“The Liquidity Coverage Ratio is a key component of the Basel III framework …. For the first time in regulatory history, we have a truly global minimum standard for bank liquidity.” Then-Bank of England Governor Mervyn King cited as Chairman of the Group of Governors and Heads of Supervision, January 6, 2013
Modern bank regulation has two complementary parts: capital and liquidity requirements. The first states that, for a given quantity and riskiness of their assets and off-balance sheet exposures, a bank must have a minimum prescribed level of equity financing. By contrast, based on the structure of their liabilities, liquidity requirements specify the quantity of a bank’s assets that must be easily convertible into cash. That is, one part of the regulatory framework restricts liabilities given the structure of assets (capital requirements) and the other limits assets based on the composition of liabilities (liquidity requirements). (See, for example, Cecchetti and Kashyap.)
While the structure of capital regulation―especially in its risk-based form―is a creation of the last quarter of the 20th century, liquidity regulation is much older. In fact, the newly implemented liquidity coverage ratio (LCR) harks back to the system in place over 100 years ago. In the United States, before the advent of the Federal Reserve in 1914, both national and state-chartered banks were required to hold substantial liquid reserves to back their deposits (see Carlson). These are the reserve requirements (RR) that remain in effect in most jurisdictions today, the United States included.
In this post, we briefly examine the long experience with RR as a way to gain insight regarding the LCR. We draw two conclusions. First, we argue strongly against using the LCR as a monetary policy tool in advanced economies with well-developed financial markets. Like RR, it is simply too blunt and unpredictable. Second, for the LCR to work as a prudential policy tool, it should probably be supplemented by something like a fee-based line of credit at the central bank.
The similarity of liquidity and reserve requirements is evident in their definitions. The Basel Committee on Banking Supervision describes the LCR as follows:
[It ensures] that banks have an adequate stock of unencumbered high-quality liquid assets (HQLA) that can be converted easily and immediately in private markets into cash to meet their liquidity needs for a 30 calendar day liquidity stress scenario. (See paragraph 1, here.)
Feinman (page 570) provides a canonical definition of a reserve requirement:
[D]epository institutions … are required to maintain reserves against transaction deposits … and other highly liquid funds.
Digging slightly deeper, the likeness becomes even more apparent: both state that the quantity of certain bank assets must exceed a specific weighted average of their liabilities. In the case of RR, reserves―defined as deposits at the central bank plus cash in a bank’s vault― must surpass a fraction of qualifying deposits. Hence, the term fractional reserve banking. Similarly, the LCR states that the sum of high quality liquid assets (HQLA)―defined to include reserves, sovereigns, and a variety of other securities presumed to be liquid―must be larger than a weighted sum of liabilities with maturity of 30 days or less.
What do we learn from the history of RR? Comparable to the usual prudential goals today, Carlson describes how the aim of requirements in 19th century America was to ensure that banks had sufficient liquidity to meet outflows in times of stress without reducing lending. In practice, banks generally met these requirements—that ranged from 15 to 25 percent of deposits—but failed to use them as a buffer under stress. Instead, when faced with an increased risk of deposit loss, bankers would increase their reserves and contract credit supply. A second lesson from RR history is that, when users of funds are confident that their banks can provide liquidity, they hold fewer liquid assets themselves. Put another way, liquidity requirements result in a shift of liquid assets into the banking system.
Despite this experience, the prudential view of reserve requirements as buffers against possible deposit withdrawals persisted into the early 20th century. The approach changed in the 1920s (see Goodfriend and Hargraves). Rather than simply guarantee the elastic supply of the currency, providing reserves and lending to meet banks’ liquidity needs, the Federal Reserve began to view the reserve requirement as a monetary policy tool they could use to alter aggregate credit conditions.
Adopting this perspective, policymakers in the 1930s proceeded to make what some view as an enormous mistake. Following the sequence of system-wide bank panics in the first years of the decade, the surviving banks sought to prove their vigor by massively increasing excess reserves―those over and above the quantity required. By 1935, excess reserves had surged by a factor of nearly 90 to $3 billion, while M1 was little changed. (Data for total and excess reserves are available here.)
Concerned that banks would use these excess reserves to go on a lending spree, the Board of Governors of the Federal Reserve began raising the reserve requirement. As they stated in their 1937 annual report (page 2):
Under the law the Board has the responsibility of changing reserve requirements in order to prevent injurious credit expansion or contraction, and the Board had acted to … prevent an uncontrollable expansion of credit in the future.
Featured by Milton Friedman and Anna Schwartz in their classic A Monetary History of the United States, 1867 to 1960, this episode highlights the risks of using RR as a monetary policy tool. As the following chart shows, between August 1936 and May 1937, the Fed raised the reserve requirement from 13 to 26 percent in three steps. (Different banks faced different requirements, but they all doubled.) Industrial production (in black) peaks around the time of the third increase. The 33-percent plunge that followed is second largest on record (exceeded only by the 54 percent collapse earlier that decade). In addition, the M2 multiplier (in blue), a measure of the willingness and ability of the banking system to create transactions money (demand deposits) from a given level of central bank liabilities, shrinks from by 20 percent. (For details and debate about this episode, see here and here.)
Reserve requirements, industrial production and the M2 Money Multiplier, 1922-39
Source: The M2 Money Multiplier is the ratio of Friedman and Schwartz M2 to the St. Louis Adjusted Monetary Base from FRED, industrial production is from FRED, and the reserve requirement is from Feinman, Table A.1.
We see three lessons in this historical experience. First, when banks choose to hold high levels of excess reserves, as they were doing in the mid-1930s, raising the reserve requirement can trigger a credit contraction and a recession. This is the same message that Carlson took away from the 19th century record: bankers do not act as if required reserves provide a buffer against runs. Faced with heightened uncertainty, they increase their holdings of excess reserves as insurance.
Second, the impact of a reserve requirement change is difficult to predict. The initial increase in August 1936—from 13 to 19½ percent—did not seem to slow things down. It was only after the two succeeding increases totaling 7½ percentage points that conditions changed sharply.
Third, like a number of prudential tools, reserve requirements have an inherent asymmetry. Tightening them, as the Federal Reserve did in 1936 and 1937, will tend to lead to a contraction in lending as banks shift their balance sheets to meet the new higher requirement. By contrast, if banks are capital-impaired or merely skeptical about borrowers’ viability in poor conditions, a cut in reserve requirements that boosts excess reserves need not result in an increase in credit.
Returning to the 21st century, what are the lessons for modern liquidity regulation? To put the discussion into perspective, we start with data on U.S. banks’ liquid assets relative to their total liabilities. Given that reserve requirements were not binding during this period, and that the LCR did not take effect until 2017, this chart shows banks’ voluntary holdings of liquid assets. First, note the steady fall in cash holdings (in purple), from 13 percent of total liabilities in the early 1970s to 3 percent in August 2008. Unsurprisingly, in September 2008, everything changes. Over the next year, banks quickly re-built their liquidity buffers so that by the end of 2009, cash reached 12 percent of liabilities. Total liquid assets peak at more than 35 percent of total liabilities, and stand at one-third today, well above previous norms.
Cash, Treasury and Agency securities as a fraction of total liabilities of U.S. commercial banks (monthly), Jan 1973-Feb 2018
Source: H.8, Board of Governors of the Federal Reserve System.
We draw two conclusions from all of this. First, as far as we know, no one in advanced economies is proposing to use the LCR as a monetary policy tool. But, if they do, varying liquidity requirements in the absence of bank distress in order to influence credit conditions looks like a very bad idea. Considering banks’ large voluntary holdings of liquid instruments, we would expect the impact to be as blunt and unpredictable as reserve requirements have been in the past.
The situation is different in economies in less advanced financial systems. Federico, Vegh and Vuletin document that two thirds of developing countries use reserve requirements as an active policy stabilization tool. The People’s Bank of China (PBOC) is a prominent example. Liu and Spiegel report that, between 2006 and 2016, in an effort to manage credit growth, the PBOC adjusted the reserve requirement no less than 40 times. However, as financial markets develop and alternatives to bank financing expand, we expect that policymakers in these countries will shy away from this tool.
Second, we remain concerned about the usability of the liquidity buffer. The challenge is that regulations bind in booms, but markets bind in busts. While supervisors will clearly allow banks to run down their LCR buffer when circumstances warrant, history suggests that banks will be reluctant. While one bank could deploy its liquid assets to meet sudden outflows, others might respond by increasing their buffers to demonstrate their strength. (Diamond and Kashyap note that a liquidity requirement could be structured to reduce run risk even if it is not usable.)
In his discussion of liquidity requirements, both their rationale and their mechanics, Stein suggests a possible solution to the problem of usability. He appeals to the Reserve Bank of Australia’s committed liquidity facility (CLF). For countries that do not have sufficient securities to meet the demand created by the LCR, the Basel Committee created a second option: central banks could offer—for a pre-set fee--secured lines of credit that would count as high-quality liquid assets (see our previous post for details).
Ultimately, solvent banks will deploy their LCR willingly in an episode of system-wide distress only if they are assured that the central bank will supply reserves elastically. Provided that banks have adequate high-quality collateral, a CLF would help reinforce their confidence.