UPDATE (2/25/2022): George Selgin informs me that in the final version (his book Floored) of the Cato working paper which I discuss below he modified the argument that I criticize that paying interest on reserves caused banks to raise their lending rates to borrowers and that he now generally agrees with my argument that paying interest on reserves did not cause banks to raise the interest rates they charged borrowers. George also points out that I did misstate his position slightly. He did not argue, as I wrote, that paying interest on reserves caused banks to raise interest rates to borrowers; his argument was that banks would accept a reduced percentage of loan applications at the prevailing rate of interest.
The economic theory of banking has a long and checkered history reflecting an ongoing dialectic between two views of banking. One view, let’s call it the reserve view, is that the circulating bank liabilities, now almost exclusively bank deposits, are created by banks after they receive deposits of currency (either metallic or fiat). Rather than hold the currency in their vaults as “safe deposits,” banks cleverly (or in the view of some, deceitfully or fraudulently) lend out claims to their reserves in exchange for the IOUs of borrowers, from which they derive a stream of interest income.
The alternative view of banking, let’s call it the anti-reserve view (in chapter 7 of my new book Studies in the History of Monetary Theory, I trace the two views David Hume and Adam Smith) bank liabilities are first issued by established money lenders, probably traders or merchants, widely known to be solvent and well-capitalized, whose debts are widely recognized as reliable and safe. Borrowers therefore prefer to exchange their own debt for that of the lenders, which is more acceptable in exchange than their own less reliable debt. Lenders denominate their IOUs in terms of an accepted currency so that borrowers can use the lender’s IOU instead of the currency. To make their IOUs circulate like currency, lenders promise to redeem their IOUs on demand, so they must either hold, or have immediate access to, currency.
These two views of banking lead to conflicting interpretations of the hugely increased reserve holdings of banks since the aftermath of the 2008 financial crisis. Under the reserve view, reserves held by banks are the raw material from which deposits are created. Because of the inflationary potential of newly created deposits, a rapid infusion of reserves into the banking system is regarded as an inflationary surge waiting to happen.
On the anti-reserve view, however, causation flows not from reserves to deposits, but from deposits to reserves. Banks do not create deposits because they hold redundant reserves; they hold reserves because they create deposits, the holding of reserves being a the cost of creating deposits. Being a safe asset enabling banks to satisfy instantly, and without advance notice, demands for deposit redemption, reserves are held only as a precaution.
All businesses choose the forms in which to hold the assets best-suited to their operations. Manufactures own structures, buildings and machines used in producing the products they sell as well as holding inventories of finished or semi-finished outputs and inputs into the production process, as well as liquid capital like bank deposits, and other interest or income-generating assets. Banks also hold a variety of real assets (e.g., buildings, vaults, computers and machines) and a variety of financial assets. An important class of those financial assets are promissory notes of borrowers to whom banks have issued loans by creating deposits. In the ordinary course of business, banks accumulate reserves when new or existing customers make deposits, and when net positive clearings with other banks cause an inflow of reserves. The direction and the magnitude of the flow of reserves into, or out of, a bank are not beyond its power to control. Nor does a bank lack other means than increasing lending to reduce its holdings of unwanted reserves.
While reserves are the safest, most liquid, and most convenient asset that banks can hold, non-interest-bearing reserves provide banks with no pecuniary yield, so holding reserves rather than interest-bearing assets, or assets expected to appreciate involve a sacrifice of income that must be offset by the safety, liquidity and convenience provided by reserves. When the Fed began paying interest on reserves in October 2008, the holding of reserves no longer required foregoing a pecuniary yield offered by alternative assets. The next safest and most liquid class of assets available to banks is short-term Treasury notes, which do provide at least a small nominal interest return. Until October 2008, there was an active overnight market for reserves — the Federal Funds market — in which banks with excess reserves could lend to banks with insufficient reserves, thereby enabling the banking system as a whole to minimize the aggregate holding of excess (i.e., not legally required) reserves.
Legally required reserves being unavailable to banks to satisfy redemption demands without incurring a penalty for non-compliance with the legal reserve requirement, required reserves provide banks with little safety or liquidity. So, to obtain the desired safety and liquidity, banks must hold excess reserves. The cost (foregone interest) of holding excess reserves banks can be minimized by holding interest-bearing Treasuries easily exchanged for reserves and by lending or borrowing as needed in the overnight Fed Funds market. In normal conditions, the banking system can operate efficiently with excess reserves equal to only about one percent of total deposits.
The Fed did not begin paying interest on reserves until October 2008, less than a month after a financial panic and crisis brought the US and the international financial system to the brink of a catastrophic meltdown. The solvency of financial institutions and banks having been impaired by a rapid loss of asset value, distinguishing between solvent and insolvent counterparties became nearly impossible, putting almost any economic activity dependent on credit at risk of being unwound.
In danger of insolvency and desperate for liquidity, banks tried to hoard reserves and increase holdings of Treasury debt. Though yielding minimal interest, Treasury notes serve as preferred collateral in the Fed Funds market, enabling borrowers to offer lenders nearly zero-risk overnight or short-term lending opportunities via repurchase agreements in which Treasury notes are sold spot and repurchased forward at a preset price reflecting an implied interest rate on the loan.
Increased demand for Treasuries raised their prices and reduced their yields, but declining yields and lending rates couldn’t end the crisis once credit markets became paralyzed by pervasive doubts about counterparty solvency. Banks stopped lending to new customers, while hesitating, or even refusing, to renew or maintain credit facilities for existing customers, and were themselves often unable to borrow reserves without posting Treasuries as collateral for repo loans.
After steadfastly refusing to reduce its Fed Fund target rate and ease credit conditions, notwithstanding rapidly worsening economic conditions, during the summer of 2008, an intransigent stance from which it refused to budge even after the financial panic erupted in mid-September. While Treasury yields were falling as the markets sought liquidity and safety, chaotic market conditions caused overnight rates in the Fed Funds market to fluctuate erratically. Finally relenting in October as credit markets verged on collapse, the Fed reduced its Fed Funds target rate by 50 basis points. In the catastrophic conditions of October 2008, the half-percent reduction in the Fed Funds target was hardly adequate.
To prevent a system-wide catastrophe, the Fed began lending to banks on the security of assets of doubtful value or to buy assets — at book, rather than (unknown) market, value – that were not normally eligible to be purchased by the Fed. The resulting rapid expansion of the Fed’s balance sheet and the creation of bank reserves (Fed liabilities) raised fears (shared by the Fed) of potential future inflation.
Fearing that its direct lending to banks and its asset purchases were increasing bank reserves excessively, thereby driving the Fed Funds rate below its target, the Fed sought, and received, Congressional permission to begin paying interest on bank reserves so that banks would hold the newly acquired reserves, rather use the reserves to acquire assets like borrower IOUs, lest total spending and aggregate demand increase. Avoiding such a potentially inflationary increase in aggregate demand had been the chief policy objective of the Fed throughout 2008 even as the economy slid into deep recession just prior to the start of the financial crisis, and the Fed was sticking to that policy.
Struggling to contain a deepening financial crisis while adhering to a commitment to a 2-percent inflation target, the Fed experimented for almost two months with both its traditional Fed Funds target and its new policy of paying interest on bank reserves. The Fed eventually settled on a target for the Fed Funds rate between zero and .25% while paying .25% interest on reserves, thereby making it unnecessary for banks with accounts at the Fed to borrow, and making them unwilling to lend, reserves in the Fed Funds market. Thanks to the massive infusion of reserves into the banking system, the panic was quelled and the immediate financial crisis receded, but the underlying weakness of an economy was aggravated and continued to deepen; the liquidity and solvency problems that triggered the crisis were solved, but the aggregate-demand deficiency was not.
In his excellent historical and analytical account of how and why the Fed adopted its policy of paying interest on reserves, George Selgin credits the idea that had the Fed not paid interest to banks on their reserves, they would have used those reserves to increase lending, thereby providing stimulus to the economy. (Update: as noted above, the argument I criticize was made in Cato Working Paper not in the published version of George’s book, and he informs me that he modified the argument in the published version and now disavows it.) Although I agree with George that paying interest on bank reserves reduced aggregate demand, I disagree with his argument that the reduction in aggregate demand was caused by increased interest charged to borrowers owing to the payment of interest on reserves.
George believes that, by paying interest on reserves, the Fed increased the attractiveness of holding reserves relative to higher-yielding assets like the IOUs of borrowers. And, sure enough, after the Fed began paying interest on reserves, the share of bank loans in total bank assets declined by about the same percentage as the share of reserves in total bank assets.
The logic underlying this argument is that, at the margin, an optimizing bank equates the anticipated yield from holding every asset in its portfolio. If the expected return at margin from bank loans exceeds the expected return from reserves, an optimizing bank will increase its lending until the marginal return from lending no longer exceeds the marginal return from holding reserves. When the Fed began paying interest on reserves, the expected return at the margin from holding reserves increased and exceeded the expected return at the margin from bank loans, giving banks an incentive to increase their holdings of reserves relative to their holdings of bank loans. Presumably this means that banks would try to increase their holdings of bank reserves and reduce their lending.
At least two problems undercut this logic. First, as explained above, the yield from holding an asset can be pecuniary – a yield of interest, of dividends, or appreciation – or a flow of services. Clearly, the yield from holding reserves is primarily the service flow associated with the safety, liquidity and convenience provided by reserves. Before October 2008, reserves provided no pecuniary yield, either in explicit interest or expected appreciation, the optimal quantity of reserves held was such that, at the margin, the safety, liquidity and convenience generated by reserves was just sufficient to match the pecuniary return from the loan assets expected by an optimizing bank.
After the Fed began paying interest on reserves, the combined pecuniary and service return from holding reserves exceeded the return from banks’ loan assets. So, banks therefore chose to increase their holdings of reserves until the expected pecuniary and service yield from reserves no longer exceeded the expected return from loan assets. But as banks increased their reserve holdings, the marginal service flow provided by reserves diminished until the marginal pecuniary plus service yield was again equalized across the assets held in banks’ asset portfolios. But that does not imply that banks reduced their lending or the value of the loan assets in their portfolios compared to the value of loan assets held before interest was paid on reserves; it just means that optimal bank portfolios after the Fed began paying interest on bank reserves contained more reserves than previously.
Indeed, because reserves provided a higher pecuniary yield and more safety, liquidity and convenience than holding Treasuries, banks were willing to add reserves to their portfolios without limit, because holding reserves became costless. The only limit on the holding of bank reserves was the willingness of the Fed to create more reserves by buying additional assets from the private sector. The proceeds of sales would be deposited in the banking system. The yield on the acquired assets would accrue to the Fed, and that yield would be transferred to the banking system by way of interest paid on those reserves.
So, if I don’t think that paying interest on bank reserves caused banks to raise interest rates on loans, why do I think that paying interest on bank reserves reduced aggregate demand and slowed the recovery from the Little Depression (aka Great Recession)?
The conventional story, derived from the reserve view, is that if banks have more reserves than they wish to hold, they try to dispose of their excess reserves by increasing their lending to borrowers. But banks wouldn’t increase lending to borrowers unless the expected profitability of such lending increased; no increase in the quantity of non-interest-bearing reserves of the banks would have increased the profitability of bank lending unless consumer confidence or business optimism increased, neither of which depends in a straightforward way on the quantity of reserves held by banks.
In several published papers on classical monetary theory which were revised and republished in my new book Studies in the History of Monetary Theory (chapters 2-5 and 7 see front matter for original publication information), I described a mechanism of bank lending and money creation. Competitive banks create money by lending, but how much money they create is constrained by the willingness of the public to hold the liabilities (deposits) emitted in the process of lending.
The money-lending, deposit-creation process can be imperfectly described within a partial-equilibrium, marginal-revenue, marginal-cost framework. The marginal revenue from creating money corresponds to the spread between a bank’s borrowing rate (the interest rate paid on deposits) and its lending rate (the interest rate charged borrowers). At the margin, this spread equals the bank’s cost of intermediation, which includes the cost of holding reserves. The cost of intermediation increases as the difference between the yields on Treasuries and reserves increase, and as the quantity of reserves held increases.
So, in the basic model I work with, paying interest on reserves reduces the cost of creating deposits, thereby tending to increase the amount of lending by banks, contrary to Selgin’s argument that paying interest on reserves reduces bank lending by inducing banks to raise interest rates on loans.
But, in a recession — and even more so in a financial crisis or panic — the cost of intermediation increases, causing banks to reduce their lending, primarily by limiting or denying the extension of credit to new and existing customers. Of course, in a recession, businesses and households demand fewer loans to finance spending plans, and instead seek credit with which to meet current obligations coming due. As banks’ costs of intermediation rise, they inevitably curtail lending, increasing the share of reserves in banks’ total assets.
While Selgin attributes the increasing share of reserves in banks’ assets to the payment of interest on reserves, a more plausible explanation of the increase is that it results from the increased intermediation costs associated with recession and a financial crisis, which more than offset the cost reduction from paying interest on reserves.
Although paying interest on reserves was a major innovation, in a sense it was just a continuation of the policy approach adopted by the Fed in 2004 when started gradually raising its Fed Funds target rate to 5.25% in June 2006, where it stood until July 2007. Combined with the bursting of the housing bubble in 2006, the 5.25% Fed Funds target produced a gradual slowdown that led the Fed to reduce its target, but always too little and too late, as the economy slid into recession at the end of 2007. So, the payment of interest on reserves, intended to ensure that the reserves would not trigger a surge in spending, was entirely consistent with the restrictive policy orientation of the Fed before the financial panic and crisis of 2008, which continued during and after the crisis. That policy was largely responsible for the unusually weak economic recovery and expansion in the decade after the crisis, when banks willingly absorbed all the reserves created by the Fed.
The specific point on which I disagree with Selgin is his belief that paying interest on bank reserves discouraged banks from increasing their lending despite the increase in their reserves. I maintain that paying interest on reserves did not discourage banks from lending, but instead altered their incentive to hold reserves versus holding Treasuries. That decision was independent of the banks’ lending decisions. The demand for loans to finance spending plans by businesses and households was declining because of macroeconomic conditions in a recessionary economy during a financial crisis and recession and the subsequent slow recovery.
Had the Fed not paid interest on reserves while purchasing assets to provide liquidity to the banking system, I am doubtful that banks would have provided credit for increased private spending. If no interest were paid on reserves, it seems more likely that banks would have used the additional reserves created by the Fed to purchase Treasuries than to increase lending, driving up their prices and reducing their yields. Instead of receiving interest of .25% on their reserves, banks would have received slightly less interest on short-term Treasuries. So, without interest on reserves, banks would have received less interest income, and incurred slightly more risk, than they actually did. The Fed, on the other hand, would have had a net increase in revenue by not paying more interest to banks than it received from the Treasuries sold by the banks to the Fed.
The only plausible difference between paying interest on reserves and not doing so that I can see is that the Fed, by paying interest on reserves, lent credibility to its commitment to keep inflation at, or below, its 2-percent target. The Fed’s own justification for seeking permission to pay interest on reserves, as Selgin (Floored, p. 18) documented with a passage from Bernanke’s memoir , was that not doing so might result in an inflationary increase in lending by banks trying to shed their excess reserves. Because I believe that expectations of inflation have a tendency to be self-fulfilling, I don’t dismiss the idea that paying interest on reserves helped the Fed anchor inflation expectations at or near its 2-percent inflation target.
Economic conditions after the financial crisis of 2008-09 were characterized by an extreme entrepreneurial pessimism that Ralph Hawtrey called a credit deadlock, conditions akin to, but distinct from, the more familiar Keynesian phenomenon of a liquidity trap. The difference is that a credit deadlock results from pessimism so intense that entrepreneurs (and presumably households as well) are unwilling, regardless of the interest rate on loans, to undertake long-term spending plans (capital investment by businesses or consumer-durables purchases by households) requiring credit financing. In a liquidity trap, such spending plans might be undertaken at a sufficiently low interest rate, but the interest rate cannot fall, bear speculators cashing in their long-term bond holdings as soon as long-term bond prices rise to a level that speculators regard as unsustainable. To me, at least, the Hawtreyan credit deadlock seems a more plausible description of conditions in 2008-09 than the Keynesian liquidity trap.
In a Hawtreyan credit deadlock, the capacity of monetary policy to increase spending and aggregate demand is largely eliminated. Here’s Hawtrey’s description from the 1950 edition of his classic work Currency and Credit.
If the banks fail to stimulate short-term borrowing, they can create credit by themselves buying securities in the investment market. The market will seek to use the resources thus placed in it, and it will become more favourable to new flotations and sales of securities. But even so and expansion of the flow of money is not ensured. If the money created is to move and to swell the consumers’ income, the favourable market must evoke additional capital outlay. That is likely to take time and conceivably capital outlay may fail to respond. A deficiency of demand for consumable goods reacts on capital outlay, for when the existing capacity of industries is underemployed, there is little demand for capital outlay to extend capacity. . .
The deadlock then is complete, and, unless it is to continue unbroken till some fortuitous circumstance restarts activity, recourse must be had to directly inflationary expedients, such as government expenditures far in excess of revenue, or a deliberate depreciation of the foreign exchange value of the money unit.
In this passage, Hawtrey, originator of the widely reviled “Treasury View” (also see chapters 10-11 of my Studies in the History of Monetary Theory) that denied the efficacy of fiscal policy as a countercyclical tool, acknowledged the efficacy of fiscal policy in a credit deadlock, while monetary policy could be effective only through currency devaluation or depreciation, though I would add that in monetary policy could also be effective by inducing or creating expectations of inflation.
The long, but painfully slow, recovery from the 2008-09 financial crisis lent credence to Hawtrey’s description of credit deadlock, and my own empirical findings of the unusual positive correlation between changes in inflation expectations and changes in the S&P 500 supports the idea that increasing inflation expectations are a means whereby monetary policy can enable an escape from credit deadlock.
Uneasy Money 