first consists of event studies and regression analysis, conducted by economists in the Federal Reserve System and at the Bank of England. Of course, without a theory, these studies are pretty meaningless. We need sound structural evidence to say anything about the effects of policy. Second, Yellen argues that we can use large-scale macroeconometric models to look at the effects:
Turning now to the macroeconomic effects of the Federal Reserve's securities purchases, there are several distinct channels through which these purchases tend to influence aggregate demand, including a reduced cost of credit to consumers and businesses, a rise in asset prices that boosts household wealth and spending, and a moderate change in the foreign exchange value of the dollar that provides support to net exports. The quantitative magnitude of these effects can be gauged using a macroeconometric model such as FRB/US--one of the models developed and maintained by Board staff and used routinely in simulations of alternative economic scenarios.Now, we know what the FRB/US model is for. In instances where no one at the Board has a clue what the effects of a policy will be, the policy (as best this can be represented in the model) is run in the FRB/US model, the computer spits out some numbers, these numbers are delivered to Janet Yellen, and she puts them in her speech. Hopefully the public buys it. Unfortunately there is no way to check whether the numbers make any sense. There are some sources like this one, that tell us a bit about what is going on in FRB/US, but we can't see the whole model, and we certainly can't run it for ourselves to see how it behaves. The best guess is that FRB/US is essentially a dressed-up version of the FRB/MIT/Penn model, developed in the 1960s. The economists at the Board who run the FRB/US model may pay some lip service to rational expectations, the Lucas Critique, and other developments in modern macroeconomics, but we would be surprised if the FRB/US model were anything more than an elaborate quantitative IS/LM model. As such, it would be ill-equipped to predict the effects of the QE2 program. Janet Yellen may tell us, as she does in her speech, that the FRB/US model predicts a 25 basis-point reduction in the 10-year Treasury yield and an increase of 700,000 in aggregate employment, but we would be foolish to attach any more significance to those numbers than we would to predictions that Janet thought up while walking her dog.
What have the effects of QE2 been thus far? Here, Yellen is quite selective. According to her, since late last summer, when Treasury yields have fallen it was because of QE2, and when yields have been rising, it is because of other factors. In other words, we have no idea whether this is working according to plan or not. In any case, it appears that the Fed cannot control long-term Treasury yields, unless they are just not intervening enough. Ultimately, of course, if the Fed were to purchase the entire stock of Treasury debt, it could certainly control long Treasury yields.
What might be a useful approach to understanding the effects of QE2? We might start by thinking of the Fed as a "shadow bank," of the type so well-described by Gary Gorton here and in his published work. Currently, the Fed is holding as assets mainly long-maturity securities - Treasury bonds, agency debt, and mortgage-backed securities, and it finances this asset portfolio by issuing currency and reserves. It is the transformation of long-maturity marketable assets into overnight reserves that looks like the asset transformation done by a shadow bank, which typically borrows overnight using repurchase agreements (repos) with the securities in its portfolio used as collateral. The Fed does not post collateral against its overnight borrowing in the form of reserves, as everyone has confidence that the Fed will not default. Now, in shadow banking, as Gorton points out, the shadow banks perform a useful service. Large institutional investors need a place to park cash overnight, and asset-backed securities are what helps support credit markets in the US. The shadow banking system is critical to how the banking system works as a whole.
Now, how useful is the shadow banking that the Fed is doing? On the asset side, the Fed's long-maturity assets are not information-sensitive (as Gorton would say). The stuff is Treasury debt, agency debt, or mortgage backed securities issued by the GSEs - nothing there of questionable quality. Presumably a private shadow bank is not going to have trouble in using any of these assets as collateral. On the liability side, the issue of reserves as liabilities is not accomplishing much. Many large institutional investors do not hold reserve accounts, and therefore cannot park their overnight liquidity with the Fed. The inconvenience of reserves (as Andy Harless points out) is reflected in the fact that reserves pay 0.25% overnight while the 3-month T-bill rate is currently 0.14%. Indeed, the Fed could potentially get more bang for the buck with QE2 if it could issue a marketable security - i.e. Fed bills, which could be used as collateral in repo transactions.
The only potential advantage the Fed has as a shadow bank is that it can issue currency (by virtue of the monopoly granted to it). But the stock of currency is determined by the willingness of people to hold it, and the unwillingness of banks to hold reserves. It's not clear whether, or how much, currency holdings will increase as a result of QE2.
Conclusions? We actually know little about what QE does. The theory has not been developed, and we can't have useful empirical evidence without a theory. It seems to me that intermediation has to be at the heart of the theory. To understand the effects of QE, we need to understand what the Fed can do as a financial intermediary that the private sector cannot, and this has a lot to do with the functioning of the so-called shadow banking system. Once we have the theory, we need to be able to measure the extent of private intermediation activity, particularly in the shadow banking sector. Unfortunately, as Gary Gorton points out, the measurement of shadow banking activity is close to non-existent. All the more reason to bring this activity under the regulatory umbrella.
"Ultimately, of course, if the Fed were to purchase the entire stock of Treasury debt, it could certainly control long Treasury yields."ReplyDelete
Could it really? The price of the only Treasury security in the world might be high, but I don't think it would be higher than the price of a German bond plus the price of a currency hedge.
Suppose the world price for riskless debt of a particular maturity is P. If the Fed offers P+e for the stuff, it can get all it wants. Some people point to the pre-Accord (pre-1951) period in the US, when the Fed appears to have been able to peg long bond yields, as evidence that control of long rates by the Fed is feasible.ReplyDelete
Things may be different now than they were in the short pre-Accord period. I think Friedman and Schwartz say the peg worked because of expectations of deflation after the war. There are also more alternatives to U.S. Treasuries for investors wanting riskless debt than there were in the late 1940s.ReplyDelete
If the world yield on riskless debt is i, getting a sustained yield greater than or less than i on U.S. debt will be pretty hard.
In any case, it could well be that a $600 purchase does not move Treasury yields at all.
Oops, I assume you understood I meant $600 billion.ReplyDelete
"Thus, we can safely say that we are still waiting for a good theory of preferred habitat."ReplyDelete
...and this proceeds from the fact that - given the tradition in which you work - assets, along with money, cannot go into the utility function. And the reason for this is because assets and money are - above all - intrinsically valueless, and therefore cannot enter the utility function, whereas consumer goods are valuable, utility-yielding, and the only items that can enter utility functions.
Off topic I know, but just wanted to verify. Does that make sense?
That's exactly the idea. You want a theory of asset prices based on the underlying payoffs on the assets, and how those assets are used in exchange. Assuming people hold money, or Treasury bills, or mortgage-backed securities because they like them does not help us understand anything.
I came across the reference to our paper (Vayanos-Vila) in your post above. We are not assuming that assets enter in the utility function, unlike what is claimed in your post. We are instead assuming that investors want to consume at specific dates, and are infinitely risk-averse over consumption. (For example, a pension fund might have inflexible commitments to pay pensions at fixed dates.) A riskfree consumption profile at a specific date can be achieved by holding the zero-coupon bond maturing at that date, but also by replicating that bond using a trading strategy that involves other bonds. For example, an investor wanting to consume in fifteen years can hold the fifteen-year zero-coupon bond, but can also invest in a portfolio involving the one- and the thirty-year bonds, and rebalance continuously. The key point is that preferences are over riskfree consumption in fifteen years and not over the fifteen-year zero-coupon bond.
Our work is definitely not the last word on preferred habitat, and there is scope for better theories. But claiming that preferences in our paper are over assets rather than over consumption is incorrect, in my opinion.