Raghuram Rajan: Breaking the QE Ratchet Effect

Raghuram Rajan on liquidity dependence, the Fed’s expanding balance sheet, and how to restore normalization as the default.

This week the Macro Musings podcast went full video! Yes, after ten years and 530 episodes, we took the plunge and joined the world of full video podcasting. To be clear, our traditional audio version of the podcast will continue but now each episode will also be available in full video on YouTube. We’ve been slowly building toward this for some time, and it felt like the right moment to expand the show’s reach and give listeners a richer way to engage with our discussions. If you haven’t yet, I hope you’ll subscribe to the YouTube channel and help us grow this next chapter of Macro Musings.

To kick things off, we had Raghuram Rajan join us for a wide-ranging and timely conversation. Raghu is the former chief economist at the IMF, former governor of the Reserve Bank of India (RBI), and now a professor at the University of Chicago and chair of the Group of 30. Among other things, we discussed his famous 2005 Jackson Hole warning about financial fragility, his Volcker-like re-anchoring of Indian inflation, his efforts to stabilize India during the taper tantrum, and his views on the long-run prospects for the Indian economy.

Most of our time, though, was spent on his research addressing one of the most pressing questions facing the Federal Reserve: what are the long-run consequences of repeated rounds of quantitative easing? We explored his work on the “ratcheting effect” of QE, the growing liquidity dependence of the financial system, and why shrinking central bank balance sheets has proven so difficult in practice. I previously covered this research in an earlier newsletter and received several follow-up questions. In the remainder of this essay, I will summarize Raghu’s arguments, restate my proposed solution, and respond to a few questions.

The QE Ratchet Effect and Liquidity Dependence

In a series of papers with Viral V. Acharya, Rahul S. Chauhan, and Sascha Steffen, Raghu shows that QE alters the liability structure of bank balance sheets. As reserves flood into the banking system, banks become more comfortable funding themselves with demandable deposits and other runnable liabilities while shrinking time deposits. The reserves created by QE are viewed as a backstop for these more liquid liabilities. The result is that even as QE raises the gross level of reserves, banks simultaneously expand claims on that liquidity. Net aggregate liquidity, therefore, is smaller than headline reserve figures suggest. When QT begins and reserves decline, those runnable liabilities do not unwind proportionately. This asymmetry—liabilities that expand during QE but do not contract during QT—is what Raghu and his coauthors call the ratcheting effect of QE.

The data strongly support this mechanism. Raghu’s work shows that demandable deposits rise during each QE episode and do not meaningfully reverse afterward. At the bank level, institutions that receive larger reserve inflows expand uninsured deposits and other runnable liabilities more aggressively. QE, in short, expands liquidity promises along with liquidity itself. When QT reduces reserves, those promises remain. Net aggregate liquidity therefore shrinks faster than headline reserve balances suggest.

Other research, such as Darst et al. (2026), reinforces this conclusion: QE may appear to flood the system with liquidity, but by reshaping funding structures and compressing private liquidity insurance, it can actually reduce effective liquidity in the economy. As a result, banks’ structural demand for reserves grows after each round of QE because the system itself has become more liquidity-dependent. In effect, the supply of reserves creates its own demand.

This dynamic looks less like a well-oiled liquidity machine and more like a liquidity blob. It grows larger, more complex, and increasingly dependent on continuous support.

If liquidity dependence is the problem, the solution lies in redesigning how liquidity is supplied, priced, and allowed to mature.

My Proposed Four-Step Solution

In the above Barron’s op-ed, I outlined a four-step approach to shrinking the Fed’s balance sheet. Those steps, however, can be understood more broadly as a blueprint for a state-contingent operating system, one that permits QE when needed at the ZLB while ensuring it is fully offset by QT once conditions normalize. Properly designed, such a framework would allow elasticity in bad times without embedding the ratchet effect into the Fed’s balance sheet.

The first step is to normalize the Fed’s ceiling facilities, the Discount Window and Standing Repo Operations. I outline some specific steps here that would reduce stigma and encourage routine use. Doing so would reduce structural demand for reserves.

The second step is to neutralize swings in the Treasury General Account (TGA), either by adopting Annette Vissing-Jorgensen plan or the Bill Nelson plan. The former relies on Treasury bill purchases and sales; the latter uses temporary repo operations to offset reserve fluctuations caused by TGA movements. Either approach would make it easier for the Fed to shrink reserves without jeopardizing interest-rate control.

The third step is a Treasury–Fed asset swap. The Treasury would issue additional short-term bills and exchange them for the longer-term bonds currently held by the Fed. Because these bills would not finance new spending, they would not count against the debt ceiling. Once the Fed holds short-term bills, it could sell them gradually or simply allow them to mature. In either case, the balance sheet would shrink more smoothly, and reserves would drain with less market disruption than if the Fed had to sell long-term bonds outright. Again, see this earlier newsletter for more details.

My fourth step directly addresses the Raghu ratchet effect. It calls for the Federal Reserve to adopt a corridor-style framework that would rely heavily on the term deposit facility. Specifically, the Fed would offer term deposits at the target policy rate, set interest on reserves slightly below it, and place the discount window and standing repo operation rates slightly above it.

By introducing maturity into the Fed’s liabilities, this structure would reduce banks’ reliance on runnable funding while preserving interest-rate control and lender-of-last-resort capacity. Because term deposits would earn more than overnight reserves, banks would have less incentive to treat reserves as a costless, open-ended backstop for short-term liabilities. Most importantly, term deposits would automatically roll off at maturity unless renewed. QT would therefore become the default once QE ended, preventing the supply of reserves from permanently ratcheting upward. This final step draws inspiration from the Riksbank operating system.

Questions on My Proposed Four-Step Solution

Below are three questions I have been asked multiple times about my proposal. My answers to these questions are provided below.

Question 1: You claim that when QE ends, term deposits would roll off unless renewed, and the balance sheet would shrink by default. Would not maturing term deposits just roll into regular reserve deposits?

Answer: First, note that under the current setup QT is initiated by the Fed in a top-down manner. Reserves fall exogenously and banks are forced to respond as best they can: hoard liquidity, cut credit, or bid up money-market rates. In short, banks adjust after reserves are removed, not before.

In my proposed setup, banks effectively initiate QT (after QE ends) on their own terms. That is, as term deposits mature, banks have to decide whether to (1) sit on the reserves they get once the term deposit matures (2) roll over the funds back into new term deposits, or (3) reduce their own balance sheets which, in turn, will allows the Fed to run off its asset holdings.

Option (1) is unlikely because, under this framework, reserves would earn less than the targeted policy rate, making them an unattractive asset to hold in large quantities. Option (2) is also unlikely to dominate because continuously rolling over term deposits preserves balance-sheet size and the associated regulatory, capital, and leverage costs, even after the demand for excess liquidity has faded.

As a result, once QE ends and the value of warehousing liquidity declines, banks will find it optimal to shrink their own balance sheets. When they do so, the demand for Fed liabilities falls, allowing the Fed to let its assets mature without reinvestment. QT, therefore, occurs automatically, not because the Fed drains reserves but because banks no longer wish to fund a large central bank balance sheet.

Question 2: Would not the Fed-Treasury asset swap force the Fed to recognize the mark-to-market loss on its bonds?

Answer: The asset swap would not create new losses for the Fed; it would simply make existing unrealized losses transparent. The swap effectively transfers duration risk from the Fed’s balance sheet to the Treasury’s. To the extent losses exist, they are already there. The swap merely makes them explicit and places them at Treasury.

Question 3: Would not the Fed-Treasury asset swap affect term premiums on treasury bonds?

Answer: There should be no direct effect on term premiums. The long-term bonds currently held by the Fed are already out of circulation. In the swap, the Fed transfers those bonds back to the Treasury, which effectively retires them and issues short-term bills in exchange. From the public’s perspective, the stock of long-term bonds in circulation does not change.

A secondary effect is possible. Once the Fed holds short-term bills, it may sell them into the market. That would shorten the duration of Treasury debt held by the public. If investors then rebalance toward longer-term bonds, term premiums could actually decline rather than rise.

Conclusion

QE may be necessary in crises, but without structural reform it leaves behind a larger, more liquidity-dependent system each time. The challenge is to preserve elasticity in bad times while making discipline the default in good ones. Raghu and his coauthors’ work shows just how consequential this design choice is.