5. Money, Collateral and Safe Assets

Manmohan Singh
Published Date:
October 2016
  • ShareShare
Show Summary Details

Between 1980 and the 2008 financial crisis, the use of collateral in financial markets rose exponentially in the US and in other financial markets. After the crisis, there has been a reduced pool of assets considered acceptable as collateral, resulting in a liquidity shortage. When trying to address this, policymakers will need to consider collateral besides the traditional money metrics.


In the traditional view of a banking system, credit and money are largely counterparts to each other on different sides of the balance sheet. In the process of maturity transformation, banks are able to create liquid claims on themselves, namely money, which is the counterpart to less liquid loans or credit. Banks create money-like assets (not money). Owing to the law of large numbers, banks have – for centuries – been able to safely conduct this business with relatively little in the way of liquid reserves, as long as basic confidence in the soundness of the bank portfolio is maintained.

In recent decades, with the advent of securitisation and electronic means of trading and settlement, it became possible to greatly expand the scope of assets that could be transformed directly, through their use as collateral, into highly liquid or money-like assets. The expansion in the scope of the assets that could be securitised was in part facilitated by the growth of the shadow financial system, which was largely unregulated, and the ability to borrow from non-deposit sources. This meant deposits no longer equalled credit (Schularick and Taylor, 2012). The justification for light-touch or no regulation of this new market was that collateralisation was sufficient (and of high quality) and that market forces would ensure appropriate risk taking and dispersion among those educated investors best able to take those risks, which were often tailor-made to their demands. Where regulation fell short was in its failure to recognise the growing interconnectedness of the shadow and regulated sectors, and the growing tail risk that sizable leverage entailed (Gennaioli, Shleifer and Vishny, 2011).

Money and the (Adjusted) Money Multiplier

Payments finality can be defined in a contract or understood to be defined in law. In the US, for example, Federal Reserve (Fed) banknotes are legal tender for all debts, both public and private. In other words, if you owe someone US$100 million, your offer to pay them in Fed notes cannot be refused unless prespecified in a contract. The market practice (and/or law) is to accept deposits in any Fed bank as final payment (they can always be converted into Fed notes) for all debts. This does not mean other financial assets cannot be accepted as payment, just that central-bank money cannot be rejected. The further advantage of central-bank “money” is that it is risk-free in nominal terms (not in real terms, as inflation is unknown). Other financial assets, even US Treasury bills, have a degree of nominal price risk. They thus raise the issue of the price of the asset in terms of central-bank money. For example, if you pay a debt of US$100 million due today in central-bank money at 10 am or 2 pm, it is the same “quantity” of deposits at the Fed. If someone is settling in T-bills or bonds, or Exxon shares, however, the price will not be the same at 10 am and 2 pm, in general. This entails the added complexity of trying to determine the “market” price, as well as opening up the opportunity to distort the market price in somebody’s favour.

The value of central-bank money in terms of the nation’s unit of account never fluctuates. You can always pay a US$50 million debt with a central-bank deposit of US$50 million. In paying a US$50 million debt in T-bills or Exxon shares, the number of bills and shares will fluctuate. Now, society might benefit from moving from settlement finality in central-bank money to settlement in Exxon shares, since Exxon shares yield, say, a 4% real return on average, while central-bank money yields a negative real return on average. So, in that case, we can imagine (in the context of Exxon shares) that all prices would be quoted in terms of Exxon shares and Exxon could issue fractional shares and coins. The general price level would then change with changes in the perceived value of Exxon shares as well as with share splits and reverse share splits, etc. Shareholders would receive their dividends in more shares (not fiat money). This is clearly not optimal. People prefer nominal claims for a reason!

Now let us imagine an economy that has not done away with the legal tender of central-bank money; however, there are many other assets in the market, such as Exxon shares, T-bills, bonds and securitised revenue streams that are very widely accepted and held. So dealers finance their inventories through borrowing and swapping securities that are “high”-yielding (or at least of a positive yield), thereby minimising the use of central-bank money, which is low-yielding. Financial market investors do not like to hold much monetary base (ie, central-bank money). They prefer to hold claims on money market funds (which have variable prices, unlike bank deposits), and various other mutual funds or securitised assets.

The money multiplier (m) says something about the efficiency of the infrastructure of the financial intermediary sector. The conventional money multiplier m is defined as the ratio of total monetary liabilities and the monetary base. The monetary base comprises central-bank monetary liabilities, that is, currency and central-bank depository liabilities. These central-bank liabilities (legal tender) are the most liquid assets in the economy.

We can think of more than one money multiplier, though, where each money multiplier refers to a specific aspect of the efficiency of financial-services provision.1 Specifically, currency in modern economies is held largely by households and nonfinancial enterprises. In order to better assess how efficient financial intermediaries are at financing lending upon a small base of liquid assets that they hold, it is illustrative to subtract banknotes from the monetary base. For US banknotes there is also demand by non-residents in high-inflation or unstable countries. The residual (monetary base minus banknotes) comprises bank deposits at the central bank, or liquid reserves. In this chapter, we call these deposits D.

The adjusted money multiplier (where only D comprises the “adjusted monetary base”) is then the ratio of the financial system’s monetary liabilities to the non-financial private sector, divided by their legal-tender liquid reserves. Intuitively, all money multipliers (M1, M2, etc) are relative to a monetary base. We are defining D as the relevant monetary base, analogous to the collateral discussion in this chapter. Figure 5.1 shows the obvious divergence in the conventional (M0) and adjusted monetary base (D) in the US during the past 50 years. The demand for currency rose steadily during that period while the demand for bank reserves remained remarkably stable in nominal terms until the current crisis. Since liabilities increased, the adjusted money multiplier rose much more sharply than the multiplier as conventionally measured.2

Figure 5.1Monetary Base and Deposits at Central Bank (1959–2011)

Source: Federal Reserve Bank of St Louis FRED database and authors’ calculations

The adjusted m becomes a measure of the efficiency of banking services provision. If the adjusted m is very high, a high degree of lending is supported by a relatively small reserve base. If m is high, then someone making a cash deposit in Peoria, Illinois, sets off a cascade of interbank lending. If adjusted m is low, then that deposit sits in the vaults of the bank for a few weeks, and then is shipped to the Federal Reserve Bank of Chicago, leading to a credit in the excess reserves of the bank.

An adjusted mi equal to 1 is equivalent to Milton Friedman’s “narrow banking” ideal: bank transaction deposits fully backed by reserves at the central bank. Thus viewed from the mirror image, adjusted m provides a shorthand metric of the degree of liquidity risk present in the financial system. While the liquidity risk of the banking system is of interest, we expect to find the most interesting dynamics in the shadow financial system. Since the early 1980s through 2008, this portion of the US financial system has accounted for almost the entire growth in US financial deepening (Figure 5.2).3

Figure 5.2US Total Credit Market Assets (Ratio to GDP)

Source: Updated by authors from US Flow of Funds

Furthermore, since only banks have access directly to central-bank deposits, the adjusted mi of the total financial system – both banks and the shadows – has increased sharply over the same period. This happened largely through the use of securitisation and collateralised borrowing. In other words, the financial system expansion has relied on the increased use of collateral as complementary “liquid” assets beyond bank reserves (D).


We next consider the efficiency of the financial system in using collateral. For this, we introduce a multiplier, c, analogous to m, defined as the ratio of all financial market liabilities (satisfying certain characteristics) to the sum of deposits at the central bank (D) plus liquid collateral held by the financial system, which we denote as “C”. Where C differs crucially from D is that, unlike D, which, at least in a floating-exchange-rate regime, is determined entirely by the central bank, C is partially market-determined.

To illustrate this point let us divide C into C1 and C2. C1 comprises a class of assets that in all states of the world are accepted as collateral, as they can either be directly converted into D or are direct obligations of a (fiscally sound) sovereign. C2 is composed of other assets deemed acceptable as collateral under normal market conditions, but lose value when markets are distressed. C1 is primarily determined by the sovereign and central bank. It is partially determined by market forces, since households and non-financial enterprises (such as pensions, insurance companies) hold C1 assets, which limits the supply available to the financial system. Here, for illustrative purposes, since we use the US flow-of-funds data, we define C1 to comprise only direct obligations of the central government (eg, US T-bills or US Treasuries). The volume of C2 is fully market-determined depending on market sentiment, counterparty fears, the length of collateral chains and market-imposed haircuts. C1 and C2 may be considered within Gorton and Ordoñez’s (2012) framework as information-insensitive and informationsensitive assets respectively. Note that debt of government-sponsored enterprises and agencies (eg, Fannie Mae and Freddie Mac) is now considered “safe” owing to an explicit guarantee ex post. However, ex ante, there were reasons to believe it was not on a par with Treasuries. In fact, a credit event was declared on Fannie and Freddie that triggered their credit default swaps.

The proximate cause of the 2008 liquidity crisis was the differentiation of C2 collateral from C1 collateral. The major central banks and treasuries responded to the crisis by both increasing the monetary base and swapping superior for inferior collateral. This led to an exponential rise and subsequent crash in the ratio of total US financial-sector liabilities to what we refer to as “ultimate liquidity” (D + C1 held by banks).

Table 5.1Definition of Terms Used
DBank deposits at the central bank (D excludes banknotes, vault cash)
C1Good collateral in all states of nature; can be converted to D at no haircut
C2Collateral that under normal market conditions is “good”, else loses value
C1 held by banksOnly banks can convert C1 to D overnight. Non-banks cannot change C1 to D
Ultimate liquidityD plus C1 held by banks (Chapter 4 shows C1 may contribute more to financial lubrication than D)

The nonfinancial sector has gone from holding bank liabilities to holding a diversified portfolio of securitised assets directly. While not backed by D, they were backed by C. As long as there is confidence in the assets comprising C, or as long as C1 remains a significant share of C, it may be assumed that these claims are “liquid”, ie, they can be converted into central-bank money at fairly short notice. In recent years, the financial system converted a huge stock of claims on future revenues (loans, cell-phone-fee receivables, etc) from illiquid claims into notionally highly liquid claims. In the process, this created a demand to securitise other claims, such as legal-damage claims, awards and lottery payouts.

So what happens in that economy when suddenly there are doubts about the underlying value of Exxon shares and other securitised revenue streams? Naturally, they lose their attractiveness as investments and as liquid assets that are used as money. Suddenly, there is deemed to be a liquidity shortage, and this intensifies when it is clear just to what extent the value of pseudo-liquid assets in the economy has expanded in relation to central-bank money. Other collateral or money may continue to be acceptable, such as US Treasuries and bunds. So there is a sudden split between cash and certain types of collateral, and everything else. Everything else ceases to be liquid.

Some Analytics of Collateral – Pre- and Post-Crisis

Before the 2008 crisis, C2 comprised an abundance of securities, and the nature and number of such securities was growing. Ultimate liquidity leverage (total liabilities/{D+C1 held by commercial banks}) was growing exponentially, as D was extremely small. There was a price differential among D, C1 and C2, but it was small, a few basis points. So the market was operating with a reasonable liquidity cushion, ie, although ultimate liquidity leverage was quite high, total liquidity leverage (or total liabilities/{D+C1+C2}) may have been deemed adequate or more than adequate.

During the crisis, the quantity and nature of securities comprising C2 changed dramatically. At the same time, there was a surge in demand for liquidity. Basically, institutions that were borrowing using C2 as collateral had to find C1 or D to avoid default as they could no longer use the securities (now distressed) that used to be in C2. This differs from some of the academic work that lumps all collateral together. Some collateral will always remain close to par or above par and not lose value (ie, C1). There was a race to acquire the highest-quality collateral, C1, which was then hoarded. The information-sensitive portion of C (ie, C2) became subject to a “lemons” problem. C2 consequently was no longer accepted at a full information market clearing price or, sometimes, at any price. Central banks became subject to a form of Gresham’s Law if they were slow or reluctant to adjust their collateral policies accordingly. The events in the eurozone during the crisis were in line with this theoretical description. Policymakers have the task of increasing the volume of C1 collateral in the market domain. One way is to accept C2 collateral (which will have a zero velocity by definition, as it will be parked at the central bank) in exchange for D or C1 at a subsidised price. Printing money via conventional quantitative easing (QE), ie, exchanging D (roughly equal to “excess reserves”, since required reserves are a trivial fraction of D) for C1, does not increase D + C1. In fact, D can be inferior to C1 if rehypothecation, or collateral velocity, is constrained, as explained in Chapter 4.

This may be a detour, in recent years there was a similar discussion regarding liquidity provision in the context of the demand for short-duration bunds, but this stemmed from the eurozone crisis and the need to skirt political/legal constraints to obtain seniority. However, short bund yields have been negative, suggesting that there has been no bending over backwards to cater to the demand for “safe” assets. It may be useful to note that Japanese short-end issuance is intertwined with other monetary-policy variables, including key foreign-exchange levels for the yen (and thus draws parallels to the pre-1981 policy in the United States). This raises, in addition, the interesting question as to whether the provision of liquidity relief through collateral substitution (C1 for C2) should be better undertaken by treasuries (fiscal agent) – owing to the fiscal risk entailed – or by central banks.4

Although we do not have a long historical series for C2, we do have series for D and C1 via the flow-of-funds data.5 As defined earlier, we call the sum of D and the portion of C1 held by banks “ultimate liquidity”. Only C1 held by banks were included in ultimate liquidity, as only banks could transform C1 into D overnight. We can see that, before the crisis, the ratio of total financial intermediaries’ liabilities – the broadest measure of financial-sector liabilities available – to ultimate liquidity was rising exponentially. This measure is currently back to a level last seen in the 1970s, before the rapid expansion of the shadows and the securitisation boom that started with mortgage-backed securities (MBSs) in the early 1980s. Total financial-system ultimate liquidity leverage rose from 4 at end-1951 to 673 at end-2006 before falling sharply to 33 at end-2011 (Figure 5.3). By comparison, ultimate liquidity leverage was 36 at end-1981. Ultimate liquidity deleveraging was effected almost entirely through an increase in D. Between end-2007 and end-2011, D rose by US$1.5 trillion and C1 held by banks rose by only US$0.1 trillion. We exclude Federal Reserve Bank liabilities and holdings of Treasuries in these calculations, as the Fed has increasingly provided support to this market in recent years. Total liabilities rose by US$1.1 trillion over the same period.

Figure 5.3Ratio of Total US Financial Intermediaries’ Liabilities to Ultimate Liquidity

Source: US Flow of Funds accounts and authors’ calculations

We next consider banks’ and non-banks’ liquidity leverages. The liquidity leverage of both banks and non-banks experienced a similar increase and dramatic decrease before and after the crisis, as seen in Figures 5.4 and 5.5. Whereas liquidity deleveraging was accomplished in the banking system through an increase in D, for non-banks this was not possible (note, the denominators of Figures 5.4 and 5.5 are different). Consequently, their liquidity deleveraging was effected through an increase in holdings of C1 and a decline in total liabilities. In the context of banks and non-banks, the substitution of D and C1 is important. Between end-2007 and end-2011, C1 held by non-banks rose by US$1.4 trillion, while total liabilities fell by US$2.7 trillion. Figure 5.3 is the sum of numerators of Figure 5.3 + Figure 5.4 divided by the denominator of Figure 5.3 only. The intuition is that non-bank financials hold the vast proportion of C1; post-crisis, the banks picked up D and the non-banks C1.

Figure 5.4Ratio of Total US Commercial Bank Liabilities to Ultimate Liquidity

Source: US Flow of Funds accounts and authors’ calculations

Figure 5.5Ratio of Total US Non-Bank Financial Intermediaries’ Liabilities to their Holdings of C1

Source: US Flow of Funds accounts and authors’ calculations

Clearly, from the balance-sheet identity, non-bank holdings of C2 assets had to fall. Among the more prominent declines were in mortgages, which fell by US$1 trillion (due to the Fed’s buying programme), and holdings of commercial paper and bankers’ acceptances, by US$0.6 trillion.

Thus, before the crisis, there was an inverse pyramid on a very small “money-and-safe-collateral base”. What has happened postcrisis is that there has been a disintermediation process. People are withdrawing from leveraged institutions; there has been a reassessment of what is acceptable collateral, an increase in the haircuts applied to that collateral, fear of insolvency, etc (eg, Greek collateral was not acceptable for repo at LCH’s Repoclear arm). Some C2 collateral was indistinguishable from C1 collateral in the good state. However, post-crisis, some of the C2-type collateral does not have a “market clearing” price. In 2011, Greek bonds were not acceptable at LCH Clearnet (UK) at any level of haircut. Although Greek restructured bonds were acceptable at the ECB, such C2 collateral did not have a market clearing price at that time and was not accepted at LCH Clearnet. Therefore, the interbank/inter-institutional market was grinding to a halt. This is exactly what Friedman’s narrow banking proposal had been designed to prevent: payments gridlock. During the crisis, not only does the interbank market dry up and the exposures to each other become restricted, but the volume of credit to the non-bank sector also falls.

To the extent that the central bank merely substitutes central-bank money for assets that have retained their value as collateral, not much liquidity relief is attained. In order to provide effective liquidity relief for the system, central-bank money and liquid collateral must be injected against illiquid or undesirable assets (C2); the supply of unencumbered collateral has to increase. This requires the government or the central bank to take some risk on their balance sheets. As part of their crisis response the treasuries of the UK, US and Norway, among others, swapped government debt for a variety of less liquid collateral, including MBSs. Not all operations provided collateral liquidity, however.

Monetary Policy and Financial Lubrication

Regulators may take solace from the fact that lower leverage and shorter collateral chains lessen financial-stability risks. However, from a monetary-policy perspective, the global financial market’s transmission mechanism is currently in the midst of an episode of grinding gears. The reduced availability of collateral and shorter “collateral chains” – resulting from constrained collateral trading – lower global financial lubrication and increase the overall cost of capital to the real economy.

In the US and Europe, both the Fed and ECB consider many information variables when determining monetary policy. The monetary base or M2 is an integral part of the “orthodox” monetary informational variable toolkit, whereby the velocity of money is considered stable.6 After the Lehman collapse and the aforementioned grinding of the gears of the transmission mechanism, and as central banks have drifted towards the “zero”-interest-rate lower bound, traditional operational variables, such as the overnight interest rate and guideposts such as the Taylor Rule, have become largely moot. Instead, there has been an increased resort to quantitative-based or QE policies (Hanoun 2012).

In considering an appropriate degree of balance-sheet expansion, it is important to recognise that what has been done so far by the key central banks may not have sufficiently substituted for the loss in financial collateral, particularly to the extent that traditional QE may have merely substituted D for C1. The state of the pledged-collateral market needs to be considered when setting monetary policy, and more consideration may need to be given to qualitative easing, such as the substitution of D or C1 for less liquid assets.7 An example is the move in 2012 by the UK’s (now defunct) FSA that allows banks to undertake liquidity swaps with insurers/pension funds whereby the banks receive gilts or highly liquid collateral (C1) against C2 collateral (eg, mortgage-backed or infrastructure bonds). There are links between pledged collateral that is intermediated by large banks and “quantitative” monetary-policy instruments. For example, in many markets, cash and unencumbered collateral substitute for each other – certain CCPs or central clearing entities are indifferent between cash and (acceptable) collateral. Through QE, central banks have tried to add towards global lubrication.

Pledged Collateral and Money Aggregates

Overall, global liquidity remains below pre-Lehman levels. This does not imply anything about the optimality of the pre-Lehman level of global liquidity. Although the Fed has discontinued publishing the M3 metric since 2006, the UK and ECB also publish the M3 measure. When M3 was published (prior to March 2006), only some repos transactions between the primary dealers and the Fed were included in this metric. Thus the extent of repo included in M3 was not complete.

Annual reports of large banks suggest that financial collateral (including collateral reuse) is sizable and comparable to monetary aggregate such as M2 or broad money. When we consider collateral use/reuse in addition to M2 or the monetary base in the US, UK and eurozone, financial lubrication was over US$30 trillion before Lehman (and one-third came via pledged collateral). Japanese Government Bonds (JGBs) are generally not used in the cross-border pledged-collateral market, hence Japan is not included here. However, if good collateral becomes scarce and repo rates across all global jurisdictions remain near zero or negative, it is likely that there will be more fungibility within good collateral. Thus, looking forward, JGBs may find their way in the eurozone market and bunds in the US market (unlikely, given the scale of JGB purchases by Bank of Japan recently).

The subsequent decline in both available collateral and associated reuse of collateral was sizable (an estimated US$4–5 trillion). This is the difference between the green and red lines in Figure 5.6. The “kinks” in the red line in Figure 5.6 show M2 expansion, due to base money via QE. As of end-2015, the overall financial lubrication has rebounded to approximately US$30 trillion, but the “mix” is more in favour of money (around 80% of financial lubrication), which not only has lower velocity than pledged collateral but much of it “sits” as excess reserves with central banks (Singh 2011). Thus a rebound in the pledged collateral market may be more effective in easing liquidity constraints in financial markets than further QE. More importantly, the cost–benefit analysis of QE may become more apparent when trade-offs will need to be made when central-bank balance sheets unwind (or do not unwind but continue to “carry” collateral until it matures). As discussed in Chapter 4, the Fed’s reverse repo, a by-product of QE, will have consequences (ie, rusting the non-bank–bank plumbing, or providing “puts” to non-banks etc).

Figure 5.6Financial Lubrication (Money and Collateral), 2007–2015

Source: Authors’ calculations and central bank websites

Safe Assets and Treasury Bills – What Determines their Supply

In the past few years, the short-term US Treasury yield curve (between 1 and 12 months) has been well below 25bp (mostly below 12bp). Meanwhile, the Fed is currently paying banks 25bp on overnight deposits. Clearly, no US bank is going to bid in the T-bill auction for its own account. This takes out quite a lot of demand. So, for the market to be clearing at such low rates, there must be sizable demand coming from somewhere – non-banks (eg, mutual funds). Non-banks’ investors, flush with liquidity, prefer fewer bonds and more T-bills; cash-rich non-banks continue to suggest there remains a shortage of bills and lobby for more. This is exactly what the Fed “twist” did (T-bill supply increased while bonds in the market decreased). However, the total volume of debt issuance is determined by budgetary needs and financing options (long versus short tenor).

Related to this discussion is the ongoing research that highlights demand for “safe assets” (which may have several definitions, including that of Gorton et al, 2012). Since 1982, the US Treasury’s “regular and predictable debt issuance strategy” had a primary goal: issue at least cost. Figure 5.7 illustrates that the Treasury is roughly meeting its objective. The correlation between bills/total issuance and 10-year-minus-6-month spread is over 0.6. (Annex 5.1 highlights the debt-management policy and structural changes since 1961.) In the current environment, a change in the composition of US debt towards the short end would shift interest-rate risk from the private sector to the public balance sheet. This increases the Treasury’s fiscal risk unless we presume that non-banks would be bailed out anyway, as happened in 2008–9. Thus, catering to the non-bank demand ex ante is a faint attempt to skirt the more obvious needs that would become apparent during a crisis.

Figure 5.7Ratio of T-bills/Total Issuance by US Treasury Since 1982

Source: Fed and Treasury; we remove Fed holding of US Treasury issues

It is also important to note that D is currently yielding zero but providing a nominal guarantee (put at par) for free. As discussed in the context of money, D and C1 are both liquid by definition. The desire to pick up yield on assets that can be put to a central bank at par is what makes C1 more attractive than D. In the past, this put was largely ignored but at zero interest rates the insurance premium imposed by the Federal Deposit Insurance Corporation (FDIC) on bank deposits is material. The FDIC (under the Dodd–Frank Wall Street Reform and Consumer Protection Act) allowed temporary unlimited deposit insurance till end-2012 on non-interest-bearing transaction accounts. Another relevant example was the Transaction Account Guarantee extension, whereby the FDIC stipulated that those opting for (extended) deposit insurance would have to pay more.

Safety of principal (ie, no interest is paid) was being offered by one of the two main US custodians, the Bank of New York (charging 13bp). If demand for 1-month T-bills is indeed relatively inelastic (see Duffee 1996), the market can clear at much lower T-bill spreads (ie, up to negative 13bp). Thus, we might raise the question why the Fed was, up until the December 2015 lift-off, paying 25bp rather than charging 13bp for accepting deposits.

The US Treasury’s Borrowing Advisory Committee (TBAC) has said that “broadly agreed that flooring interest rates at zero, or capping issuance proceeds at par, was prohibiting proper market function. The Committee unanimously recommended that the Treasury Department allow for negative yield auction results as soon as logistically practical.” Some other elements of the TBAC report are also interesting.

Panel 5.1:Floating-Rate Note “Puts” – are They Forthcoming?

At the time of the discussions leading up to the Fed–Treasury Accord of 1951, which ended an extended period of artificially suppressed interest rates on Treasury bonds, there was much internal debate about the potential deleterious impact on bondholders from a “surprise” rise in rates. There was also concern about a potential buyers’ strike and/or fear that a new market equilibrium would entail a sharp spike in rates. This discussion was conditioned by the similar situation faced by the US Treasury in 1919, after it promised to stabilise bond prices during and after World War One. This policy caused conflict with certain Fed policymakers and the eventual losses on Liberty bonds were still remembered by Congress and the Treasury in 1951, 30 years later. As a consequence, at the time of the announcement of the Accord, buyback options were offered by the Treasury, that is, the US Treasury offered to swap the outstanding stock of long-term debt with new long-term debt with higher coupons (coupled with restrictions on sales before maturity). The idea was to cushion the market from capital losses.

Might the US Treasury go down a similar path again in conjunction with an eventual Fed exit strategy? In the current environment, markets have witnessed a 30-year secular decline in bond market yields. Serious market turbulence might result – significantly greater than that associated with the February 1994 “surprise” rise in rates initiating a tightening cycle – were the market to believe it were embarking on a steady (or rocky) rise in rates from near zero to a “neutral” Fed funds rate of 400bp and a “normal” 5% yield on two-year US Treasuries. These early TBAC proposals in 2012 for floating-rate notes (FRNs) seems an obvious option to cushion the transition for the market. As an indication that the eventual unwinding and normalisation of the yield curve will take time and inflict pain on holders of fixed-income debt, the market appears already to be requesting such “puts”. In this context, it is useful to quote from a recent TBAC report (January 31, 2012):

… ways to explore the viability of Treasury issuing floating rate notes (FRNs). In particular, the presentation assessed potential client demand, optimal maturity, reference index, and reset frequency. The structural decline in the stock of global high-quality government bonds, coupled with an increase in demand for nonvolatile liquid assets, should make US government-issued FRNs extremely attractive. Pricing for a hypothetical two-year FRN was estimated to be in the arena of 3 month Treasury bills plus 8 basis points.

On January 29, 2014 after much deliberations, US Treasury issued FRNs linked to the 3 month Treasury bill rates. Some readers will also note that the Fed’s RRP programme since September 2013, which was uncapped at about US$2 trillion on December 16, 2015 (at the time of the Fed’s lift-off from zero), also supplies short-term safe assets but skirts the US Treasury debt issuance schedule.

In this context, it is useful to recall that the US Treasury discontinued 30-year bonds in the early 2000s. The Treasury did not factor in the demand for duration coming from pension funds and insurers. Due to the primary surplus in the Clinton years, the US Treasury continued to largely embrace its debt issuance strategy. In fact, with no new 30-year bonds, the 30-year swap curve turned negative as pensions and insurers were short of the 30-year bonds. So, although at present some non-bank sectors continue to demand more T-bill issuance, it is not clear why the US Treasury has to accommodate such a lobby (Panel 5.1). In this context, it is worth noting that, under the monetary policy rubric, the Fed’s RRP programme is similar to issuance of overnight T-bills! It should be noted that the role of government policy – in reshuffling debt issuance – is diminished when debt is high/capped (Greenwood, Hanson and Stein 2010. Their suggestion of “replacing the entire stock of T-bills with maturity greater than 100 days, with T-bills of an average duration of 58 days” will entail rollover risk. In the present environment, the long end offers free money in real terms. Academia continues to ask for more government supply of safe assets as a public good, but its models do not factor in the “reuse” or velocity of good collateral! (Gourinchas and Rey 2016).


“Monetary” policy is currently being undertaken in uncharted territory and may change some fundamental assumptions that link monetary and macrofinancial policies. Central banks are considering whether and how to augment the apparently “failed” transmission mechanism and in so doing will need to consider the role that collateral plays as financial lubrication. Swaps of “good” for “bad” collateral may become part of the standard toolkit. Since the Lehman crisis, central banks have interposed themselves as risk-taking intermediaries in the market and/or have circumvented the transmission mechanism by acting directly on specific long-term interest rates through quantitative interventions.

If so, the fiscal aspects and risks associated with such policies – which are virtually nil in conventional QE swaps of central-bank money for treasuries – are important and cannot be ignored. Furthermore, the issue of institutional accountability and authority to engage in such operations touches at the heart of central-bank independence in a democratic society.8

These fundamental questions concerning new policy tools and institutional design have arisen at the same time as developed countries have issued massive amounts of new debt. Although the traditional bogeyman of pure seigniorage financing – that is to say massive monetary purchases of government debt may have disappeared from the dark corners of central banks – this does not imply that inflation has been for ever arrested. Thus, a central bank may “stand firm” yet witness rises in the price level that occur to “align the market value of government debt to the value of its expected real backing.”9 Hence current concerns as to the potential limitations fiscal policy places on monetary policy are well founded and indeed are novel only to those unfamiliar with similar concerns raised for decades in emerging and developing countries as well as in the “mature” markets before World War Two.10

Many recent financial regulations have largely focused on building significant equity cushions and reducing leverage at large banks, and have not fully incorporated the non-bank–bank funding nexus. Much of the empirical work undertaken by researchers and policymakers using variants of equity and leverage from the banking sector has neglected to include this hidden fragility of non-bank funding in its analytical frameworks; thus, their work (and associated empirical results) is susceptible to model misspecification. Until recently, non-bank funding to banks was assumed to be sticky and mainly in the form of household deposits. However, the size and elasticity of the “non-bank–bank funding” nexus is also an important component that should be considered (Borio and Disyatat 2011). Analytically, bank credit to ultimate borrowers is funded either by the equity of the banking system or by the savings that non-banks (ie, households, pension funds and insurers) provide to the banking system. The traditional view of a banking system is that total debt funding from non-banks is relatively “sticky”. However, this stickiness was a flawed assumption and ignored much of non-M2-related funding (see Figure 3.1 in chapter 3).

Annex 5.1: US Treasury Debt-Management Strategy Since the 1960s11

Before 1982, the US Treasury had issued debt on a “tactical” basis that did not follow a predictable pattern and often caught investors off guard and also adversely impacted on markets. During the early 1960s, the Treasury issued every quarter for cash and to retire maturing debt. Maturities were selected after surveying market participants’ demand for various tenors. During the 1970s, due to the sizable deficit in 1975, significant tactical offerings were disrupting the market. The Treasury decided to change the framework for debt issuance to a more “regular and predictable” timetable so that the investors could plan in advance.

Also, before 1982, the Treasury had sometimes announced other policy objectives in addition to least-cost financing. For example, in the 1960s, Treasury issuance would be influenced by the desire to increase (or maintain upward pressure on) short-term interest rates to prop up the value of the US dollar, or contain long-term interest rates to spur economic growth. As shown in Figure 5.1.A, prior to 1982 there were too many policy objectives that did not result in least-cost financing (ie, there is no discernible relationship between issuance of US T-bills relative to total debt issuance and the cost of long-term/short-term funding). The structural break in the series from 1960 to 2011 in the early 1980s (Chow test) provides support to the inception of “regular and predictable”.

Figure 5.1ABills/total Issuance Relative to 10-year Yields Minus 6-Month Yields

Source: US Treasury; data removes Fed holdings of the US Treasury

However, since 1982, the “regular and predictable” issuance became a precondition to the least-cost financing objective. The correlation between the ratio of US T-bills/total debt issuance and the relevant funding costs is over 0.6. This strategy, however, gives the Treasury some flexibility to alter the timing of the auction schedules. For example, in 1985, STRIPS (separate trading of registered interest and principal securities) were introduced. These enhanced demand and contributed to lower financing costs. In 1991, there was a shift from bills to longer-term notes; also in 1993, seven-year notes gave way to thirty-year bonds; in 1998, three-year notes were replaced by five-year notes. Also during the early 2000s, the 30-year bonds were discontinued in light of large and persistent federal budget surpluses and a significant reduction in financing requirements. Thus, given the “regular and predictable” issuance schedule since 1982 and the bias towards least-cost financing, taking rollover risk at taxpayer expense may not be feasible today (especially when the cost of long-term debt is almost zero in real terms).


    BorioClaudio and PitiDisyatat2011Global Imbalances and the Financial CrisisBIS Working Papers No. 346 available at

    DudleyWilliam C.2011Securing the Recovery and Building for the Futureremarks at the United States Military AcademyWest PointNew YorkNovember17.

    DuffeeGregory R.1996Idiosyncratic Variation of Treasury Bill YieldsJournal of Finance51(2) pp. 52751June.

    GarbadeKenneth D.2007The Emergence of Regular and Predictable as a Treasury Debt Management StrategyFederal Reserve Bank of New YorkEconomic Policy ReviewMarch.

    GennaioliNicolaAndreiShleifer and RobertVishny2011A Model of Shadow BankingNBER Working Paper No. 12115 available at

    GortonGary B.Stefan M.Lewellen and AndrewMetrick2012The Safe-Asset Share” available

    GortonGary B. and Guillermo L.OrdoñezGuillermo L.2012Collateral Crises” (January) NBER Working Paper No. w17771January.

    GourinchasPierre-Olivier and HelenRey2016Real Interest Rates, Imbalances and the Curse of Regional Safe Asset Providers at the Zero Lower BoundECB Forum on Central BankingSintraJune.

    GreenwoodRobinSamuelHanson and JeremyStein2010A Comparative-Advantage Approach to Government Debt Maturityworking paperHarvard University.

    HanounHerve2012Monetary Policy in Crisis – Testing the Limits of Monetary PolicyBank for International Settlement speechJanuary27 available at

    LackerJeffery M2009Government lending and Monetary Policy”, speech at the National Association for Business EconomicsWashington Economic Policy conferenceAlexandria, VirginiaMarch2.

    LeeperEric M. and Todd B.Walker2011Perceptions and Misperceptions of Fiscal InflationBIS Working Paper No. 364.

    PlosserCharles I.2012Fiscal Policy and Monetary Policy: Restoring the Boundariesspeech to the US Monetary Policy ForumUniversity of Chicago Booth School of BusinessNew YorkFebruary2 available at.

    RicksMorgan2011Regulating Money Creation After the CrisisHarvard Business Law Review1 p. 75.

    SchularickMoritz and Alan M.Taylor2012Credit Booms Gone Bust: Monetary Policy, Leverage Cycles and Financial Crisis, 1870–2008American Economic Review Vol. 102 No. 2.

    ShirakawaMasaaki2009Unconventional Monetary Policy – Central Banks: Facing the Challenges and Learning the Lessons” remarks at a conference co-hosted by the People’s Bank of China and the BISShanghaiAugust8.

    SimsChristopher2003Limits to Inflation Targeting” available at

    SinghManmohan2011Velocity of Pledged Collateral – Policy and Implications”. IMF Working Paper 11/256.

    SteinJeremy2012Interview, Study CenterGerzenseeSwiss National Bank available at…/Newsletter_January_2012.pdf

Let us therefore define money multiplier as the ratio of L(i), total monetary liabilities of financial institutions satisfying characteristics “i”, and the monetary base. Or mi =L(i)/monetary base.

Thus, L (total monetary liabilities of financial institutions)/M0 < L/D.

For noncommercial bank assets, we use flow of funds, and subtract Monetary Authority (L109) and US commercial Banks (L110) from Financial Business (L108).

During 2008–9 both the US Treasury and the Fed had MBS purchase programmes; in other words, such actions muddle the fiscal costs to central-bank balance sheets.

Flow of Funds Tables L108 and L109, “Total Liabilities of Financial Business minus Total Liabilities of Monetary Authority” (the Financial Business table in FOF is the sum of all types of financials, including the Fed, so we needed to subtract Fed) – see Figure 5.3. For Figure 5.4, we use FOF’s Table L110, “Total Liabilities of US Chartered Commercial Banks”. For Figure 5.5, we use Tables L108, L109 and L110, “Total Liabilities of Financial Business minus Total Liabilities of Monetary Authority minus Total Liabilities of US Chartered Commercial Banks” (Financial Business table in FOF is the sum of all types of financials, including the Fed; we subtract the Fed and commercial banks to arrive at Non-bank financials).

Ricks (2011) makes a legal distinction between fiat money and other money-like instruments.

Some Fed policymakers, for recent views along lines favouring new purchases of MBSs rather than Treasuries. Similar actions could in principle be undertaken by the Treasury, where the attendant fiscal risk might be more appropriately managed and budgeted. Other FOMC members have argued against resuming MBS purchases and for returning quickly to a Treasuries-only policy.

See Lacker (2009), Plosser (2012) and Shirakawa (2009) for considerations of this question.

In models of the fiscal theory of the price level, such as Leeper and Walker (2011), inflations may have fiscal origins without any necessary debt monetisation.

Problems that may arise when the fiscal authorities do not support central banks are discussed in Sims (2003).

This annex primarily builds on the findings of the NY Fed’s Economic Policy Review (Garbade 2007).

    Other Resources Citing This Publication