Even though the overnight interest rate has been stuck near zero for 20 months, are there options available to the Federal Reserve or the U.S. Treasury to bring longer-term yields down further? I have been looking into this question with Cynthia Wu, an extremely talented UCSD graduate student. We present our findings in a new research paper, some of whose results I summarize here.
Our starting point was a framework developed by Vayanos and Vila (2009), who interpret the term structure of interest rates as arising from the behaviour of risk-averse arbitrageurs. This model is one way to capture formally the portfolio balance channel that Fed Chairman Bernanke indicated is central to the Fed’s understanding of how nonstandard monetary operations might affect the economy. Vayanos and Vila’s framework has previously been applied to our question by Greenwood and Vayanos (2010) and Doh (2010). One of our contributions is to develop specific measures of how the available supplies of Treasury securities of different maturities might be expected to influence the pricing of level, slope, and curvature risk of the term structure. Although I began as a sceptic of the claim that bond supplies would make much difference, we found pretty strong evidence that historically they have. For example, we found that over the 1990-2007 period, we could predict the excess return from holding a 2-year bond over a 1-year bond with an R2 of 71% on the basis of the level, slope, and curvature of the yield curve along with our 3 Treasury supply factors.
One of the challenges plaguing this kind of research is the problem of endogeneity. There may be a correlation between bond supplies and interest rates, but is that because bond supplies affect interest rates, or because the Treasury or the Fed are responding to interest rates in deciding which maturities of Treasury securities to sell or buy? Our solution to this problem is to pose the empirical question in terms of a conditional forecast. Suppose you already knew today’s level, slope, and curvature of the term structure of interest rates, and in addition to those values, I tell you today’s 3 Treasury supply factors. How would the latter cause you to change your forecast of next month’s interest rate for any given maturity? Our finding is that the Treasury factors make a statistically significant contribution across the yield curve.
We can summarize the implications of that forecast in terms of the following scenario. Suppose that the Federal Reserve were to sell off all its Treasury securities of less than one-year maturity, and use the proceeds to buy up all the longer term Treasury debt it could. For example, in December of 2006, this would have required selling off about $400 B in bills and notes or bonds with less than one year remaining, with which the Fed could have effectively retired all Treasury debt beyond 10 years. The figure below summarizes the implied average change in forecast for the 1990-2007 period as a result of this change for interest rates of various maturities. Yields on maturities longer than 2-1/2 years would fall, with those at the long end decreasing by up to 17 basis points. Yields on the shortest maturities would increase by almost as much. While our estimates imply that the Fed could make a modest change in the slope of the yield curve, it would not make any difference for the average level of interest rates.
We then extended the framework to the case when, as at present, short-term interest rates are as low as they could go. Even though short term interest rates have been near zero since the end of 2008, longer term yields have continued to vary from week to week, as shown in the solid lines in the graph below. Our interpretation is that these fluctuations in longer-term yields come from investors’ beliefs that short-term interest rates are not going to be stuck at zero forever. We suppose that investors attach a probability to escaping from the zero lower bound at various future dates, and that, when we do, short-term rates and the rest of the yield curve will revert to a dynamic behaviour similar to that exhibited prior to 2007.
We were then able to describe interest rate dynamics since the beginning of 2009 in terms of the historically estimated parameters along with three new coefficients, which correspond to the average short-term interest rate as long as we’re stuck at the zero lower bound, the average new short-term interest rate once we escape from the zero lower bound, and a fixed probability of escaping in any given week. The red dashed lines in the figure above represent the predicted values from this model. This simple framework seems to do a pretty reasonable job of explaining interest rate movements over the past couple of years.
Moreover, the framework gives us the information we need to assess the effects of nonstandard open market operations under a zero-lower-bound regime. The figure below shows how our model implies that the forecasting relation described above would be different under the zero lower bound. The experiment here is the same as before– the Fed sells off all its short-term Treasury holdings and buys an equivalent amount of long-term debt. However, under the zero lower bound, the effect on short-term interest rates all but disappears as a consequence of investors’ beliefs that near-zero short-term interest rates are likely to persist for some time. Quantitative easing– buying the longer-term securities with newly created interest-bearing reserves– would have the same effect in our framework.
Hence our estimates imply that whereas an asset swap by the Fed could not reduce interest rates in normal times, under the present situation, it would succeed in driving overall interest rates lower. To take an illustration, the Fed’s combined $1.1 trillion in mortgage-backed securities plus $300 B in new longer term Treasury purchases might have succeeded in driving 10-year yields 50 basis points lower than they would have otherwise been.
Although our estimates imply that the Fed could do more than it already has, in many ways the U.S. Treasury is the more natural institution to implement such a policy. According to the theoretical framework that motivated our measures of the Treasury risk factors, the average slope of the yield curve arises from the preference of the U.S. Treasury for doing much of its borrowing with longer term debt. For reasons presumably having to do with management of fiscal risks, the Treasury is willing to pay a premium to arbitrageurs for the ability to lock in a long-term borrowing cost. If the Treasury has good reasons to avoid this kind of interest-rate risk, it is not clear why the Federal Reserve should want to absorb it.
But, according to our estimates, if the Fed wanted to absorb more of this risk, it could reduce the slope of the yield curve further by doing so.
The full paper is available here.