Missing the Forest for the Trees

In many ways, Milton Friedman's ideas were avant-garde and helped shape modern monetary economics. In his famous 1967 AEA Presidential Address, Friedman introduced the natural rate hypothesis and argued that inflation expectations ultimately limit the ability of monetary policy to influence real economic activity. This insight laid much of the intellectual foundation for modern central banking. Friedman was also an early advocate of large scale asset purchases, as evidenced by his call in 2000 for the Bank of Japan to do QE to end Japan’s deflation. He further anticipated modern thinking on monetary transmission by embracing the portfolio balance channel long before it became vogue. And while Friedman never explicitly advocated NGDP targeting, he eventually came to view “the broadest framework . . . of the work that I and others have done in analyzing monetary experience” as a “theory of nominal income,” a perspective that many proponents of NGDP targeting, like myself, also share.

Another example of Friedman's ability to anticipate future developments can be found in his 1969 essay The Optimum Quantity of Money. In it, Friedman argued that the opportunity cost of holding money should be driven to zero so that households and firms do not inefficiently economize on liquidity. Put differently, Friedman viewed the opportunity cost of holding money as a form of implicit tax on liquidity. The higher this tax, the greater the incentive to devote real resources toward avoiding money holdings.

This understanding became known as the Friedman Rule. Advocates of the ample reserve system often invoke it as a justification for maintaining a large Fed balance sheet while paying interest on reserve balances at market rates. Former Fed Chair Ben Bernanke recently made this point at a Brookings conference when asked about the size and importance of the Fed’s balance sheet:

[I]t should be big, again on good economics, because it’s a very good way to manage the short-term interest rate and assure that banks have enough reserves. And that’s a good thing for financial stability.

It’s a good thing. Milton Friedman would like it because it keeps the cost of transactions low. You know, money should have a zero opportunity cost and all those things… So I think, from an economic point of view, that the balance sheet is pretty much okay where it is.

As a fan of Milton Friedman, I am sympathetic to this argument. It captures an important insight. If liquidity is taxed, households and firms will devote real resources toward avoiding it. Good monetary policy should seek to minimize those costs.

However, applying the Friedman Rule to the Fed’s balance sheet is more complicated than Bernanke’s remarks suggest. The Friedman Rule is fundamentally about the cost of liquidity, not the quantity of reserves. Yet advocates of the ample reserve system often move from the proposition that liquidity should not be taxed to the conclusion that the Fed should maintain a permanently large balance sheet. That conclusion does not necessarily follow.

In addition, the Friedman Rule is only one margin among many that should guide balance-sheet policy. The size of the Fed’s balance sheet also affects market structure, regulatory burdens, interbank lending, Treasury market functioning, central bank independence, and other considerations. Focusing narrowly on the opportunity cost of reserves risks missing the balance-sheet forest for the Friedman-rule trees, as illustrated above.

In this newsletter, I explain why this is the case. I begin by showing how an ample reserve system can satisfy the Friedman Rule and why, at times, it may fail to do so. I then consider alternative ways to satisfy the Friedman Rule before turning to the broader set of tradeoffs that should also inform the size of the Fed’s balance sheet.

To illustrate how an ample reserve system can satisfy the Friedman Rule, consider the equation below. It frames the opportunity costs of liquidity in terms of the potential tax on reserves borne by banks:

The first term, Market Rate – IORB, measures the opportunity cost of holding reserves rather than alternative safe assets. If reserves earn less than, say, Treasury bills, then banks face an implicit tax for holding liquidity. This spread can be viewed as the implicit tax rate on bank reserves.

The second term, Reserve Holdings, is the tax base. It measures the quantity of reserves on which this implicit tax is imposed. Taken together, the equation implies that the total tax burden for banks holding reserves depends on both the opportunity cost of reserves and the quantity of reserves held.

From this perspective, the Friedman Rule can be understood as a call to minimize the tax rate on bank reserves. If the Fed pays interest on reserves at a rate equal to the return on comparable safe assets, then the opportunity cost of holding reserves is zero. The implicit tax on liquidity disappears.

Enter the ample reserve system. Under this approach, the Fed supplies enough reserves to more than satiate banks' demand for liquidity while paying interest on those reserves at a rate comparable to other safe assets. In effect, the Fed floods the banking system with reserves, placing it on the flat portion of the reserve demand curve. That, in turn, pushes short-term market rates toward IORB. As a result, banks can hold reserves without sacrificing yield relative to other safe assets. From this perspective, the ample reserve system provides a practical implementation of Friedman's insight. It minimizes the opportunity cost of bank liquidity by using a large balance sheet to align market rates with the interest paid on reserves.

This is the argument Bernanke and other defenders of the ample reserve framework have in mind. If liquidity should not be taxed, then supplying ample reserves while paying a market rate of interest on them seems like a reasonable solution.

The Friedman-rule defense of the ample reserve system rests on a simple proposition: if reserves earn a market rate of return, then banks face no opportunity cost from holding liquidity. But this conclusion only follows if reserves are otherwise costless to hold.

The problem is that the post-GFC regulatory framework has reintroduced a tax on liquidity through a different channel. Even when reserves earn a market rate of interest, leverage requirements and other balance-sheet constraints can make them costly to hold.

To see this point, consider a slightly modified version of the reserve-tax equation introduced above:

The significance of this new term, Regulatory Cost(R), is that the ample reserve system may not always satisfy the Friedman Rule. The traditional argument assumes that paying IORB at a market rate eliminates the opportunity cost of holding reserves. But if expanding the Fed’s balance sheet also increases the regulatory costs associated with holding reserves, then the effective tax on liquidity does not disappear. It simply changes form.

An example of the Regulatory Cost(R) term is the Supplementary Leverage Ratio (SLR). It is a post-2008 regulation that requires large banks to hold a minimum amount of capital against their total assets, regardless of risk. The SLR treats reserve balances much like other assets on a bank’s balance sheet. Consequently, when the Fed creates reserves through QE, those reserves consume balance-sheet capacity even though they are risk-free assets.

The Federal Reserve itself implicitly acknowledged this reality in 2020 when it temporarily excluded reserve balances and Treasury securities from the SLR after a surge in reserves and deposits placed pressure on bank balance sheets. Notably, the exemption did not alter the interest paid on reserves. It reduced the regulatory cost of holding them.

This concern becomes even more important once one recognizes the ratchet effect associated with repeated rounds of QE. As shown by Raghuram Rajan and his coauthors, QE can create a form of liquidity dependence in which reserve demand rises alongside reserve supply and does not fully reverse when the Fed shrinks its balance sheet. Bill Nelson has pointed to a complementary channel: supervisory expectations regarding prudent liquidity management may also ratchet upward over time, leading banks to maintain larger reserve buffers than they otherwise would. These developments suggest a broader concern about a supply-driven ample reserve system. By continually accommodating elevated reserve demand—whether driven by liquidity dependence, supervisory expectations, or both—it may cause reserve balances and the Fed's balance sheet to grow faster than the underlying economy over time.

This possibility creates a deeper challenge for the Friedman-rule defense of ample reserves. If regulatory costs rise with reserve holdings, then a larger balance sheet increases both the quantity of reserves subject to the reserve tax and the effective tax rate itself. In that case, the very mechanism used to eliminate the tax on liquidity can gradually recreate it through the expansion of the Fed’s balance sheet.

More fundamentally, regulatory costs are only one consequence of maintaining a large balance sheet. As discussed later, the same large balance sheet that drives market rates toward IORB may also suppress interbank lending, create liquidity dependence, weaken lending to the real economy, and undermine the Fed's independence. This raises a broader question: are there other ways to satisfy the Friedman Rule without relying on permanently abundant reserves?

Fortunately, the Friedman Rule does not uniquely imply an ample reserve system. There are at least three alternative ways to reduce the tax on bank reserves without relying on a permanently large Fed balance sheet.

First, the Fed could aim to reduce reserve demand and operate with a smaller balance sheet. Much of the Friedman-rule defense of ample reserves focuses on reducing the opportunity cost spread (the tax rate) while holding reserve quantities (the tax base) constant. But the quantity of reserves is itself a policy choice. If banks can safely operate with fewer reserves, then shrinking the Fed's balance sheet reduces the quantity of reserves subject to the reserve tax. Recent calls by Darrell Duffie, Stephan Miran, Lorie Logan, Bill Nelson and others to reduce the structural demand for reserves point in this direction. While reducing reserve demand does not eliminate the importance of paying a competitive return on reserves, it does weaken the claim that a permanently large balance sheet is necessary to satisfy the Friedman Rule.

Second, the Fed could adopt a voluntary reserve-targeting regime, as recommended in a 2008 Fed staff study. Under this approach, banks would voluntarily choose reserve targets and earn a market return on reserve balances held within a band around those targets. This would eliminate the opportunity cost of liquidity while allowing the Fed to operate with a smaller balance sheet. It would also encourage the revival of interbank lending. Banks holding reserves above their target would earn a lower return on excess balances and therefore have an incentive to lend them to other institutions. Conversely, banks falling short of their target would have an incentive to borrow reserves rather than incur a penalty.

Third, the Friedman Rule can be approximated in a demand-driven operating system anchored by robust, business-as-usual ceiling facilities. If banks can obtain reserves on demand at rates close to market rates, then they need not hold large precautionary reserve balances (the tax base) and face little or no opportunity cost (the tax rate) in accessing liquidity when needed. In this framework, liquidity is not pre-funded through ample reserves. Instead, it becomes latently ample through reliable access to central bank facilities.

The broader point is that the Friedman Rule is often interpreted as an argument for abundant reserves. A better interpretation is that it is an argument for abundant liquidity. Once that distinction is recognized, demand-driven operating systems become a natural alternative to the ample reserve system.

A final point is that even if an ample reserve system satisfies the Friedman Rule, it only addresses one margin in determining the optimal size of the Fed’s balance sheet. But the Fed’s balance-sheet policy is not a single-margin optimization problem. The Fed must weigh the benefits of reducing the opportunity cost of reserves against the costs that emerge elsewhere in the system. At least four additional margins deserve attention.

By supplying reserves in such abundance that banks have little need to borrow or lend reserves, the ample reserve system reduces the role of interbank funding markets in allocating reserves across institutions. Over time, interbank lending can atrophy weakening price discovery and market discipline. The tradeoff is straightforward: lower liquidity costs versus a diminished role for private liquidity redistribution. This margin has become a motivating factor for many central banks moving toward a demand-driven operating system.

As noted above, Raghuram Rajan and his coauthors argue that QE can create a form of liquidity dependence in which reserve demand rises alongside reserve supply and does not fully reverse during QT. If so, ample reserves may not merely satisfy liquidity demand but help create it. The result is a financial system that becomes increasingly dependent on high reserve levels and a Fed balance sheet that tends to grow faster than the underlying economy.

Research by William Diamond and his coauthors find that large reserve injections can crowd out bank lending by raising the marginal cost of expanding bank balance sheets. To the extent that ample reserves encourage banks to hold more liquidity and rely on shorter-term funding, they may reduce incentives for the maturity transformation that underpins traditional bank lending to the real economy.

A larger balance sheet may also create political vulnerabilities. Interest paid on reserves can be portrayed as a subsidy to banks, operating losses can invite criticism, and large asset holdings can blur the line between monetary and fiscal policy. As the Fed’s balance sheet grows, so too may the political pressures surrounding its use.

Taken together, these margins underscore the broader point of this essay. The Friedman Rule addresses an important question: how can the opportunity cost of liquidity be minimized? But it does not answer every question relevant to the size of the Fed’s balance sheet. The optimal operating framework must balance reserve efficiency against its effects on market functioning, liquidity fragility, bank lending, and central bank independence. Focusing on only one of these margins risks missing the balance-sheet forest for the Friedman-rule trees.

Ben Bernanke is right that the Friedman Rule offers an important insight: liquidity should not be taxed. But it does not follow that a permanently large Fed balance sheet is the only—or even the best—way to achieve that objective. As I have shown, an ample reserve system may not always satisfy the Friedman Rule once regulatory costs and balance-sheet constraints are taken into account. Moreover, even when it does, the Friedman Rule is only one margin among many that should guide balance sheet policy. Interbank lending, liquidity fragility, bank lending to the real economy, and central bank independence all deserve consideration as well. The broader lesson is that the Friedman Rule is best understood as an argument for abundant liquidity, not abundant reserves. Once that distinction is recognized, alternative operating frameworks become possible Put differently, the Friedman Rule identifies an important tree. Determining the optimal size of the Fed's balance sheet requires understanding the entire forest.