Few monetary policy tools have generated as much argument as quantitative easing, the practice by which central banks buy up large quantities of long-term government bonds. Critics have blamed QE for fueling the post-pandemic inflation surge and saddling central banks with enormous losses when they later had to raise interest rates. Defenders point to the recessions that QE helped shorten. So who’s right, and under what conditions?
A new NBER working paper by economists at the International Monetary Fund attempts to sort out these questions by building a detailed economic model and running it through a series of stress tests. Their conclusion is nuanced: QE tends to be a valuable tool during deep recessions and can actually improve the government’s overall fiscal position, but it becomes considerably riskier when deployed in milder downturns.
The question behind the model
The research team, led by Tobias Adrian of the IMF and including Christopher Erceg, Marcin Kolasa, Jesper Lindé, and Pawel Zabczyk, wanted to reconsider how QE should be evaluated. In public discussions, central bank losses on their bond portfolios have received substantial attention. When the Federal Reserve, European Central Bank, and Bank of England raised interest rates rapidly starting in 2022, the long-term bonds they had accumulated during the pandemic lost value, and the short-term interest they now owed on reserves exceeded what they earned on those bonds.
The authors argue this narrow accounting misses the bigger picture. What matters, they contend, is the combined fiscal position of the government and central bank taken together, because central bank profits and losses are ultimately transferred to (or covered by) the treasury. If QE lowers government borrowing costs, boosts tax revenue by stimulating the economy, and reduces the real value of outstanding debt through modestly higher inflation, those benefits can far outweigh the losses on the central bank’s balance sheet.
Building an economy on a laptop
To test this idea, the team built what economists call a DSGE model, essentially a simulated economy governed by mathematical equations describing how households, firms, and the government behave. Their version has some distinctive features designed to capture recent experience.
One feature is segmented bond markets. In the model, some households can only save through long-term bonds, not short-term ones. This means that when the central bank buys long-term bonds, it genuinely changes conditions those households face, rather than being neutralized by arbitrage. Another feature is a nonlinear relationship between economic slack and inflation, so that inflation responds mildly when the economy is weak but much more aggressively when it starts running hot. A third feature is what the authors call “cognitive discounting,” a way of assuming households don’t fully believe every promise about the distant future, which limits how much a central bank can accomplish just by talking about future policy.
The researchers calibrated the model to match U.S. economic data, including how the economy has historically responded to interest rate changes, government spending shocks, and asset purchases. They deliberately chose a conservative setting for how much QE moves the economy, meaning their estimates likely understate rather than overstate its benefits.
Deep recessions: QE looks like a bargain
The team first simulated a severe recession in which the ideal policy interest rate would be well below zero but is stuck at the “effective lower bound” of zero for a prolonged period. Think of the years following the 2008 financial crisis, when unemployment stayed high for a long time.
In this environment, a QE program equal to 10 percent of GDP produced a peak output boost of roughly 0.8 percent and lifted inflation by about 0.2 percentage points. Cumulative output losses over the simulation shrank by about 13 percent. If the QE program were closer to the 20 percent of GDP that many central banks actually deployed during COVID, the improvement would be substantially larger.
On the fiscal side, the results are striking. Government debt fell by about 6 percent of GDP after five years relative to a scenario without QE. The improvement came from several sources working together: higher tax revenue from a stronger economy, lower interest payments on government debt, cheaper borrowing costs, and a slight erosion of existing debt from higher prices. The central bank itself made modest profits, since it was buying long-term bonds yielding more than the near-zero rate it paid on its own liabilities.
The contrast with conventional fiscal stimulus is sharp. When the researchers compared QE to a government spending increase calibrated to produce the same output boost, they found the spending program raised debt by about 3 percent of GDP, a total swing of roughly 10 percentage points versus QE. The spending program generated less tax revenue than QE, partly because it drew more people into the labor force, holding down wages and the tax base.
Shallow slumps: A riskier proposition
The picture changes when QE is used in what the authors call a “shallow liquidity trap,” where the economy is close to potential but inflation is stuck below target and interest rates are near zero. This roughly describes conditions in Europe and Japan for stretches before the pandemic.
Here the same 10 percent of GDP QE program does less work. The output boost is smaller and shorter-lived because the central bank starts raising interest rates sooner, offsetting some of QE’s effect. More concerning, the risks change dramatically. If a shock hits and the economy recovers faster than expected, the central bank may need to raise rates aggressively to contain inflation, and QE can end up amplifying overheating rather than smoothing the recovery.
The nonlinear Phillips curve in the model plays a big role here. Once the economy heats up past a certain point, inflation accelerates rapidly, meaning a QE program that seemed reasonable in a shallow slump can push inflation noticeably higher than intended. The authors interpret this as consistent with concerns raised by economist Athanasios Orphanides and others that QE contributed to the post-COVID inflation problem.
The commitment aspects of QE compound these risks. Central banks typically pair asset purchases with pledges to keep rates low even after purchases end. If the economy recovers faster than expected, this commitment can prevent the central bank from raising rates as quickly as it otherwise would, amplifying the overheating. When the authors simulated a scenario where the central bank delayed liftoff by a year despite an upside surprise, the inflation overshoot grew substantially.
What about those central bank losses?
The paper devotes considerable attention to the question that has driven much of the public criticism of QE: what if the central bank takes big losses on its bond portfolio? The researchers ran stochastic simulations that stress-tested QE against the kind of macroeconomic volatility observed between 1960 and 2019, a period that includes the Great Inflation.
Even in this challenging environment, and even when the model was set up unfavorably (with an already-low term premium and a large QE program of 20 percent of GDP), the probability that QE would worsen the consolidated fiscal position was small. Central bank losses could indeed be substantial, sometimes exceeding 2 percent of GDP over five years, but they were typically outweighed by the fiscal benefits elsewhere.
The authors argue this shows that focusing narrowly on central bank losses is misleading. “Central bank losses from QE,” they write, “are a poor metric of the fiscal implications of QE.”
Caveats and cautions
The researchers are careful to note limits on their conclusions. Their model doesn’t capture financial stability risks that may arise when central banks unwind QE through “quantitative tightening.” It also doesn’t model whether large central bank losses could damage the institution’s credibility or political independence in ways that would matter over the long term.
They also caution that the sensitivity of long-term interest rates to QE may weaken when yields are already low, meaning additional purchases might do less to stimulate the economy while still exposing the central bank to losses if rates later rise. And they emphasize that QE should probably be paired with fiscal support rather than deployed alone.
The practical implication the authors draw is that central banks should think about refining how they implement QE, particularly by including “escape clauses” that allow them to raise rates promptly if inflation pressures emerge, rather than binding themselves to prolonged low-rate commitments. In deep recessions the case for aggressive QE looks strong on both macroeconomic and fiscal grounds. In milder slumps, size matters, and so does flexibility.
