“The problem with QE is it works in practice, but it doesn’t work in theory.”
Former FRB Chairman Ben S. Bernanke, Brookings Conversation, January 16, 2014.
Last week’s 12th annual U.S. Monetary Policy Forum focused on the effectiveness of Fed large-scale asset purchases (LSAPs) as an instrument of monetary policy. Rarely is such a public policy discussion so timely and enlightening (for full disclosure, we are founding members of the USMPF panel).
While there are notable disagreements, the report and discussion reveal a broad (if not universal) consensus on key issues:
- In a world of low equilibrium real interest rates and low inflation, policymakers could easily hit the zero lower bound (ZLB) in the next recession.
- At the ZLB, the Fed should again use a combination of balance-sheet tools and interest-rate forward-guidance to achieve its mandated objectives of stable prices and maximum sustainable employment (see our earlier post).
- Yet, significant uncertainties about the impact of balance-sheet expansion mean that LSAPs may not provide sufficient stimulus at the ZLB.
- Fed policymakers should undertake a thorough (and potentially lengthy) assessment of alternative policy tools and frameworks—ranging from negative interest rates to a higher inflation target to forms of price-level targeting—to ensure they remain as effective as possible.
The remainder of this post discusses the challenges of measuring the impact of balance-sheet policies. As the now-extensive literature on the subject implies, balance-sheet expansions ease financial conditions. However, as this year’s USMPF report emphasizes, there is substantial uncertainty about the scale of that impact. This, combined with political-economy concerns, suggests that policymakers should be cautious about relying on these tools. Consequently, while the hurdle for altering the Fed’s flexible inflation-targeting framework should be high (see, for example, the comments by FRB Cleveland President Loretta Mester), changes that would reduce the likelihood of relying on these balance-sheet measures in the future merit serious consideration.
To begin, let’s ask what we have learned about balance-sheet policy since the Federal Open Market Committee (FOMC) lowered the policy rate close to zero in December 2008, and then proceeded to alter the mix and size of its asset holdings.
Shortly after this process started, when teaching a new course on the financial crisis and the policy response, one of us illustrated the state of affairs using an analogy to a pharmaceutical trial. It is only after they test a new remedy on patients that doctors can calibrate the dosage, determine the appropriate length of treatment, gauge the nature of any side effects, and assess the consequence of terminating therapy. The same was true in 2008: U.S. central bankers could only speculate about mechanisms through which these untested polices would influence finance conditions and the real economy. They had little knowledge of the appropriate size of an intervention. People could only speculate about possible side effects. And, no one knew what would happen when the policies were tapered, halted or reversed. Naturally reluctant to run monetary policy experiments on an economy, Fed officials resorted to these balance-sheet instruments only in extremis and as a last resort to counter the most severe financial crisis in 75 years.
In the intervening decade, researchers have produced a large number of high-quality studies examining the efficacy of central bank balance sheet tools aimed at measuring both the size of their impact and the mechanisms through which they influence financial conditions and the real economy. Over time, we have learned to distinguish two types of LSAPs: quantitative easing (QE) that expands the central bank’s balance sheet and targeted asset purchases (TAP) that alter its composition. The latter seem particularly powerful as a device to substitute for private intermediation in a crisis (see Gertler). But, because of their distributional effects, these quasi-fiscal actions pose risks to central bank independence (see here).
In comparison to TAP, QE is a more traditional monetary policy tool and is thought to operate in two ways: both by signaling that affects the expected future level and volatility of interest rates (or as a commitment device to enhance the impact of the central bank’s forward interest rate guidance); and through a portfolio-balance channel (that affects the relative prices of imperfectly substitutable assets, such as long- and short-dated Treasury debt). Even so, Bernanke’s famous quip about theory and practice of QE highlights persistent uncertainties about balance-sheet policy.
Recent summaries of the existing literature—including those of Borio and Zabai, Fischer, Gagnon, and Williams—conclude that QE lowered long-term Treasury bond yields substantially. The consensus is that the cumulative effect of QE was to reduce the 10-year Treasury yield by about 100 basis points (see the USMPF report). This is a large number: based on past experience, it would take a 400-basis-point decline in the federal funds rate to achieve the same outcome (see Chung et al). The presumption is that financial conditions generally—including the risk premia on equity and private credit—eased commensurately with the Treasury yield, helping to stimulate the economy.
However, unlike the case of policy interest rate changes, there are a number of obstacles to measuring the impact of QE with any precision. First, the U.S. sample is tiny. If we count generously, there are only four data episodes, three expansionary and one contractionary. These are: QE1 from November 2008 to August 2009; QE2 from November 2010 to June 2011; QE3 from September 2012 to October 2014; and the “Great Unwind,” which began less than six months ago and will likely proceed well into the next decade. Trying to disentangle the impact of these policies from all the other economic developments and fiscal policy choices over these lengthy time intervals is a daunting task. Not only that, but the three QE episodes themselves were quite different in their speed, composition and intended impact.
Further complicating matters, what we want to know is the overall impact of the policy—both the part that was anticipated and the surprise. We can measure the instant impact of an announcement, but to the extent that policymakers succeed in being transparent, market prices will incorporate expectations of future balance-sheet shifts. So, if we focus only on the announcements, we will underestimate the impact of the policy. We note that this applies even to QE1, which began well after the massive expansion of the Fed’s balance sheet in the weeks following Lehman’s September 2008 bankruptcy (and a decade after Japan had experimented with QE).
Finally, policymakers’ announcements carry two possible signals that we need to distinguish. On the one hand, by announcing further easing, Fed officials can reassure people that growth will do well. Alternatively, by stating that they plan additional stimulus, officials could be revealing that they view the economy as being in worse shape than previously believed. In the first case, bond yields could rise, while in the second they would fall (see, for example, Evans and Nakamura and Steinsson).
Reflecting these measurement challenges—particularly the first one—researchers resort to event studies. That is, they look at a narrow time window (usually no more than a few days, and often much less) around policy developments to extract the “signal” from the bond market “noise.” In their new USMPF report, A Skeptical View of the Impact of the Fed’s Balance Sheet, David Greenlaw, James Hamilton, Ethan Harris, and Kenneth West highlight two aspects of the “event study” methodology that cause the resulting estimates to overstate the cumulative impact of QE:
- Selection: Studies based on a small sample of QE “events” that are selected based on the size and direction of bond yield movements arbitrarily exclude contrary observations.
- Persistence: Use of a brief time window presumes that policy’s impact persists and ignores the possibility that the initial market response will unwind.
Using various means to correct for these deficiencies, Greenlaw et al conclude that previous studies significantly overstate the persistent impact of QE. To take one example, out of a sample of 2,374 business days from November 2008 to December 2017, the authors identify 255 “Fed days.” These include days in which there were FOMC policy announcements, the release of FOMC minutes, and policy speeches from the Fed Chair. The authors hypothesize that if event studies are capable of measuring QE’s impact, then yields should decline on Fed days during the three expansionary episodes and then rise during the subsequent period of tapering and exit.
The following chart depicts the cumulative change in the 10-year Treasury yield on Fed days: in contrast to the hypothesis, the yield tends to rise in the three shaded QE episodes (after dropping at the start of QE1). And, while the yield initially rose (as predicted) in response to Chairman Bernanke’s taper comments in May 2013 (the “taper tantrum”), it then drifted lower, even after the taper began (a phenomenon the authors label the “shrinkage shrug”). Consequently, aside from two key turning points, the cumulative yield shift does not match the authors’ hypothesis. In contrast, the second chart—for the “non-Fed days” that account for nearly 90 percent of the sample—shows that the yield generally trended lower through the QE episodes. (Using another approach—based on a news-story definition of “Fed days”—the authors find comparable patterns.)
Cumulative change of the nominal and inflation-adjusted 10-year Treasury yields on “Fed days” (basis points), November 2008-December 2017
Note: Shaded areas denote QE periods. Source: Exhibit 4.2 in Greenlaw et al, A Skeptical View of the Impact of the Fed’s Balance Sheet.
Cumulative change of the 10-year Treasury yield on “non-Fed days” (basis points), November 2008-December 2017.
Note: Shaded areas denote QE periods. Source: Exhibit 4.6 in Greenlaw et al, A Skeptical View of the Impact of the Fed’s Balance Sheet.
The USMPF report’s results are really an indictment of the event study methodology employed in previous studies of QE, not of the policy itself. This is the point that Federal Reserve Bank Presidents Dudley (New York) and Rosengren (Boston) emphasize in their comments on the report. They highlight three ways in which these methods may overlook the stimulative impact of QE. First, to the extent that investors anticipated QE expansions, any disappointment on announcement could have raised the Treasury yield. Second, because Fed policy is both transparent and data-dependent, economic data releases that raise QE expectations would register the QE impact on “non-Fed days” rather than “Fed days.” In the extreme, where QE actions are fully anticipated, “Fed days” would show no impact whatsoever. Third, measures of the yield level do not capture the impact of QE on yield volatility, which also influences the risk premia on other financial assets.
So, what now? As the USMPF authors make clear, policymakers should seek ways to strengthen the impact of QE. The stated purpose of QE is to lower long-term interest rates. Specifically, policymakers are trying to lower term premia on long-term Treasury bonds, which will then filter through the financial system lowering the risk premia on long-term private financing. The Fed does this by reducing the scale and weighted-average maturity of outstanding U.S. government debt held privately. Now, the Treasury’s borrowing plans can undo this, blunting the impact of QE (see Greenwood et al). One solution is to better coordinate in a way that gives the Fed the “last mover” advantage in adjusting the volume and duration of Treasury debt to be absorbed by the private sector. Another would be to concentrate the Fed’s holdings in short-duration assets, permitting a larger duration extension when desired. A third would be (as in QE3) to make the timing of QE open-ended (“QE Infinity”).
Other, more extreme, possibilities—like the Bank of Japan’s “QE with yield curve control” that makes the size of the central bank’s balance sheet endogenous—also exist. But we know much less about the efficacy of these policies, and are especially wary of the possibility for market disruption when exit approaches. Since the Treasury Accord of 1951, no major central bank has exited from a commitment to cap a long-term bond yield. We view such a yield-capping policy as a mix of QE with fiscal stimulus—aka helicopter money—for which the implementation hurdle should be very high (say, a persistent or deteriorating deflation at the ZLB).
If, after efforts to strengthen QE, its impact remains insufficient, some will argue simply for increasing the dosage. That is one way to interpret the impact of the ECB’s asset purchase program, which has now absorbed all but 10 percent of the free float of German government bonds, making their prices more sensitive to central bank purchases or sales (see the USMPF comments of ECB Governing Council Member Benoit Cœuré). In our view, however, expanding QE is not cost-free: aside from altering the effectiveness of markets, political dissatisfaction with a large central bank balance sheet poses a real threat to Fed independence, inviting further moves to fund Federal spending directly from the Fed’s revenues.
All this makes us uncomfortable. We strongly believe that the Fed should not hesitate to use QE aggressively should the need arise. But, confidence that QE will be the silver bullet that can slay the next recession would be severely misplaced. In lieu of such complacency, there are some things the Fed should do now. First on the list is to emphasize the symmetry of their flexible inflation-targeting framework. Most importantly, we hope that Chairman Powell and his colleagues highlight that, after years of undershooting the target, an extended period of mild overshooting will not trigger an outsized (recession-inducing) policy tightening.
Looking further ahead, we argue elsewhere that the hurdle for changing the overall framework should be high, and should include sufficiently broad political support to sustain the independence of the Fed. Nevertheless, we believe that the FOMC should establish a system in which they periodically review their policy framework to ensure that they are reacting in a timely way to changes in the economic and financial environment. The transparency of this process will be critical to securing the support of the private sector and the Congress for any fundamental changes.