Three Views of QE: Irrelevance, Insurance, and Irreversibility

Why QE Had Modest Effects After 2008 but Powerful Ones in 2020–2022

Quantitative Easing (QE) is back in the news. So, too, is the large size of the Federal Reserve’s balance sheet. The public’s renewed interest in these topics has been sparked by President Trump’s nomination of Kevin Warsh to be the next Fed Chair. Warsh is a vocal critic of both QE and the Fed’s expansive balance sheet, and he has called for a “regime change” at the Fed on these issues.

I am broadly sympathetic to Warsh’s concerns about the size of the Fed’s balance sheet, and in previous newsletters I have outlined several steps to carefully reduce it. In this piece, however, I will focus on QE itself. That choice is motivated in part by my recent podcast conversations with Tyler Muir and Scott Sumner, where QE featured prominently. It is also motivated by a bit of serendipity: my wife recently unearthed the original version of the image shown above, which I first created back in 2014.

I sketched that image—originally in pencil and now enhanced with AI—during the debates over why QE1-QE3 failed to spur a more robust recovery from the 2007–2009 recession. Beyond its attempt at humor, the image captures a deeper puzzle: the stark divergence between the economic outcomes associated with QE during and after the Great Recession and those associated with the pandemic-era QE. What explains this difference?

To answer that question and address some of the concerns raised by Kevin Warsh, I will draw from my recent conversations with Tyler Muir and Scott Sumner. Specifically, I will outline three distinct views of QE that emerged in those discussions: irrelevance, insurance, and irreversibility. Before turning to them, though, it is helpful to first outline the conventional wisdom on QE.

The Conventional Wisdom on QE

Large-scale asset purchases (LSAPs) by a central bank operating at the zero lower bound were first tried by the Bank of Japan (BoJ) from 2001–2006. The term Quantitative Easing (QE) was aptly applied to these LSAPs since the BoJ’s efforts were genuinely focused on expanding the quantity of the monetary base. The BoJ’s focus on the liability side of its balance sheet contrasted with the Fed’s focus on the asset side of its balance sheet during QE1-QE3. Former Fed Chair Ben Bernanke tried to highlight this distinction by calling the Fed’s LSAPs “credit easing” though this framing never stuck as he later acknowledged:

I also tried, without success, to name the program “credit easing,” to distinguish it from the Bank of Japan’s earlier foray into asset purchases (Bernanke, 2009). I argued that “credit easing” focused on removing duration from bond markets, in contrast to BOJ-style quantitative easing, which had the primary goal and metric of increasing the high powered money stock.

The distinction between asset-side QE and liability-side QE is more than a historical footnote. It reflects two different views of how central bank balance-sheet policies are supposed to work. After the Great Recession, the Fed’s asset-side framework became the conventional wisdon on QE, while the liability-side perspective was largely ignored. As I will argue in the section on irreversibility, this liability-side view can help explain why QE4 had a much larger macroeconomic impact than QE1–QE3.

Under the asset-side view of QE, LSAPs are supposed to operate primarily through two channels: the portfolio balance channel and the signaling channel. The portfolio balance channel works by reducing the supply of longer-duration or riskier assets held by the public, inducing investors to rebalance their portfolios toward close substitutes, which raises asset prices and lowers yields across markets. The signaling channel operates by reinforcing expectations that short-term interest rates will remain low for longer than previously anticipated, further easing financial conditions and stimulating spending.

Despite the conventional wisdom on QE, Ben Bernanke famously quipped in 2014 that “QE works in practice, but it doesn’t work in theory.” Although offered tongue-in-cheek, his remark captures a deeper tension: in standard benchmark models, LSAPs should have little to no effect on real economic outcomes. It is this theoretical challenge to the asset-side view of QE that motivates the next section.

The Irrelevance View of QE

The notion that QE should have no meaningful economic effect goes back to Neil Wallace’s 1981 paper, “A Modigliani–Miller Theorem for Open-Market Operations.” The basic idea is that, for a given fiscal policy, changes in the Fed’s balance sheet simply reshuffle government liabilities and do nothing to reduce aggregate risk in the economy. For example, suppose the Fed were to purchase risky corporate bonds from the public. It might be tempting to conclude that the private sector now bears less risk on its balance sheet. However, the Fed is part of the government, which in turn is backstopped by taxpayers. As a result, the private sector still ultimately bears the risk.

Michael Woodford and Gauti Eggertsson (2003) extend that logic further to show what they call the irrelevance proposition for QE: balance sheet expansions have no effect on real activity unless they credibly and persistently affect expectations about the future path of monetary policy. Absent such expected permanence, QE itself is irrelevant.

Taken together, the irrelevance view implies that the conventional asset-side channels of QE—portfolio balance effects and signaling—are largely illusory in benchmark models. Asset purchases merely reshuffle government liabilities without changing aggregate risk or real incentives. The same logic also applies to liability-side QE: expansions of the monetary base can matter only if they are expected to be permanent and greater than the economy’s real demand for base money. Absent those conditions, QE should have little macroeconomic effect.

But are these benchmark models the right lens through which to understand QE in practice? Tyler Muir, one of the leading experts on QE, says no. His views are considered next.

The Insurance View of QE

Tyler Muir noted on the podcast that the benchmark models underlying the irrelevance view abstract from the institutions that actually set asset prices. In practice, financial markets are not priced by a representative household, but by a relatively small set of financial intermediaries—banks, dealers, hedge funds, and other leveraged investors—whose balance sheets are limited and whose risk-bearing capacity varies over time.

In this setting, QE works by reallocating risk away from constrained financial intermediaries precisely when they are least able to bear it. When the Fed purchases long-duration or risky assets during periods of stress, it removes those assets from intermediary balance sheets, relaxes binding risk constraints, and compresses risk premia. Because intermediaries are the marginal price setters in bond markets, this redistribution of risk has large and immediate effects on asset prices, even if the associated fiscal risk is borne more diffusely and over time by households. QE really does matter!

Moreover, in a series of important papers, Muir and his coauthors show that QE1–QE3 cumulatively lowered the 10-year Treasury yield by roughly 115 basis points. Only about 40 basis points of this decline can be attributed to the direct, mechanical effect of the Fed’s asset purchases. The remaining 75 basis points reflects what Muir characterizes as a form of state-contingent insurance for bond markets (see his discussion of it in the video above). Because investors expect the Fed to intervene aggressively in adverse states, long-term bonds become safer ex ante, lowering yields even in normal times outside of crises. This insurance channel is large, persistent, and demonstrates that LSAPs operating through the asset side of the Fed’s balance sheet are economically meaningful.

Muir’s work should bring some solace to Ben Bernanke’s worries about the theoretical underpinnings of QE. Still, the question remains as to why the LSAPs of the pandemic period, often called QE4, seem to pack far much more of a macreoconomic punch than QE1-QE3. This question is taken up next.

The Irreversibility View of QE

One way to illustrate the different macroeconomic outcomes between QE1-QE3 and QE4 is to look at the level of aggregate demand they generated relative to an estimate of a neutral path. Fortunately, the Mercatus Center provides just such a benchmark in its NGDP Gap measure. Based on forecasts, this measure captures the level of nominal GDP—or, equivalently, total dollar income—that households and firms expected at a given point in time relative to what actually occurred. The expected path can be interpreted as a neutral benchmark for aggregate demand and is shown in the figure below.

The figure highlights a stark contrast between the two QE episodes. Following the collapse in aggregate demand during the Great Recession, NGDP never returned to its pre-crisis trend path. Instead, expectations gradually adjusted downward, converging only slowly to a new, lower level of NGDP. In sharp contrast, the 2020–2022 period saw aggregate demand not only return to its expected path, but overshoot it. Recall that NGDP is total dollar spending and over the medium run its level is a policy choice.

Now both periods had QE programs. So what explains the difference? One way to answer that question is to revisit the QE gallows humor image above: the Fed effectively shot the buffalo slug!

So what does “shooting the buffalo slug” mean? In my discussion with Scott Sumner, we emphasized that QE4 effectively demonstrated the power of level targeting. During the pandemic, the federal government was determined to restore the level of aggregate nominal income, not merely its growth rate. Massive fiscal transfers—accommodated and financed by QE4—that the public viewed as a permanent wealth transfer made that commitment credible. And, in fact, the level of nominal income was restored and then overshot.

Level targeting, however, carries an important implication: it requires a commitment to policy irreversibility. To restore a missed level path of nominal income, markets must expect that some portion of the Fed’s balance sheet expansion will not be undone. Put differently, for QE to meaningfully raise the price level and nominal income, it must be perceived as at least partially permanent.

This irreversibility view of QE brings us back to the earlier work of Woodford and Eggertsson (2003). They argued that expansions of the monetary base can matter only if they are expected to be both permanent and large relative to the economy’s real demand for base money. Under those conditions, inflation expectations rise, real interest rates fall below their equilibrium level, and aggregate demand accelerates. The same mechanism can be understood through a forward-looking quantity-theoretic lens: a monetary base expansion that is expected to persist will ultimately raise the price level and, with it, nominal income.*

At this point, it is important to note that from a consolidated government budget constraint perspective, a permanently higher monetary base must ultimately be validated by fiscal policy. If future fiscal policy is expected to fully reverse the expansion through higher taxes or spending cuts, QE will not be perceived as irreversible. Consequently, for the irreversibility channel of QE to operate, fiscal policy must be non-Ricardian. Such an arrangement, in economic substance, resembles an implicit helicopter drop.

To summarize, QE4 succeeded where QE1–QE3 struggled because it finally fired the buffalo slug: a credible, fiscally backed, and irreversible commitment to restore the level of nominal income, operating primarily through the liability side of the Fed’s balance sheet rather than through the specific assets it purchased.

Conclusion: What I Would Say to Kevin Warsh About QE

If I were to summarize the lessons from QE1–QE4 and address Kevin Warsh’s concerns I would say the following:

• QE should remain an important contingency tool, not a standing policy.
  QE has proven valuable in zero lower bound environments and during acute financial stress, but it is not something the Fed needs—or should want—to deploy under normal macroeconomic conditions.

• QE must be understood through both sides of the Fed’s balance sheet.
  Asset-side QE matters as Tyler Muir’s work shows. But the macroeconomic punch of QE4 came primarily through the liability side.

• The liability-side power of QE should be automated through a level target.
  The irreversibility channel works best when it is not discretionary. A clear NGDP levle target (my preference) or price-level target would create a systematic way to implement QE when needed without requiring ad hoc interventions.

• QE must be paired with a credible exit strategy.
  Precisely because QE can be powerful, it must be deployed in a way that allows for complete and orderly quantitative tightening once conditions normalize. I have laid out how to do this in previous newsletters.

Taken together, these points suggest that if QE is used sparingly, guided by clear rules, and paired with credible balance sheet normalization plans, it can remain a legitimate part of the Fed’s policy toolkit.

*It is worth noting that Woodford and Eggertsson (2003) were not the only ones making such arguments. Scott Sumner has written a nice paper on the Princeton School of Macroeconomics who made similar points. This group included Ben Bernanke, Gautti Eggertson, Paul Krugman, Lars Svennson, and Michael Woodford. In the QE gallows humor image up top, Paul Krugman is depicted with a t-shirt that refers to his seminal 1998 paper in this literature.