Central Banking (Lecture Notes, TU Berlin, Summer 2019)

An introduction to

central banking (Lecture notes, TU Berlin, summer 2019)

U. Bindseil*

1. Economic accounts and financial system 2. Central bank 3. Conventional monetary policy 4. Unconventional monetary policy 5. Financial stability 6. Lender of last resort 7. International monetary system List of References

*European Central Bank, Directorate General Operations and Technical University Berlin. Views are not necessarily the ones of the ECB

Chapter 1: Economic accounts and financial system 1.1 Introduction 1.2 Real economic sectors and economic transactions 1.3 Banks and central banks 1.4 An example: the German stylised financial accounts of 2013

1.1 Introduction

This chapter introduces a system of financial accounts to allow for a representation of real and financial flows. Transactions are represented within a closed economy consisting of the following sectors: (i) households; (ii) banks; (iii) non-financial corporations and the government; (iv) the central bank. These sectors have financial claims and liabilities towards each other, and some of the sectors, namely households, non-financial corporates, and the government, are the main holders of the real assets of the economy. The basic types of assets and liabilities are: real goods, gold and other precious metals (which can also be classified under real goods), banknotes, deposits, bonds, loans, and equity. Not all of these assets and liabilities will be present in every of our illustrations as we will always choose the simplest representation to make our point. Why starting with a chapter recalling such basic issues? Money is a financial asset and liability. The central bank is to a large extent defined by its balance sheet and central bank money is the central bank’s basic liability. Both monetary policy implementation and lender of last resort issues relate to liquidity flows or “flows of funds” issues within balance sheets. Recalling the logic of financial flows at the most basic level is therefore the best basis for understanding the nature of central banking.

1.2 Real economic sectors and basic types of transactions

The first economic sector “on earth” is the household, who initially owns potentially n real goods and who has no debt. According to United Nations (2008), “a household is a group of persons who share the same living accommodation, who pool some, or all, of their income and wealth and who consume certain types of goods and services collectively, mainly housing and food”. We can characterise this household in terms of its holdings in the n real assets of the economy, of which gold is the n-th asset, i.e. its asset quantity vector X’ = {x1, x2, …xn}. For that we need to have specified what the n goods are, and in what units we express them. For example, the household owns: X = {x1 chicken, x2 cows, … xn gold coins}. If we are interested in value aggregation and wealth, we in addition need to characterise each asset position by its value, expressed in a unit of account, which we call numeraire. The numeraire is one of the n goods, and the price of each good expressed in the numeraire is the number of units one needs to offer in the market of the numeraire good to obtain one unit of it.

In principle every pair of goods has one relative price, say p(Chicken, Cow) is the number of Chicken needed to buy one cow. Obviously p(cow, chicken) = 1/p(chicken, cow). Also p(chicken, cow) = p(gold coin, cow) / p(gold coin, chicken). With n goods, we have n-1 independent relative prices. Assume now that we chose the n-th good, a well specified gold coin (like the Florentine ducat with a fine gold content of 3.44 grams), as numeraire. Let’s simplify notation writing the relative price P(Gold coin, chicken) simply as p1 and P(Gold coin, cow) as p2, etc. Call P’={p1, p2, … 1} the vector of prices expressed in the n- th good. The total value of the households’ assets is then P’X and we can write the balance sheet of the household as follows. That the total value of household assets is equal to equity means that the household has no external financing, i.e. it has no liabilities but its own wealth.

Household accounts Real Assets P’X Household Equity P’X

One heroic implicit assumption in the above approach (but common in economics) is the one of universal and immediately executable prices. Reality differs from this ideal: • Illiquidity. Many real and financial markets are illiquid meaning that the assets are rarely traded, and that if one wants to purchase a certain asset in a relatively short period of time (within a minute, within a day, etc.), then one normally will have to pay (significantly) more than if one has a lot of time to purchase it. Similarly, or often even worse, if one wants to sell the good in a short period of time, one will have to accept a significant price reduction, relative to the price one may be able to achieve if one has a lot of time. Immediacy of execution has a price, measurable e.g. in the form of bid ask spreads offered by dealers (therefore one says that the dealer offers “immediacy services”). As everyone can easily verify, in many used good markets (antiques, cars and other consumer durables), dealer bid ask quotes are around 30% to 50%. In the most efficient parts of financial markets, e.g. US Government “on the run” bonds, bid ask spreads are below 1 basis point (0.01%). Most other markets are somewhere in between. • Asset specificity. A machine may have been tailor-made to exactly fit into the production process of a unique factory. Therefore, the value of the machine for the factory is much higher than for any other use, even with an unlimited time to sell. Also human capital may be specific, as an employee may be an expert in a certain unique production process, or a manager in terms of knowing and being able to manage the psychological idiosyncracies of the members of a specific team. The crucial role of asset specificity for economic organisation was worked out in particular by Williamson (1984). Both distinct matters will play a key role in chapters 5 and 6.

Introducing a (non-financial) corporate sector and the state We now introduce two other sectors which we treat most of the time as one: corporates and the state. All those are distinct from households, in the definition of United Nations (2008), these are “legal or social entities …whose existence is recognized by law or society independently of the persons, or other entities, that may own or control it. Such units are responsible and accountable for the economic decisions or actions they take, although their autonomy may be constrained to some extent by other institutional units; for example, corporations are ultimately controlled by their shareholders”. The simplified financial accounts of an economy with a corporate and government sector can be represented as follows, after introducing financial claims into the economy (financial claims and liabilities in bold, real items in normal font). Indeed, the corporate sector will hold real assets, and as counterpart financial liabilities, representing the means through which the corporates have been funded. There is one major difference between the corporate and the state sector: in the case of the corporate sector, also equity is a financial liability, i.e. the equity is owned by e.g. households. In the case of the state, equity is genuine equity not owned by anyone (such as household equity is not owned by anyone). The definition of the corporate sector that we consider here is closest to what United Nations (2008) calls an “enterprise”, namely “an institutional unit as a producer of goods and services”.

Household Real Assets HE-G-CE-CD-SD Household Equity HE Gold G Company Equity CE Company Debt CD State Debt SD Corporates Real assets CE+CD Corporate Equity (“financial equity”) CE Debt CD State

Real assets SE + SD State Equity (“real equity”) SE Debt SD

Why is there a separate corporate sector, separated from households? Economic production and therefore welfare can be spectacularly expanded by establishing “capitalist” firms, which have as liability both debt and equity (held by investors) in some maybe optimal proportion, and that as counterpart own a part of the real assets of the economy. Why there are “capitalist” firms is discussed for instance in Coase (1937) or Williamson (1984). Alternatives to pool ownership for larger scale industrial endeavours would be for instance the labour owned firm or state-owned production like in a socialist economy. Both alternatives work to some extent, but the capitalist firm has overshadowed all alternatives in its efficiency – at least for a large part of productive activity.

The amount of real assets in the economy does not change due to the creation of the corporate sector. A part of the real assets moves into the ownership of the corporate sector, and the household is compensated by receiving financial claims such that neither the net wealth nor the balance sheet length of the household changes. Of course, over time the creation of the real sector will make a difference: (i) it leads to a higher productivity and therefore to a steeper growth trend of the real assets held by the corporate sector (and hence the total amount of real assets of society); (ii) individual households will have different individual exposures to real assets, to equity and to debt, and therefore also their wealth will evolve over time differently, depending on how they positioned themselves.

The corporate sector’s need for real assets is obvious as far as traditional industries are concerned. For example, 19th century growth industries like mining, canal transportation, railways, clothing, breweries, etc. all obviously had needs to heavily invest into real assets. In the case of modern NFC sectors like IT or services, the financing needs arise for the purpose of establishing intellectual assets or the necessary brand name capital. Significant work is needed before the assets obtain value (e.g. thousands of programming hours before a complex software runs smoothly and can be deployed to clients). In these cases, it is not really existing real assets that are transferred from the household to the corporate. Instead, the firm uses its funds to rent “real” human capital.

In contrast to corporates, we can imagine the state as originating from one extraordinary large and powerful household who (i) is able to impose taxes on others, i.e. to appropriate wealth; (ii) Issues occasionally or systematically debt; (iii) provides public services to society (financed through taxes and fees). The raison-d’etre of the state, and the optimality of the state, is the subject of Public Economics. Generally speaking, the state should provide “public goods”, i.e. goods with natural monopoly properties in which economies of scale in production are positive without limits, such as for security and defence, the legal system, the core of the monetary system, and some parts of the infrastructure. Moreover, the

state may regulate market failures (externalities in production and consumption), and address acknowledged irrationality in human behaviour (e.g. enforce education and prohibit drugs). All this requires a stock of assets and employees. While the feudal state can really be considered as one enormous rich and powerful household, democracies could be considered being “owned” in a non- financial sense by the people. In the definition of United Nation (2008): “Government units are unique kinds of legal entities established by political processes that have legislative, judicial or executive authority over other institutional units within a given area. Viewed as institutional units, the principal functions of government are to assume responsibility for the provision of goods and services to the community or to individual households and to finance their provision out of taxation or other incomes; to redistribute income and wealth by means of transfers; and to engage in non-market production.”

1.3 Banks and central banks We now introduce banks into the financial accounts. What do banks do? They undertake various activities, as summarized in the following list. Some banks were specialized to a subset of these activities, while others cover many of them (“universal banks”). Historical origins of banks and of the various banking functions are explained for example in Kindleberger (1984, 71-152). Summarized briefly: • Merchant bank: a bank that supports merchants and in particular far distance trading in payments and settlements. A merchant bank may have a network of branches across countries to allow for efficient cross border settlement and to allow minimizing the species payment and transport. • Exchange bank: a bank specialized on exchanging precious metals and currencies against each other. The heterogeneity of coins in circulation and the key role of coins and precious metals in general in payments up to the 19th century explains their past importance. • Deposit bank: a bank to offer deposits to non-banks. In the oldest form, the bank simply keeps gold/silver coins of the customer in safe custody, allowing the depositors to make payments amongst each other through transfers that the deposit bank settles in the depositors’ accounts, avoiding the need to physically move gold or silver coins between transacting parties (allowing for higher “physical” efficiency and safety). • Note issuing bank. A bank that issues banknotes. Today only central banks issue banknotes, but in the 19th century in many countries more than one (private) bank was allowed to do so. • Discount bank: a bank purchasing a dealing with bills of exchange and other short term claims • Industrial bank: A bank financing industries and infrastructures • Investment bank: a bank specialized on financial market and corporate finance operations, as well as wealth management. The term emerged in the context of US bank regulation of the 1930s. • Universal bank: a bank doing many of the above activities.

United Nation (2008) uses for banks in our sense the term “Deposit-taking corporations except the central bank” and defines those as entities with “financial intermediation as their principal activity. To this end, they have liabilities in the form of deposits or financial instruments (such as short-term certificates of deposit) that are close substitutes for deposits. The liabilities of deposit-taking corporations are typically included in measures of money broadly defined.”

We introduce now a deposit and a note issuing bank into the financial accounts, which however only holds the gold in custody (i.e. this bank offers payment and security services). The two liabilities are actually identical in terms of financial accounts representation – they are only distinct in terms of technicalities of transfer and earmarking of bank liabilities to the claim holders. This bank will have to finance its running costs through fees it imposes on its clients.

Households Real Assets HE-G-C Household Equity HE Gold G-B-D Bank deposits D Banknotes B Claims to corporates C Corporates / state

Real assets C Liabilities to households C Banks

Gold D+B Deposits of HH D

Banknotes issued B

However, banks also soon became merchant- and industrial banks, i.e. they would use the assets obtained through deposit and banknote issuance to finance investments. Assume here for the moment that (i) banks only lend to corporates and the state, and not back to households; (ii) banks still hold gold at a certain ratio α of their total assets, essentially as a self-chosen or imposed liquidity reserve; (iii) corporates and the state do not hold deposits. Assume moreover that the gained ability of the banks to provide credit creates new opportunity for the corporate balance sheet to expand, say because bank credit can finance projects that direct financing from the household never can because of the insufficient monitoring expertise of households. In this case, the new financial accounts look as follows. The corporate uses the fresh bank credit to partially acquire more real asset from the household.

Households Real Assets (HE -G-C) - (1- α)(D+B) Household Equity HE Gold G-B-D + (1-α) (D+B) = G-α(D+B) Bank deposits D Banknotes B Claims to corporates C Corporates / state

Real assets C + (1- α)(D+B) Liabilities to households C

Bank credit to corporates / state (1- α)(D+B) Banks Gold α (D+B) Deposits of HH D

Credit to corporates / state (1- α)(D+B) Banknotes issued B

Commodity money, financial assets and IOU economy To understand the origins of central banking, it is worth recalling the merits of money in general and of financial money in particular from the perspective of contemporaneous authors. Already Aristotle (Politics, Part I, section IX) had noted that the inefficiencies of a barter economy can be overcome to some extent by the designation of one real asset as medium of exchange – typically coined silver or gold. In the early 18th century, this was described for instance by John Law (1705, Chapter 1) as follows:

“Before the use of money was known, goods were exchanged by barter, or contract; and contracts were made payable in goods. This state of barter was inconvenient, and disadvantageous… In this state of barter there was little trade, and few arts-men…. Silver as a metal had a value in barter, as other goods; from the

uses it was then apply’d to. … … What is meant by being used as money, is, that silver in bullion was the measure by which goods were valued: the value by which goods were exchanged: and in which contracts were made payable. He who had more goods than he had use for, would choose to barter them for silver… Silver being capable of a stamp, princes, for the greater convenience of the people, set up mints to bring it to a standard, and stamp it; whereby its weight and finess was known, without the trouble of weighting or fining…. As money increased, the disadvantages and inconveniences of barter were removed; the poor and idle were employed, more of the land was laboured, the product increased, manufactures and trade improved, the landed-men lived better, and the people with less dependence on them.”

The use of a precious metal as money has various efficiency limitations, in particular for larger scale payments: structural and cyclical scarcity of the precious metal, heterogeneity due to imperfect coinage and usage, fragmentation of units used, weight, risk of theft and therefore cost of storage and transport. Therefore, trade has, at an early stage, also relied extensively on credit. A financial asset is a claim of one economic subject towards another, for whom it is a financial liability. Financial contracts typically refer to unconditional or conditional cash flows to be paid in the future, whereby “cash flow” meant in the past settlement with species. The most basic financial asset is an “IOU” for “I owe you” – i.e. a promise to pay, which can be expressed in a numeraire good or any other specific good. IOUs can help to partially address the inefficiency of both a barter and of a species-based economy. In the words of Thornton (1802, 75) the benefits of credit are many:

“The day on which it suits the British merchant to purchase and send away a large quantity of goods may not be that on which he finds it convenient to pay for them.” Without credit, “he must always have in his hands a very large stock of money; and for the expense of keeping this fund (an expense consisting chiefly in the loss of interest) he must be repaid in the price of the commodities in which he deals.” Credit sets him “at liberty in his speculations: his judgement as to the propriety of buying or not buying, or of selling or not selling… may be more freely exercised”.

The problems of an IOU-financial system with many agents and therefore multiple claims and liabilities lengthening agents’ balance sheet are: (i) Complexity to keep a record of all the claims and liabilities; (ii) Credit risk and costs to monitor all claims; (iii) Possible contagion in case of late payments or credit events. This raises the question of netting claims, and/or eventually settling them in real money. Two steps have to be distinguished: financial claims netting without any increases of exposures to specific names and with “novation”, which implies such increases of claims to specific debtors. The potentials of netting without and with novation have been described for example by Kindleberger (1984, 440) in the specific context of the European Payments Union (EPU), but apply universally to any multilateral netting and settlement issue. Multilateral netting is in any case a complex practical issue, and it is unlikely that many agents can spontaneously coordinate on it in a pre-modern environment. Netting through novation moreover requires that agents are willing to accept changes to the identity of their debtors – which will only be the case if the new debtors are systematically better than the old ones. If there is enough species to settle transactions, then of course such a situation would not arise. But merchants may have insufficient species reserves or transferring species as a means of payment may be very costly.

A way to avoid both the problems of species payments and of an IOU system, is to create financial liquidity through a single high credit quality, multiple-unit IOU which is accepted by all as means of payment and store of value, and which therefore plays the same role as species in achieving settlement finality of bilateral trades, without however any of its inconveniences. By definition, if this IOU has the highest possible credit quality, then novating financial claims towards it is always an improvement, i.e. can be regarded as “settlement” of the claim. Issuing these universal prime IOUs can be done in various ways, the only constraint being that the issuer must be considered to be of the highest achievable credit quality (such that novation is always accepted). The status of having the highest possible credit quality

can be supported by a credible commitment of convertibility into species (i.e. into a real good), and this was considered necessary throughout most early central banking history. The issuer of this universal prime IOU may be the state, a public bank, or possibly a state-sponsored private bank. In the words of Aglietta and Mojon (2012, 3-4):

“Because debt must be settled in other forms of debt, there is a hierarchy of debts and for that matter of the institutions that issue them. The central bank is the bank that issues the debt in which all other debts are settled. … the ultimate liquidity in a payment system can be a commodity minted by the sovereign (or a foreign currency), or it can be the liability of a financial institution empowered by society as a whole or by its highest political authority, the sovereign. This institution is a central bank.”

To illustrate the mechanics of central financial money creation, consider a simple economy with n households (or n “merchants”), which each initially have equity R and real asset holdings R. Assume the households initially do not have a suitable medium of exchange, which limits their ability to trade with each other. The following financial accounts are obtained if the central bank is a 100% reserve bank.

Household/Merchant i´s accounts (i=1…n)

Real Assets R Household Equity R + S Silver S-A Banknotes/Deposits A Silver-assets only central bank Silver nA Banknotes/Deposits nA

While this scheme may improve the convenience of payments because it replaces species payment either by banknote or by giro payment, it does not solve the issue of netting and settling the multiple cross household IOUs if the amount of precious metals in the economy is structurally insufficient or subject to cyclical fluctuations. The availability of medium of exchange to ensure efficient payment and settlement is only increased if the central bank extends its balance sheet further by adding non-money assets, be they real or financial, i.e. by mixing into its assets elements of the previous schemes. The following scheme assumes that the central bank in addition purchases some government debt (amounting to B per household) and by providing some collateralised credit to each household (C).

Household/Merchant i´s accounts (i=1…n)

Real Assets R - B Household Equity R + S Silver S -A Credit from central bank C Banknotes/Deposits A+B+C Government Real assets nB Government debt nB Central bank expanding the monetary base Silver nA Banknotes/Deposits n(A+B+C) Government debt nB Collateralised lending to privates nC

The asset mix and total amount of assets will have to respect the need of the central bank to remain solvent and liquid, implying that the share of liquid assets should be sufficiently large (i.e. and nothing is as liquid in this context as silver species, as this is what the central bank commits to pay out to its creditors any time) and that the credit riskiness of the portfolio should be contained – through an adequate average quality of non-silver assets, and sufficient diversification. 1.4 An example: the German stylised financial accounts of 2013

The real economic accounts as they are provided by statistical offices or central banks provide a good picture of the actual relevance of the various financial, but also real assets. The following system of financial account reflects the German 2013 financial and real wealth, in absolute terms and relative to GDP (German GDP in 2013 amounted to EUR 2821 billion EUR; see DeStatis, Deutsche Bundesbank, 2016, “Sektorale und Gesamtwirtschaftliche Vermögensbilanzen, 1999-2015”, Statistisches Bundesamt, Wiesbaden). Note that “Real assets” are all assets that are non-financial, i.e. including intellectual rights etc.. Human capital is not counted.

German 2013 financial accounts - In percent of GDP (2013 DE GDP = 2.8 trillion euro) Entire national economy Real goods 454% Equity (net national wealth) 475% Machinery and equipment 43% Buildings and structures 271% intellectual property 18% Land without buildings 14% Land underlying buildings 107% Net claims towards RoW 21% Total 475% 475% Non-financial corporates Real goods 143% Debt 79% Cash and deposits 14% Equity 161% Securities 61% Other financial claims 21% Total 239% 239% State Real goods 61% Debt (includig deposits) 82% Cash and deposits 9% Equity 14% Securities 17% Credit and other claims 11% Total 98% 96% Financ. corp. (banks, insurance, etc) Real goods 7% Debt (includig deposits) 325% Cash and deposits 61% Equity 89% Securities 186% Credit and other claims 161% Total 414% 414% Households Real goods 243% Debt (includig deposits) 57% Cash and deposits 71% Equity 364% Securities 43% pension- anms other claims 64% Total 421% 421% Sums from the four sectors Real goods 454% Debt (includig deposits) 543% Cash and deposits 156% Equity 629% Securities 306% Other financial claims 257% Total 1173% 1171%

The Bundesbank is included in the financial corporates. Note that the difference in equity between the consolidated balance sheet at the top and the aggregated balance sheet below reflects the equity that is actually part of the “securities” and “other financial claims” on the asset side of sectors, i.e. it is a financial equity that can be netted out when establishing the net financial wealth of the economy. More generally, the aggregate balance sheet contains the financial assets of the different sectors and the related liabilities, which by definition have to be equal, but for the net claims to the rest of the world of

EUR 0.6 billion. It is also interesting to note what real assets actually consist in: they are dominated by buildings and related structures, and the land underlying buildings. Machinery and equipment is only a relatively small part of the stock of real assets (43% of GDP, out of a total value of real assets of 454% of GDP). Intellectual property is even less (18% of GDP). The very different equity ratios of the sectors are striking. Households are by far the least leveraged, followed by NFCs. Most leveraged are financial corporates. Another striking difference is the relevance of real assets versus financial assets, with the highest share of real assets held by the state, followed by households and NFCs with both having broadly one half of their balance sheet in the form of real assets. Finally, financial corporates have obviously the by far lowest share of real assets. The shares of financial assets is not per se an indicator of financial instability, as also real assets can lose value. The following table shows the shares of real assets and the share of equity in total balance sheet length of the four sectors.

NFCs State FCs HHs Asset composition: Share of real assets 56% 73% 2% 47% Liability composition: Share of equity 68% 41% 22% 87%

The following table from Ynesta (2008, table 1 and 2), gives an overview of household wealth allocation in four countries. The share of real wealth (normally mostly real estate) is between 43% (US) and 65% (UK). The financial asset allocation is even more diverse across countries: deposits constitute between 12% (US) and 50% (JP). Equity constitutes between 32% (US) and 10% (UK) of financial wealth, bonds are between 11% (DE) and 1% (UK). Insurance and pension claims are another very important financial wealth position of households: between 29% (JP) and 59% (UK). Table: Household wealth – shares in % (financial wealth shares sums up to 100%). From Tables 1 and 2 from Ynesta (2008), excerpt, most figures refer to 2006 Currency Securities Shares / Mutual Rest, incl. life Share of non- and other than equity fund insurance, financial assets deposits shares shares pension funds in HH wealth Germany 34 11 13 12 31 64 UK 26 1 10 4 59 65 US 12 7 32 14 35 43 Japan 50 3 15 4 29 46

What about human capital? For 1991, researchers (Georg Ewerhart, Heinz Lampert) have estimated total human capital in Germany to amount to 21 trillion DM, which compared at that time to 13 trillion DM fixed assets. Human capital is financed to a large extent in an "archaic" way, i.e. without involvement of money (e.g. parental care) or out of regular cash flows (direct payment for education by parents). The relatively low share of financing of human capital is illustrated by the relatively low value of education and student loans, at least in Germany. Sometimes financing of human capital may be hidden, like if a family takes a mortgage loan to finance the education of children. Also the transfer of human capital use from the household to the corporate sector is not done through a purchase of human capital, as this would be illegal, but normally through a renting for periods of time (e.g. a one years labour contract). This does not change the fact that the use rights of human capital are transferred (at least for a given number of hours per day) from the household to the employer.

Chapter 2: Central bank 2.1 Introduction 2.2 Central bank in a gold standard 2.3 Central bank in a paper standard 2.4 Collateral constrains 2.5 Logic of the central bank balance sheet from the perspective of monetary policy implementation 2.6 Disaggregating sectors 2.7 “Plain money” and “Full reserve banking” 2.8 Central bank electronic money

2.1 Introduction The pre-1800 origins of central banking are still debated (Bindseil, 2018). What characterises today’s central banks? (i)Today’s central banks are the only issuer of banknotes in the jurisdiction; (ii) They are, with few exceptions, publicly owned institutions (in 1900, still the majority of central banks were owned privately, even if they were subject to tight regulations); (iii) They are normally given by law a clear public mandate (“achieve price stability”); (iv) They can, in today’s paper standard, create legal tender (central bank money) almost without constraints through credit operations or purchases of securities. Their ability to create legal tender implies that they are not subject to liquidity risk, they are never forced to default, and exposures to them in the domestic currency will be considered credit risk free. (v) They deal only with other banks and with the government, i.e. they do not accept themselves deposits from corporates and households (which they tended to do in the 19th century). (vi) They often have a financial stability mandate. The following table summarises the relationship of the modern central bank with the other sectors in terms of credit provision and deposit taking. Practices changed over time, in particular modern central banks withdrew from accepting deposits from corporates, and they also stopped providing credit to Governments.

CB Assets CB liabilities Sector CB Credit provision CB Securities holdings CB deposit taking Households No No No NFCs No Yes – but rarely Not today (yes pre-WWI) Government No. Yes in war times, Yes - standard Yes Banks Yes (against collateral) Yes – but rarely Yes

Introducing the central bank means essentially splitting up the former banking sector into two more specialised sub-sectors: one “commercial bank” or simply “bank” sector accepting household deposits and providing credit to corporates, and one “central bank” essentially issuing banknotes, accepting deposits from banks (and the Government) and providing credit to banks.

2.3 A central bank in a paper standard In a pure paper standard, the central bank no longer needs to hold gold. Instead, it may, besides loans to banks, want to hold e.g. some corporate and government bonds. If so, the financial accounts could look as follows (but SCB may also be set to zero). It is assumed that the central bank buys corporate/state bonds in the secondary market from households. Therefore, on the household asset side, such “open market operations” of the central bank are reflected in a change of their deposits with banks. The banks compensate this by adjusting accordingly their recourse to central bank credit.

Households

Real Assets HE- G - D- SHH – B Household Equity HE Gold G Bank deposits D Banknotes B Corporate/state bonds SHH Corporates / state

Real assets D+B+SHH Debt securities issued SHH+ SCB Bank credit to corporates / state D+B-SCB Bank

Lending to corporates D+B -SCB Household deposits D Credit from central bank B -SCB Central Banks

Corporate / state bonds SCB Banknotes issued B Credit to banks B- SCB

The following table provides another stylized representation of the balance sheets of the different economic sectors. Note that it does not distinguish between financial equity and debt. However, it distinguishes between financial equity and real equity (equity that is the financial asset of no-one).

Real Financial claims towards → (financial liabilities of ↓) assets 1. 2. 3. 4. 5. Total HH Corporate Governt. Banks CB assets

1. Household RA(1) F(1,1) F(1,2) F(1,3) F(1,4) F(1,5) RA(1)+Σi(1,i) 2. Corporates RA(2) F(2,1) F(2,2) F(2,3) F(2,4) F(2,5) RA(2)+Σi(2,i) 3. Government RA(3) F(3,1) F(3,2) F(3,3) F(3,4) F(3,5) RA(3)+Σi(3,i) 4. Banks (RA(4)) F(4,1) F(4,2) F(4,3) F(4,4) F(4,5) RA(4)+Σi(4,i) 5. CB (RA(5)) F(5,2) F(5,3) F(5,4) RA(5)+Σi(5,i) Real equity RE(1) (RE(3) RE(5)

Total liabilities Σi(i,1) Σi(i,2) Σi(i,3) Σi(i,4) Σi(i,5) RE(5) RE(5) RE(5)

Note that the following equalities hold: - Σ real assets = Σ real equity, i.e. total real assets of economy are equal to total real equity. - Σ Σ financial assets = Σ Σ financial liabilities (sum of all financial assets across all sectors equals sum of all financial liabilities across all sectors) - Σ assets of one sector = Σ liabilities of one sector

Two sectors are assumed to actually make choices: first the household chooses to diversify its real assets into financial assets and determines the extent of this diversification, as well as the related use of the three types of financial assets; second the central bank decides on the split up of its monetary policy operations between outright securities holdings SCB and credit operations with banks (as residual, B-SCB). All other balance sheet position are expressed in terms of these four choice variables, plus the initial household endowment HE. The household demand for specific financial assets is potentially unstable, as households may want to rebalance their financial asset portfolio e.g. for reasons of credit risk fears, or speculative position taking.

This is taken up in the following version of the financial accounts. It simplifies the previous one in terms of merging gold into real assets. In particular, households may: withdraw deposits from banks (maybe for seasonal reasons, or out of fear on the stability of banks), and hold more banknotes. This is captured in flow d below. Moreover, they may be unwilling to roll over debt securities at maturity (e.g. out of credit fears regarding the issuers), and instead hold more deposits – as captured in flow s below. How do these flow filter through the entire financial system depends inter alia on whether the central bank is ready or not to buffer out the impact of the shifts of household demand for financial assets or not. In the following system of financial accounts, it is assumed that the central bank is willing to react to shock d by tolerating more central bank credit to banks, and to shock s by substituting lending to banks by outright holdings of securities. While it has been best practice for central banks to react in this way to shocks of kind d, it is controversial weather central banks should react like this to shock s (be a “market maker of last resort”). A simple system of financial accounts with two types of liquidity shocks Households / Investors

Real Assets HE – D – SHH - B Household Equity HE Deposits Bank D - d + s

Debt securities SHH - s Banknotes B + d Corporate / Government

Real assets D+B+SHH Debt securities SCB +SHH

Credits from banks D+B-SCB

Bank

Lending to corporates D+B-SCB Household deposits D - d + s Credit from central bank B -SCB + d – s Central Bank

Debt securities SCB + s Banknotes B + d Credit operations with banks B -SCB + d – s

In this specification, the central bank is the key built in liquidity stabilizer of the financial system. The central bank is able to issue banknotes demanded by households/investors and to re-compose its assets in line with the possible need to stabilize the overall demand for different financial assets. In the system of accounts above, none of the financial shocks relating to the instability of the household demand for the various financial assets reaches the corporate / government sector. This is a positive scenario, as funding shocks reaching the real sector imply forced deleveraging or defaults, both of which are costly for society (although a total shielding of the Government and corporate sector from funding risks is also no solution, as it would undermine market discipline and incentives, and would lead to a zombification of the economy). If the central bank could not see itself responsible for addressing the run on the corporate sector bond market, i.e. for the shock s, then the banking system could itself act as LOLR for the corporate sector by recycling the deposit inflow as illustrated below:

Corporate / Government

Real assets D+B+SHH Credits from banks D+B-SCB + s

Debt securities SCB +SHH -s Bank

Lending to corporates D+B-SCB + s Household deposits D + s Credit from central bank B -SCB

Above, it is assumed that the banking system is fully elastic in extending its balance sheet. However, in reality, quick adaptation of the balance sheet of the bank will come at extra cost: (i) e.g. the staff has to get extra paid for “night shift work”; (ii) expensive additional resources may be needed in the form of consultants; (iii) the additional risk taking associated with the rapid acceptance of extra exposure needs to be compensated, etc. The banks will charge these extra costs to the corporate, which will be at the expense of the profits (or, in stock terms, of the capital) of the corporate. What would happen if the banks also refuse to use the deposit inflow to increase lending to the real economy, i.e. both banks and the central bank are unwilling to help the corporate sector. In this case, the problem arises that corporate assets are illiquid and specific to some extent, i.e. they have been made specific to the uses by the corporates (in the sense of Williamson, 1984). Therefore, when being sold back to households in the short term to produce the liquidity needed to pay off the corporate debt that cannot be rolled over, they lose value and create financial losses, as captured below. The letter “f” stands for fire sale losses. To capture the effects, we need to introduce a positive equity of the corporate sector, held by households. Then we can capture the losses as follows. Note that we did not reproduce the financial accounts of the financial system (bank and central bank) because they are not affected. (They would be affected if equity of corporates would be insufficient to buffer fire sale losses, and then also the fixed income liabilities of corporates would suffer losses).

Households / Investors

Real Assets E – D – SHH - B – CE + s Household Equity E – f Deposits Bank D

Debt securities SHH - s Corporate equity CE - f Banknotes B Corporate / Government

Real assets D+B+SHH +CE - s – f Corporate equity CE -f Credits from banks D+B-SCB

Debt securities SCB +SHH – s

Compared to the previous scenarios (either the central bank or the banks acting as LOLR to the corporate sector), society now suffers from asset fire sale losses of f. The corporate needs to sell not only asset of a value s, but of a value s+f, generating liquidity of s. These losses reduce the corporate’s equity, and eventually the household equity (as the household holds the corporate equity). This is a first example of how liquidity and solvency interact, and how illiquidity can cause damages to society.

2.4 Collateral constraints The quantity and quality of central bank eligible collateral limits the borrowing potential of banks from the central bank. Limits arise from (i) restricted eligibility (e.g. excluding particularly non-liquid and non- transparent bank asset classes and setting a minimum credit quality for the collateral obligor), (ii) conservative collateral valuation (to reduce the risk of assuming too high collateral values), (iii) haircuts (to cater for losses in value during the liquidation period after a counterparty default), or (iv) quantitative collateral limits (to address concentration and correlation risks). To simplify, assume the following bank balance sheet with two liabilities, household deposits and central bank credit, and two assets, loans and securities. Assume also that the central bank imposes a haircut of h1 on loans to corporates and of h2 on securities, with 1>h1>h2>0. Collateral value after haircuts (or central bank credit potential) is for loans (1-h1)L and for securities (1-h2)S. Bank Loans to corporates L Household deposits D - d Securities holdings S Credit from central bank L+S-D +d

The maximum borrowing of this bank from the central bank is (1-h1)L+(1-h2)S. The actual borrowing from the central bank, CB, must not exceed this, i.e.:

L+S-D +d ≤ (1-h1)L+(1-h2)S

This implies that when the bank suffers from a deposit outflow exceeding d*= (1-h1)L+(1-h2)S – (L+S-D) then it will hit the collateral constraint, and it may become illiquid and default, unless it is able to conduct fire sales with sufficiently low losses (although the bank above actually has no equity buffer and therefore strictly speaking cannot afford any fire sale losses), or the central bank is willing to extend collateral availability by reducing haircuts.

For example, in the case of the euro area, out of EUR 30 trillion of aggregated bank assets, the value of central bank eligible collateral after haircuts that could be used at any moment in time may be around EUR 5 trillion.

2.5 Logic of the central bank balance sheet from the monetary policy perspective The central bank balance sheets shown so far were simplifications with regards to two important aspects that now need to be differentiated further. First, so far the banknote item was the only balance sheet item outside monetary policy operations and the banks’ sight deposits with the central bank. However, in reality, banknotes are only one amongst various so-called “autonomous” central bank balance sheet factors. Autonomous factors may be defined as all those central bank balance sheet items which are neither monetary policy operations nor the deposits of banks with the central bank. They are factors that are not under direct control of the monetary policy implementation function, although they may have an impact on the amount of deposits of banks with the central bank, and hence on the money market. In the central bank balance sheets above, banknotes were the only autonomous factor. In reality, there are a number of other autonomous factors, which can all be integrated into the financial account model (essentially by introducing other sectors who interact with the central bank). For example: (i) Governments may be allowed to use the central bank for their (volatile) cash holdings. When the Government collects taxes on a certain day of the month, its deposits with the central bank may increase steeply. (ii) The central bank may intervene in foreign exchange markets and increase or decrease its foreign reserves holdings. (iii) The central bank

may purchase financial assets not for monetary policy, but for investment purposes (iv) the IMF may draw on credit lines it has with each central bank. The figure below illustrates the time series patterns of autonomous factors in the Eurosystem balance sheet. Liquidity providing autonomous factors are shown with positive, liquidity absorbing autonomous factors are shown with negative sign. For government deposits and banknotes, the charts on the right side zoom into the period 2016-mid 2017. The following observations may be made: - Net foreign assets were rather stable from 1999 to 2009, but then exhibited some volatility and trend growth. The volatility is actually also related to the ECB’s swap operations with foreign central bank related to the provision of USD to domestic euro area banks. These may also be considered as monetary policy operations and would then have to be taken out of a net foreign assets time series which is purely supposed to contain non-monetary policy positions. Foreign exchange policy operations only took place once for a sizable amount – namely in September and November 2000. The interventions were so small that they can hardly be detected. - Government deposits were fluctuating around EUR 50 billion for the first 10 years of the euro. They increased and became more volatile during the crisis years when Government treasurers preferred to keep higher shares of their cash with central banks also for credit risk reasons (despite the less attractive remuneration, compared to deposits with banks). Again government deposits increased after the launch of the large scale asset purchase programme in 2015, reflecting that banks started to apply very low rates to corporate and Government deposits (as low as e.g. -60 basis points, against a remuneration of Government deposits at the level of the deposit facility rate of – 40 basis points). The zoomed chart shows the erratic nature of the day- to-day changes of government deposits, with however some monthly seasonality. - Banknotes exhibit high trend growth with idiosyncratic moves around the euro cash change-over in 2002. During the crisis years (2008-2012) banknotes moved to slightly above their long term trend. The zoomed chart shows the seasonal patterns of this time series: peak around Christmas / year end, with weekly and monthly regularities that allow for being exploited in forecasting. Autonomous factors – Eurosystem (1999-2007)

As illustrated in the following partial financial accounts, the starting level and fluctuations of any of these four autonomous factors affect the necessary recourse of banks to central bank credit, which can matter both from a monetary policy perspective and from a bank funding / financial stability perspective.

Bank

Lending to corporates D+B-INV Deposits D - GD + FR Credit from central bank B +GD – INV – FR Central Bank Investment portfolio INV Banknotes B Foreign reserves FR Government deposits GD Credit to banks B +GD – INV – FR

Second, we need to differentiate between different types of monetary policy instruments in the central bank balance sheet. Monetary policy instruments are the tools used by the central bank to reach its operational target. Central banks use mainly three such tools: standing facilities, open market operations, and reserve requirements. • Open market operations may be defined as central bank financial transactions with banks at the central bank’s initiative, whereby two subtypes can be distinguished: (i) Outright purchases or sales of assets (normally debt securities) from banks; (ii) Lending (or “credit”, “reverse” or “temporary”) operations, conducted for instance through an auction (or “tender”). • Standing facilities are central bank financial transactions at the initiative of banks, on the basis of a commitment of the central bank to enter such operations at certain conditions. Three variants have to be distinguished, of which the first two are liquidity-providing and the third liquidity-absorbing: (i) A discount facility: banks can sell certain short term paper to the central bank at any time, whereby the discount rate specified by the central bank is applied to calculate the price on the basis of the securities’ cash-flows. (ii) A “Lombard” facility: banks can borrow at any time against eligible collateral at the rate specified by the central bank, with overnight maturity. (iii) A deposit facility: banks can deposit at any time funds with the central bank on a specific account where it gets remunerated at a specific rate. • Reserve requirements oblige banks to hold in a certain period (per day, or on average over an e.g. two weeks or one-month period), a certain minimum level of sight deposits (“current accounts”, “reserves” and “central bank liquidity” are used as synonyms for banks’ sight deposits with central banks in monetary policy implementation jargon) on their account with the central bank. The fulfillment is measured only on the basis of end of day snapshots (i.e. intra-day levels of reserves are not relevant). The size of the reserve requirement of a specific bank is normally set as a function of specific items of its balance sheet which need to be reported on a monthly basis. For instance, in the case of the European Central Bank, the requirement for each bank amounts to 1% of its liabilities to non-banks with a maturity below two years. Even if reserve requirements are zero, there still is a sort of reserve requirement in the sense that banks need to hold at day end at least a zero balance on their deposit account with the central bank. We therefore need to show explicitly this deposit account of banks with the central bank. Below, we will assume that the central bank offers a Lombard and a deposit facility, but that it does not impose reserve requirements. When the control of short term interest rates will be modelled, the differentiation between (i) outright open market operations; (ii) credit open market operations; and (iii) standing facilities will be necessary. The following figure reflects this slightly more differentiated representation of the central bank balance sheet, ordered according to the three main types of balance sheet items.

A central bank balance sheet ordered according to the types of balance sheet items Central bank Autonomous factors: liquidity providing Autonomous factor: liquidity absorbing Net Foreign assets Banknotes Investment portfolios Government deposits

Monetary policy operations: liquidity providing Monetary policy operations: liquidity absorbing Open market op. - outright purchases Term deposit collection / repo Open market op. - credit to banks Issuance of debt certificates Borrowing facility Deposit facility

Deposits of banks

What was labelled B (originally for “banknotes”) in the financial account models above would now be defined as “autonomous actors”, with:

Autonomous factors = Banknotes + Government deposits - Foreign reserves – Investment portfolios

Defined in this way, autonomous factors are netted as central bank liability item. If we define “monetary policy operations” as the sum of all monetary policy operations netted as a central bank balance sheet asset item, we can write the balance sheet identity as follows:

Deposits of banks = Monetary policy operations – Autonomous factors

If the central bank imposes reserve requirements, then deposits of banks have to be equal to (at least) those reserve requirements. This implies that central banks have to adjust in principle their monetary policy operations to the fluctuations of autonomous factors such that: Monetary policy operations = Reserve requirements + Autonomous factors One can call the left-hand side of this equation “supply” and the right-hand side of this equation “demand” for central bank deposits. The deposit supply by the central bank has to be such as to suffice both for reserve requirements and the liquidity absorption through autonomous factors. Moving again down to the level of the individual central bank balance sheet items, one can call “liquidity absorbing” all central bank balance sheet liability items (except bank deposits), and “liquidity providing” all central bank balance sheet asset items. This is because bank deposits with the central bank are a central bank liability item, and therefore any increase of another central bank liability item leads – all other items being assumed unchanged – to a decrease of bank deposits, while any increase of a central bank asset item leads – again all other items being assumed unchanged – to an increase in central bank deposits. For example, the autonomous factor “Investment portfolios” is a liquidity providing autonomous factor, while “issuance of debt certificates” is a liquidity absorbing monetary policy operation.

The liquidity deficit of the banking system vis-à-vis the central bank

The liquidity deficit of the banking system vis-à-vis the central bank may be defined in two ways. First, as original liquidity deficit: the need to provide funding in the form of monetary policy operations, i.e. including both credit operations and outright securities purchases (i.e. both types of central bank open market operations). Second, as liquidity deficit post monetary policy outright operations: the remaining need for the central bank to provide funding to banks in the form of central bank credit operations.

Consider the following example of a central bank balance sheet in which banks hold not only the required amount of deposits with the central bank (i.e. required reserves, RR), but also some excess reserves, XSR.

A central bank balance sheet to define the liquidity deficit Central bank Autonomous factor B Monetary Policy Operations Credit monetary policy operations B+RR+XSR-S Outright monetary policy portfolio S Deposits of banks RR+XSR

The original liquidity deficit is equal to B+RR. Note that excess reserves are not counted as a component of the liquidity deficit. The liquidity deficit post monetary policy outright operations is equal to B+RR-S.

2.6 Disaggregating sectors So far, we have assumed that every economic sector (households, corporates, the government, the banks, the central bank) is in itself homogenous, which is somewhat equivalent to assume that each of the individual components of the sectors are equal and can be captured by one representative agent and that there are no noteworthy flows within the entities of a sector. This also implies that shocks (liquidity and solvency) would affect all individual households and firms equally within each sector. In reality, all sectors – except the central bank and the Government, which are constituted each by a single entity - are internally heterogeneous. In the case of households: some have taken loans and are thus leveraged, while others are not; households have different asset compositions. Some own real estate (as typical largest real asset position of households) while others have almost only financial assets. Also, some households may have a lot of equity (are rich) while others may have almost none (are poor). Corporates and banks have different liability structures, i.e. different shares of credit, bond issuance, and equity. The bank credit and bond parts may have different maturity structures, etc. We start with providing two illustrations relating to liquidity flows.

Deposit shifts between individual banks and the interbank market Today’s banking crises are typically not about shifts of deposits into banknotes, but about shifts of deposits between banks. We therefore need to adjust the previous financial account systems by introducing two separate banks. This will also be the precondition for modeling interbank markets. We assume that the banks are identical ex ante and each represent one half of the banking system. The interbank market position between the two types of banks is set initially to Y, with bank 1 lending to bank 2. This could be the case because bank 1 has comparative advantages in deposit collection, while bank 2 has comparative advantages in originating and managing loans to corporates. We simplify the model above in the sense that we no longer consider securities issuance as a funding source for corporates, but introduce two new flows to the model. Flow k reflects a deposit shift between banks initiated by households. The flow may be triggered by one bank suddenly offering a higher remuneration rate, or by rumors about one bank having solvency problems. Flow y affects the interbank market lending, and may reflect a change of business strategy by the lending bank, or that the lending bank believes that the borrowing bank is in trouble and that therefore credit riskiness of loans to it is perceived as too high. Flows k and y both imply funding losses for bank 2, which consequently has to extend its central bank borrowing. Note that if k+y>B/2, bank 1 will be in excess liquidity, i.e. without any recourse to central bank credit, bank 1 will have a claim on the central bank of k+y-B/2. In this case, the central bank balance sheet expands by the latter amount.

Household deposit and interbank lending shifts – with two separate banks Households Real Assets E - D -B Household Equity E Deposits Bank 1 D/2 + k Deposits Bank 2 D/2 - k Banknotes B Bank 1 Lending to corporates D/2 + B/2 - Y Household deposits / debt D/2 + k Deposits with CB max(0,-B/2+k+y) Credit from central bank max(0,B/2–k–y) Credit to Bank 2 Y -y Bank 2 Lending to corporates D/2 + B/2 + Y Household deposits / debt D/2 - k Credit from central bank B/2 + k + y Credit from Bank 1 Y - y Central Bank Credit to banks B/2+k+y +max(0, B/2–k–y) Banknotes B Deposits banks max(0,-B/2+k+y)

These accounts allow defining three important concepts. • By allowing liquidity flows k and y to be compensated by heterogeneous changes of the recourse by individual banks to central bank credit, without this yet lengthening the central bank balance sheet (as B/2-k-y≥0), the central bank provides relative intermediation to the banking system. • Once liquidity flows are such that some banks deposit excess funds with the central bank, thereby lengthening the central bank balance sheet, (which happens in the financial accounts above when B/2-k-y<0), while other banks are particularly dependent on the central bank, we speak of absolute central bank intermediation of the banking system. Normally absolute central bank intermediation can be avoided by setting a sufficient spread between the (higher) rate at which banks can borrow from the central bank and the (lower) rate of remuneration of excess deposits. Interbank lending allows the banks to avoid the costs associated with this spread. • By choosing to conduct its credit operations as “full allotment” operations, i.e. providing whatever the banks ask for, the central bank can provide these intermediation services passively, i.e. it does not need to take any particular initiative for it. The limit to intermediation despite full allotment is central bank collateral availability. Widening the collateral set in a crisis specifically for the sake of allowing for more intermediation, or for providing confidence to markets that banks have large liquidity buffers, would be examples of active intermediation measures.

Finally, note that interbank lending can be either positively or negatively correlated with household deposit shifts, whereby the former case is detrimental, and the latter case supportive to financial stability. A positive correlation is likely if the main underlying factor are news on a poor performance of the bank and possible solvency problems. A negative correlation will occur if household deposit shifts are driven by factors which are independent from the actual credit quality of the bank (e.g. “noise” reflecting payment flows) and if the banks have no mutual credit risk concerns such that the interbank market can serve as an elastic buffer compensating exogenous liquidity flows. A negative correlation generally prevailed e.g. in the euro area between 1999 and 2007, a positive correlation was experienced particularly during the sovereign debt crisis, in which Greek, Spanish, Portuguese or Irish banks experienced both a cut off of interbank lending and deposit outflows. The following figure represents graphically the above stated system of accounts.

Credit money created by banks To represent credit money creation in our system of financial accounts, we start from the simplest case of a financial account system with two banks and with all financing to the real economy being done through the banking system. Beforehand, recall Tobin’s famous critical presentation of banks’ credit money creation as a climax of any introductory course on monetary economics (Tobin, 1963, 1): “Perhaps the greatest moment of triumph for the elementary economics teacher is his exposition of the multiple creation of bank credit and bank deposits. Before the admiring eyes of freshmen he puts to rout the practical banker… The banker is shown to have a worm’s-eye view, and his error stands as an introductory object lesson in the fallacy of composition. From the Olympian vantage of the teacher and the textbook it appears that the banker’s dictum must be reversed: depositors entrust to bankers whatever amounts the bankers lend. To be sure, this is not true for a single bank: one bank loan may wind up as another bank’s deposit. But it is, as the arithmetic of successive rounds of deposit creation makes clear, true of the banking system as a whole. Whatever their other errors, a long line of financial heretics have been right in speaking of “fountain pen money” – money created by the stroke of the bank president’s pen when he approves a loan and credits the proceeds to the borrower’s checking account.” For instance Werner (2014, 1), one of the promoters of plain money (see section 2.7 below) summarises the old question and the three schools in monetary economics about the ability of banks to create money as follows: “According to the financial intermediation theory of banking, banks are merely intermediaries like other non-bank financial institutions, collecting deposits that are then lent out. According to the fractional reserve theory of banking, individual banks are mere financial intermediaries that cannot create money, but collectively they end up creating money through systemic interaction. A third theory maintains that each individual bank has the power to create money ‘out of nothing’ and does so when it extends credit (the credit creation theory of banking).” Werner (2014) also presents an empirical test to conclude that the third school is right (“money supply is created as fairy dust produced by the banks individually, out of thin air”, p. 1). Below we will use the financial accounts model to try to understand the issue. Denote by C´1 (C2) the credit money creation by bank 1 (bank 2) to the households. We assume that households keep the money in the form of deposits with banks, but not necessarily with the same bank. Concretely, the household would split up its additional credit money holdings equally across the two banks, regardless of which bank provided the credit. Financial accounts with two banks and credit money creation (assuming │C2-C1│

Deposits Bank 1 D/2+(C1+C2)/2 Credit from bank 1 C1 Credit from bank 2 C2 Deposits Bank 2 D/2+(C1+C2)/2 Bank Equity F Banknotes B Corporate / Government

Real assets D+B+F Credits from banks D+B+F Bank 1

Lending to corporates D/2 + B/2 +F/2 Household deposits / debt D/2+(C1+C2)/2 Lending to households C1 Credit from central bank B/2+(C1-C2)/2 Equity F/2 Bank 1

Lending to corporates D/2 + B/2+F/2 Household deposits / debt D/2+(C1+C2)/2 Lending to households C2 Credit from central bank B/2+(C2-C1)/2 Equity F/2 Central Bank Credit operations B Banknotes B

On the basis of these accounts, what could be the possible limits to credit money creation by banks? Assuming that credit claims to households are not central bank eligible collateral (or at least are subject to haircuts), the collateral constraint is a natural break to over-proportional credit money creation by a single bank. If credit claims to households can be pledged as central bank collateral, then the haircuts imposed by the central bank would still constitute a limitation for the single bank.

What happens instead if C1=C2, i.e. if the banks engage in parallel credit money creation? In this case, consider the constraint of the bank in terms of recourse to central bank credit to derive the maximum credit creation:

(1-h)(D/2 + B/2 +F/2) ≥ B/2 + r(D/2 + C/2) <=> (1-h)(D + B +F) ≥ B + r(D + C) <=> D+C = [(1-r)(D+F)-hB]/r

 CMax = [(1-h-r)D – hF - hB]/r = [-h(D+F+B)+(1-r)D]/r

For example, if F=1, B=2, D=7, then CMax = [-10h+7(1+r)]/r. The central bank can even enforce CMax=0 by setting (h,r) as follows: [-10h+7(1-r)]/r = 0 => h* =7/10(1+r). Of course it should be taken into account that banks will anyway not go to their limit as they prefer to have some liquidity buffer.

Even more importantly, in reality, limits to credit money creation however arise anyway out of the preferences of the household. Bank credit money creation is costly in the sense that the banks will require a higher remuneration rate for the claims towards the households than what they offer in terms of remuneration rate of deposits (banks have to cover their operations costs and compensate their financial risk taking). Therefore, households will have a demand for credit money only to the extent that they see a particular utility attached to it justifying the costs.

It is important to note that the credit money expansion by the bank has no impact on the central bank balance sheet in the financial accounts system above as long as the difference in the pace of additional credit provision by the two banks is not too large, i.e. as long as │C2-C1│

2.7 “Plain money” and “full reserve banking” A number of monetary economists claim that deposit money creation by banks is one of the major causes of monetary and financial instability and recurring crises. Therefore, in their view, all money creation should be undertaken by the central bank. Banks would need to refinance through the central bank. They could also still refinance in capital markets, assuming that this form of bank funding is sufficiently remote from creating “money”. For example, Benes and Kumhof (2012, 4) analyze Irving Fisher’s Chicago Plan (a sort of full reserve banking proposal) and summarize its features and advantages as follows.

“The key feature of this plan was that it called for the separation of the monetary and credit functions of the banking system, first by requiring 100% backing of deposits by government-issued money, and second by ensuring that the financing of new bank credit can only take place through earnings that have been retained in the form of government-issued money, or through the borrowing of existing government-

issued money from non-banks, but not through the creation of new deposits, ex nihilo, by banks. Fisher (1936) claimed four major advantages for this plan. First, preventing banks from creating their own funds during credit booms, and then destroying these funds during subsequent contractions, would allow for a much better control of credit cycles, which were perceived to be the major source of business cycle fluctuations. Second, 100% reserve backing would completely eliminate bank runs. Third, allowing the government to issue money directly at zero interest, rather than borrowing that same money from banks at interest, would lead to a reduction in the interest burden on government finances and to a dramatic reduction of (net) government debt, given that irredeemable government-issued money represents equity in the commonwealth rather than debt. Fourth, given that money creation would no longer require the simultaneous creation of mostly private debts on bank balance sheets, the economy could see a dramatic reduction not only of government debt but also of private debt levels.

Werner (2016) argues that by not understanding the problematic implications of deposit money creation, “the economics profession has failed over most of the past century to make any progress concerning knowledge of the monetary system”.

Huber (1999, 5-6), one of the key German supporters of “plain money”, summarizes the plain money proposal as follows: “The plain money proposal says: Give the central bank unimpaired full control of the total money supply on the legal basis of a general prerogative of money creation. In other words, have the entire money base - cash as well as non-cash money - exclusively issued by the central bank. This implies the abolition of the banking sector’s capability to create non-cash money in the form of sight deposits. Today, there is a mixed money base made up of one kind of money created by the central bank and another kind of money (sight deposits) created by the banks. Plain money still implies a two-tier banking system, but it does not mean having a mixed money base any longer, instead, just one kind of money from a single source, easy to understand, to handle and to keep control of. Plain money does not necessitate particular changes of institutional and market structures. Simply, banks would be credit brokers and no longer be credit creators. They would lose today‘s seignorage, the extra profit from the creation of non-cash money. Apart from that, the normal profitability of the banking business will remain untouched. Banks would be able without any restrictions to continue to carry out every kind of business they do now, e.g. managing deposits and transfers of their clients, granting loans to whomever they consider creditworthy, investing in finance market papers such as bonds or equity shares for their clients and for themselves, offering any variety of financial products, etc.”

Huber (1999, 18) also explains the financial account implications of plain money: “the credit claims of a bank on the loan-taking clients remain; the cash liabilities of a bank to the account-maintaining clients disappear, and the cash claims of the account-maintaining clients on the bank disappear equally; in exchange for the latter a credit claim of the central bank on the bank appears. These credit claims would be part of the assets on the balance sheet of the central bank, corresponding to the sums of non-cash money being registered on the liability side (neither of which are the case today).” Also KPMG (2016) studied plain money (or as they call it “sovereign money system”) in a report commissioned by the Icelandic Prime Minister’s Office. In this report, financial account representations are shown, and a survey is provided of political initiatives (e.g. in Switzerland, Iceland, UK, US) to study and possibly introduce plain money, and of academic literature.

In the financial accounts below, we interpret “plain money” as meaning that sight and savings deposits need to be held at the central bank and we use a basic numeric example.

Households Real Assets 6 Household Equity 14 Sight deposits 6 Retail mortgage loans 2 Banknotes 1 Bank bonds 1 Bank equity 1 Corporate/state bonds 1 Corporate / Government

Real assets 8 Bonds issued 1 Bank credit to corporates / state 7 Commercial banks Credit to corporates / state 7 Sight deposits of HH 6 – 6 Retail mortgage loans 2 Bonds issued 1 Central bank credit 1 + 6

Bank equity 1 Central Bank

Credit to banks 1 +6 Banknotes issued 1 Sight deposits of households 6

The central bank becomes a much more important financier of the commercial banks, i.e. it becomes an intermediary between depositors and banks. The central bank balance sheet will lengthen, and central bank collateral constraint will become much more binding.

If instead we interpret plain money as full reserve money, as in the Chicago Plan, the following accounts would be obtained.

Commercial banks Credit to corporates / state 7 Savings and sight deposits of HH 6 Retail mortgage loans 2 Bonds issued 1 Required reserves +6 Central bank credit 1 + 6

Bank equity 1 Central Bank

Credit to banks 1 + 6 Banknotes issued 1 Required reserves of banks 6

The implications on the central bank balance sheet are similar to plain money: it lengthens considerably and credit risk taking and collateral constraints are likely to become more relevant. The central bank collateral framework becomes a crucial allocation mechanism of the financial system and central banks would need to accept almost the entire assets of banks as eligible, and impose only moderate haircuts. This implies de facto a centralization of the credit allocation process, which arguably is unlikely to make the economy more efficient. Regarding the other claims of the promoters of plain money and full reserve banking, they cannot be judged from our simple financial accounts representation, but a dose of skepticism may be warranted. For example it seems doubtful that a significant additional income can be generated for the public sector if we believe that the banking sector is broadly competitive and therefore did not appropriate significantly oligopolistic rents. Also, the central bank may need to lower its central bank lending rate to maintain financial conditions unchanged when introducing plain money.

2.8 “Electronic central bank money” (CBDC) offered to non-banks. This idea goes partially in the same direction as plain money, but would be voluntary and is motivated from the perspective of efficiency of the means of payment and store of value, and not out of skepticism

regarding the stability properties of bank money creation. CBDC (“central bank digital currency) would be brought into circulation in the same way as banknotes: on demand by households and corporate users, who could freely convert bank deposits into CBDC. Indeed, the internet and the use of mobile devices have transformed the possibilities how money can be stored and exchanged. Already today, a large share of money transfers is undertaken electronically, via internet banking, card payments or the use of e- money. Furthermore, new technologies for digital currencies have become available. Normally, electronic money transfers are based on commercial bank money, i.e. money is transferred from one commercial bank account to another. Only central bank account holders, i.e. commercial banks and a few other institutions (e.g. market infrastructures, governments), can transfer central bank money electronically. Non-banks can currently hold and transfer central bank money only in the form of banknotes. CBDC is, like banknotes, a direct claim on the central bank. Several central banks have recently published on this topic (The People’s Bank of China, the Bank of England and Sveriges Riksbank).

The following concrete advantages for the use of CBDC may arise: • It is more convenient and efficient to hold money in digital form than in the form of cash. • CBDC is more secure than commercial bank money, from a credit and settlement risk perspective. • People’s preference for money in digital form could lead to an undesired increase in the usage of virtual currencies in the absence of CBDC. Virtual currencies may create risks to price and financial stability. • The provision of CBDC is cheaper for the central bank than the provision of cash. • Promoters of plain money (see section 2.7) argue that increased reliance on central bank money has various macroeconomic advantages, such as higher fiscal income for the state and a more stable financial system.

CBDC could be implemented in two forms: (1) Offering deposit accounts with the central bank to all households and corporates. From a technological perspective, this would not be very or innovative, but just a matter of scaling the number of accounts offered. (2) Alternatively, the central bank could offer a digital token currency that would circulate in a decentralized way without central ledger, i.e. without the central bank knowing who currently holds the issued tokens. This would be more innovative, but certainly feasible as well with more recent technology developed for crypto-currencies. Relative advantages of (1) compared with (2) are its relative simplicity and that money laundering is not facilitated.

If households substitute banknotes with CBDC, then central bank and commercial bank balance sheets do not really change. However, if households substitute commercial bank deposits with CBDC, then this would imply a funding loss for commercial banks, i.e. lead to “disintermediation” of the banking sector. In particular sight deposits largely used for payment purposes could be expected to shift at least to some extent into riskless CBDC, leading to a loss of commercial banks’ funding of equal size. Banks would have to try to offer better conditions on their deposits in order to protect their deposit base as much as possible – but this would imply higher funding costs for banks and a loss of commercial bank “seignorage”. The central bank could aim at limiting the attractiveness of CBDC, through fees, or through a lower remuneration rate than the short-term market rate. In addition to the structural loss of funding for banks, there is also a financial stability issue. CBDC makes it significantly easier for non-banks to shift funds out of banks in the case of emerging general credit risk fears towards the banking system (although already today customers can easily shift their funds from one bank to the next). As mentioned, some authors have perceived the occurrence of CBDC as positive for financial stability. For example, Dyson and Hodgson (2016) argue that CBDC “can make the financial system safer: Allowing individuals,

private sector companies, and non-bank financial institutions to settle directly in central bank money (rather than bank deposits) significantly reduces the concentration of liquidity and credit risk in payment systems. This in turn reduces the systemic importance of large banks. In addition, by providing a genuinely risk-free alternative to bank deposits, a shift from bank deposits to digital cash reduces the need for government guarantees on deposits, eliminating a source of moral hazard from the financial system.”

Below we show the creation of CBDC in a financial account system. For the sake of simplicity, we assume that claims to NFCs / the state are always the same, regardless of whether they are bonds issued or loans. The absolute numbers in the accounts could of course be changed/rescaled (or replaced by variables). The creation of CBDCs has been split into two parts: CBDC1 which substitute banknotes and CBDC2 which substitute deposits with banks. The accounts also reflect that the central bank would instead of increasing only its lending to banks, it would also increase its outright security holdings. In theory, one could imagine that the central bank would absorb e.g. corporate and government bonds from existing stocks of investors, and that this will make it possible for banks to issue new bonds that investors will happily take to close the gaps created by central bank purchases. This case is captured by the bond purchase flow S below. Then, if S=CBDC2, the eventual difference for banks would consist in being funded more through capital markets than through deposits (due to the introduction of CBDC).

Households Real Assets 6 Household Equity 14 Sight deposits 6 - CBDC2 Retail mortgage loans 2 CBDC CBDC1 + CBDC2 Banknotes 2 - CBDC1 Bank bonds 2 + S Corporate/state bonds 2 - S Corporates / state

Real assets 8 Bonds issued / loans 8 Commercial Banks corporates / state bonds/loans 8 Sight deposits of households 6 - CBDC2 Retail mortgage loans 2 Bonds issued 2 + S Central bank credit 2 + CBDC2 - S Central Bank Credit to banks 2 + CBDC2 - S Banknotes issued 2 - CBDC1 Purchases of Corp/state bonds S CBDC CBDC1 + CBDC2

It is important to note that bank profitability will have been undermined: (i) Central bank lending tends to be more expensive than deposits, which normally can be funded at less than the short term risk free interest rate. Therefore bank profitability will suffer and banks will have to deleverage to some extent. (ii) if the central bank decides to address this issue through purchases of securities, this does not help either: capital market funding is even more expensive than central bank credit. To compensate for the implied tightening of monetary conditions, the central bank may have to lower its policy rate, for a given inflation target and for a given growth rate. This would reduce the positive effects on central bank income. Another issue, also arising under the “plain money” proposal, is collateral scarcity of banks, because central bank credit has to be substantially increased. In this case, central banks who had so far a narrow collateral framework may have to broaden their framework to also accept non-Government securities and loans to NFCs as collateral to secure the enlarged structural credit provision. Recently, central banks have devoted growing efforts to analyse central bank digital currency, as documented by e.g. CPMI/MC (2018) and Sveriges Riksbank (2018).

Chapter 3: Central bank conventional monetary policy

3.1 Short term interest rates as operational target of monetary policy 3.2 Basic natural rate logic of monetary policy 3.3 Complicating the natural rate logic 3.4 Transmission of monetary policy from the operational to the ultimate target 3.5 Basic central bank techniques to control short term interest rates 3.6 Tender procedures, maturity structure and size for credit operations 3.7 Central bank collateral framework

3.1 Short term interest rates as operational target of monetary policy

The operational target of monetary policy is an economic variable, which the central bank wants to control, and indeed can control to a very large extent on a day-by-day basis through the use of its monetary policy instruments. It is the variable for which (i) the policy decision making committee decides the level in each of its meetings; which (ii) gives guidance to the implementation officers in the central bank what really to do on a day-by-day basis in the inter-meeting period, and (iii) serves to communicate the stance of monetary policy to the public. Today, there seems to be consensus among central banks that the short-term inter-bank interest rate is the appropriate operational target in normal times. In normal times, it is the sole variable that captures the stance of monetary policy.

Type of operational target: There are essentially three main types of operational targets: (i) A short-term interest rate (pre-1914, and post 1990 the dominating type of operational target, and in between also playing at least implicitly a key role). (ii) A quantitative, reserve related concept – officially in the US the operational target in the period 1920-1983 (in numerous variants, but never totally clear). (iii) A foreign exchange rate, for central banks which peg their own currency strictly to a foreign one.

Already Thornton (1802, 254), who is usually praised as the most advanced monetary policy theorist of the early 19th century, views central bank policy as “Bank rate” (Bank of England discount facility rate) policy, and analyses how Bank rate policy should be conducted. The idea that Bank rate needs to follow the real rate of return of capital in order to allow controlling the expansion of money and hence inflation, was probably first spelled out by him:

“In order to ascertain how far the desire of obtaining loans at the bank (the Bank of England) may be expected at any time to be carried, we must inquire into the subject of the quantum of profit likely to be derived from borrowing there under the existing circumstances… We may, therefore, consider this question as turning principally on a comparison of the rate of interest taken at the bank with the current rate of mercantile profit. The bank is prohibited, by the state of (usury) law, from demanding, even in time of war, an interest of more than five per cent, which is the same rate at which it discounts in a period of profound peace. It might, undoubtedly, at all seasons, sufficiently limit its paper by means of the price at which it lends, if the legislature did not interpose an obstacle to the constant adoption of this principle of restriction.”

A key point of Thornton is that the bank rate is always an adequate and sufficient tool of central bank policy to prevent over-issuance of money and hence inflation (except if the central bank is constrained in the use of this tool, e.g. by usury law). Thornton’s concept of a “rate of mercantile profit” indeed looks much like the “natural rate” of interest described in 1898 by Wicksell (1898/1936, p. 102) as follows:

“There is a certain rate of interest on loans which is neutral in respect to commodity prices, and tends neither to raise nor to lower them. This is necessarily the same as the rate of interest which would be determined by supply and demand if no use were made of money and all lending were effected in the form of real capital goods. It comes to much the same thing to describe it as the current value of the natural rate of interest on capital.”

The following arbitrage diagram (see also Richter, 1990), with two goods (wheat and money) at two points in time (today and tomorrow) and with the relevant relative prices, shows that, under stable prices, the rate of interest on money, i, has to correspond to the real rate of interest, r, which can be thought as independent from the “monetary sphere” of the economy. P1 is the price of wheat prevailing today, and P2 is the expected spot price of wheat tomorrow.

Arbitrage diagram between real and nominal interest, and inflation rate

Wheat P1 Dollar today today

1+r 1+i

Wheat Dollar P2 tomorrow tomorrow

(Note: this chart can obviously be expanded by one additional period – it then also allows to capture term premia t: (1+I1)^2=(1+i1)(1+E(i2))+t. The same decomposition can be applied to the long term real interest rate)

Arbitrage logic allows establishing some basic relationships between prices by moving within the diagram from one good to another via different paths. Indeed, the possibility to move from “Wheat today” to “Wheat tomorrow” through two different ways allows to establish the following arbitrage equation: 1+ r = P1(1+i) / P2

Defining (1+π) as P2/P1, this corresponds to the Fisher equation: (1+r) = (1+i)/(1+π)

The latter equation is approximately equivalent to i=r+π, i.e. in equilibrium the nominal interest rate should be the sum of the real rate and the (expected) inflation rate. Where does the real rate of interest come from? The real rate of interest is simply a relative price between “wheat tomorrow” and “wheat today”, which, like all relative prices for goods that can be transformed into each other, depends on the production function (intertemporal in this case) and on relative preferences (consumption today versus consumption tomorrow, from today’s perspective of course). As we may die before being able to consume tomorrow, we normally prefer consumption today to consumption tomorrow (or we may simply be impatient), while the intertemporal production function is normally producing positive returns. Hence, real interest rates are normally positive, although in recent years, negative real rates seem to have been observed (see the next sub-section).

Starting from an initial state in which actual and expected inflation corresponds to the central bank’s target and stable inflation expectations, E(πt+1) = πt = π*, the central bank could preserve this state if it manages to keep the money (nominal) interest rate, it, always equal to the expected real rate of return on capital E(rt) plus π*. Wicksel’s above quoted statement on the natural rate of interest assumes as starting point zero inflation expectations, so that the natural interest rate equals the expected real rate of return on capital goods, E(rt). If however inflation expectations are positive, then the relevant concept is the “non-accelerating interest rate”, which is the rate that is neutral not to the price level, but to the rate of change of the price level, and this rate is equal to E(rt)+E(πt). The expected real rate of return on capital goods, E(rt), will in fact vary over time, as its underlying factors will vary. Therefore, the central bank has to adjust also the nominal interest rate across time, if it wants to achieve stability of the inflation rate at its target level over time.

3.1 Basic natural rate logic of monetary policy The nominal interest rate it is contractually fixed at the point in time t and is the nominal interest rate on money covering the period [t,t+1]. The real rate rt covering [t,t+1] is not yet fixed at t, nor is πt = (Pt+1- Pt)/(Pt)) The general idea of the dynamics triggered by a perceived arbitrage opportunity is as follows:

• it > it* = E(rt) + E(πt) => it is profitable to sell real goods and hold more money investments => excess supply of real goods today => disinflationary impulse => actual inflation will fall below expected inflation: πt < E(πt)

• it < it* = E(rt) + E(πt) => it is profitable to buy more real goods for real investment projects, hold less money investments (or be short in money, i.e. borrow money), => excess demand for goods today => inflationary impulse => actual inflation will turn out to be above expected inflation: πt > E(πt) While this intuition is clear, it is not obvious to fully specify this dynamic process in a two point in time arbitrage diagram1. Modern macroeconomic monetary theory aims at capturing such dynamics. The central bank can choose its inflation rate target, and under normal circumstances can also achieve it in the medium term. For example, the central bank could conclude that π* = 2% is the optimal inflation rate (on models of the optimal rate of steady state inflation, see e.g. Schmidt-Grohe and Uribe, 2010, who conclude that low inflation targets such as today’s widespread 2% inflation targets are sensible).

An important constraint to the control of πt and the choice of π* is the zero-lower bound to nominal interest rates. Nominal interest rates cannot get significantly negative because banknotes can normally not have a negative remuneration – i.e. they always have a zero remuneration (see section 4.2). Therefore, if the central bank would try to set negative nominal interest rates, economic agents could undermine this by withdrawing all deposit money and holding it in the form of banknotes. Therefore, the choice of the inflation target π* must respect the constraint that rt + π* > 0. It is important to note that the real rate of return on capital rt can be negative, for instance if the economy no longer grows, but shrinks. In this case, a positive target inflation rate can “lift” the non-accelerating rate of interest into positive territory.

The problem created by the zero-lower bound to nominal interest rates is that it could be the reason for an economy to fall into a so-called “deflationary trap”. Indeed, in case the non-accelerating interest rate it* would be negative (it*=E(rt)+E(πt)<0), but the central bank cannot set a nominal interest rate (it) below zero, then monetary policy will be dis-inflationary. That means that inflation and inflation expectations will fall further, making zero interest rate policies even more dis-inflationary, etc. This is why some authors (e.g. Ball, 2014) have concluded that in a world of low growth dynamics and low real rates of

1 It is interesting to note that a similar issue arises in FX - interest rate parity arbitrage. The main difference is that the exchange rate is a price that reacts immediately (as it is set in the most liquid financial market), while the price level in an economy reacts sluggishly to news.

return, it is preferable to choose a higher inflation target (e.g. 4%) as a buffer against negative shocks that could push the economy into the deflationary trap.

3.2 Complicating the basic natural rate logic The equilibrium relationship above reflects a number of simplifications. In the real world, one needs to distinguish the ex-ante and ex post view, acknowledge exogenous shocks and sticky adjustment, and therefore give up the idea of simple and obvious dichotomy between the real and money sphere. Moreover, one needs to acknowledge that term premia are unstable, and that empirically decomposing nominal rates is challenging. Five issues can be distinguished.

(a) Different concepts of the real rate First, the system will most of the time be outside steady state equilibrium. Adjustment dynamics are non-trivial and will invoke more challenging economic modelling. Prices and real rates of return on capital are hit constantly by exogenous shocks. This implies that one actually needs to differentiate between the expected (ex-ante) and the actual (ex post) real rate of return on capital, E(rt) and rt (e.g. the actual rate of return on wheat is affected by the unpredictable weather conditions over the next 12 months). Moreover, when non-anticipated price pressures (relative to expected prices) occur, adjustment of prices is typically sticky. Amongst other things, this implies that the real rate of return on capital needs to be distinct from the real rate of return on money investments – in particular ex post. Indeed, the fact that ex ante it=E(rt)+E(πt) does not imply that ex post it = rt + πt. The real rate of return on money investments is equal to (ex post) it – πt. The real rate of return on capital is (ex post) rt. There is a third concept that needs to be distinguished, which is the ex-ante real rate of return on money investments which is it – E(πt). This is actually the most frequently meant concept when the term “real interest rates” is used in the media and academic papers. In an ex ante arbitrage steady state equilibrium, this should be equal to E(rt). However, in reality, ex ante adjustments to reach an arbitrage equilibrium may be imperfect and slow (e.g. the “time to build” argument), so that it is necessary to distinguish between the expected real rate of return on capital (E(rt)) and the expected real rate of return on money investments (it-E(πt)). The following table summarises the four concepts of real rates that eventually need to be distinguished, as they will, for the reasons mentioned above, not be identical (but in steady state equilibrium). Note the assumption that the nominal interest rate on money is identical ex post and ex ante. This holds as long as debtors do not default.

Four concepts of the real rate of interest Ex ante Ex post capital good investment E(rt) rt money investment it-E(πt) it- πt

(b) More than one real good Second, in reality there is not only one good (“wheat”) which is at the same time a consumption and an investment good, but there is a wide range of goods with very different properties. Investment goods are supposed to determine the real rate of return on capital, while consumer goods determine inflation. Consumer and investment good prices are eventually linked, but in reality, such links will be imperfect and lagged.

Third, nominal funding costs of the real economy are not identical to the short-term nominal interest rate that the central bank sets. Nominal funding costs of the real economy can be estimated by producing a weighted average of funding rates, the weights reflecting the share of that type of funding in the total funding of the real economy. For example, for the euro area, Table 4.5 of the statistical annex of each ECB Economic Bulletin contains a detailed split up of lending rates for new and outstanding loans to various obligor classes (household consumer credit, household mortgage loans, loans no non-financial corporates, etc.) with volumes known from the Monetary Financial Institutions (MFI) statistics. Corporate and sovereign bond yields can be collected from information systems such as Reuters and Bloomberg. The weighted average nominal lending rate of the economy can be thought to reflect three main factors: (i) The quasi-risk free short term interbank interest rate which is normally controlled precisely by the central bank; (ii) The term spread in the risk-free benchmark yield curve; (iii) The various instrument specific liquidity and credit risk premia. The challenge for the central bank is then no longer limited to the estimation of the expected real rate of return on capital goods only (such as to shift the nominal short-term interest rate across time in parallel to this), but in addition to adjust across time for the varying spread between the weighted average funding costs of the real economy and risk free short- term rates. Moreover, if the real rate of return on capital is low (as it is likely the case in a crisis), and in addition credit and liquidity spreads are high, then it is likely that the central bank will reach the zero lower bound with its operational target without being able to make monetary policy expansionary. To express this generalisation formally, define: • τ as the term spread summarising the slope of the risk-free yield curve, i.e. the difference between the risk-free rate at the average duration of real economic projects (say five years) and the short end of the risk-free yield curve. • λ as the spread between the weighted funding costs of the real economy and the risk-free yield with the same duration, i.e. capturing credit and liquidity premia.

What does τ and λ imply for the setting of short term nominal interest rates by the central bank? Does the central bank has to set it*= E(rt)+ E(πt)- τt -λt? It seems at least plausible that if e.g. in a liquidity crisis, λ shoots up significantly, it would need to be compensated by a corresponding lowering of it the short- term interest rate to ensure that monetary conditions remain unchanged.

The following chart illustrates this point for the euro area. As of 2012, the ECB seemed to have hit the ZLB and funding costs of euro area households and NFCs appeared stuck at around 3%, with credit and liquidity spreads clearly higher than at the beginning of 2008. But then, through unconventional monetary policies, the ECB managed to create a new trend of falling funding rates, benefitting from further declines in all three components (short term risk free rates, term spread, and liquidity and credit spread). The chart provides two risk free rates, namely the 10 year and the 6-year EONIA swap rate. These maturities correspond broadly to the average maturities of loans to household for housing purposes and loans to NFCs, respectively.

Figure: Borrowing costs of euro area NFCs and households and selected components

The following chart present the relevant credit and liquidity spreads: Loans to households minus 10 years OIS, and NFC loans minus 6-year OIS. The chart suggests that liquidity and credit risk premia have remained unusually high (say compared to 2006/2007) in the euro area. Before the break-out of the crisis in August 2007, they were mainly in the range 0.5% to 1%. They increased after Lehman towards around 2%, to drop again relatively quickly. In the context of the sovereign debt crisis, they increased again and stayed for five years in the range of 2% to 2.5%. This will have impacted on the neutral, i.e. non-accelerating central bank interest rate. Combined with low growth rates, it explains why the ECB had to undertake considerable non-standard monetary policy easing measures.

Figure: spread between on one side borrowing costs of households and NFCs and on the other side risk free rates of comparable maturity

In sum: funding costs of the real economy (i.e. households and NFCs) depend not only on the short-term risk-free rates as they are usually controlled by central banks, but also on term, liquidity, and credit spreads. These spreads vary over time, implying corresponding needs for the central bank to adjust the nominal interest rate even for a given real rate of return on capital (and given adequate and stable inflation expectations). A particular issue during financial crisis is the increase of liquidity and credit

spreads, i.e. bank intermediation spreads, which adds to the potential ZLB problem. However central banks can also influence spreads through non-conventional measures, namely forthcoming LOLR policies (to moderate credit/liquidity spreads) and long-term bond purchases (to moderate the term spread).

(d) Quantity constraints in credit markets Fourth, it has to be kept in mind that the actual availability of credit to the real economy cannot necessarily be measured by contemplating interest rates alone (e.g. Stiglitz and Weiss, 1981). Indeed, funding markets for some indebted companies can break down completely due to an increase of uncertainty and information asymmetries (see also chapters 6 and 7). The role of quantitative funding constraints has been recognised as a relevant element of monetary conditions by central banks, and therefore central banks have started to collect systematically survey data to be able to monitor this element of the transmission mechanism. For example, the ECB collects on a quarterly basis qualitative and quantitative bank lending data (see the “Euro area bank lending survey”, European Central Bank 2013d) and data on the access of SMEs to funding (“Survey on the access to finance of small and medium-sized enterprises in the euro area”, European Central Bank 2013c).

(e) Empirical estimation issues The expectations theory of the term structure of interest rate (longer term rates as a geometric average of expected short term rates) does not capture comprehensively movements of the long-term yields. Decomposing long term rate changes into expectations on short term rates and varying term premia is a demanding exercise requiring both modelling, assumptions taking and sufficient financial price data (see e.g. Abrahams et al., 2016). A consistent finding is that the variability of term premia for nominal interest rates can be of a similar order of magnitude as variations in expected future nominal short- term interest rates. The same holds for the inflation component of nominal rates (i.e. its split up into expected inflation and inflation term premia, or inflation risk premia) and for its real rate component (split up into expected real rates and real rate term premia). Term premia may be regarded as a residual item capturing, beyond the term risk premium, also other effects not captured by the measurement of expected short term rates. In any case, it is another important component to be kept in mind (and ideally to be well understood) when analysing the transmission mechanism of monetary policy, and the determination of the adequate level of the operational target of the central bank across time.

Estimating the real rate of interest. Laubach and Williams (2015) review different approaches for estimating real interest rates: • The simplest way would be to calculate past average values of ex post real rates (average nominal short term interest rates minus average inflation rates). to allow for persistent changes is to compute multi-year averages of past values. • More sophisticated statistical approaches use time-series filtering techniques that try to separate longer-term trends from short-term variations. • Yields on inflation linked bonds can be used to extract future expected real rates. However, forward rates include a term premium that contaminates the measurement of the market perception of the natural short-term interest rate. • The Laubach-Williams (2003) model uses a multivariate model that explicitly takes into account movements in inflation, output, and interest rates. The natural rate of interest is implicitly defined by the absence of inflationary or deflationary pressures. Recent estimates of the natural rate using the Laubach-Williams (2003) model suggest that the natural rate fell from a 1980-level of around 3.5% in the US and 2.8% in the euro area to levels below 1% and to below 0%, in 2015, respectively.

These five issues (a)-(e) are the reason for why the theory of optimal short term central bank interest rate setting is complex, diverse and inconclusive, and also why central banks have large economics departments. Modern New-Keynesian economics relies, as a starting point, on Wicksellian ideas, and tries to capture short term dynamics (e.g. Woodford, 2003, Gali, 2008). The Taylor-rule and New Keynesian model are based on frictions linked to sticky prices which are the driver of transitory dynamics. However, both are certainly too simplistic prescriptions of what is relevant in complex real- world environments, not to mention the true system dynamics in periods of financial turmoil and ZLB considerations. The New Keynesian agenda has received important extensions, qualifications and in some cases outright criticism from various perspectives (e.g. Cochrane, 2011).

Two extreme examples from German history in which the basic arbitrage logic was ignored In retrospect, one can identify episodes in which the central bank was obviously way off a reasonable interest rate policy, and thereby triggered fatal dynamics of the purchasing power of money. German monetary history of the 20th century provides two outstanding illustrations. Inflationary central bank interest rate policies are best illustrated by the application of the 5% discount rate by the Reichsbank from 1914 to 1922. Applying the arbitrage logic above, this discount rate was clearly far too low as of 1915. Indeed, the inflation rate reached 35% already in 1915 (essentially due to the extreme public demand shock associated with war mobilisation) and remained at similar or higher levels until it exploded in 1922 and 1923. The approach “(i) borrow money; (ii) buy and hold real assets, (iii) sell real assets in the future” was therefore a consistent profit-making opportunity without interruption for eight years. Actual inflation rates were certainly limited by the price controls for basic goods during the war years and would otherwise have exploded even earlier. Stolper (1940/1969) notes about credit provision during this period that (p. 83): “Persons who were resourceful and had the necessary banking connections to procure a maximum of commercial credit had nothing to do but invest the money without delay in ‘physical values’ in order to amass a gigantic fortune in no time at all. The most typical example of such practice and, in general, of the trend towards capital accumulation, was Hugo Stinnes…. He began to buy up at random and in large numbers the most varied businesses, including banks, hotels, paper mills, newspapers and other publishing concerns.” Hugo Stinnes, who was named “King of Inflation” in the Weimar Republic, died in 1924 as one of the richest and most influential German industrialist, with ownership in more than 4500 companies. His empire however ran into liquidity troubles when the Mark was stabilized, and it fell apart in 1925.

Reichsbank discount rate and inflation in Germany, 1914-1923 (Source: Bundesbank, 1976, 6) Reichsbank discount rates Consumer price Inflation 1915 5% 35% 1916 5% 33% 1917 5% 25% 1918 5% 38% 1919 5% 58% 1920 5% 113% 1921 5% 28% 1922 5%-10% 1025% 1923 10% >10^9 %

Deflationary central bank interest rate policies are exemplified by the maintenance of high nominal interest rates in the deflationary context of Germany in 1930-32, as shown in the following table. One may find various reasons why the Reichsbank kept discount rates so high despite deflation (namely defending the gold standard and convertibility of the Reichsmark as prescribed according to International Treaties like the Dawes Plan and the Young plan, despite capital flight and a debt overhang

due to reparation debt). However, having explanations for these interest rate policies does not change the conclusion that they were highly deflationary, illustrating the Wicksellian cumulative process in the opposite direction than during the period 1914-1923.

Reichsbank discount rate and inflation in Germany, 1929-1932 Reichsbank discount rates Consumer price Inflation 1929 6.5%-7.5% 0.0% 1930 4%-5% -5.5% 1931 7%-15% -9.3% 1932 4%-6% -10.4%

3.3 Transmission of monetary policy from the operational to the ultimate target We do not aim to cover monetary macroeconomics in any depth here, but only touch upon it very briefly to provide the link to monetary policy implementation.

Transmission of monetary policy The monetary macroeconomics literature distinguishes a number of so-called transmission channels of monetary policy, i.e. how changes of the operational target impact the financial system and the economy such as to eventually reach the ultimate target of monetary policy – say price stability. The most basic transmission channel that central bankers tend to have in mind today is the interest rate channel based on the Wicksellian arbitrage logic explained above: If it > E(rt) + E(πt) => πt ↓; If it < E(rt) + E(πt) => πt ↑. The following further transmission channels are often mentioned in the literature, whereby the first three channels are closely linked to Wicksell’s arbitrage logic.

• Exchange rate channel: it↓ => capital outflows => exchange rate ↓ => Exports ↑, Imports ↓, => Aggregate demand ↑ => πt ↑

• Equity/housing price channel (Tobin’s Q): it↓ => Value of discounted cash flows from asset ↑, Asset prices ↑ and therefore above replacement costs => Investment ↑ => Aggregate demand ↑ => πt ↑

• Wealth channel: it↓ => Value of discounted cash flows from asset ↑, Asset prices ↑ => Wealth ↑ => Consumption ↑ => Aggregate demand ↑ => πt ↑

• Balance sheet channel: it↓ => Asset prices ↑ => Equity of banks, firms and households ↑ => balance sheet constraints to expand activity ↓ => balance sheet expansion => Asset prices ↑ and aggregate demand ↑ => πt ↑. The theory and empirical assessment of transmission channels is the key issue of monetary macroeconomics. Deciding on interest rate changes relies on predictions of transmission, such as to achieve to the best possible extent the ultimate target across time. Non-conventional monetary policies will partially rely on the same transmission channels, but partially also on additional ones.

Monetary policy strategy Another key concept of monetary macroeconomics is the choice of the monetary policy strategy, which includes in particular choices relating to the ultimate target or monetary policy. • Single or dual mandate: for example, the ECB has a single mandate to achieve price stability, while the US Fed has, according to the Federal Reserve Act as amended in 1972, the statutory objectives for monetary policy of maximum employment, stable prices, and moderate long-term interest rates (although these seem to be three objectives, reference is made to a “dual” mandate). • How to specify more precisely the ultimate target: for example, the ECB decided that "Price stability is defined as a year-on-year increase in the Harmonised Index of Consumer Prices (HICP) for the euro area of below 2%." Moreover, the ECB has chosen to operationalize further this objective by aiming at an increase of the HICP of “close to but below 2%”, and to do so with a medium-term orientation.

Some, like Ball (2014), have suggested that it would be better to set the inflation target to 4%, at least in the possible new world of secular stagnation. • Monetary aggregates and intermediate targets: In the past, proposals have been seen (Milton Friedman, 1982) to make a narrow monetary quantity the ultimate target of the central bank. One variant brought forward by Friedman was the idea to set an open market operations operational and ultimate target. A somewhat less radical variant is what Deutsche Bundesbank defines as follows on its website: “Monetary targeting is a monetary policy strategy that aims to promote price stability through the intermediate goal of monetary growth. This strategy is closely associated with the name of the Deutsche Bundesbank, which pursued a policy of monetary targeting between 1975 and 1998.” The difference between the two approaches is that in the first, monetary aggregates are an ultimate goal, while in the latter it is a so-called “intermediate” target. • Exchange rate target: in case of an absolute currency peg, the ultimate target is the exchange rate, and all other variables, including the inflation rate, have to be accepted to fluctuate essentially according to the conditions in the anchor currency. In other words, in case of a currency peg, the operational and ultimate target collapse into one. • Nominal GDP targeting. At least since Clark (1994), nominal GDP targets have been considered as an alternative monetary policy strategy to inflation targeting. Recently, Williams (2016) has advocated nominal GDP targets as they would have a number of advantages in a world with lower growth and lower natural interest rates.

3.4 Basic central bank techniques to control short term interest rates Starting from the balance sheet logic explained above and adopting the interbank overnight rate as operational target of monetary policy, we now illustrate two basic techniques of short term interest rate control through monetary policy operations. One-sided standing facility based monetary policy implementation In this approach, the banking system takes systematic recourse to one central bank standing facility and the interest rate of this systematically used facility determines short-term interbank interest rates as a result of arbitrage. This is actually the simplest way to steer short term interest rates. Two variants are obviously possible, depending on whether the banks are kept systematically dependent on the borrowing facility or on the deposit facility provided by the central bank. To implement this technique in its first variant, to be called here ceiling approach, the central bank needs to do the following: First, taking into account the level of autonomous factors, ensure (if needed through liquidity absorbing outright operations, such as the issuance of central bank debt certificates) that the banking system has a systematic post-outright operations liquidity deficit vis-à-vis the central bank. Second, do not conduct any open market credit operations. Third, set the rate of the borrowing facility to the level of the intended policy target rate, i.e. iB=i*. In case of changes of the target, simply change iB accordingly. The following set of financial accounts represents this technique. It is key that for any possible fluctuation of d, RR+B+d will remain positive.

Financial account of one-sided standing facility based monetary policy implementation Household Real assets E-D-B Equity E Banknotes B +d Deposits bank D -d Corporate /State Real assets D+B Loans from banks D+B Bank Loans to corporate D+B Deposits of HH D – d Reserves of banks (incl RR) RR CB borrowing facility RR + B + d Central bank Borrowing facility RR + B + d Banknotes B + d Reserves of banks (incl RR) RR

This technique ensures that the interbank overnight rate will systematically correspond to the rate of the borrowing facility chosen by the central bank: Overnight interest rates cannot exceed the borrowing facility rate because it would always be cheaper for a bank to go to the borrowing facility than to the interbank market. Also, overnight rates cannot fall below the borrowing facility rate since at the margin the banking system needs to borrow at the latter rate, and hence no bank should be willing to lend at a lower rate. The approach relies on sufficiency of central bank eligible collateral, as otherwise the ceiling is not necessarily effective in constraining market rates. The implied permanent structural recourse to a liquidity providing standing facility may also be reduced somewhat by the central bank by holding an outright portfolio of securities (this could also be a way to address collateral scarcity problems). This approach was largely used by central banks in the 19th century, when banks had to take structural recourse to the discount facility and the discount rate determined market rates.

The second variant of this approach, to be called the floor approach, was used by the Federal Reserve or by the Bank of England after 2009, and is now considered by these two as the most likely new normal. To implement this technique, the central bank needs to do the following: First, taking into account the level of autonomous factors, ensure through sufficiently high liquidity injecting outright monetary policy operations that the banking system has a systematic post outright operations liquidity surplus vis-à-vis the central bank; Second, do not conduct any open market credit operations; Third, set the rate of the deposit facility (or the remuneration of excess reserves) to the level of the intended policy target rate, i.e. iD=i*. In case of changes of the target, simply change iD accordingly. The following financial accounts illustrate this approach (households are unchanged; the funding of the corporates/state is now a pool of bank loans and bonds issued). It is assumed that the central bank chooses the size of its outright portfolio SCB such that D + B > SCB > B + RR.

Corporate / State Real assets D+B Loans from banks /corporate bonds issued D+B Bank Loans to corporate D+B - SCB Deposits of HH D – d Reserves held with CB (incl. RR) RR Central bank deposit facility SCB – B - RR - d Central bank

Corporate bonds SCB Banknotes B + d Reserves of banks (incl RR) RR Deposit facility SCB – RR – B - d

Note that sometimes central banks have implemented one-sided facility approaches with two facilities offered in the same direction. For example, during the gold standard, central banks often steered interest rates between two liquidity providing facilities, with “Lombard rate > i > discount rate” and the Fed since 2015 has applied a similar floor system with “Interest rate on excess reserves” (IOER) > i > Reverse repo rate”. These systems require that the more attractive of the two facilities is somehow constrained in terms of access (Discount facility possibly through scarcity of available eligible paper, IOER through limiting access to banks, excluding non-banks). In any case, these are variants of one-sided standing facility-based systems.

Symmetric corridor approach with open market operations volume set by the central bank Under this approach, the central bank offers both a borrowing and a deposit facility, sets the rate of these two facilities symmetrically around the target interest rate and steers the scarcity of reserves such that there is an equal probability that at day end (or at end of reserve maintenance period), the banking system will need one or the other facility. This was the approach of the European Central Bank and many other central banks until the Lehman collapse in October 2008. Its idea is simple: the central bank injects through its open market operations, e.g. through sales or purchases of securities, an amount of reserves such that the aggregate banking system is equally likely at day end to be short or long of reserves. Then, the equilibrium interbank interest rate is the mid-point of the corridor set by standing facilities. Changes of the level of the target interest rate (monetary policy changes) are carried out by moving the corridor set by the standing facilities and the target rate (in its middle) in parallel up or down, while not changing the scarcity of reserves. To capture the technique more precisely, assume the following daily timeline of events, as also summarized in the figure below: 1) In the morning, the central bank adjusts its securities position S by means of open market operations, such that S=B+RR. B is the level of autonomous factors in the morning, RR are the required reserves. As S=B+RR, in the morning, the total bank reserves R will be equal to RR. 2) At mid-day, a trading session takes place, in which, in a competitive market (assume a high number of banks trading, with some of them short of liquidity and others long), the interbank rate is set as the weighted average of the two standing facility rates, the weights being the perceived probabilities of the banking system being short or long at day end. 3) In the afternoon, the true level for autonomous factors (B+d) is revealed, i.e. the stochastic variable d gets realized. We assume E(d)=0 and d having a symmetric density function. 4) Accordingly, the banks need to take recourse to one or the other facility.

Daily timeline of central bank operations and interbank trading

(2) Interbank trading session in which (4) End-day recourse overnight market rate i is determined to standing facilities

(1) CB conducts open market operation, i.e. adjusts its (3) Autonomous factor shock d securities holdings S

In the financial accounts, the symmetric corridor can be represented as follows.

Bank Loans to corporate D-RR Deposits of HH D - d Reserves with central bank (incl RR) RR CB deposit facility max(0, -d) CB borrowing facility max(0, d) Central bank

Securities (SCB) RR + B Autonomous factors B +d Reserves of banks (incl. RR) RR CB borrowing facility max(0, d) CB deposit facility max(0, -d)

How exactly will the interbank interest rate i be determined? The basic idea is that for risk-neutral banks, arbitrage requires that the overnight interbank market rate is equal to the expected end of day marginal value of reserves, which itself is a weighted average of the two standing facility rates. The weights are the probabilities associated with the needs to take recourse to either of the two facilities. If the banking system is “short” of reserves at day end, because of higher than expected banknotes in circulation, banks will have to take recourse to the borrowing facility. If the banking system is “long” of reserves at day end, because of lower than expected banknotes in circulation, banks will have to take recourse to the deposit facility. This arbitrage condition is summarized in the following equation.

i = P("short")iB + P("long")iD

= P(S ≤ RR + B + d)iB + P(S > RR + B + d)iD

= iD + P(S ≤ RR + B + d)(iB − iD )

Substituting S = B + RR allows to write: i = iD + P(0 ≤ d)(iB − iD )

If d is symmetrically distributed around zero, then: i = iD + 0.5(iB − iD ) = (iB + iD ) / 2

The recourse to the standing facilities will simply be equal to d, with the recourse to the borrowing facility being max(d,0) and the recourse to the deposit facility being max(-d,0). If one assumes that

d ≈ N(0,σ d ) , then one may also express the equation above as ( Φ() being the cumulative standard normal distribution):  S − RR − B  = + Φ−  − i iD  (iB iD )  σ d 

Based on this, one can calculate the effect on interest rates of S deviating from RR+B. If the central bank would like to implement an asymmetric corridor approach (say iD=0, iB=1 and i*=0.75), it could base its liquidity supply S on this logic. The figure below illustrates the relationship between i and S in case of RR=0; B=1 ; iD=0; iB=1. Assuming first that the central bank applies a symmetric corridor approach and that σd=1, the central bank needs to set its open market operations volume to 1 to achieve its operational target i*=0.5. If now volatility of autonomous factor shock e.g. doubles so that σd=2, then still an unchanged open market operations volume is adequate. This means that in a symmetric corridor approach, the central bank does not need to calibrate open market operations as a function of autonomous factor volatility, but only as a function of expected values of autonomous factors. Assume now instead that the central bank would be operating in an asymmetric corridor approach and would

aim at an interest rate of 0.8 in the [0,1] corridor. From the previous equation, we can derive the adequate open market operations volume S* to achieve this interest rate

−  i * −i  = + − σ Φ 1 D  S* RR B ( d )    iB − iD 

The central bank would therefore have to choose an OMO volume of 0.15 (“tighten liquidity”) to achieve its operational target i*=0.8. However, if autonomous factor volatility doubles, the central bank now needs to tighten further liquidity through a change of the OMO volume to -0.65 (i.e. the central bank now needs to conduct liquidity absorbing open market operations, e.g. issue debt certificates). This illustrates that any asymmetric approach requires the central bank to take into account, when calibrating open market operations, second order moments of autonomous factors, which is quite annoying and which explains why (i) central banks tended to apply a symmetric corridor approach; (ii) why central banks do not implement changes of their operational target level through a “liquidity effect” (i.e. change OMO volumes relative to autonomous factors) but through a parallel transfer of the standing facilities corridor. Some academic authors had imagined that changes of the interest rate target would be implemented through liquidity changes, i.e. through changing S (for example Hamilton, 1996).

Central bank “liquidity management” is the use of open market operations to control the scarcity of reserves of banks with the central bank in a way to have their price (the overnight interest rate) close or equal to the operational target level. In practice, a corridor approach needs specification along the following five key dimensions: (i) Reserve requirements of a certain size with averaging or no averaging (in the former case normally less than daily open market operations, in the latter case daily open market operations); (ii) Symmetric versus asymmetric setting of the operational target within the corridor (and associated to this, symmetric or asymmetric liquidity management); (iii) Width of the corridor (e.g. 50, 100, or 200 basis points); (iv) Open market operations: fixed rate full allotment or allotment amount set by the central bank; (v) Size of liquidity deficit of the banking system that is covered with central bank short term credit operations.

Under none of the three approaches presented above (ceiling, floor, symmetric corridor), interest rate changes are implemented through changes in liquidity conditions engineered through changes of reserve supply through open market operations.

Example: The Eurosystem applied until 2008 reserve averaging over a one-month maintenance period and weekly credit open market operations, with the overnight rate as an implicit operational target and standing facilities at +/- 100 basis points around the operational target. Until 2007, control of overnight rates was precise, but for the end of maintenance period volatility that can be seen in the chart as periodic upwards or downwards spikes. In 2004, the ECB changed the exact timing of the monthly reserve maintenance periods such as to normally exclude that decisions to change the operational target become effective within the reserve maintenance period, as this destabilized bidding in open market operations and the reserve fulfillment path. Also, the ECB in 2004 reduced the maturity of its weekly main refinancing operation from two weeks to one week, such that instead of two, always one such operation is outstanding. Overall, the ECB’s system was effective and sufficiently transparent and did not impair the money market. However, the end of maintenance period spikes could have been addressed with some innovations that the Bank of England applied in 2006 (narrow end of reserve maintenance period corridor and carry-over of reserve imbalances). After 2008, the ECB maintained a corridor approach, but the banking system was permanently in excess liquidity, so that overnight rate were most of the time close to the deposit facility rate. Also, the ECB moved the deposit facility rate into negative territory, which however per se made no difference.

Overnight interest rate control by the ECB, 1999-2017

3.5 Tender procedures, maturity structure and size for credit operations

Tender procedure for credit open market operations Central banks provide credit normally through some mechanism that ensures the desired control and equal but competitive access of counterparties. Over the 20 years of its operations, the ECB has conducted both fixed rate and variable rate tenders, with and without discretion in the allotment decisions. For studies of the design of central bank tender procedures for credit operations, see e.g. Välimäki, 2003; Ayuso, and Repullo, 2003, or Bindseil, Nyborg and Strebluaev, 2009.

Generally, a preference should be given to automated tender procedures, i.e. tender procedures in which the allotment decision by the central bank is an automatism. This makes the tender more transparent and predictable. There are two options regarding automated tender procedures2: In fixed- rate-full-allotment, the interest rate is set by the central bank and the banks get whatever they bid for (assuming that they have sufficient collateral). The difference to a standing facility is that the tender is only offered at specific points in time. The fixed-rate-full-allotment approach has been applied by the Eurosystem to almost all credit operations since 2008. The second automated tender procedure is variable-rate-fixed-volume as applied for example by the Eurosystem in all its longer-term refinancing operations (LTRO) between 1999 and 2008 (the tender volume is pre-announced before the bidding starts, and the marginal rate is therefore automatically set by the intersection between the vertical supply curve and the demand curve). The main advantages of the fixed-rate-full-allotment approach are: (i) Fixing the rate seems a logical consequence (in the intra-policy meeting period) if the overnight rate is the operational target. (ii) It is even simpler than variable-rate-fixed-volume (no need for bidders to think about at what rate to bid or about how other bidders will bid, no need for the central bank to predict liquidity needs and to calibrate allotment amounts). (iii) In case the new Basel III regulation makes banks’ demand for excess reserves unpredictable as some have argued, the fixed-rate-full- allotment approach allows banks to address autonomously the issue. Advantages of the variable-rate- fixed-volume approach are: (i) It solves a specific potential problem of the fixed-rate-full-allotment approach that banks cannot co-ordinate bids in a way to ensure that the aggregate bid matches the actual aggregated liquidity needs – even if a common forecast of aggregate needs is available. At least, the aggregate bid in fixed-rate-full-allotment operations will contain some noise term which will affect liquidity conditions (Välimäki, 2003). This problem will be material if the relative liquidity shocks are large and difficult to distinguish at the individual bank level from aggregate shocks. (ii) Some may argue that the fixed-rate-full-allotment approach makes life of banks too easy in terms of central bank reliance (however, the mechanisms against such excess reliance could be more targeted than introducing variable rate tenders for this purpose). (iii) Some authors have argued that the variable rate tenders are more market oriented and allow the central bank to extract more information on markets and banks. Overall it seems premature to conclude for short term credit operations in a universal manner in favour of one or the other automated tender procedures, although the relative advantages of the fixed-rate- full-allotment approach may have grown because of the difficulties for the central bank to predict the demand for excess reserves resulting from new Basel III regulation (see CGFS/MC, 2015). Variable-rate- fixed-volume tenders should in any case be used for longer term credit operations, i.e. those going beyond the forthcoming meeting of the policy decision making body.

2 Examples for tender procedures without automatic allotment (i.e. discretionary allotment) are: (1) fixed rate tender with ex ante uncertain allotment quota (i.e. central bank fixes allotment quota after receiving bids, somewhere below 100%), used e.g. by the ECB in its main refinancing operations in 1999 and 2000; (2) variable rate tenders with ex ante unknown allotment volume (i.e. central bank chooses volume, and hence intersection between demand and supply curves, after receiving the bids).

Frequency and duration of credit operations; number of credit operations outstanding in parallel An effective but parsimonious approach can probably be based on two regular operations. With some daily liquidity buffer available, this could be a combination of one week and three months credit operations (as the Eurosystem practiced it between 2003 and 2007, whereby the three-months operations were conducted on a monthly basis so that always three of them were outstanding). In the absence of a buffer, daily operations with overnight maturity are needed in addition. Relevant factors for choosing an overall time architecture of regular tender frequencies and maturities are: (i) Time series properties of autonomous factors and of their forecast errors. (ii) Total size of expected average liquidity deficit to be covered with credit operations. The larger the liquidity deficit, the more operations of decent size can be outstanding in parallel. (iii) Perception of roll-over risks by banks: if allotments per counterparty are considered uncertain (as it is the case under variable rate tenders) banks may have a preference for limiting roll over risks, suggesting benefits of keeping the size of single operations limited. In any case, parsimony should remain a guiding principle. Also, misconstructions should be avoided in which the tender sizes drift apart, as it can happen when overlapping credit operations (e.g. weekly operations with two-week maturity) are used to address temporary autonomous factor fluctuations.

Size of liquidity deficit of banks towards central bank to be covered by short term credit operations The size of the liquidity deficit of banks towards the central bank that needs to be covered by short term credit operations is determined by the sum of outright portfolio holdings and long-term credit operations, relative to autonomous factors. The larger the liquidity deficit to be covered by short term credit operations, the (i) higher the roll over risk for banks (unless the operations are conducted as fixed- rate-full-allotment, which minimizes roll over risks); (ii) the shorter (everything else equal) the average duration of central bank assets (with its consequences on the risk-return characteristics of central bank balance sheets and the central bank’s contribution to maturity transformation); (iii) the more broad- based the direct transmission of policy rates to market rates (which may be important in less efficient markets); (iv) the larger the buffers for short term changes in autonomous factors that ensure that short term credit operations remain of meaningful size; (v) the larger the potential for relative central bank intermediation within the LOLR function of the central bank. These various dimensions create trade-offs which need to be looked at carefully for any specific environment.

Amount of reserves provided via longer term credit operations The optimal amount of reserves provided via longer term credit operations should also be considered within specific trade-offs, and in particular: (i) for a given amount of total credit operations, banks perceive longer term operations to reduce roll-over risks; (ii) longer term operations may be appreciated by banks in the post crisis world as they support the compliance with liquidity ratios; (iii) a larger share of longer term credit operations in all credit operations reduces the flexibility of the central bank to react to (non-anticipated) large autonomous factor changes.

3.6 Central bank collateral framework We will encounter the central bank collateral framework further in the context of the central bank as lender of last resort (chapter 6). However, the central bank collateral framework is not only relevant in the context of the LOLR. It is also key to implementation of regular monetary policy, protecting the central bank from losses in the conduct of monetary policy, and it to some extent influences the price of financial assets and thereby potentially the allocation of resources in society, as Nyborg (2017) has recently claimed. Bindseil, Corsi, Sahel and Visser (2017) provide an “economic ABC” of the collateral framework and recall that central bank collateral issues have a long history, and their relevance,

including beyond the protection of the central bank, has been noted for a long time – also outside central banks. Consider few examples only from German history:

The lender of last resort (LOLR) literature, such as Bagehot (1873) and King (1933), contain various references to central bank collateral. But also outside the LOLR question, the importance of collateral frameworks for real economic activity and even foreign policy has been noted. Consider the following examples from Germany central banking history (obviously similar examples can be found for any other country):

1) Already the very first German charter for a public bank namely for Nürnberg in 1498, prescribes that lending must be against “collateral, guarantee and insurance” (in German: “pfand, pürgschaft und versicherung nehmen”, see Bindseil, 2018a)

2) Two hundred years later, the Regulation of the Leipzig Banco of 1698 would be more explicit by also specifying eligibility criteria for collateral, valuation, and haircuts (Bindseil, 2018):

• IX.1 (Eligibility): “The Banco can provide advances against all jewelry, gold, silver-tableware, tin, copper, etc. and various current merchandise, however subject to the conditions below.”

• X.2 (Haircuts): “On mobile goods the following loan values should be applied: for gold and silverware three quarters of the correct value but excluding labor cost; on usual merchandise one half, or in case their quality is confirmed two thirds; on non-usual merchandise and jewelry one third. In the case the jewels have good weight, the Banco may, depending on circumstances, grant him one half.”

• XI.1 (Valuation): “Every collateral needs to be valued by one or two persons that have been explicitly mandated to do so by the Banco and who have best expertise and need to have been sworn in under the Royal authority, and in particular for precious objects need to attest the value. …”

3) Again 200 years later, the role the “Lombardverbot”, the ineligibility of Russian bonds as collateral in Reichsbank Lombard loans (Kindleberger, 1984, 227) would lead to major capital flows between Russia, France and Germany and would be one step amongst many towards WWI. “In the German Foreign Office about this time, Herbert Bismarck, the Chancellor’s son, proposed forbidding bank advances on Russian securities as an act of economic warfare […] on 10 November 1887, an order was issued forbidding banks to lend on Russian securities – the famous Lombardverbot. Prices of Russian bonds fell further in Germany; some were bought back by Russian investors, a great many by French.” The Lombardverbot was eventually lifted only on 26 October 1894. It however had a lasting impact on foreign debt allocation within Europe and contributed to making French investors the by far largest group of victims of the ultimate Russian default arising with the Red revolution.

4) Hjalmar Schacht mentions in his book „Die Stabilisierung der Mark“ (1928, 162) a number of measures undertaken by the Reichsbank in 1925 to support long term credit to agricultural enterprises. One of these measures was that “the Reichsbank accepted a considerable additional number of covered bonds as collateral for Lombard loans of the Reichsbank and increased the maximum loanable amount from 50% to 75% [i.e. lowered haircuts from 50% to 25%]. The result was a surprising upwards movement of covered bond valuations, which supported the inflow of substantial additional new funds also from other investors to agriculture.” (see also the Annual Report of the Reichsbank for 1925, page 9).

5) A remarkable reference to central bank collateral policies can also be found in Hjalmar Schacht’s Harzburg speech of 11 October 1931, maybe the most political and most controversial speech of a central banker ever. Schacht (1953, 367) argues there that the Reichsmark would be “a currency that no longer serves the real economic activity, but to hide the illiquidity of our financial institutions and of the public sector […]. Out of fear to make the public nervous, one does not tell that the collateral portfolio of the Reichsbank now only consists to a very small degree of actually eligible bills.” It is of course an irony that the Reichsbank of 1931 is today remembered as having been too conservative and as having done the mistake, if any, to not lend freely enough to banks and to have thereby precipitated the banking crisis of July 1931. It is even more of an irony that Schacht himself was rather ruthless in extending asset eligibility rules under his second Reichsbank term (1933- 1939) and in instrumentalizing it for purposes of financing Germany’s massive rearmament program in the run up to World War II.

Central banks themselves have obviously been aware for a long time of the importance of the collateral framework, and some central banks had a tradition on transparency regarding both the eligibility criteria and the use of collateral. For example, in “Die Reichsbank – 1876-1900”, more than one third of 300 pages is devoted to collateral issues (see English translation Reichsbank, 1910). Lists of eligible securities have also been published by central banks for a long time (see e.g. Reichsbank, 1935). Also, the use of eligible collateral has been documented by central banks. For example, Reichsbank (1926, 82-83) contains a detailed listing of collateral used in Lombard credit on an annual basis from 1876 to 1924 (“Bestände an Lombardforderungen und ihre Verteilung auf die verschiedenen Unterpfänder”).

The economics of central bank collateral frameworks have been discussed for example in Chailloux et al. (2008), Bindseil (2013), Nyborg (2017), and Bindseil, Corsi, Sahel and Visser (2017). Key principles are (again, the role of collateral in the context of the LOLR will be treated in chapter 7): Sufficiency of collateral to implement monetary policy through credit operations, i.e. collateral scarcity should not lead to a distortion of rates or constrain the access of the banking system as a whole to the necessary amount of central bank credit; Sufficient access of the targeted parts of the banking system, i.e. need to ensure that collateral availability is such that it the business models of banks that the central bank would like to participate to central bank operations provide for assets that are eligible; the eligible asset set should only allow for assets fulfilling certain properties, like: legal certainty, minimum liquidity, simplicity, ease of pricing (though market prices or reliable theoretical prices); Avoid that the collateral eligibility premium is so high that collateral decisions could influence relative asset prices in a way to affect resource allocation in the economy: A larger collateral set supports a lower collateral eligibility premium and hence reduces the risks of such distortions. Further principles follow in the context of the LOLR role of the collateral frameworks as explained in chapter 7, such as: limit pro-cyclicality of central bank collateral, support an adequate amount of leveraging of the financial system, prevent moral hazard, etc. The following table shows how the eligible collateral set and its use by banks evolved in the case of the ECB with three data points. Around 14 trillion of securities were eligible collateral in 2017, against a balance sheet length of the euro area banking of around 30 billion and securities holdings of banks of around EUR 5 billion. Central government securities constitute more than 50% of eligible collateral. The use of securities as central bank collateral is not really proportionate to eligibility: use of Government debt is under-proportional, while the use of covered bonds and ABS is over-proportional.

Central Regional Asset- Uncovered Covered Corporate government government backed Other Total bank bonds bank bonds bonds securities securities securities Eligible securities end 2004 3,980 215 905 1,256 680 318 292 7,646 end 2011 6,036 407 1,882 1,537 1,224 981 683 12,751 2017 Q1 7,033 432 2,075 1,360 1,489 626 865 13,879 Pledged securities (use of collateral) end 2004 252 58 169 213 27 45 19 784 end 2011 255 82 269 288 96 358 58 1,406 2017 Q1 295 75 105 336 63 336 75 1,285

The risk control framework for central bank collateral essentially consists in the haircut schedule and possible limits on the use of certain types of collateral. Gonzalez and Molitor (2009) and ECB (2015) present methodologies for deriving a central bank risk control framework for credit operations, such as haircuts, daily valuations, and margin calls. The table below presents an excerpt of the ECB haircut scheme, showing 3 out of the 5 maturity buckets. The haircut scheme is a mapping of three features of each security into a haircut, namely (see ECB Press Release of 18 July 2013; ECB, 2011, Chapter 6): Rating: BBB rated assets have higher haircuts than A-AAA rated ones (assets with ratings below BBB are normally not eligible at all); Residual maturity: the longer the residual maturity of bonds, the higher the price volatility and hence the higher the haircut; Institutional liquidity category of assets: The ECB has established six such categories, which are supposed to group assets into homogenous institutional groups in terms of liquidity. Of course, any such grouping will be a simplification.

ECB’s haircuts (in %) for different securities classified in liquidity categories according to issuer types, for three buckets of residual maturity and for two rating classes (ECB, 18 July 2013, annex to press release) Category I Category II Category III Category IV Category V Categ.VI* Issuer types → Central Local Gvt Covered Unsecured ABS Credit Government debt; Jumbo bonds; corp. bank debt claims to Assets ↓ debt covered bonds bonds instruments corporates 0-1Y; A-AAA 0.5 1.0 1.0 6.5 10.0 12.0 3-5Y; A-AAA 1.5 2.5 3.0 11.0 10.0 21.0 ≥ 10Y; A-AAA 5.0 8.0 9.0 17.0 10.0 45.0 0-1Y; BBB 6.0 7.0 8.0 13.0 22.0 19.0 3-5Y; BBB 9.0 15.5 22.5 32.5 22.0 46.0 ≥ 10Y; BBB 13.0 22.5 27.5 37.5 22.0 65.0 *For credit claims: referring to nominal value, as applicable to most credit claims accepted by Eurosystem

Methodology for haircut determination. The haircut setting of the ECB (and presumably of other central banks) follows the principle of risk equivalence, i.e. post haircut, it should not matter from the central bank risk taking perspective which type of asset a bank brings as collateral. In case of counterparty default, the collateral submitted by that counterparty needs to be sold. This takes some time and, in case of less liquid markets, a sale in the shortest possible time would have a negative impact on prices. The ECB classifies each security in one liquidity category, which is associated with an orderly liquidation period, i.e. a period for which it can be assumed that the sales has no impact on asset values. The haircut should depend, on the price volatility of the relevant asset and on the prospective liquidation time. An additional haircut may address uncertainty regarding the initial value of the asset. The higher the haircuts (against valuation uncertainty before counterparty default or against value changes after

counterparty default), the better the central bank is protected, but the higher are the collateral needs for a given volume of central bank borrowing. This trade-off needs to be addressed by setting a certain confidence level against losses. The Eurosystem, for example, according to ECB (2004), set haircuts to cover 99 per cent of price changes within the assumed orderly liquidation time of the respective asset class. Later, it adjusted this method to cover with 99% the Expected Loss, which is the expected loss conditional on exceeding the 99% confidence level (i.e. haircuts were made more conservative).

Example: Assume that an asset has a four weeks orderly liquidation period, and that the four weeks 2 price change due to general volatility of the risk free yield curve is N(0,σM ); the uncertainty on the true 2 asset value at the pre-default valuation is N(0,σV ); the liquidation price uncertainty stemming from spread changes (if it a BBB asset, then the volatility of the BBB-AAA spread) and credit migration risks (the risk that the asset gets downgraded from BBB to e.g. BB with the associated price decline) is 2 N(0,σS ). Assuming independently distributed factors, total uncertainty on liquidation values will be 2 2 2 2 N(0,σM +σV +σS ). Call σT the variance of total liquidation value uncertainty of the asset. If the risk tolerance of the central bank has been defined as “preventing with 99% probability that the asset value at liquidation falls short of the last valuation post haircut”, then haircuts need to be set for each asset at -1 2 2 σT,iΦ (0.01), where Φ() is the cumulative standard normal distribution. If for example σM =4%; σV =2% ; 2 and σS =2%, then the adequate haircut is 6.60%. Some assumptions underlying the approach taken in this example deserve refinement in practice: (i) the assumption that the uncertainties affecting the eventual liquidation value follow a normal distribution; (ii) the assumption that the different factors are independently distributed; (iii); the assumption that past time series provide an accurate picture of the behaviour of these risk factors after a counterparty default (which often will occur in a general crisis situation).

Chapter 4 Unconventional monetary policy 4.1 Introduction 4.2 Negative interest rates 4.3 Non-conventional credit operations 4.4 Outright purchase programmes 4.5 Distinguishing between conventional, non-conventional, and LOLR policies

4.1 Introduction In chapter 3 the basic Wicksellian logic was introduced: accordingly, there is a “neutral” or “non- accelerating” short term risk free interest rate i*, such that if it > it* => πt ↓; If it < it* => πt ↑ or, in words, if the actual short-term risk-free rate is below the neutral level, inflation will increase, while in the opposite case inflation will decrease. In the most basic version, the neutral rate is simply the sum of the expected growth rate and the expected inflation rate, i.e. it* = E(rt) + E(πt). If however the key issue are the funding costs of the real economy, and not just an abstract risk free interest rate, then it is more correct to define the neutral interest rate as: it* = E(rt) + E(πt) –τ – λ, with τ being a measure of the term spread and λ being a measure of the liquidity and credit risk spreads between the average short term funding costs of the real economy and the short term risk free interest rate. In particular the latter will increase in a financial crisis beyond normal levels and needs to be addressed through an additional easing of monetary policy.

In a financial crisis and the associated economic slowdown and starting from a low to medium structural growth economy as Japan or Europe, expected growth will easily be zero or even negative. If in addition, credit and liquidity spreads increase by 100 or 200 basis points relative to normal level, such as it happened in the recent crisis, and expected inflation is also close to zero, then the neutral interest rate it* will be negative, meaning that an inflationary impulse will require either negative interest rates, or the combination of zero/negative interest rates with “non-conventional” measures that will exert downward pressure on τ and λ. Downward pressure on τ can be achieved through forward guidance3 (concretely: committing to hold rates low for long) and through outright purchase programs of long term fixed rate securities to compress term spreads. Downward pressure on λ can be achieved through so-called credit-easing measures, including purchases of less liquid and more credit risky securities, and strengthening the lender of last resort (LOLR) support to the banking system such as to reduce perceived funding liquidity risks of banks. In this chapter, such non-conventional monetary policies will be discussed, whereby policies relating to the LOLR will be dealt with in chapter 7.

Non-conventional monetary policy measures typically are considered to have some potential negative side effects, while short term interest rate policies have not (this is why non-conventional measures are not used when they are not needed, i.e. when it* > 0). The negative side effects are likely to be subject, for every measure, of increasing marginal intensity. This fact will also be the basis for combining optimally the different measures to achieve the one and only adequate overall stance of monetary policy. We can think of each non-conventional measure having (i) a fixed set up / transition cost (need to analyse, specify, decide, communicate new measure); (ii) an increasing marginal cost from “distortions” it creates. If such cost schedules are available for each non-conventional measure, and a certain degree of additional accommodation is needed when the ZLB has been reached, then the central bank can in principle derive an optimal set and combination of measures.

3 A more detailed treatment of forward guidance can be found in e.g. Filardo and Hofmann (2014), Campbell Evans, Fisher and Justiniano (2012), and Swanson (2017).

It appears that central banks have assessed the relative costs of the different unconditional measures differently: for example, the Fed and the Bank of England have not hesitated to conduct large scale asset purchase programs as of 2009 but have preferred to not try negative interest rates. In contrast, the ECB has taken a while before launching a true “quantitative easing” type of asset purchase program but did not hesitate so much move interest rates into negative territory. In the case of the ECB, the exact combination of QE [“EUR 2 trillion with average maturity of 9 years”] and NIRP [“-40 basis points for at least 2 years”] should reveal the relative assessment of the ECB on side effects. In theory, other combinations of the two measures (e.g.: “-120 bp NIRP for two years, no QE”, etc.) might have allowed the same total accommodation, but were not chosen. Of course, the negative side effects of non- conventional measures always depend on circumstances, i.e. may be different from one jurisdiction to the other, or from one episode to the other.

The reasoning above assumes that the choice and specification of non-standard measures can basically be mapped into one single number: the additional accommodation needed beyond the zero-lower bound. However, one may question this, and instead see non-trivial issues in the interaction of non- standard measures that imply that one cannot just add up the accommodation that each measure brings. Then, the art of combining non-standard measures, including their sequencing, becomes even more subtle.

4.2 Negative interest rate policy (NIRP) Four European central banks have applied NIRP in recent years, namely those of Denmark, Switzerland, Sweden and the euro area (for a survey of the implementation of NIRP by these central banks, see e.g. Bech and Malkhozov, 2016). In addition, the Bank of Japan introduced NIRP in early 2016. In principle, the rationale for applying negative rates under some circumstances is obvious from the Wicksellian logic above. Blanchard et al (2014, 8) believe that in 2008, the policy-adequate short term interest rate would have been as low as between -3% to -5%, i.e. if central banks would have been able to implement negative rates at these levels, the crisis would have been more short-lived (avoiding the large scale economic contraction and associated welfare damage) and further non-conventional monetary policies (such as large scale asset purchase programs) with their complexities and side effects would not have been needed. Strong supporters of negative interest rates as obvious policy tool are for instance Buiter (2009) and Rogoff (2016), who all also reflect on how to make negative rates possible. They put particular emphasis on solving the first and most obvious of the following four possible challenges relating to the implementation of negative intertest rates, namely the one arising from banknotes in circulation.

(1) Effective lower bound created by the zero remuneration of banknotes in circulation Could negative rates lead to an explosion of the demand for banknotes, as banknotes have a zero remuneration? Indeed, it could be argued that all economic agents (banks, investors, households) can escape negative interest rates by substituting negatively remunerated financial assets with banknotes (which have zero remuneration). This is a powerful and obvious argument against deeply negative interest rates, and the only solution to it would be do discontinue the existence of physical banknotes (or find a way to be able to impose negative remuneration on them). There is little doubt that this would be technically feasible and efficient in advanced economies. However, critics argue that this would create a tool for central banks to expropriate savers even better (by imposing negative interest rates) and that discontinuing banknotes would also destroy, a la George Orwell’s “1984”, the freedom that is perceived to be associated with anonymous payments. These arguments are likely to prevail for the time being in most countries, and therefore the issue of banknotes will likely limit the scope for negative interest rates to the levels reached over the last years, i.e. not lower than -100 basis points. This seems to be the level at which banknote demand could start to have substantial momentum and

undermine the effectiveness of negative interest rates. As in the case of a central bank digital currency undermining household deposits with banks seen in chapter 2, a ballooning of central bank money holdings of households would imply that banks would lose large amounts of deposits and become more and more dependent on central bank credit. This would deplete collateral buffers and could put banks under severe liquidity stress, making it unlikely that bank lending rates will decline, i.e. undermining the effectiveness of NIRP. Below, we will note the issue that banks may actually not want to pass on negative rates to household deposits to not trigger this run on deposits. However, then, banks’ profitability suffers, as discussed further under point 2 below. In any case, the banknote hoarding argument also applies e.g. to banks, who could, in an environment of excess reserves, such as prevailing typically in the negative interest rate countries, start to hoard cash.

In the financial accounts below, we assume that both households and banks started to hoard cash as a consequence of negative interest rate policies. Household could also do so to escape from negative yielding bonds. Efficient arbitrage would allow asset holders to fully escape from negative asset remuneration, fully undermining the transmission of negative central bank rates to asset yields. These financial accounts also show the case that the cash hoarding goes so far to switch back the banks’ excess reserves into a liquidity deficit, implying growing needs of central bank credit, eventually creating potential liquidity stress on banks.

Households Real Assets HE - D-B Household Equity HE Bank deposits D - x Banknotes B + x Corporates / state

Real assets D+B corporates / state debt (loans and securities) D+B Commercial Banks

Credit to corporates / state D+B -SCB Deposits of HH D - x Excess reserves with CB max(0,SCB- B – y – x) Credit from CB max(0,-(SCB- B – y – x)) Banknotes y Central Bank

Securities of corporates and state SCB Banknotes issued B + y + x

Credit from CB max(0,-(SCB- B – y – x)) Excess deposits of banks max(0,SCB- B – y – x)

(2) Effective lower bound created by the non-transmission of negative interest rates to the economy due to negative effects on the profitability of the banking system It has been argued that specifically negative rates undermine bank profitability and undermine the transmission of negative rates as banks would be unable to pass on negative rates to retail depositors. For example, Bech and Malkhozov (2016, 39-40) note: “The key exception in terms of transmission has been banks’ reluctance to pass negative rates through to retail depositors. This reaction was motivated by the concern, shared by some central banks, that negative deposit rates would lead to substantial deposit withdrawals. In Switzerland, banks have responded to lower lending margins in some business lines by adjusting other selected lending rates upwards. In particular, Swiss banks have raised the lending rate on mortgages, even as government and corporate bond yields fell in line with the money market rates … The Swiss experience points to a fundamental policy tension if the intention of negative policy rates is to transmit negative interest rates to the wider economy. If negative policy rates do not feed into lending rates for households and firms, they largely lose their rationale.”

A number of central banks have acknowledged the particular effects of negatively remunerated excess reserves combined with an unwillingness/inability of banks to pass on negative rates to retail deposits through an innovation in their operational framework: so-called excess reserves tiering systems which exempt a part of the excess reserves from the application of negative interest rates. The idea is that a tiering system would allow the combination of (i) negative rates still effective at the margin and therefore passed on to money and capital markets, and (ii) the exemption of parts of the excess reserves moderating negative effects on bank profitability that could weaken the effectiveness of NIRP. By disconnecting the two, the transmission of NIRP to bank lending rates could be improved, and the effective lower bound (“ELB”, at which further rate cuts are no longer effective in terms of lowering funding costs of the real economy) could be lowered. The need to further study tiering systems and to possibly have blueprints for them available for future possible use is therefore one concrete lesson from the recent (or ongoing) lower-bound experiences. If tiering systems allow pushing down the ELB by 25, 50 or even 100 basis points, then they really would make a difference in a world in which the ELB is considered a major macroeconomic issue (as prominent macroeconomists have argued).

(3) Can financial markets at all work normally under negative interest rates? Before the introduction of negative rates, there were some fears that money and other key financial markets can at all function with negative interest rates. As also noted by Bech and Malkhozov (2016, 37), steering short term interest rates into negative territory has not been particularly challenging, nor did financial markets change their behaviour in negative territory. One may add that the combination of NIRP and asset purchase programmes also pushed longer term bond yields into negative territory, e.g. for Switzerland, Japan and Germany for the entire risk free yield curve beyond 10 years. Again, there was no indication of negative effects on market functioning.

(4) General counterproductive effects of low/negative interest rates Finally, a number of critical authors have argued that central banks’ low (and by implication, also negative) interest rate policies are ineffective or, at the very least, have major negative side effects that central banks tend to underestimate. These authors also seem to suggest that acknowledging the problem of low interest rate policies could lead to the conclusion that central banks should increase nominal interest rates without delay. The main arguments are as follows. • Low interest rates would weaken the life-time income prospects of savers, and therefore lead to more saving and less consumption, and are actually negative for aggregate demand. • Low interest rates create bubbles and therefore contribute to creating the next crisis and undermining the efficiency of resource allocation. • Low interest rates and elastic central bank liquidity supply tend to weaken hard budget constraints because of their supportive effect with regard to funding market access for indebted companies, households and the state. They therefore would lead to zombification and low growth, creating a vicious circle. • Low nominal interest rates weaken bank profitability as they destroy the interest rate margin arising normally from the non-remuneration of sight deposits. With decreasing marginal returns of financial intermediation, this implies that a lowering of central bank rates towards zero necessarily leads to a wider intermediation spread and deleveraging, and therefore only reaches the real economy to a limited extent (essentially issue (2) above).

Bindseil, Domnick and Zeuner (2015) and others have refuted these arguments. The European Systemic Risk Board (2016) has devoted an extensive study on macroprudential issues related to low interest rates. Overall, it seems that issues arise if economic agents do not accept a new reality of low real and nominal interest rates and therefore either continue making unsustainable return promises to investors, or try, through unsound risk taking, to generate returns that are unrealistic. Also, if agents

did not see the low interest rate environment coming and therefore took positions (or run a business model) that in the low interest rate scenario undermine their solvency, a transition issue arises that needs to be addressed in a way that minimises damage for society while keeping in mind moral hazard issues.

In sum: negative interest rates may be viewed as an obvious continuation of Wicksellian interest rate policies when the neutral level of interest rates falls into negative territory, as it has become more likely in an environment with low growth potential and high central bank credibility as inflation fighters. In this sense, NIRP could be classified as a conventional monetary policy approach, reducing the need for non- conventional policy measures in the narrow sense with their possible more problematic side effects (such as large-scale asset purchase programmes). That NIRP is effective has been demonstrated by its strong effects on both capital market rates and bank lending rates (the latter at least in the euro area). At the same time, two “effective lower bounds” have to be acknowledged, namely (i) the one where banknote demand would explode; (ii) the one in which bank profitability would be undermined in a way that a further lowering of central bank interest rates no longer leads to decreases in bank lending rates, as partially observed in Switzerland. While the former has to do with storage and insurance costs of banknotes, the latter also depends on the willingness and ability of banks to pass on negative rates to different types of depositors and the amount of excess reserves that banks hold with the central bank at negative interest rates. While the two issues are partially linked (through the decision of banks to pass on or not negative rates to depositors), the two ELBs are not necessarily the same. Both ELBs could be overcome through a discontinuation of banknotes – which parts of society seem however to reject.

4.3 Non-conventional credit operations

Central banks have taken a variety of measures during the crisis to make their open market operations more convenient. Some of these measures relate to the lender-of-last-resort (LOLR) function, but even those are relevant from the monetary policy perspective, in particular in case the zero lower bound is close. Indeed, strengthening the LOLR implies to reduce funding stress to banks, which reduces pressure on them to deleverage or to increase the role of expensive funding sources, and therefore contributes to maintain the readiness of banks to provide credit to the economy at a moderate mark up to short term risk free rates (Bindseil, 2013 and chapter 6 here).

First, central banks have lengthened the duration of their lending operations to banks, with the ECB going as far as four-year credit operations. Banks may consider a sequence of short-term borrowings from the central bank as inferior, from a liquidity risk perspective, to one longer-term borrowing operation. Consider three reasons for this: (i) Banks could perceive as uncertain the conditions under which central banks will provide short-term funding in the future (rates, access conditions, etc.). (ii) Even if the central bank commits to keep conditions for short-term access stable, e.g. it commits to full allotment at a given rate for its short-term operations for the next twelve months, banks may, as a matter of principle, find revolving short-term central bank refinancing less certain than twelve-month refinancing. (iii) Banks may be subject to some liquidity regulation, which treats longer-term refinancing from the central bank more favourably.

Second, central banks have replaced auction procedures to allocate central bank credit with ‘fixed rate full allotment’ (FRFA) operations. The ECB has done so in October 2008 and ever since then has applied to this simpler allotment procedure, which has in particular the following advantages.

• It is even more automatic than the variable-rate tender with pre-announced volume. This is per se a positive feature, as automatism means simplicity and transparency and hence fewer potential mistakes by the central bank and the banks. • In a liquidity crisis, the reduction of banks’ uncertainty about the results of the tender assuages liquidity risk. • Aggressive bidding in variable-rate tenders is avoided and with it high and more volatile marginal interest rates, which could imply an undesired stance of monetary policy (and unintended signals). • The central bank no longer needs to estimate which allotment amount would ensure that market rates remain close to target rates. Carrying out fixed-rate full allotment tenders is almost equivalent to setting standing facility rate at the level of the target rate, with the only difference that an open market operation is not continuously open. This should bring short-term rates to the level of the fixed tender rates.

Third, central banks have widened the access of counterparties to their credit operations. When interbank markets break down, then those financial institutions are in trouble which are not central bank counterparties, but which could normally manage their day-to-day funding needs through credit operations with banks and capital market access. Allowing direct central bank access makes them independent from the functioning of interbank markets.

Fourth, central banks have introduced “targeted” credit operations which make favourable lending terms (or access in general) conditional on some desirable behaviour of banks, such as providing more lending to the real economy. The ECB has done this through its so called TLTRO I and TLTRO II operations, the Bank of Japan through its “Loan support programme” (LSP) and the Bank of England through its “Funding for lending scheme” (FLS).

Fifth, central banks have started to provide credit in foreign currency, notably in USD. The ECB and the Bank of Japan have done so since the end of 2007, with small interruptions. In case USD spot and swap markets are impaired, this ensures that banks have sufficient USD funding to meet their obligations in USD.

Finally, widening the central bank collateral set applicable to credit operations is also both a monetary policy and a LOLR measure, and will be discussed in more detail in chapter 6. However, as Bindseil (2013) argues, it is also an unconventional monetary policy measure as it supports the ability of banks to continue providing credit and lowers the intermediation spread between short term risk free rates and bank lending rates. At the ZLB, compressing this spread or at least counteracting its raise can be decisive to prevent the economy to glide into a deflationary trap.

4.4 Outright purchase programmes All major central banks at some stage of the crisis that started in August 2007 established outright purchase programs for financial assets. The following seven objectives of such measures can be identified.

(1) Anti-deflationary threat to “purchase all real assets in the world” Besides hyperinflation, the biggest trauma to the monetary theorist and the monetary practitioner is the deflationary trap, as experienced to some extent by Japan since the burst of its asset price bubble in the early 1990s, and as feared during 2002-2003 and again in the aftermath of the 2008 crisis and in Europe also in 2014-16. Even if an economy has fallen into the deflationary trap, and conventional interest rate policies (including NIRP) may have become ineffective, the central bank’s should be able to

purchase all assets of the world with the money that it can issue without constraints. At some stage, when the central bank launches such a universal purchase program, the other economic agents will become less willing to sell all their assets (including equity, commodities, etc.), they will require ask for prices, and hence the purchasing power of the currency will fall. This will be anticipated, and a credible announcement to purchase huge amounts of assets could immediately defeat deflationary expectations.

(2) Creating excess reserves and waiting for the money multiplier to kick in Large scale outright purchase programmes imply that large parts of the banking system end in a liquidity surplus position towards the central bank, such as it happened in the US and the UK from 2009 to 2013. This may be at least a welcome side-product of asset purchase programmes as it facilitates central bank liquidity management and the control of the overnight rate (which will be close to the deposit facility rate, or to the rate of remuneration of excess reserves), and as a situation of general excess reserves may support the perception of financial stability. Beyond these objectives, Excess reserves targets play a role in the “money supply” approach to monetary policy implementation, as promoted in the official communication of the Bank of Japan between 2001 and 2016.

(3) Bringing long term risk free yields down through purchases of long –term assets The transmission of monetary policy goes through longer term rates, as most economic decisions (e.g. building a house or a new plant) depend on longer term rates. Longer term rates reflect an average of expected short term rates, plus a term premium (according to the expectations hypothesis of the term structure of interest rates). There may be two particular reasons why the central bank may want to lower long-term risk-free yields: First, the central bank could consider that due to a crisis and implied financial market distortions, the term premium has been distorted and is too high. Second, the central bank may have reached the zero-lower bound in short term interest rates and may want to distort down through long term asset purchases the term spread, to reduce further the effective monetary conditions. This has been the more relevant argument in recent years and has been key to the Fed and the Bank of England programs that started in 2009. It could be argued that the most neutral such operation is purchases of Government paper at longer maturities. Alternatively, it could be argued that it is more neutral to buy a broad set of long term debt instrument from different issuer classes, maybe proportionally to outstanding market capitalisation.

The Bank of Japan switched to defining its operational target of its outright purchase programme in 2016 toward controlling the 10-year interest rate. According to its press release of 10 September 2016: “The Bank decided to introduce "QQE with Yield Curve Control" … The guideline for market operations specifies a short-term policy interest rate and a target level of a long-term interest rate. … The Bank will purchase Japanese government bonds (JGBs) so that 10-year JGB yields will remain more or less at the current level (around zero percent). With regard to the amount of JGBs to be purchased, the Bank will conduct purchases more or less in line with the current pace -- an annual pace of increase in the amount outstanding of its JGB holdings at about 80 trillion yen -- aiming to achieve the target level of a long-term interest rate specified by the guideline.”

(4) Compress undue credit and liquidity spreads In a financial crisis, risky assets’ prices may be unduly depressed due to asset fire sales and the absence of optimist buyers looking for bargains. More generally, the usual arbitrage between asset classes may no longer work because of liquidity and capital constraints, systemic uncertainty, and self-fulfilling fears. In such an environment, the central bank can through purchases of the unduly depressed assets directly affects parts of the universe of yields relevant for monetary policy transmission, and thereby ease funding costs and constraints. Of course, central banks should not compress spreads below a

“reasonable” risk premium in view of the higher credit risk and lower liquidity of a security. Assessing what is an appropriate spread is of course challenging.

(5) Taking risks into the central bank balance sheet and easing capital constraints of banks The central bank may reduce total risk in banks’ balance sheets by buying outright credit and market risky assets from them. Therefore, if banks feel constrained in terms of economic or regulatory capital, outright purchases by central banks may attenuate these constrains and thereby influence positively their lending behaviour and in this sense ease effective monetary conditions felt by the real sector. In deciding on asset purchases, the central bank should keep in mind that it is unlikely to be competitive in assessing complex credit risky assets. Therefore, while accepting some credit risk, it may want to take it via relatively straightforward assets, such as corporate bonds, commercial paper, equity, conventional standard asset backed securities (e.g. plain vanilla residential mortgages backed securities) or covered bonds, and also diversify sufficiently.

(6) Substituting banks’ illiquid with liquid assets to improve funding liquidity of banks Purchasing illiquid assets outright improves liquidity of banks, in particular if these assets were previously not eligible as central bank collateral, or only at a high haircut.

(7) Directly supporting the funding liquidity of the issuing credit institutions or the real sector By purchasing in the primary market bonds from issuers (unsecured bank bonds, covered bank bonds, corporate bonds, etc.), the central bank can support in the most direct and obvious way the funding liquidity of these institutions. Central bank purchases of NFC debt, if done in the primary market, directly refinance the real sector and thus can, at least partially, offset the unwillingness of banks to provide their usual lending and liquidity services. A general problem with the direct refinancing of the real sector is the limited knowledge of the central bank about commercial banking business. If banks do no longer lend and the central bank has to take over this function, then the human expertise in banks on such credit business is idle, while the central bank lacks this expertise.

Impact of purchase programmes on yield levels There is a growing empirical literature that tries to estimate the effects of large scale asset purchase programmes on the risk-free yield curve and its further transmission to other interest rates and the real economy (a comprehensive recent study covering the programmes of the US, UK, Japan and the euro area is Agostini et al, 2016). Effects on long-term interest rates of recent large scale asset purchase programmes are generally believed to be in the area of up to 100 basis points. In combination with NIRP, this would mean that these two policies together would have brought (in the case of the euro area) effects on long term funding rates of up to (or even above) 140 basis points, which obviously means substantial further easing (NIRP also contributes to reduce long term rates as expectations on future short-term interest rates decrease).

When looking more precisely at the effects of purchase programmes on asset prices and long term yields, it is important to distinguish between the following three effects (D’Amico and King, 2011, were the first to investigate theoretical and empirical aspects of flow- vs stock effects of the US Fed’s asset purchase programmes): - Stock effect: If we assume that there are static demand and supply elasticities for different types of securities (based on investors’ static preferred habitats), then one would expect that the eventual stock of securities purchased in a programme will determine the price impact. - Announcement effect: As asset prices in principle reflect at any moment in time all available information, it can be expected that most of the impact on prices and yields materialises immediately when the central bank announces an asset purchase program. The announcement

effect should be an anticipation of the stock effect, and not of the flow effect. The announcement effect will mainly depend on (i) the degree to which the announcement has not been anticipated (for example, when the ECB’s PSPP was announced, markets hardly moved as it had been anticipated); (ii) the credibility of the central bank (determined e.g. by its history to meticulously implement what it promises); (iv) how remote in the future the promised measures are (with non-perfect central bank credibility, more remote measures will have a lesser announcement effect than measures which are relatively nearby), (v) the clarity of the announcement. - Flow effect: if one perceives the price of an asset to be driven essentially by the daily demand and supply conditions and if we believe that agents’ ability to bridge prices across time through intertemporal arbitrage is limited, then one should expect that the daily flows of purchases would drive prices. The strength of flow effects will therefore depend on (i) the pace of purchases (purchased volume per unit of time); (ii) the efficiency and flexibility of market makers and investors to do intertemporal arbitrage and warehouse positions accordingly; (iii) the speed at which investors are able or willing to adjust their stocks, which also depends on who in particular holds the assets (a pension fund vs a bank in its trading book); (iv) the time between the announcement of the programme and its start (more time allows investors to prepare for selling assets and dealers to accumulate stocks waiting for the central bank).

Buying with too short lead times and at a too high pace distorts markets, in the sense of letting yields undershoot temporary more than necessary. It also implies that the central bank will over-pay. Buying with two long lead times and with a too low pace delays unnecessary the desired easing of financial conditions. Interestingly, in the case of limited central bank credibility, stronger flow effects may be more positive in the sense that they contribute to a faster adjustment of prices towards the new equilibrium price, i.e. faster effectiveness of monetary easing, without this implying that the central bank purchases above equilibrium prices. In this sense, a less credible central bank should buy at a higher pace and start quicker than a credible central bank.

4.5 Distinguishing between conventional, non-conventional, and LOLR policies

Central banks may, despite the presumption that non-conventional measures have negative side effects and conventional measures have not, tolerate that the accommodation provided by non-standard measures is more than implied by the existence of the ZLB: For example, the Fed’s policy normalisation has consisted in first withdrawing accommodation through a number of standard interest rate increases, before starting to do so via reducing its stock of “QE” securities. This would be preferable because rate hikes would be easier to dose than the impact of changes of QE stocks on the stance of monetary policy. Moreover, it would be easier to switch direction with regard to changes of interest rates, than in terms of changes to a QE securities stock. Also, in 2007 and 2008 central banks undertook various non- standard measures without having yet reached the zero-lower bound. Those may have related to LOLR measures, which may be beneficial for society regardless of having reached the ZLB or not (because the LOLR may save viable and solvent, but temporarily illiquid projects).

While the LOLR will only be discussed in chapter 6, it is interesting to already here try to put order into the three types of policy reasons for central bank market operations. The following chart puts the LOLR into context with conventional and non-conventional monetary policies and assigns central bank financial operations and instruments to any of the three (overlapping) areas. The different types of

operations are inscribed in the circles in different colors: Credit operations are in normal font (including collateral), purchase programs are underlined, and interest rate policies in white.

Short term interest rate control NIRP Colla-- Emergency teral “QE” APP liquidity assistance Credit easing TLTRO APP FRFA

- NIRP (negative interest rate policy) can be classified as “conventional” monetary policy, as it is in some way just a continuation of central bank short term interest rate policies. Still, it has something unconventional as it had never been done before 2013. - FRFA (fixed rate full allotment – see below) is both a LOLR and an unconventional monetary policy instrument. It reduces banks’ funding stress, and thereby also supports their willingness to lend. - Credit easing purchase programs could have more or less LOLR content, depending on how acute the funding challenges of the securities issuing industry are. - Collateral: is the one and only element in the intersection of the three circles: it is necessary to conventional monetary policy credit operations, and in exceptional circumstances (liquidity crises and/or zero lower bound problem), broadening the collateral set supports funding liquidity of banks, which attenuates the crisis and supports the banks’ lending. - Control of short term interest rates is the classical pure form of conventional monetary policy - TLTRO (targeted longer-term refinancing operations) and QE (“quantitative easing”) types of asset purchase programs are pure non-conventional monetary policy operations. - ELA (Emergency liquidity assistance) is by definition outside monetary policy. At the same time, ELA may prevent contagion of a narrow liquidity issue to the rest of the financial system, which would have repercussions for monetary policy transmission. In this sense, ELA decisions may often be non- neutral for monetary policy.

Chapter 5: Financial stability

5.1 Introduction 5.2 Default probability and cost of default 5.3 Illiquidity because of adverse selection and market breakdown 5.4 Bank runs and investor strikes 5.5 Increase of haircuts (“margin”) in collateralised financial market transactions 5.6 Increase in bid-ask spreads in a dealer market and fire sales costs 5.7 Interaction between crisis channels

5.1 Introduction A financial crisis is normally associated with a liquidity crisis, which has typically a funding and a market dimension. In a funding liquidity crisis the willingness of households and investors to maintain exposures to indebted entities is destabilised, triggering funding stress on economic agents who depend on external refinancing. In a market liquidity crisis, turnover in securities markets drops, bid-ask spreads go up and much higher discounts have to be accepted when selling assets, in particular if sales volumes are large and if sales have to be conducted rapidly because cash is urgently needed. Typically, asset liquidity and funding liquidity crises appear in tandem, which will be explained below. Liquidity crises may lead to the inability of debtors to fulfil their contractual obligations and hence to their default, with additional economic damage. Most authors analysing financial crisis highlight the role of strong downward revisions of asset values as crisis trigger (from Bagehot, 1873, to e.g. Kindleberger and Aliber, 2011). In the crisis that started in August 2007, the inversion of an unsustainable trend in real estate prices and related securitisations (and associated malpractices) in the US played this role. Typically, the following changes of asset prices have played an important role to trigger financial crisis: • Housing prices: houses typically constitute a major component in the wealth of households. While on aggregate, the household sector is moderately leveraged, a significant share of households tends to leverage when purchasing real estate. In particular, during boom phases, leverage ratios (loan to value ratios) become high and are built on the assumption of ever increasing prices. In history, housing pricing bubbles have been repeatedly observed and in particular recently. • Commodity prices: some emerging market economies and industries are based considerably on commodity extraction and export. The high volatility of commodity prices accordingly triggered at various occasions financial crisis, in particular in emerging market economies. • Changes in interest rates: the valuation of many assets is based on the discounting of expected cash- flows. Therefore, strong increases of long term interest rates may lead to significant downwards revisions of asset prices. Of course, this may also depend on the drivers of the increase of interest rates and the nature of the assets. If for example the increase is only due to a generalised increase of inflation expectations, then it may be neutral for asset prices if the cash flow generating capacity of the firm’s assets growth proportional to the higher nominal rate used for discounting. If however the assets of the firm consist in long term fixed coupon bonds, then a positive inflation shock reduces asset values. For indebted entities, solvency issues can also be triggered by changes of refinancing costs implied by increases of the interest rate level. This applies in particular if asset values are not or negatively affected by the change of interest rates, while funding has been done through floating rate instruments (or roll overs short term fixed rate instruments). • Re-assessment of technologies and competition: The value of a firm’s asset may be based on the outlook of the sales prospects of its products. In particular, in dynamic growth industries, bad news can therefore trigger significant downwards revisions of asset values. The burst of the tech bubble in 2001 was one such example.

Strong declines of asset values have various negative effects on economic agents. Solvency declines, with various implications as described in more detail in this chapter. Lower solvency can undermine the ability to access funding sources and the willingness and the ability to undertake risky projects. In the case of banks, an additional factor comes into play: a decline in capital puts at risk compliance with capital adequacy regulations, adding urgency to deleveraging through the shrinking of lending or through asset fire sales. In the case of household, lower wealth reduces consumption, which will have recessionary effects.

A number of historical quotes, which keep all their relevance, illustrate the mechanisms of liquidity crisis and their similarity across time. Already Thornton (1802) noticed the problem of liquidity hoarding and bank runs, and how they relate to a lack of trust.

“That a state of distrust causes a slowness in the circulation of guineas, and that at such time a great quantity of money will be wanted in order to effect only the same money payments, is a position that scarcely needs to be proved… When a season of extraordinary alarm arises, and the money of the country in some measure disappears, the guineas, it is commonly said, are hoarded. (p. 99)”…“If any one bank fails, a general run upon the neighbouring ones is apt to take place which if not checked in the beginning by pouring into the circulation a large quantity of gold, leads to very extensive mischief” (p. 180).

Bagehot (1873, all of the following from chapter VI “Why Lombard Street Is Often Very Dull, and Sometimes Extremely Excited”) argues that while liquidity crises can be triggered by various exogenous events, their consequences tend to be similar:

“Any sudden event which creates a great demand for actual cash may cause, and will tend to cause, a panic in a country where cash is much economised, and where debts payable on demand are large. …. Such accidental events are of the most various nature: a bad harvest, an apprehension of foreign invasion, the sudden failure of a great firm which everybody trusted, and many other similar events, have all caused a sudden demand for cash. And some writers have endeavoured to classify panics according to the nature of the particular accidents producing them. But little, however, is, I believe, to be gained by such classifications. There is little difference in the effect of one accident and another upon our credit system. We must be prepared for all of them, and we must prepare for all of them in the same way—by keeping a large cash reserve.”

Bagehot also highlighted the systemic nature of liquidity crises:

“Most persons who begin to think of the subject are puzzled on the threshold. They hear much of 'good times' and 'bad times,' meaning by 'good' times in which nearly everyone is very well off, and by 'bad' times in which nearly everyone is comparatively ill off. And at first it is natural to ask why should everybody, or almost everybody, be well off together? Why should there be any great tides of industry, with large diffused profit by way of flow, and large diffused want of profit, or loss, by way of ebb?”

This chapter provides a set of simple partial models of liquidity crises. Chapter 7 adds the lender of last resort issue in more detail, which is of particular relevance for the central bank. Understanding the logic of liquidity crises is a precondition for understanding the role of the central bank in stopping the escalation of liquidity crises and in addressing their economic consequences. Typically, financial crisis imply an increase of financial intermediation spreads and an economic recession, which both will imply, as explained in chapter 4, a need to lower central bank interest rates in order to prevent disinflation. Quickly, financial crisis will push central banks to their zero lower bound and therefore also require the conduct of (costly) non-conventional measures. Therefore, alertness to financial crisis and readiness to act forcefully to address their monetary policy implications will be crucial for central banks, beyond the narrow LOLR issue treated in chapter 7.

5.2 Default probability and cost of default The default risk of borrowers is opaque, and it is the core competence of banking to assess and monitor credit risk. Credit risk increases in financial crisis when the value of assets owned by the borrower declines and/or becomes more volatile. Consider the following leveraged corporate or financial institution, with ε being a random variable impacting asset values.

A leveraged firm Assets Liabilities Assets A+ε Debt D Equity E+ε

While the level of debt is given, asset values are subject to a random shock depending on various uncertain events. Assuming that ε is a normally distributed random variable over the relevant horizon with expected value 0 and standard deviation σε, then the probability of default (PD) of the company, in the sense of the probability that its asset values will be below the value of debt, could be estimated as (with Φ( ) being the cumulative standard normal distribution):

 A − D  PD = P(E + ε < 0) = P(A + ε < D) = Φ−   σ ε  For a number of reasons, this is a strong simplification. First, asset values are not really normally distributed. Second, variables such as σε are not directly observable. Third, time is continuous and there is no unique horizon to consider. Fourth, it is not clear that default occurs exactly when A+ε touches D, as in fact default is eventually triggered by illiquidity. Modelling default risk while taking into account all of these issues has been undertaken e.g. by Crosbie and Bohn (2003) applying Merton’s structural credit model (Merton, 1974).

Rating agencies provide comprehensive statistics on how their population of rated debtors actually performed. For example, Standard & Poors publishes regularly a statistical default study (e.g. Standard & Poors, 2013). Accordingly, the following default rates applied in the period 1981 to 2013:

Annual AAA AA A BBB BB B CCC default rates Min 0.00 0.00 0.00 0.00 0.00 0.25 0.00 Max 0.00 0.38 0.39 1.01 4.22 13.84 48.94 W.average 0.00 0.02 0.07 0.21 0.80 4.11 26.87

Differences between good years and bad years are obvious, and the time series of annual default shows a sort of financial cycle with peaks in default frequency around 1991, 2001, and 2008. The following table shows the rating distribution of corporate issuers (i.e. including banks) at end 2007 and at end 2016; for 2016, also a split up into banks and NFCs is provided.

S&P Global corporate rating distribution, end 2007 and end 2016 (source: S&P global ratings reports) Rating: AAA AA A BBB BB B CCC/C 2007 2% 9% 22% 24% 18% 23% 2% 2016 - all 0.2% 5% 20% 26% 19% 27% 3% 2016 – Banks (1323 names) 0.3% 9% 25% 31% 18% 15% 2% 2016 – NFCs (4779 names) 0.2% 2% 12% 26% 22% 34% 5%

One may note that the business model of banks seems to suggest a somewhat higher average credit rating than the business of NFCs. For example, Fitch (Fitch credit outlook 1Q17) provides an overview of ratings of sovereigns, weighted by outstanding debt. Accordingly, 36% of global sovereign debt rated by Fitch was in the range A-AAA, 21% in the BBB area, 16% BB and 27% at B or CCC level.

It is also important to consider what losses debtors make in case of default. If all debtors rank pari- passu, and if the default point would be exactly equal to the point when A+ε=D, then the “Loss-given- default” (LGD, = 1 – recovery ratio) of investors should in fact be equal to zero. However, in reality, rating agencies typically mention out of their empirical studies LGDs of around 50% on average (e.g. Moody’s, 2008 provides LGDs for secured, unsecured senior, and subordinated bonds of 40%, 63%, 69%, respectively, considering a sample of around 1800 bonds that defaulted between 1989 and 2007). This implies that either default on average occurs only when A is already clearly below D (i.e. creditors typically would not notice in time that the company has a solvency problem), or that the default event itself is costly, which is indeed plausible since default typically means that organisational and human capital gets lost and specific physical assets need to be liquidated at fire sale prices (e.g. a sophisticated machine may have to be sold at its raw material value).

Empirical estimates of default costs in the corporate finance literature vary between 10% and 44% (see e.g. Glover, 2011, and Davydenko et al 2012). The fact that default is costly is one crucial aspect in assessing the merits of the central bank to prevent default due to illiquidity. We will see below that in settings of asymmetric information which are typical for the high uncertainty prevailing in financial crisis, credit markets can break down such that default can occur even when A > D. In these cases, it seems plausible that the central bank can contribute to social welfare if it assures liquidity.

Higher credit riskiness due to lower capital has been identified for a while as an issue for monetary policy transmission in the literature on the “credit channel”. Lower capital of corporates means higher agency costs in the lending between banks and corporates. Lower capital of banks means higher agency costs between holders of bank liabilities and banks. Higher agency costs result from the fact that the alignment of incentives between debt and equity owners is likely to suffer for lower levels of capital (e.g. Holmström and Tirole, 1997). Moreover, indebted corporates and banks will aim at overcoming the implied frictions in accessing credit by aiming at deleveraging, also causing economic contraction.

Consider the following four key concepts of debtor trouble and how they relate: • Defaulted: a missed payment obligation • Illiquid: inability to identify money for fulfilling (forthcoming) payment obligations • Insolvent: debt exceeds assets, and therefore equity is negative • Over-indebtedness: Relating to insolvency but less linked to a strict threshold. Overindebted companies may still have positive capital, but insufficient capital to grant them healthy and sufficiently cheap market access, so that in the medium term an insolvency/illiquidity issue looms.

A corporate can default despite being solvent when it is illiquid. It may be illiquid because it needs to roll over debt but because of a systemic liquidity crisis situation all possible lenders stopped lending (and want to hoard cash). A corporate may be insolvent but does not default yet, because no debt comes due and needs to be rolled over yet. Or it may be insolvent but creditors did not realise this yet and are willing to roll over their claims. Eventually, a corporate which is insolvent will very likely end up being illiquid and will eventually default because it does not make sense for creditors to give fresh loans to an insolvent debtor.

Debtors can also be solvent conditional that they do not face funding constraints forcing them to undertake asset fire sales, but they may be at the same time insolvent conditional on having to undertake asset fire sales of x (at some specified horizon). This reflects the problem explained in section 1.2 that fair book values of assets are normally higher than their (short-term) liquidation value. Assume that an indebted company has total assets of 1, and that these assets are ordered from the most liquid to the least liquid. Assume that L(x) is the liquidity generated by liquidating (at some horizon, say one week) the share x of assets that is most liquid. Then obviously we should assume L(0)=0, L(1)<1, dL(x)/dx>0 and d2L(x)/d2x < 0 (reflecting the ranking from the most liquid to the least liquid assets). We can also define similarly the fire sale loss function F(x) = x-L(x) which are the fire sale losses generated by selling a share of assets x, starting with the most liquid assets. The function f(x)=dF(x)/dx is the marginal fire sales loss function indicating the size of fire sale losses resulting from selling the asset ranked at x, and q(x)=dL(x)/dx the marginal liquidity generated by selling the asset ranked x. We can illustrate the asset liquidation function of the company as in the left-hand side of the chart below.

Assuming the company has equity E, and has to generate through fire sales a cash flow L1 at horizon T1 (because it needs to repay a debt instrument at maturity and is unable to roll it over or find another form of financing), and noting that for every T the function LT(x) is non-decreasing and therefore invertible, such that we can write the inverted function as xT(L), the company is solvent conditional on the need to generate a cash flow L1 at time horizon T iff:

FT1(xT1(L1)) < E

The function f(x) above is under the assumption of a certain time horizon of the fire sales. We can express this by adding an index T which indicates the time horizon of the fire sale losses in days, i.e. fT(x). If T1>T2, then for every x, we expect that fT2(x) ≥ fT1(x). To distinguish fire sale losses due to time constraints from fire sale losses due to asset specificity, one could say that fAS(x) = lim(T→∞)[ fT(x) ] is the liquidation loss function isolating the role of asset specificity, allowing then to call fT,TR(x) = fT(x) - fAS(x) the strictly time pressure related losses relating to asset fire sale at a given time horizon T. The following chart on the right-hand side provides an example of fire sale loss functions for the same company at different time horizons. To ensure funding stability, the bank (or any debtor) should ideally be able to ensure liquidity at all time horizons. For example, if one debt position of an amount L1 matures in 1 week, and another one of L2 in four weeks, then both conditions F1Week(x1Week(L1))) < E and F1Month(x1Month(L1+L2))) < E should be fulfilled in order to ensure that the debtor is stable and can communicate to its current or potential future creditors that it will be fine anyway (even conditional on no roll-over of funding). Chapter 6 will provide a more precise model in which, for a specific functional form of L(x) and F(x), precise conditions for funding stability will be derived. For the moment, it is useful to retain that (i) solvent debtors can be sub-classified into those which are solvent regardless of assumptions taken with regards their ability to roll over some debt instruments maturing at some horizon, and those which are solvent only conditional on accessing fresh funding; (ii) the latter may create multiple equilibrium situations, as further explained in chapter 6; (iii) both a negative asset value and a deterioration of asset liquidity can push a lender from being unconditionally solvent into being solvent only subject to funding renewal.

L(x) f(x)→

F(x)

5.3 Illiquidity because of adverse selection and market breakdown The potential of information asymmetries to impact negatively on financial markets and thereby to potentially harm the funding of the real economy has been identified as a key issue in the economic literature for a while, see for example Stiglitz and Weiss (1981) or Bolton and Freixas (2006). The following simple model is a variant of the model of Flannery (1996). It assumes a one-period economy with many potential borrowers, of whom some are "Good" and others are "Bad". An adverse selection problem in the spirit of Akerlof (1970) arises. The model assumes that projects require an investment of one monetary unit, which needs to be obtained by the entrepreneur from a bank through a loan. At the end of the project period, Good loan applicants’ projects will be worth VG>1, which is sufficient to repay loans provided that the contracted interest rate is not higher than VG-1 (this value is what is left after having repaid the capital and the interest on the loan is paid out of it). Bad applicants will not pay back one cent (if they had obtained a loan). The number of borrowers and their quality are exogenous: the proportion of Good borrowers is δ, while the proportion of Bad borrowers is (1-δ). Banks have a costless but imperfect technology for assessing the creditworthiness of a loan applicant. Assume that this technology consists in a signal, which can be either SG or SB . If the borrower is Good, then with probability p > δ, a Good signal ( SG ) is captured, and the bank may lend. With probability (1-p), the signal Bad ( SB ) is captured, and no lending will take place. If the borrower is Bad, instead, then, the signal Bad will be received by the bank with probability p, and the signal Good with probability (1-p). This allows applying Bayes Law in the sense that:

P(S ∩G) P(S G) = G ⇒ P(S ∩G) = pδ G P(G) G P(S ∩ B) P(S B) = B ⇒ P(S ∩ B) = p(1− δ ) B P(B) B

P(S ∩ B) P(S B) = G ⇒ P(S ∩ B) = (1− p)(1− δ ) G P(B) G P(S ∩ G) P(S G) = B ⇒ P(S ∩ G) = (1− p)δ B P(G) B A perfect technology is one with p = 1. The figure summarises the probabilities of the four possible cases.

Matrix of probabilities in lemons market model Good signal Bad signal Good borrower δ p δ(1-p ) Bad borrower (1-δ)(1-p) (1- δ) p

The bank needs to fulfil, in competitive equilibrium, the no-loss constraint or, equivalently, an equilibrium rate of return. The profit it achieves if it lends to a Good borrower must on average compensate the credit losses it experiences with Bad borrowers. Define the minimum interest rate that a bank needs to set in order not to make losses as i*. As a loan is only provided in case a Good signal was obtained, non-zero pay offs to the bank occur only in the two subcases (Good signal column in the table above), and therefore the no loss-constraint can be written (assuming that banks are funded at a zero interest rate level):

i δ p + (-1)(1-δ) (1-p ) ≥ 0  i* = (1- δ)(1-p)/(δp)

Lending will take place as long as VG – 1 ≥ i*, since otherwise, the Good borrowers would no longer be profitable. For instance, for δ =0.5, if p=0.5 (no useful signal), then the interest rate i* needs to be 100%, if p=0.8. the interest rate needs to be 25%, if p=0.95, it needs to be 5.3%, and if p=1, then it needs to be zero (assuming indeed that the bank is otherwise perfectly efficient).

The model allows identifying three intuitive triggers that can cause a break-down of lending: (i) a decline of δ, the share of Good loan applicants in the population, (ii) a decrease of VG, the return generating power of Good loan applicants, (iii) a decrease of p, the power of the screening technology.

All three effects are likely to occur in the case of a major negative economic shock, such as an asset bubble burst. Therefore, the adverse selection model explains why negative economic shocks trigger liquidity crises, and why these effects can be so abrupt: in principle, a minor further deterioration can determine that the market collapses, because suddenly the critical condition no longer applies.

Before a deterioration of economic conditions such as captured in (i) and (iii) leads a complete funding market breakdown, it is already felt in the form of an increasing equilibrium lending rate i* for a given level of funding costs of banks (which were assumed to be zero). This will affect the transmission mechanism of monetary policy in the sense that for unchanged central bank policy rates (for an unchanged operational target of monetary policy), suddenly higher funding costs to the real economy materialise. In so far, the effective monetary conditions can tighten due to the deterioration of any of the two parameters δ, p without any actual change in monetary policy by the central bank. For otherwise unchanged conditions, the central bank needs to react to such a tightening of the transmission mechanism by lowering central bank interest rates.

5.4 Bank runs and investor strikes

Bank runs have been a major financial stability issue at least since the 19th century, in the early 1930s e.g. in the US and Germany, and again in the recent crisis, such as e.g. on Northern Rock, or on Greek and Cypriot banks during the respective crises there. The latter runs mainly materialised through electronic transfers of deposits to accounts with non-domestic euro area banks, i.e. at no time there were queues in front of the banks to withdraw cash. In the early 1930s there had been a multitude of bank runs, which led to the conclusion to establish deposit insurance schemes.

Bank runs have been modelled e.g. by Diamond and Dybvig (1983) and Rochet and Vives (2004). The particularity of bank runs is their self-fulfilling property: once a run on a bank starts, it can lead to the default of the bank, confirming the individual wisdom of those who were first in the queue to withdraw their money. We will illustrate in chapter 7 with a very simple but powerful strategic bank run model that a bank can essentially be in three states in terms of stability of its short-term liabilities: • a) Funding stability: if there is a single no-run equilibrium • b) Multiple equilibria: there are two equilibria, one in which depositors run, and one in which they do not run. The depositors’ behaviour can in principle switch from one to the other. This situation arises for solvent banks with however a too weak overall liquidity-solvency profile. As it implies that with some probability, a run will start and destroy the bank, implying fire sale losses to society as a whole, this is not a satisfactory state of the world. • b) Single run equilibrium: depositors and other short-term investors should run in any case. This equilibrium applies to insolvent banks.

A switch from state a) to state b) can occur if: (i) asset liquidity deteriorates; and/or (ii) asset values fall implying a fall of equity. A switch from state a) or from state b) to state c) will occur if asset values fall such that equity gets negative. In crisis situations, both factors tend to materialise with higher probability than usually, in particular for banks with more limited solvency and liquidity buffers, or with an unlucky asset concentration, i.e. an asset concentration towards those assets which by bad luck suffer from value losses and a decline in liquidity.

5.5 Increase of haircuts (“margin”) in collateralised financial market transactions Collateralising financial claims is a widespread financial market technique. It is used in particular in the following three types of operations. • Interbank repo operations (i.e. collateralised interbank lending). • Lending of banks to non-banks: e.g. mortgage loans to household and commercial real estate developers; loans of banks to corporates; loans of banks to hedge funds, etc. • Collateralisation of the value of derivatives transactions, be they “over-the counter” (OTC) or via central clearing counterparties (CCPs). The importance of margining for the generation of a liquidity crisis and for the depreciation of asset values via fire sales results from the fact that margin requirements are a limitation to leverage, both for home-owners and for leveraged investors using as collateral the assets they invest in. Unfortunately, haircuts tend to increase substantially during financial crisis, implying that feasible leverage shrinks, implying often the need for fire sales.

It is easy to explain why in a financial crisis haircuts tend to increase dramatically. Assume that the cash investor (or the party having an exposure out of an OTC derivative transaction) wants to maintain the probability of a loss conditional on counterparty default at a certain confidence level, i.e. say a loss should have a probability not higher than β (e.g. β=1%). Then, the investor should set the haircut at the level of the “Value-at-Risk” (VaR) from a unit exposure to the collateral.

Concretely, VaR is calculated on the basis of the following factors (for a definition of VaR and its role in financial risk management see any standard risk management text, such as e.g. Hull, 2018): • The holding period T of the asset (how much time is needed to liquidate an asset without the liquidation leading to a negative market price effect?) • The volatility σ of the asset price at some standard time horizon, say one day. If daily price innovations are independently and identically normally distributed, then price changes over the horizon T will have a volatility of σ T . • The confidence level for the VaR limit (in our case of the no-loss requirement), i.e. with what probability the VaR should not be exceeded. For normally distributed price changes, the confidence level β can be translated into a multiplier of volatility using the inverted cumulative standard normal distribution, i.e. Φ-1(β). For instance Φ-1(0.1%) = -3.10; Φ-1(1%) = -2.33, and Φ-1(5%)=-1.64. The VaR reflecting a conditional daily price volatility σ, a liquidation horizon of T days, and a confidence level β is then simply equal to: VaR = Φ −1 (β )σ T In a financial crisis, all factors in this formula will support an increase of haircuts: • Cash providers may seek a higher confidence level of protection as their capital buffers may have suffered due to the crisis, implying a need to run down overall risk taking: β ↑⇒ Φ −1 (β ) ↑ . • Asset price volatility increases: σ ↑ • It takes longer to liquidate assets without the liquidation having additional price effects: T↑. Increases in haircuts will mean less borrowing potential for a given set of assets – and maybe too little borrowing potential to prevent a need to fire sell assets, i.e. to shrink balance sheets. Again, this will imply also a tightening of monetary conditions and various negative second round effects.

5.6 Increase in bid-ask spreads in a dealer market In a dealer market, where dealers quote binding bid-ask prices (for certain specified amounts), the bid- ask spread is the most obvious measure of asset liquidity. Bid-ask spreads can dramatically increase during financial crises, meaning that asset sales become costlier or even prohibitive. There is an extensive microstructure literature explaining why this is so. Bid-ask spreads in a dealer market with information asymmetries have been modelled in particular by e.g. Kyle, 1985. Consider below a basic version of a model explaining why asset liquidity as measured through the bid ask spread in a dealer market will deteriorate in a financial crisis. The model relies on the following set of assumptions. 2 • The true value of assets, vt, innovates every day according to vt =vt-1+εt with εt being N(0,σ ε ) • The bid ask spread z is set every day symmetrically by the market maker around the estimated true value vt of the asset. However, the market maker only knows the true value of the asset on the previous day, and therefore on day t, the spread is [vt-1-z/2, vt-1+z/2]. Market makers are competitive and deliver their services without operating cost. The market makers commit to trade at the bid-ask quotation for a total value of q. • Noise traders (i.e. uninformed traders) trade every day for an amount W =W(z)≥0 and W(0)>0, dW/dz<0, whereby it is equally split between demand and supply. We also require that d(zW(z))/dz<0.

• The insider knows vt on the same day. Whenever the true value is outside the bid-ask spread, i.e. whenever vt < vt-1-z/2 or vt > vt-1+z/2, the insider exploits the dealing commitment of the trader. Assume that the market maker becomes aware of an insider transaction only once he notes the imbalance in demand and supply, and that this is the case when the insider has traded a volume of q. Then, the market maker stops trading for that day, and re-opens next day again with a symmetrical bid ask spread around the new asset value.

It is easy to show how, under these assumptions, bid-ask spreads and trading volumes reflect the 2 existence of insider information, as captured by σ ε (price volatility is equivalent here to insider information, as the insiders are always ahead one period in time with regard to knowing the fair value of the asset). The competitive dealers will set z such as to have expected profits of zero. Expected profits have two components. First, the profit of dealers generated by the noise traders (“EPN”) is: EPN = zW(z)

Second, the expected losses inflicted on the market maker by the insiders (“ELI”) are ( fε () being the probability distribution of value innovations): −z / 2 ∞  + + − ELI =  ∫(x z / 2) fε (x)dx ∫(x z / 2) fε (x)dxq  −∞ z / 2  In competitive equilibrium, z has to be set such as to bring expected profits to zero, with the first element in the expression compensating the second, i.e. EPN – ELI = 0. Taking as an example that 2 2 W(z)=1/(1+z) and q=3 then for σ ε =1 the competitive bid-ask spread will be 0.5, while for σ ε =5 it will be 3.6. This dealer market model with insiders explains simply the well-known inverse relation between information intensity of assets and asset liquidity, measured by the bid-ask spread. The model also explains why in a financial turmoil, where uncertainties and hence the potential for insider information dramatically increase, bid-ask spreads increase, possibly in an extremely rapid and strong way, leading to an associated collapse of market liquidity, and hence an increase of asset fire sale discounts.

As we have seen above, in a financial crisis, asset liquidity deteriorates, and leveraged entities have to shrink their balance sheet regardless of whether they were funded in collateralised, or unsecured interbank and capital markets. This creates potentially a vicious circle: asset fire sales to obtain liquidity further depress asset prices, such that, in the worst case, the asset fire sales depress market prices of assets remaining on the balance sheet by so much, that, after the fire sales, solvency and liquidity problems are even worse and require even more fire sales, and so on. Moreover, the asset fire sales of one institution impact negatively on the mark-to-market value of other institutions’ assets (and hence imply losses), leading to negative externalities. Cifuentes, Ferrucci and Shin (2005) develop and simulate a model of the asset fire sale channel of systemic contagion and related negative externalities. They also analyse stability effects of capital and liquidity requirements in a system of interconnected banks for given portfolio choices, when mark-to-market rules are in place.

5.7 Interaction between crisis channels The following chart illustrates the interaction between the various crisis channels, and how solvency, liquidity and default relate to each other. As mentioned, the key trigger of a financial crisis is normally a decline in asset value and/or an increased volatility and uncertainty about asset values. • Asset value declines may lead to immediate insolvency, which normally will lead to an inability to roll over funding (a run) and eventual default. • In case there is no immediate insolvency, still the various liquidity channel components set in, like loss of funding stability; increasing funding costs due to increased credit risk premia, increase of haircuts, increase of bid-ask spreads, decline in the ability to monitor and to overcome the adverse selection problem (as all discussed above through simple models). • Banks’ and corporates’ funding stress possibly triggers asset fire sales by them. • These fire sales may prevent or not the immediate default of the bank or corporate. Even if default can be prevented as enough liquidity is generated through fire sales, the firm can have generated losses through the fire sales such as to now have insufficient or negative capital. • Both defaults and fire sales without eventual default create further asset value declines and a further increase of asset value uncertainty – reflected in the red feedback loop arrows in the chart.

In addition, once lending to the real economy becomes scarce and expensive, an economic downturn will likely occur and will lead to additional losses via asset price declines and impairment of loan portfolios (as the default frequency of debtors increases). The resulting economic and financial dynamic may be fatal and calls for external circuit breakers such as in particular Government authorities, regulators, and the central bank. Central banks, which can in theory provide unlimited liquidity to banks by lending freely, have the power to suppress the liquidity part of any vicious crisis circle, making the central bank (unwillingly) the key player deciding on the fate of banks and other leveraged entities.

Asset value declines Increased uncertainty and volatility of asset values

Probability of Repo Bid-ask Monitoring insolvency ↑ haircuts ↑ spreads ↑ ability ↓

Funding liquidity problems

Asset fire sales

Insolvency

Illiquidity

Default & liquidation Shrinkage and survival

Chapter 6: The central bank as lender of last resort

6.1 Introduction 6.2 Why should central banks be lender of last resort? 6.3 Three forms of the LOLR 6.4 Overall propensity of central bank to be the LOLR 6.5 Central bank collateral as key LOLR parameter in simple bank run model

6.1 Introduction During financial crises, not only the monetary policy transmission mechanism gets disrupted implying that the central bank needs to make more supportive its monetary policies to ensure that adequate monetary conditions continue to apply to the economy. In addition, the central bank is called to play a fundamental role in terms of providing liquidity support to banks in view of frozen asset, interbank and capital markets, i.e. to act as “lender of last resort” (LOLR). Obviously, the monetary policy and LOLR functions are closely interlinked: in a financial crisis, perceived real rates of return on capital will normally fall; if in addition, bank intermediation spreads dramatically increase, then a strong deflationary shock will likely occur even if the central bank lowers central bank interest rates to close to zero. An active lender of last resort under such circumstances contributes to limiting the increase of intermediation spreads, and therefore also contributes to an effective monetary policy, i.e. preventing to fall into a deflationary trap.

Today’s thinking on the LOLR function is still strongly inspired by 19th century experience, and in particular Walter Bagehot’s outstanding oeuvre Lombard Street of 1873 (see also e.g. Goodhart, 1999 for an overview of issues in modern interpretations of the 19th century literature and Goodhart and Illing, 2002, for a collection of essential papers including more recent ones; see Rochet and Vives, 2005 for a recent model of the LOLR function). Consider three key insights of 19th century experience which still appear valid today.

Lend by every means consistent with the safety of the central bank. Jeremiah Harman, director of the Bank of England, summarized in a hearing of the Lords’ Committee in 1832 the Bank's actions in the panic of 1825 as (found e.g. in Bagehot 1873): “We lent… by every possible means, and in modes that we never had adopted before; we took in stock of security, we purchased Exchequer bills, we made advances on Exchequer bills, we not only discounted outright, but we made advances on deposits of bills to an immense amount; in short, by every possible means consistent with the safety of the Bank;… seeing the dreadful state in which the public were, we rendered every assistance in our power.” First, it is useful to note that the statement is about extra liquidity injection into the financial system at the benefit of all banks under circumstances of a collective financial market liquidity crisis, not about emergency liquidity assistance to individual banks (as often wrongly assumed). Second, Harman explains the Bank of England's action as having been creative and pro-active, i.e. to have innovated to find the best ways to support funding liquidity of financial institutions, the only constraint to creativity being the need to preserve the “safety of the Bank”. Third, Harman insists that the safety of the central bank must be preserved, i.e. central bank risk taking must be limited. The fact that the Bank of England conducted all sorts of “unconventional” lending and purchase operations while this being perceived as being compatible with its own safety also remains a key element of today’s view on the LOLR (see also the point on risk endogeneity below).

Inertia Principle. Bagehot (1873) himself states this principle according to which the central bank should maintain its risk control framework stable, and accept to not tighten it as a reaction to the deterioration of asset liquidity in a crisis situation: “If it is known that the Bank of England is freely advancing on what in ordinary times is reckoned a good security and on what is then commonly pledged and easily convertible, the alarm of the solvent merchants and bankers will be stayed. But if securities, really good and usually convertible, are refused by the Bank, the alarm will not abate, the other loans made will fail in obtaining their end, and the panic will become worse and worse.” In contrast to Harman, Bagehot here does not emphasise the pro-active nature of the measures taken, but the fact that the central bank must remain “inert”. While the emphasis is hence somewhat different, also Bagehot can be said to turn in this analysis around the duality of liquidity support and risk taking.

Risk Endogeneity. Bagehot also provides an additional perspective on liquidity support and central bank risk taking by arguing that supportive liquidity provision could be necessary to minimize the central bank’s eventual own financial risks because such measures would be the only way to prevent a financial meltdown and accompanying losses for the central bank. Bagehot explicitly writes: “(M)aking no loans as we have seen will ruin it (Bank of England); making large loans and stopping, as we have also seen, will ruin it. The only safe plan for the Bank (of England) is the brave plan, to lend in a panic on every kind of current security, or every sort on which money is ordinarily and usually lent. This policy may not save the Bank; but if it do not, nothing will save it.” In other words, the riskiness of exposures would itself be endogenous to the central bank measures, and hence more liberal central bank policies could imply lower central bank financial risk taking than more conservative policies.

6.2 Why should central banks be lender of last resort? In the following we distinguish five, partially related, reasons for the central bank to act as the lender of last resort in a financial crisis and to provide elastic credit, even if this leads to higher and more concentrated exposures. Of course, the existence of justifications for the central bank to act as LOLR does not imply that there are no draw-backs of a too supportive liquidity provision, such as e.g. to create moral hazard, to support businesses that should be stopped as they generate social losses (“zombie banks” to use the term of Caballero et al, 2008), or to prevent the necessary price adjustments in markets for certain assets. The trade-offs at stake will be discussed in more detail later. a. Negative social externalities of funding liquidity stress and default due to illiquidity. Negative externalities potentially justify the intervention of public authorities. As argued by Brunnermeier, Crocket, Goodhart, Persaud, and Shin (2009), the most important negative externality of bank default stems from the fire-sale spiral induced by liquidity problems of individual banks: “In order to deal with such liquidity problems prior to failure, and in the course of liquidation after failure, the bank in difficulties will often be forced to sell assets (fire sales). But such sales will drive down the current market price of the same assets held on other banks' books, when these are valued on a mark-to-market basis. (. . . ) In short, there is an internal amplifying process (liquidity spirals) whereby a falling asset market leads banks, investment houses, etc., to make more sales (deleveraging), which further drives down asset prices and financial intermediaries' assessed profit and loss and balance sheet net worth." By lending to banks against collateral and thereby eliminating the need for asset fire sales, the central bank can prevent the downward spiral and negative externalities of fire sales. Typically, central bank measures avoiding asset fire sales will help preserve solvency and reduce probabilities of default of counterparties and issuers, as will be illustrated more concretely through a bank run model below. Asset fire sales are not the only form of negative externalities of bank funding stress and illiquidity induced default that have been mentioned in the literature. Other forms of negative externalities are the spreading of depositors' panic in the form of a generalized bank run

(such as observed in various countries in the early 1930s), and the generalized drying up of funding and market liquidity in the financial system as a consequence of hoarding driven by the experience that claims, including collateral, can get stuck in a default. b. Unlike leveraged private entities, the central bank is not threatened by illiquidity in domestic currency. Central banks have been endowed with the monopoly and freedom to issue the legal tender, i.e. central bank money. Therefore, they are never threatened by illiquidity in their own currency and it seems only natural that, in case of a liquidity crisis when all agents attach a high price to liquidity, the central bank remains more willing than others to hold (as collateral or outright) assets which are less liquid. This argument does not rely on the existence of negative externalities. Even if the central bank were a purely profit-oriented enterprise, its exemption from liquidity stress should make it ready to take over illiquid assets in a crisis. Typically, asset prices will tend to recover after an acute liquidity crisis, and therefore the central bank may generate extraordinary profits. The fact that bank and corporate defaults are costly in themselves even without externalities, as they destroy organizational capital and normally block the efficient use of the underlying resources at least for a while, should also be seen in the context. If a bank or a corporate are threatened by illiquidity (and associated default) in a financial crisis, and if in the case of default the (presumably positive) organizational capital would be destroyed, then saving this capital is part of the `rent' that can be achieved through cooperation between the liquidity-stressed economic agent and the one that has unlimited liquidity. It is important to repeat that preventing costs of default in this sense through central bank liquidity does not invoke negative externalities, market failures and the public nature of the central bank. c. Haircuts are a powerful risk mitigation tool if credit risk is asymmetric and the collateral provider (repo borrower) is more credit risky than the cash investor (repo lender). The power of haircuts is limited if cash taker and cash lender are equally credit risky since although haircuts protect the cash provider, they expose the cash taker to unsecured credit risk which increases with the haircut level (Ewerhart and Tapking, 2008). Anecdotal evidence suggests that haircuts applied in repos between banks of similar credit quality tend to be rather low, while haircuts applied to other market participants, for example hedge funds, tend to be higher. Thus, banks would never question haircuts imposed by the central bank (repo lender), because the central bank cannot default. d. The central bank may have superior knowledge on the state of stressed banks compared with other market participants and may have special legal privileges in recovering claims (beyond collateralization). The first may relate to its information rights and practice as banking supervisor, and the fact that as a public entity, information access of the central bank does not need to be constrained. In contrast, if a competitor or a vulture fund would be candidates for providing emergency lending, it is difficult to imagine that they could be given easily full access to all relevant information. This is particularly relevant when decisions need to be taken in a very short time period. Moreover, the central bank, as public entity entrusted with special public functions and legal privileges, may be able to secure its claims resulting from LOLR operations more effectively than a private, ordinary lender. e. The LOLR as unconventional monetary policy at the ZLB: the central bank may achieve its mandate to maintain price stability, vs. falling into deflationary trap, by supporting low bank intermediation spreads through the LOLR in a crisis situation in which it hits the ZLB with its conventional monetary policy operations. Achieving its mandate is welfare improving (unless one believes that a long term soft deflation a la Japan makes no difference in terms of real economic activity).

6.3 Forms of the LOLR The central bank LOLR function can take three forms: (a) LOLR built into the regular operational framework of the central bank; (b) LOLR added through changes of the framework and additional LOLR

operations for all banks in crisis times; (c) emergency liquidity assistance to individual banks. Consider these three one after the other.

(a) LOLR built into the regular operational framework The following elements determine the LOLR content of the regular operational framework. • As mentioned earlier, collateral availability provides a first natural limit to central bank credit at the individual bank level. The size of the collateral set should be viewed in relation to the liquidity deficit of the banking system to be covered by central bank credit operations. For example, in the case of the Eurosystem, the nominal value of eligible securities is around EUR 13 trillion, of which around EUR 5 trillion is held by banks, against a EUR 0.5 trillion liquidity deficit of the euro area banking system to be covered by credit operations. This implies that a representative bank could extend recourse to central bank credit approximately 10 times relative to proportionality before hitting collateral constraints. • The size of the liquidity deficit to be covered by credit operations determines the potential relative central bank intermediation. Relative central bank intermediation occurs when liquidity constrained banks crowd out less constrained banks from central bank credit without the latter yet ending in a liquidity surplus towards the central bank. The spread between the interest rates charged on central bank liquidity providing operations and the remuneration of excess reserves does not provide counter-incentives against relative central bank intermediation. • The spread between the central bank lending rate and the rate at which excess reserves are remunerated determines the cost of absolute central bank intermediation as it kicks in once the strong banks are crowded out completely from central bank credit operations. • Active stigmatisation or de-stigmatisation through central bank communication will impact on the propensity of banks to rely on the LOLR. • It matters who is able to benefit from direct LOLR. The perimeter of institutions directly benefitting from the LOLR elements in the normal operational framework depends on the counterparty framework of the central bank.

(b) Readiness of central bank to add LOLR content to the operational framework in crisis times The impact of the LOLR on bank behavior will not be limited to the LOLR content of the operational framework in normal times. What matters as well (for a bank that is not completely myopic) is the bank’s liquidity in a scenario of financial market stress. Anticipating this case also includes building expectations on the readiness of the central bank to add LOLR content in crisis times, e.g. through making more LOLR- supportive any of the above-mentioned seven parameters that determine the LOLR content of the operational framework and my conducting any type of additional open market operations with LOLR content. Expectations will be determined by historical experience and forward looking central bank communication, which banks may find credible or not.

(c) Readiness of central bank to provide emergency liquidity assistance (ELA) to individual banks ELA can be defined as a non-rule based LOLR activity for the benefit of individual banks. Of course, also ELA needs to take place within some legal framework, within the mandate of the central bank and ideally in a consistent manner. Limitations to ELA provision can result from: • ELA collateral requirements (normally ELA collateral sets should be wider than standard credit operations’ collateral set). • Pricing of ELA, i.e. what surcharge relative to monetary policy credit operations is imposed (some surcharge is typically applied). • Preservation of systemic financial stability may be set as precondition for granting ELA. The higher the hurdle set by the central bank in declaring a systemic financial stability interest before granting

ELA, the less a bank can rely ex ante on it, in particular when deciding to take idiosyncratic and concentrated liquidity risks, and in particular if a bank is small. • Limitations on the duration of ELA (ELA is typically assumed to be of limited duration). • Possible requirement that ELA is only granted if the central bank is protected in addition by a government guarantee. Beyond additional risk protection, this may be considered useful as it requires an elected government to confirm its backing of ELA operations (but it should not delay very urgent and obvious ELA provision by the central bank). • ELA counterparty set: As by definition ELA is ex ante not subject to clear rules and constraints, also the question arises to what extent non-banks may benefit from it. While by construction it seems wrong to ex ante define the perimeter of expansion of the ELA counterparty set, it seems useful for central banks to think through the conceptual, procedural and legal issues that would arise if during a crisis such ELA would be considered.

6.4 Overall propensity of central bank to act as LOLR Assume now for a moment that the relative contribution of the three LOLR forms is not the issue, as one would simply assume that the optimal mix is chosen, including on all the underlying details. Still one could ask what the total LOLR provision by the central bank should be. It is useful to think first about two extreme LOLR choices of the central bank. • Maximum LOLR: accept in the normal-times OF all assets of banks as collateral at fair values. This would allow solvent banks to finance all their assets with the central bank, if desired, and no solvent counterparty could ever default for liquidity reasons. Furthermore, the width of the standing facilities corridor is set to zero and there are no surcharges for over-proportional borrowing. The other two forms of LOLR (expanding the LOLR content of the OF in crisis times and ELA) would not really be needed as all these could add is already covered in the normal OF. • Minimum LOLR: the central bank implements monetary policy only against risk free assets, which may either be defined as central government paper, or as highly liquid AAA-rated paper. It largely covers its asset side through outright holdings of the risk-free asset. The central bank may conduct at the margin some repos against risk-free assets, but in a bilateral way in which it chooses its counterparties and always goes for the most secure ones. In this OF, banks have no discretionary access to the central bank at all to close possible funding gaps, i.e. the OF has zero LOLR content. Moreover, the central bank would fully pre-commit to never change the LOLR content of its OF nor to ever provide ELA. Central bankers believe that the optimal LOLR is in between these two extremes (and regulation will have to play an additional role to achieve an overall optimum for society). The LOLR strengthens the ability of the financial system to provide maturity and liquidity transformation as services to society. At the same time, putting some limits to the LOLR role is beneficial for society, to have some protection against information asymmetries and moral hazard, to avoid relying excessively on the abilities of supervisors and auditors, and generally to preserve stronger incentives to maintain funding market access and thereby market discipline. Proponents of a tight approach may argue that a supportive LOLR will lead to as many financial crises as a very tight one, but crisis will be messier because when they occur the financial leverage will be much higher (“four-wheel vehicles make you get stuck in areas which are more difficult to access when you need to be rescued”).

Assume for a moment that we capture in a unit interval [0,1] the supportiveness of the LOLR framework of a central bank and let the most restrictive framework described above be represented by 0 and the most forthcoming framework by 1 (again, it is of course a simplification to assume that designing the LOLR is a one-dimensional problem). One can imagine mapping the LOLR unit interval into at least five effects of interest which should not be expected to be identical, although often this seems to be implicitly assumed:

(1) Social welfare is the ultimate measure of interest and can be equated e.g. with the extent to which the LOLR framework contributes to financial conditions leading to maximum economic growth in the medium to long term, i.e. through the financial and economic cycle. For example Keister (2015) maps the LOLR supportiveness into social welfare, and Bindseil and Jablecki (2013) map it into growth. They show that it is likely that the relationship is a concave function with interior maximum (i.e. an intermediate LOLR approach maximizes growth). (2) Central bank risk taking is normally expected to increase monotonously in the [0,1] LOLR unit interval, whereby it will typically lift off from very close to zero only beyond a certain threshold. Bindseil and Jablecki (2013) also provide an example in which the relationship is a convex function with interior minimum, illustrating Bagehot’s intuition that sometimes “only the brave plan is the safe plan” for the central bank, i.e. more forthcoming lending in a crisis reduces central bank risk taking relative to a restrictive approach. (3) Leverage of banks and their ability to provide liquidity and maturity transformation should increase monotonously with the supportiveness of the LOLR. Regulation may limit leverage to lower levels. (4) Financial fragility will probably first decrease, and then increase across the LOLR unit interval, suggesting that a measured LOLR can stabilize the financial system while a too liberal one could eventually lead to particularly deep financial crises. (5) Market discipline and funding market functioning can be thought of as either falling monotonously, or as mirroring the financial fragility curve, i.e. it would benefit from some moderate LOLR, but is undermined if the LOLR is excessive. Bindseil (2013) shows that when asset liquidity deteriorates after an exogenous shock, then the LOLR can preserve funding market access for solvent banks, but not for insolvent banks, while a restrictive LOLR will imply a run also on solvent banks. In this sense a more supportive LOLR can allow for a more effective market mechanism than a very restrictive one.

The exact shapes of all five curves will depend on various environmental parameters. The following figure provides an example of how the five curves may look like (the shapes that could be considered most likely under conditions of developed financial markets). In this example, the x-value of the highest effectiveness of market discipline is strictly positive, but below the x-value of the minimum financial fragility, which itself is below the point of maximum social welfare (or maximum long term growth).

Figure: Possible relationships between LOLR content of the OF and five impacted variables (social welfare, central bank risk taking, leverage, financial fragility, market discipline).

The devil lies in the detail, and beyond these general questions on the overall LOLR provision and the contribution of the three different forms of LOLR, a multitude of more specific issues emerges in the LOLR field. Consider below a number of such issues.

Bagehot’s “only the brave plan is the safe plan” and risk endogeneity Ever since Bagehot it is acknowledged that the central bank cannot make its LOLR choices as if it was an atomistic investor not influencing the properties (e.g. default probabilities) of the system. Often, being more forthcoming as a LOLR after a negative financial stability shock (e.g. broaden the eligible collateral set to include less liquid assets) will decrease financial risk taking by the central bank, instead of increasing it. Risk endogeneity should lead to a more forthcoming LOLR, i.e. the welfare-maximizing LOLR framework will be more supportive than the one obtained if risk endogeneity is ignored.

Moral hazard and central bank losses A popular theme in papers on the LOLR is moral hazard, but the concept often remains vague. One pragmatic view is that moral hazard only materializes in the context of the LOLR if the central bank faces actual losses from its credit operations. This interpretation also has the advantage that it would reduce the complexity of the LOLR design problem by one dimension and map something vague and complex (moral hazard) into something concrete and more measurable (central bank risk taking – even if somewhat complicated by endogeneity). If central banks are worried about moral hazard, they could tighten risk control measures (in normal times, to not be pro-cyclical) so that the probability of central bank credit losses declines even further.

Excessive stigmatization of the LOLR? Sometimes central banks worry that banks attach excessive stigma to taking recourse to different forms of LOLR. For example, recourse to the Discount Window is considered to remain stigmatized in the US although the Fed would want to change this since 2002. Also, in a number of credit open market operations of central banks during the financial crisis, aversion of banks to participate materialized such that the accommodation that the operations aimed at could not be fully achieved. Excessive stigmatization seems to go in the opposite direction of moral hazard. Central banks should therefore have tools in hand to adjust in both directions the willingness of banks to come to LOLR operations. Stigmatization through communication is difficult to control and to revert.

LOLR and liquidity regulation Some have argued that the only solution to the central banker’s Angst from on one side being exploited by banks (moral hazard), and on the other side of excessive stigmatization of LOLR operations (preventing banks to use them), would be a liberal LOLR combined with tight liquidity regulation. A good combination of appropriate incentives to take recourse to the LOLR and well-designed regulation likely allows achieving the best outcome for society. Managing liquidity risk is a core activity of banking, and it seems unlikely that “centralizing” this subtle activity through liquidity regulation can be done without efficiency costs. Therefore, liquidity regulation must not be overburdened, and central banks providing some well-designed and rule based economic counterincentives to what they deem an excessive reliance on the LOLR is an important contribution to reduce the burden put on regulation. In fact, the combination of regulation (in all its details) and the LOLR (in all its details as well) will jointly determine (i) the ability of banks to provide liquidity and maturity transformation as services to society; (ii) financial stability; (iii) central bank risk taking and banks’ possible moral hazard.

6.5 Central bank collateral as key LOLR parameter in a simple bank run model The following simple bank run model integrates the central bank LOLR function explicitly. In its continuous variant (Bindseil, 2013), the model allows moreover to capture the role of bank capital and its value added relative to long term funding. Throughout this section, we consider the following stylized bank balance sheet. The total length of the balance sheet has been set to unity. Assets are now grouped into homogeneous classes in terms of asset liquidity and fire sale discounts. There are three types of liabilities (with e in [0,1], t in [0,1] and e+t in [0,1]).

A stylised bank balance sheet to analyse funding stability of a bank Bank Liquid assets Λ Short term debt 1 (1-t-e)/2 Illiquid assets 1-Λ Short term debt 2 (1-t-e)/2 Long term debt (“term funding”) t Equity e

Total assets 1 Total liabilities 1

The stylized balance sheet is sufficient to capture one key issue of banking: how to ensure the confidence of short term depositors of the bank such that they do not easily switch to a fear mode in which they start withdrawing deposits, triggering self-fulfilling destructive dynamics ending in bank default. Confidence may be sustained in particular by two means. First, the bank may limit the role of short term funding. However, in general, households and institutional investors prefer to hold short term debt instruments over long term debt instruments and equity and request a higher return rate on the latter two types of claims. In other words, long term debt and capital is associated with higher funding costs for the bank, or, put differently, maturity transformation is one of the key contributions of banking to society (see Financial Services Authority, 2009, 68). Second, the bank may aim at holding sufficient amounts of liquid assets, both in the sense of being able to liquidate these assets in case of need, and to pledge them with the central bank at limited haircuts. However, on average, liquid assets generate lower returns than illiquid ones. Consider now in more detail the representative bank.

Bank asset market liquidity and central bank collateral treatment. In the first, simplest variant of the model, there are only two types of assets with extreme liquidity properties: • A share Λ (0≤Λ≤1) of assets is fully liquid (i.e. it can be sold without any fire sale losses). • A share 1-Λ of assets are totally illiquid, i.e. if one would try to fire sell them, one would not generate a cent of liquidity, but only generate losses. At the same time, it is assumed that the central bank applies, when accepting bank assets as collateral, a homogeneous haircut h on all assets. In other words, the central bank haircut and collateral framework is not sensitive at all to asset liquidity. We summarize these assumptions in the charts below. Obviously in this case it never makes sense to fire sale the illiquid assets as this would generate zero liquidity but maximum losses. The illiquid assets should instead be pledged. At the same time, to generate maximum liquidity, it makes sense to fire sell the liquid assets and to not pledge them with the central bank.

L2=(1-h)(1-Λ) L1 h =

Λ haircut

0 0 Λ 1 Λ 1

Bank liabilities. Four types of liabilities are distinguished: (i) Short term liabilities are equally split to two ex-ante identical depositors; (ii) Long term debt does not mature within the period considered and are ranked pari passu with short term debt; (iii) Equity is junior to all other liabilities and is also a stable funding source; (iv) Central bank borrowing is zero initially but can substitute for outflows of short term liabilities in case of need.

Time line. The model is based on the following time line: • Initially, the bank has the balance sheet composition as shown in the figure above. • Short term depositors play a strategic game with two alternative actions: to run or not to run. “Running” means to withdraw the deposits and to transfer them to another account, accepting a small cost capturing the transaction cost of withdrawing the deposits, which we notate by ε. • It is not to be taken for granted that depositors can withdraw all their funds. If one or both of the depositors run, then at least one or several of the following will apply: (i) Substitution of deposit outflows with central bank credit, assuming that the bank has sufficient eligible collateral. (ii) Liquidation of assets: the bank may also do asset fire sales of assets when liquidation values exceed collateral values after haircuts. (iii) If it is impossible to pay out the depositors that want to withdraw their deposits, Illiquidity induced default may occur. If illiquidity induced default occurs, all (remaining) assets need to be liquidated, and corresponding default related losses occur. • If the bank was not closed due to illiquidity in the previous stage, still its solvency is assessed by the supervisor and if capital is negative, the bank is resolved. Again, it is assumed in this case that the full costs of immediately liquidating all assets materialize.

Figure: Time-line of bank liability structure model

Cheapest Strategic “run” sustainable liability game played

structure chosen between Enough by banks: (t,e) depositors Solvency? Λ, h given liquidity? yes yes

no Default, no Liquidation liquidation

We use a simple equilibrium concept, namely the one of a Strict Nash No-Run (SNNR) equilibrium. The decision set of depositor i (i=1,2) from which he will choose his decision Di consists in {Ki, Ri}, whereby “K” stands for “keeping” deposits and “R” stands for “run”. Call the pay-off function of depositor i: Ui=Ui(D1,D2). Note that the strategic game is symmetric, i.e. U1(K1,K2)=U2(K1,K2), U1(K1,R2)=U2(R1,K2), U1(R1,K2)=U2(K1,R2), U1(R1,R2)=U2(R1,R2). This allows to express in the rest of the paper conditions only with reference to one of the two players, say depositor 1. A Strict Nash equilibrium is defined as a strategic game in which each player has a unique best response to the other players’ strategies (see Fudenberg and Tirole, 1991, 11). A Strict Nash No-Run (SNNR) equilibrium in the run game is therefore one in which the no-run choice dominates the “run” choice regardless of what the other depositors decide, i.e. an SNNR equilibrium is defined by U1(K1,K2)>U1(R1,K2) ∩ U1(K1,R2)>U1(R1,R2)

To identify the cheapest sustainable funding structure, we define now as a liquidity stress strategy (LSS) of a bank a mapping of the assets of the bank into either their use as fire sale reserves or as use as collateral for recourse to the central bank. In the chosen simple case, the choice of the LSS is trivial for the bank: obviously liquid assets should be fire sold, and illiquid assets should be pledged as collateral. This keeps liquidity generation capacity at a maximum and fire sale losses at the minimum (zero). It is shown below that a SNNR applies if the liquidity generating power of the bank assets is at least equal to the deposits of one depositor, and equity is non-negative: Λ+(1-Λ)(1-h) ≥ (1-t-e)/2 and e≥0

Under this condition of liquidity sufficient for paying out one depositor and equity being positive (and assuming that Λ+(1-Λ)(1-h) < (1-t-e), i.e. liquidity is insufficient for paying out both depositors), the bank run game takes the following form. The shaded area is the one in which the bank is liquidated. The areas with bold lines around are the equilibrium solutions of the run game. That (K1, K2) is the unique solution can be shown by directly applying the definition of the SNNR equilibrium. In establishing the pay-offs in case of liquidation, we take here a simplifying assumption, namely that the central bank will liquidate the assets that were posted to it as collateral in a way to “consume” the value buffer created by the haircut, i.e. it will not hand back anything to the insolvency administrator of the bank. This assumption implies that the pay-offs can be calculated in scenarios of liquidity-induced defaults by taking the pay-outs before the moment of default – nothing needs to be added to this as nothing more will be distributed after default. In the present case, the bank balance sheet will look as follows at the moment of default when both depositors would have decided to run (which they should not in the present case as it is not an equilibrium solution of the strategic game). The bank will have fire-sold all its liquid assets, and it will have pledged all its non-liquid assets. If this is not enough to pay out the short-term depositors, then obviously the assumption that the central bank will “consume” the haircut when liquidating the asset also implies that the losses in collateral liquidation will exceed and consume the previous equity of the bank. We call DDi the deposits still held by depositor i in the bank at default. We call Ri the deposits recovered (extracted by depositor i before default).

Bank balance sheet at default if both depositors would run in case Λ+(1-Λ)(1-h) < (1-t-e) Bank Liquid assets (all sold) Λ-Λ Short term debt 1 (1-t-e)/2 - (Λ+(1-Λ)(1-h))/2 Short term debt 2 (1-t-e)/2 - (Λ+(1-Λ)(1-h))/2 Illiquid assets (all pledged to central bank) 1-Λ Long term debt t Equity e Central bank funding (1-h)(1- Λ)

Bank run game with (1-t-e) > Λ+(1-Λ)(1-h) ≥ (1-t-e)/2 and e≥0, i.e. single no-run equilibrium Depositor II keep deposits Run Keep deposits (1-t-e)/2 (1-t-e)/2 - ε Depositor I (1-t-e)/2 (1-t-e)/2 Run (1-t-e)/2 (Λ+(1-Λ)(1-h))/2 - ε (1-t-e)/2 – ε (Λ+(1-Λ)(1-h))/2 – ε

Now consider the case in which liquidity is insufficient to pay out one depositor, but equity is still non- negative. Two equilibriums emerge now, one in with both depositors stay with their deposits, and one, inferior, in which they both run, causing default.

Bank run game with Λ+(1-Λ)(1-h) < (1-t-e)/2 and e≥0, i.e. multiple equilibria Depositor II keep deposits Run Keep deposits (1-t-e)/2 Λ+(1-Λ)(1-h) - ε Depositor I (1-t-e)/2 0 Run 0 (Λ+(1-Λ)(1-h))/2 - ε Λ+(1-Λ)(1-h) – ε (Λ+(1-Λ)(1-h))/2 – ε

Finally, consider the case in which there is sufficient liquidity to pay out one depositor, but negative solvency. The bank is eventually liquidated in this case regardless of what depositors do, and therefore it is always better to run to get out as much money without a haircut before it is too late. For the sake of simplicity of the presentation in the figure below, we denote the recovery ratio for depositors in case both depositors do the same as r, and the one a depositor faces if only the other depositor runs as r’. One can show that 0 ≤ r’< r < 1.

Bank run game with Λ+(1-Λ)(1-h) ≥ (1-t-e)/2 and e<0, i.e. single run equilibrium Depositor II keep deposits Run Keep deposits r(1-t-e)/2 (1-t-e)/2 - ε Depositor I r(1-t-e)/2 r’(1-t-e)/2 Run r’(1-t-e)/2 r(1-t-e)/2- ε (1-t-e)/2- ε r(1-t-e)/2- ε

Central bank collateral to restore financial stability in a financial crisis Assume that asset liquidity (Λ) and asset values can change over time, and also the central bank may change haircuts over time. Moreover, asset values V can change, which is reflected in a change of equity: ∆e = ∆V and thereby potentially leads to a violation of the solvency condition. For example, if initial equity is 0.2 and total assets 1, then a decline of asset values by more than 20% depletes equity and therefore pushes the bank into a single run equilibrium. Asset value deterioration can also lead to a deterioration of the liquidity condition, as it leads to a proportional shrinkage of liquidity relative to short term debt. We can now ask: to what extent can the collateral framework of the central bank as captured by h make a difference for funding stability? First, obviously h does not impact on solvency. Therefore, whenever e<0 we are unavoidably in the case of the single bank run equilibrium. However, h can make the difference for meeting the liquidity condition. We can calculate the maximum haircut compatible with a single no-run equilibrium from the condition of sufficient liquidity.

For given liability costs it and ie, one can now calculate the „cheapest stable funding structure“. In this simple setting, bank will however tend to not issue any equity and achieve their competitive stable funding structure only through sufficient long term funding. This is the limitation of the simplest model setting: it does not allow modelling capital as a safeguard against fire sale losses, and therefore does not contribute to a better understanding of the full liability structure including equity. Consider the following example: if Λ=0.3 and h=0.8 then the funding structure maximising short term deposits is the one in which (1-t-e)/2 = 0.44, i.e. t+e = 0.12 and e≥ 0. For any (it , ie) with it < ie , the cheapest funding structure will be e=0, d=0.88, and t = 0.12

Collateral policies as monetary policies at the ZLB Broadening the collateral set in a liquidity crisis may be a key monetary policy measure, in particular when conventional monetary policy has hit the zero-lower bound. The simple bank run model above allowed to show that when asset liquidity deteriorates, then banks need to move to a more expensive bank liability structure. If they do so quickly enough, the bad bank run equilibrium may not materialize. But a more expensive capital structure means that the spread between the short-term risk free interest rate (controlled by the central bank) and the actual bank financing costs, and thus bank lending rates increase. The central bank could maintain financial conditions unchanged by lowering the short-term risk-free interest rate. This is however not an option if the zero lower bound has been reached and there is therefore a risk that an increase in the spread between short term risk free rates and bank lending rates will push the economy into a deflation. The central bank could therefore broaden its collateral framework so as to make the old, cheap bank liability structure stable again.

We illustrate the last point further with the following bank balance sheet.

Liquid assets Λ Short term debt 1 d/2 Illiquid assets (all pledged to central bank) 1-Λ Short term debt 2 d/2 Equity 1-d

Total liabilities 1 Total liabilities 1

This bank has stable short term funding if Λ+(1-Λ)(1-h) ≥ d/2 => d* = 2(1-h+hΛ), with d obviously capped at 1. Assume that the financial conditions FC are equal to bank lending rates, and these are equal to the average funding costs of the banks, as banks would be perfectly efficient, i.e. would not have any administrative costs and be within a competitive sector. Also assume that the funding costs of deposits is equal to i, the risk-free short-term interest rate, which is the monetary policy interest rate set by the central bank. Assume that the cost of equity funding is equal to the sum of the short-term risk-free lending rate, i, plus the equity risk premium, ω (the equity risk premium contains also term, credit and liquidity risk premia, i.e. the factors τ and λ introduced in chapter 4). Therefore (FC = “funding costs”, or “financial conditions”) are:

FC = sd + (s+ω)(1-d) = s+ ω-dω

By substituting the highest possible share of deposits which ensures stable funding: If d* = 1: FC =s If d* <1:4 FC = s+ ω-ω2(1-h+hΛ) <=> FC = s + ω (2h(1-Λ) -1)

In words: financial conditions tighten (funding costs increases) with (i) short term risk free interest rates, (ii) the equity risk premium, (iii) the haircuts imposed by the central bank, (iv) the share of illiquid assets. Therefore, when the zero lower bound is reached for conventional monetary policy s, then haircuts can contribute to achieve the adequate monetary policy (i.e. the adequate financial conditions), and a decrease in haircuts can be a measure necessary to compensate for an increase of the equity risk premium or a deterioration of asset liquidity. Of course, such a lowering of haircuts for monetary policy purposes should not imply that the role of haircuts to protect the central bank from losses should be

4 Note that the multiplier of ω will be positive out of the assumption that d*<1. Indeed d* = 2(1-h+hΛ) < 1 => - 2h(1-Λ) < -1 => h(1-Λ) > 1/2.

forgotten, i.e. the central bank will face a trade-off between monetary policy objectives and risk objectives.

The model with continuous asset liquidity – power functions In this model variant (Bindseil, 2013), one assumes that (i) assets are continuous in terms of liquidity properties, that (ii) they are equally ranked from both the fire sale loss and central bank haircut perspectives, and that (iii) both haircuts and fire sale discounts have the functional form across the assets of a power function, i.e. haircuts are h(x)=xδ and fire sale losses are d(x)=xθ with δ>0 and θ>0. This continuous, power function based approach to asset liquidity and central bank haircuts has a number of advantages: (i) it allows to differentiate between the roles of equity and long term debt; (ii) it is more realistic than the assumptions taken on assets so far; (iii) the power function is easy to manipulate in the context of our model.

Consider first the case that the central bank would not at all act as lender of last resort, i.e. the only source of liquidity generation in case of a run is to fire sale assets. If the fire sales power parameter θ is close to 0, then the fire sale discounts increase and converge quickly towards 1. If in contrast θ is large (say 10), then discounts are close to zero for most assets and only start to increase in a convex manner when approaching the least liquid assets. If a certain share x of the bank’s assets has to be sold, then the fire sale discounts will have to be booked as a loss and reduce equity. Assuming that the bank starts with the most liquid assets and sells a share x of total asset, the fire sale loss will be xθ+1/(θ+1). Recall that empirical estimates of default costs in the corporate finance literature vary between 10% and 44% (see e.g. Glover, 2011, and Davydenko et al, 2012; Shleifer and Vishny, 1992, provide illustrative examples of discounts in fire sales relative to orderly sales between 15% and 70%, depending on the industry and context ). In fact this cost can be interpreted as the liquidation cost of assets, captured in the parameter θ. Liquidation of all assets will lead to a damage of 1/(1+θ), and sales proceeds (cash generated) will be θ/(1+θ). If default cost is 10%, this would mean that θ = 9, and if default cost is 44%, then θ = 1.27. For a value of default costs in the middle of the empirical estimates of say 25%, one obtains θ=3.5

In the chart below, we illustrate this approach by showing the distribution between liquidity generation and asset fire sale losses under the assumption of a power function of fire sale costs and ranked asset from the most to the least liquid, for the range of the empirical estimates of costs of default, i.e. for 10% (implying θ = 9) and 44% (implying θ = 1.25). Moreover, we show the power function to replicate the estimated recovery ratio in the case of the Lehman Brothers of 28%, i.e. default costs of 72% under the assumption that before default, Lehman would have had just zero equity (see Michael Fleming and Asani Sarkar, April 2014, “The Failure Resolution of Lehman Brothers”, Liberty Street Economics).

5 Note that a loss of job can be interpreted similarly to an asset fire sale. The specialization of the person to its previous job becomes useless, and it takes time before finding a new job (and until becoming good in it). “Employment” is defined as the transfer of the “human assets” to a corporate for a fixed period of time. In case of default, the human capital is fire sold back to the household. Losses for society result from the loss of specialization (idiosyncracy of the asset in the sense of O.E. Williamson) and the transaction costs and time before finding a new use. How these losses are allocated within society depends on law (to the household, to the defaulted company insolvency value, or to the Government / tax payer). To the extent that they are allocated to the defaulted corporate, they will be part of the default related damage done on total asset value.

Fire sale losses and liquidity generation in case total firm liquidation costs would be… 10% (θ = 9) = min. empirical 44% (θ = 1.25): maximum 72% (θ = 0.4): Lehman Brothers estimate empirical estimate

Assume that in the case of a bank run, the bank does whatever it takes in terms of asset liquidation to avoid illiquidity induced default. The total amount of liquidity that the bank can generate through asset fire sales is θ/(θ+1). Therefore, illiquidity induced default will materialise only if deposit withdrawals eventually exceed this amount. Two default triggering events need to be considered, as already highlighted in the time line chart above. Indeed, even if the bank has survived a liquidity withdrawal, it may afterwards be assessed as insolvent and thus be liquidated at the request of the bank supervisor. As noted above, for a given liquidity withdrawal x, the fire sale related loss is xθ+1/(θ+1). Default due to insolvency occurs if this loss exceeds initial equity.6

It can be shown (see e.g. Bindseil, 2013, proposition 2) that a single no-run equilibrium exists if and only if (and assuming again the bank liability structure shown at the beginning of section 7.5):

θ/(θ+1) ≥ (1-t-e)/2 and e ≥ ((1-t-e)/2)θ+1/(θ+1).

The first part of this condition is similar to the discrete case: to ensure financial stability in the case of absence of central bank credit, the liquidity generating capacity of the bank needs to correspond at least to the deposits of one of the two depositors. The second condition could not be captured in the discrete case, and states that the fire sale damage generated by generating through fire sales the liquidity needed to pay out one of the two depositors must not exceed the bank’s equity.

What is the cheapest sustainable liability structure in this model? For given θ, competing banks will always go to the limit in terms of the cheapest possible liability structure as determined by the conditions in the strategic depositor game, such that the no-run equilibrium is still maintained as SNNR equilibrium. Assume that the cost of remuneration of the three asset types are re for equity, rt for term funding, and 0 for short term deposits. Also assume that re > rt > 0. What will in this setting be the composition of the banks’ liabilities? The objective of choosing a liability composition will be to minimize the average overall remuneration rate subject to maintaining a stable short term funding basis. The two minimum conditions to be fulfilled are θ/(1+θ) = (1-t-e)/2 and e =((1-t-e)/2)(1+θ) /(1+θ). These conditions

6 Note that it is assumed that equity is never sufficient to absorb the losses resulting from a bank default, i.e. it is assumed that e ≤ 1/(θ+1). Of course, one could also calculate through the opposite case, but it is omitted here as it does not seem to match reality.

can be solved for a unique optimum t* and e*, and hence for the average necessary remuneration rate of bank liabilities t* rt + e* re.

Now consider the case in which also the pledging of collateral with the central bank is possible. To obtain outcomes in which the banks rely both on fire sales and haircuts in their liquidity stress strategy, we obviously need δ>θ (otherwise it is always superior to only pledge and never to fire sale). It can be shown in the non-trivial case that the bank’s liquidity stress strategy will always foresee the share z of most liquid assets to be fire sold, while the rest, the 1-z less liquid assets, will be pledged with the central bank. The condition for an SNNR was provided in proposition 5 of Bindseil (2013):

Let z in [0,1] determine which share of its assets is foreseen by the bank to be used for fire sales (i.e. the less liquid share 1-z of assets are foreseen for pledging with the central bank). Let k=h(z) be the fire sale losses from fire selling the z most liquid assets and let y=f(z) be the total liquidity generated from fire selling the most liquid assets z and from pledging the least liquid assets (1-z). Then a single no run equilibrium exists if and only if and .

In contrast to the discrete model variant, this variant allows explaining the full capital structure, including the distinction between long term debt and equity, and the roles of these two funding sources can be shown to depend on the relative cost of the two and the relative size of δ compared to θ.

The model and its solution is illustrated in the chart below. The vertical line z separates the liquidity- ranked asset space into the part that will be fire-sold (assets on the left of z, i.e. most liquid assets) and the part that will be pledged (assets on the right of z, i.e. least liquid assets). The bank foresees in its liquidity stress strategy to fire sale the assets [0,z] and to incur fire sale losses of C, and generate liquidity through fire sales equal to A+B. Moreover, in this strategy the bank pledges the assets ]z,1] and generates through this liquidity equal to D. Therefore, total liquidity generated (which must at least be equal to the deposits of one depositor) is A+B+D and total fire sale losses are C (which must not exceed equity e).

To calculate the relevant surfaces, we only need to apply that the integral of x^a in [0,z] is equal to (z^(a+1))/(a+1). The 6 surfaces are therefore:

A = z - z^(δ+1)/(δ+1) C = (z^(θ+1))/(θ+1) B = z – A - B = (z^(θ+1))/(θ+1) – (z^(δ+1))/(δ+1) D = 1 - 1/(δ+1) - A

haircuts function D h(x)=xδ A ←

B Fire sale loss function h(x)=xθ ↓ C F

Example: assume now a variant in which banks’ liabilities consist only in equity and short term debt. Bank A

Assets 1 Depositor 1 (1-e)/2 Depositor 2 (1-e)/2 Equity e

Also assume that initially θ=1.4 and δ=0.5 and e=0.2 so that each depositor has deposits of 0.4. One can now calculate that with z=0.5, one obtains liquidity generating power L= 0.49 and associated fire sale losses FSL= 0.08 (This is done easiest by putting the formulas of the surfaces A, B, C, D, E, F from the chart above into excel). This allows for a single no-run funding equilibrium.

Finally, we will now show how the bank run model can be integrated into a complete set of financial accounts. For this purpose, we assume that two depositors are equally exposed through deposits to two banks. A third household holds the equity of the banks. The banks each finance one NFC, i.e. in some sense the banks are pass throughs to the NFCs and one would obtain the same result if one would consolidate each bank-NFC couple into an NFC financed directly through deposits and equity. It is thereby also implicitly assumed that the bank can discontinue any time the lending to the corporate. For the sake of simplicity, we assume that the household has no demand for banknotes. If it had, then each bank would have to borrow B/2 from the central bank and for that purpose would pledge the necessary amount of illiquid assets, considering the haircut. This would modify the calculus of our simple model, without however amending the logic. We start again from a stable bank funding situation with some θ, δ, e. Now, corporate 2 however suffer from a negative asset value shock leading to a drop of asset value by say e%. Assuming exactly this asset drop will keep calculus rather simple (but also other values could be assumed of course). In the absence of corporate equity, this loss is passed on to bank 2 in the form of write downs on the value of loans. Therefore, bank 2 is now a bank with zero equity and each

short depositor contributing 50% of the bank’s liabilities. One can easily show that there is no longer any z that exists for which liquidity of 0.5 (if we rescale the bank’s balance sheet to an asset value of 1) can be generated and the associated fire sales losses are zero. Assume now in addition that the central bank refuses to restore the single no-run equilibrium by lowering haircuts, and that the run equilibrium eventually materialises. Impact of run and implied asset liquidation. Bank 2 will have to be liquidated, whereby we assume here that this also implies liquidation of the corporate, and that actually in liquidation the bank and its corporate are quasi-consolidated and generate asset fire sale losses as predicted for the bank. This means that out of the remaining asset value of (1-e), another (1-e)/(1+θ) disappear in the context of the re-use of formerly specific assets. What therefore remains as asset value if 1–e–(1-e)/(1+ θ) = (1- e)θ/(1+θ) Transfer of assets and liability from dead bank and dead corporate 2 to bank and corporate 1. One may assume that corporate 1 takes over the productive assets of corporate 2, but because of asset specificity (and various other costs associated to liquidation, such as cost of stand-still, legal costs, etc.) what ends up with corporate 1 is only …, while the rest, 0.34, is a further welfare loss for society (after the initial loss of wealth of 0.2 related to the asset value shock of corporate 2). The funding of the corporate 1’s expansion will take place as shown in the financial accounts below. We assume also that recognition of losses of depositors occurs in this phase 3. Household 3’s equity, and Households 2 and 3’s deposit positions will have declined, and therefore total wealth of society (in the form of total household equity, being equal to total asset values in this case) will have suffered, both reflecting the initial drop in asset values (that reduced bank 2 equity to zero) and the fire sale losses.

Household 1 Real assets HHE – (1-e) Equity HHE - (1-e)/(1+ θ))/2 Deposits bank 1 (1-e)/2 + ((1-e)θ/(1+θ))/2 Deposits bank 2 (1-e)/2 –((1-e)/(1+θ))/2–((1-e)θ/(1+θ))/2 Household 2 Real assets HHE – (1-e) Equity HHE - (1-e)/(1+ θ))/2 Deposits bank 1 (1-e)/2 +((1-e)θ/(1+θ))/2 Deposits bank 2 (1-e)/2 –((1-e)/(1+θ))/2–((1-e)θ/(1+θ))/2 Household 3 Real assets HHE – 2e Equity HHE – e Bank equity 2e – e Bank 1 Loans to corp 1 1 + (1-e)θ/(1+θ) Deposits of HH 1 (1-e)/2 + ((1-e)θ/(1+θ))/2 Deposits of HH 2 (1-e)/2 + ((1-e)θ/(1+θ))/2 Equity e Corporate 1 Real assets 1 + (1-e)θ/(1+θ) Loans from bank 1 1 + (1-e)θ/(1+θ) Bank 2 Loans to corp 2 1–e –(1-e)/(1+ θ) –(1-e)θ/(1+θ) Depos.HH1 (1-e)/2–((1-e)/(1+θ))/2–((1-e)θ/(1+θ))/2 Depos.HH2 (1-e)/2–((1-e)/(1+θ))/2–((1-e)θ/(1+θ))/2 Equity e – e Corporate 1 Real assets 1–e–(1-e)/(1+ θ) –(1-e)θ/(1+θ) Loans from bank 2 1–e –(1-e)/(1+ θ) –(1-e)θ/(1+θ) Central bank Credit 0 Banknotes 0

The following six key conclusions can be drawn from this very simple bank run model: • Both asset value and asset liquidity deterioration can trigger a run. • Insufficient liquidity leads to multiple equilibria, while negative equity always implies a run. • Liquidity cannot prevent runs in case of negative equity. Therefore, the LOLR (i.e. captured in this simple setting by a decrease in central bank collateral haircuts) will not stop a run if equity is negative. However, LOLR action can restore a single no-run equilibrium when only an asset liquidity deterioration and/or an asset value deterioration occurred, as long as equity remains positive. • Tightening collateral rules can destabilize banks by pushing them into the multiple equilibrium case. • In the multiple equilibrium case without the run equilibrium materializing, banks will feel incentivized to adjust their capital structure such as to restore the single no run equilibrium case. This typically leads to a more expensive capital structure, i.e. to more expensive bank intermediation and hence, everything else equal, a monetary tightening. If monetary policy has reached the ZLB, then in this case increasing collateral availability can be an effective monetary policy tool. • If banks through competition and myopic behavior tend to converge to the cheapest sustainable liability structure, then very small shocks of the three types mentioned above can destroy funding stability. It may then be useful to impose liquidity regulation on banks.

Chapter 7: International monetary frameworks 7.1 Introduction 7.2 International gold standard 7.3 Fixed exchange rate system - paper standard. 7.4 Bretton Woods system 7.5 Devaluation 7.6 Foreign reserves 7.7 Flexible exchange rate system 7.8 European monetary union

7.1 Introduction This chapter introduces the main flow of funds mechanics of various international monetary frameworks. It will be shown how the frameworks “digest” capital and current account imbalances and see what limits the systems may encounter. Most of the sections of this chapter are devoted to forms of fixed exchange rate systems (only section 7.6 treats the case of flexible exchange rates). International monetary frameworks often aim at supporting fixed exchange rate systems, or they imply by construction fixed rates, such as the international gold standard. As we will see, under fixed exchange rates, the central bank loses its otherwise unconstrained LOLR powers in its domestic currency. In addition, with fixed exchange rates, central banks lose the power to do independent monetary policy, as monetary policy will be determined by the need to be consistent with the fixed exchange rate. For example, Obstfeld and Rogoff (1995, 74) have therefore concluded that “for most countries, it is folly to recapture the lost innocence of fixed exchange rates”. So why do countries or central banks at all want to have fixed exchange rates or bind themselves to gold such as to lose parts of their power? Why has e.g. the EU launched the euro project and why has China shown so much commitment over the last decades to keep relatively stable its exchange rates by allowing its foreign reserves to fluctuate considerably?

Mainly four motivations for not giving up the idea of fixed exchange rates may be provided: 1) Effective fixed exchange rates allow achieving the network benefits from a more universal money. Two monetary areas linked through a fixed exchange rate system are at least in part like one large unified monetary area. Exchange rate stability contributes to reduce uncertainty and transaction costs (as currency dealers do not need to be compensated for risk taking or for being occasionally exploited by insiders). In particular for a small country, it can be welfare improving to give up its own monetary freedom and to link its currencies to make its own economy benefit from a larger de-facto monetary area. 2) Binding a currency to gold or to another stable currency may allow to obtain credibility as it provides a commitment that can be monitored and that anchors expectations. A commitment to a certain inflation rate is not observable on a day-by-day basis as inflation is linked to policy measures only in a noisy and lagged way. The success of binding a currency to another one (or to gold) can be monitored on a continuous basis. 3) Establish an order to prevent “beggar my neighbour” foreign exchange policies or “exchange rate wars”. For some episodes, observers have felt that exchange rate policies in flexible exchange rate systems have been used to achieve devaluations of own currencies to make domestic industries more competitive and thereby stimulate domestic growth– at the expense of trading partners who experience exactly opposite effects in a sort of zero sum game. A fixed exchange rate system, in particular with some agreed rules and a governance framework, could be seen as a way to overcome incentives for such non-cooperative behaviour which at the end makes everybody worse off.

4) Variable exchange rate systems could be perceived as having a tendency to “overshoot”, i.e. volatility of exchange rates is not just reflecting changing real factors, but additional, endogenously created volatility, e.g. relating to speculation.

For these four reasons, countries have often chosen to try to fix their currency to a precious metal or to other currencies. Different forms of fixed currencies systems exist, as we will see below (e.g. peg to metal; unilateral peg to another currency; multilateral agreement like the Bretton woods framework; EMS; monetary union).

7.2 International gold standard Flow of funds representation of current account and capital account imbalances To represent the flows-of-fund under an international gold standard, we assume that both countries are initially identical. In addition, consider the following for each sector: • Households are not leveraged, and hold deposits in equal shares in both countries, apart from extra holdings in the home bank reflecting their giving up of a part of their gold. Household initially held all the gold of the economy (like all real assets of the economy), but a part of it, “G”, they were ready to give up and hold instead extra deposits with their home bank. Therefore, they hold initially deposits with their home bank at a level D+G and with the foreign banks at a level D – this assumption being without loss of generality. Households hold banknotes equal to B in their domestic currency. • Corporates are identical across the two countries and financed exclusively through bank loans. They hold the real assets given up by the household (with the exception of gold). • The two banking systems are also identical initially. Each banking system has assets of 2D+B and is financed through bank deposits (2D+G) and through central bank credit (B-G). • Each of the central banks has a balance sheet length of B, equal to banknotes issued. In terms of assets, this is matched partially by the gold holdings (G) and partially credit to banks (B-G).

Cross border economic flows (financial flows and those relating to trade) are captured in the balance of payment, which are the financial accounts of a country within the international context. Here, the flow of funds analysis will be limited to basic trade and financial flow transactions. For example the following two events, which also affect the international accounts of a country, will not be captured: • Changes of asset (and liability) valuation. Changes of values of cross border asset and liability prices obviously will affect the international accounts of a country. In a flexible exchange rate system, many such valuation changes will stem from exchange rate adjustments. However, changes of values will not be limited to exchange rate adjustments. For example, a cross border claim in the form of equity fluctuates in value even if exchange rates do not change. Such changes will change the net foreign wealth position of a country. • Transfers, remittances, indemnities. In the balance of payment context, important transactions of this type are international transfers through donations (e.g. a rich country provides development aid to a poor country by giving real goods of money), remittances (a Pakistani accountant working in Abu Dhabi transfers money every month to his family in Pakistan); A country winning a war imposes an indemnity to the country that lost it; A country grants debt relief to another.

We will focus on basic capital and current account transactions of the following types: • Capital transactions, captured as CA in the accounts. We define here simplistically a capital account transaction as one which, without the transfer of real goods (other than gold) as counterpart, changes net financial claims between private sectors of countries. In the gold standard, they are settled in central bank gold. We assume that household B is behind the capital move, namely that

household B opens an account with bank A and then shifts a part of its bank deposit from country B into country A. There are different ways how to imagine those transfers to actually take place. For example, we can imagine that household B withdraws banknotes at B bank, and then goes to the B central bank to cash in the banknotes against gold (as the gold standard means that the central bank promises such conversion), physically travels to A-country, and pays in the gold in a A-bank to be credited with deposits. A-bank does not want to keep the gold, and therefore brings it to the central bank where its account gets credited. A bank also does not want to hold deposits at the central bank, and therefore reduces its central bank credit with these deposits. Capital account transactions do not change the net foreign position of a country (but in a fixed exchange rate system typically changes the net cross border position of the private sector, with opposite changes of the public sector as represented by the central bank). • Current account transactions CU. Here we assume that household A sells a real asset to household B. (We could also have modelled current account transactions through a transaction involving corporates). Household B instruct his bank to credit the account of household A with A-Bank. Again, we can imagine different ways how this transaction is implemented. At the end, it will have the impact on accounts as shown in the accounts below. Current account transactions normally change the net foreign position of a country.

System of financial accounts in gold standard, reflection of current and capital account transactions, assuming that gold reserves are sufficient.

Household A Household B Deposits A-bank D + G + CU Equity E Deposits A-bank + CA Equity E Banknotes B Deposits B-bank D + G – CA - CU Real assets (E-D-B) - G - CU Banknotes B Real assets (E-D-B) - G +CU Corporate A Corporate B Real assets (D+B) Loans (D+B) Real assets Loans from Bank B (D+B) (D+B) A bank B bank Loans to corp (D+B) Deposits D +G +CA+CU Loans to corp (D+B) Deposits D +G -CA-CU CB credit - G -CA- CU CB credit - G +CA+CU CB A CB B Credit to bank B - G -CA-CU Banknotes B Credit to bank B-G +CA+CU Banknotes B Gold reserves G +CA+CU Gold reserves G -CA-CU

Obviously, these accounts can also be shown in the form of an asset liability matrix, such as introduced in section x.

Asset of: → Liability of: ↓ Tot Financial Tot Real assets ↓ Hh A HH B Corp A Corp B Bank A Bank B NCB A NCB B Assets Assets

Real Equity→ 2E E E

Household A (E-D-B) - G - CU D + G + CU B D+B+G+CU E

Household B (E-D-B) - G + CU + CA D+G-CA- CU B D+B+G-CU E

Corp/Gvt A (D+B) D+B

Corp/Gvt B (D+B) D+B

Bank A D+B D+B D+B

Bank B D+B D+B D+B

CB A G +CA+CU B-G-CA-CU B B

CB B G -CA-CU B-G +CA+CU B B

Tot. Fin. Liabs D+B D+B B+D B+D B B

Tot. Liabs E E D+B D+B B+D B+D B B

An example in which the balance of payment is in equilibrium We can recall here some basics of the balance of payments logic. For example, if CA = - CU, then under fixed exchange rates (including the gold standard), the central bank gold (or foreign exchange) reserves do not change. This means that in net terms, capital flows exactly finance the net transfer of goods. There are single transactions that represent both types of transactions at once: For example: a machine is exported from corporate A to corporate B, but not paid yet, such that a financial claim from corporate A to corporate B is created at the same time when the real good passes the border. This transaction takes place entirely in the corporate sector balance sheets. In the accounts below, the value of the transaction is denoted by x, with x = CU = - CA. (Of course, the same transaction could have taken place also between households, or between one household and one corporate, etc.)

Corporate A Corporate B Real assets (D+B) – x Loans (D+B) Real assets (D+B) + x Loans (D+B) Financial claim +x Financial liability -x

It is also worth briefly recalling that capital and current account balances can have various reasons: • Smoothing of consumption path of households: for example, one country may have a particularly low birth rate, and it is then natural that the savings of this country’s household are for some years partially invested abroad to allow to transfer consumption into the future. • Growth rates, and hence real rates of return may be higher in one country than in another. Therefore, also the rates of return on capital investments should be higher in that country. Real assets should move into this country for production purposes, and the flow of real goods should be financed by capital inflows. In this case (and similarly in the previous one), the balance of payment may be balanced as Capital and Current accounts tend to compensate each other. But the net foreign position of the country changes. • There is also the case where capital and current account imbalances both have the same sign and therefore contribute jointly to a payment imbalance. For example, although not under a gold standard, emerging market economies like China in the first decade of 2000 had both large capital inflows and large current account surpluses. Capital goods were imported to China, but exports in consumer goods were so strong that the current accounts were in surplus. Foreign reserves

ballooned in China during this period (under the gold standard its gold reserves would have ballooned). Greece in 2010 seem to have represented the opposite case, i.e. it had both a capital account and a current account deficit. • Capital accounts can be driven by capital flight and then the amplitude of the capital account easily exceeds the one of the current account. This was illustrated by numerous emerging market crisis (Mexico in the 1980s, Thailand, Indonesia and Russia in the late 1990s), and also in the euro area crisis. Hans Werner Sinn (2011) insisted that TARGETII balances (see below) reflected that the “periphery countries” were “living beyond their means” but this seemed to equate the balance of payment to the question of the savings rate of the economy (while consumption is only one part of the current account, and the current account is only one part of the balance of payment). As we will see, capital flight is analogous to a bank run. For a bank it may also take long to build up deposits, but in case of a run these deposits evaporate very quickly. Generally, observers may find current and capital accounts of specific countries in specific periods, as well as cumulated external positions of these countries, as something economically sensible and welfare improving, or as reflecting undesirable imbalances with the potential for financial destabilisation and corresponding welfare damage. For example, the large short term foreign indebtedness of Germany that built up in the second half of the 1920s and that created the subsequent run on Germany in 1931 was assessed early as problematic (see e.g. annual reports of the Reichsbank in the 1920s). Similar cases were often observed until recently with emerging market economies. Recently, US President Trump criticised German current account surpluses as unnatural and problematic, while the Bundesbank defended them as reflecting the German age pyramid and hence the need for the German society as a whole to save through the temporary accumulation of a net external claim.

Alternatives to settlement in gold Returning now to the initial case in which the balance of payment is not balanced, i.e. CA ≠ - CU, it should be noted that the resulting net claim does not necessarily need to be settled through a physical gold transfer, but it can also be settled through the creation of a foreign reserves claim of central bank A towards central bank B. One could for example imagine that also under the gold standard, the two central banks have a settlement agreement in which cross border bank transfers are settled through a counterbalancing central bank position, up to a certain limit beyond which settlement in gold is required. This case is illustrated in the accounts below.

CB A CB B Credit to bank B –G -CA-CU Banknotes B Credit to bank B -G +CA+CU Banknotes B Gold reserves G Gold reserves G Gold claim to CB B CA+CU Gold liab. to CB A CA+CU

Still another alternative is that a claim of central bank A towards bank B is created. Again, the central bank has built up foreign reserves, but this time in the form of claims towards foreign banks (and not to the foreign central bank).

A bank B bank Loans to corp (D+B) Deposits D +G +CA+CU Loans to corp (D+B) Deposits D +G -CA-CU CB credit B - G -CA- CU CB credit B - G Deposit of CB A CA+CU

CB A CB B Credit to bank B –G -CA-CU Banknotes B Credit to bank B - G Banknotes B Gold reserves G Gold reserves G Claim to Bank B +CA+CU

Returning to the base case that the international transactions are settled in gold and affect central bank gold reserves: what is the limit such a transaction will eventually run into? Two limits can be considered: • First, the limit with regard to the share of banknotes that needs to be covered by gold reserves, according to the central bank law. For example, the Reichsbank was subject, according to its mandate established by the Dawes Plan in 1924, to a 40% gold coverage ratio for banknotes. • Second, assuming that gold coverage ratios have been given up, when gold reserves are fully exhausted, such as in the case of the Reichsbank in July 1931. Then, eventually the gold convertibility has to be given up.

Performance of the gold standard The gold standard worked fine during the period 1876-1914, but poorly in the interwar period. The poor interwar performance of the gold standard is explained by e.g. Eichengreen (1992) with the global scarcity of gold and non-collaborative international behaviour in the context of larger capital flow volatility due to the political and financial instabilities. Indeed, the interwar gold standard was in a context of (i) the WWI experience that one cannot rely on a universal commitment of Governments to maintain gold parity (e.g. German mega devaluation by 10^9, with only very few like US and UK not devaluing), nor on the commitment of states to pay their debt (e.g. Russian mega-default of 1917); (ii) unsolved problems of international debt imbalances (war and reparation debt); (iii) political instability, like the rise of fascism and communism; (iv) limited willingness of central bank collaboration, e.g. through inter-central bank loans, i.e. including the absence of an ILOLR (to which the next chapter will turn). With increasing uncertainties after the outbreak of the global financial crisis, central banks were even keener to each hold sufficient gold reserves to be protected against future outflows, probably implying that on average central banks kept interest rates too high in their competition for the global gold stock, triggering deflation and depression. The general weaknesses of a gold standard from a domestic perspective obviously also continued to apply: essentially the absence of an unconstrained monetary policy to achieve price stability and dampen the negative effects of exogenous shocks, and limitations in the LOLR.

7.3 Fixed exchange rate system - paper standard Flow of funds representation of an international paper standard The following financial accounts show the case of a fixed exchange rate system in a global paper standard. We assume that country A is a large country, which does not care about the exchange rate, while country B is a smaller country that does care and that ensures that its currency is pegged 1:1 to the large country A. For example, country A could be the euro area and country B Romania. For being able to defend the currency peg, the central bank of country B (CB B) needs foreign reserves. In the accounts below, we assume that these have a level F and are held in the form of deposits in currency A with A-banks, and that they originated in the past from capital account inflows into country B, which still materialize in the accounts in the form of lending of A-banks to B-corporates of F. What happens now in the case of international monetary flows, relating to new current account and capital account imbalances? Assume for example a current account transaction in which a household in country A sells a car and exports it to a household in country B. • Household A requires payment on its account at bank A and Household B requests his Bank B to make an international payment to the account of household A at Bank A. • Bank B needs for that deposits in A-land. If it has none, it will go to the FX market and will offer deposits with itself, i.e. B-deposits, and demand deposits with some bank in country A, so that it can then transfer those to the bank in A of the exporting household.

• If the market is otherwise in equilibrium, then this FX market transaction of bank B will bring the FX market in disequilibrium, which would push up the price of currency A measured in units of currency B. Central bank B committed to prevent this, and therefore central bank B enters the FX market as counterpart to B-bank. It sells deposits with A-bank to B-bank and debits the current account of B- bank with itself. However, since B-bank needs to restore zero deposits with central bank B, it will increase its credit from central bank B by taking recourse to central bank B credit operations.

Fixed exchange rate system, current and capital account transactions, sufficient foreign reserves (F) Household A Household B Deposits A-bank D + CU Equity E Deposits A-bank CA Equity E Banknotes B Deposits B-bank D – CA-CU Banknotes B Real assets (E-D-B) - CU Real assets (E-D-B) + CU Corporate A Corporate B Real assets (D+B) Loans (D+B) Real assets (D+B) Loans from Bank B (D+B)-F Loans from Bank A F A bank B bank Loans to corp A (D+B) Deposits 2D +CA+CU Loans to corp B (D+B) - F Deposits D -CA-CU Loans to corp B F CB credit B CB credit B – F +CA+CU Deposits CB B F –CA-CU CB A CB B Credit to bank B Banknotes B Credit to bank B – F +CA+CU Banknotes B Foreign reserves F -CA-CU

Exhausted central bank foreign reserves and the ILOLR What if central bank B’s foreign reserves are exhausted, i.e. if CA+CU>F? Eventually, the central bank and the government of country B have to restore macroeconomic conditions that stop and revert the flows that exhausted the foreign reserves (e.g. increase interest rates, strengthen the competitiveness through supply side reforms, etc.). However, such measures typically require some time to be effective. There are two short term options to gain the necessary time: either Central bank B finds an international LOLR to replenish its foreign reserves, or central bank B “defaults” on its promise to fix the exchange rate. In the latter case, the system moves towards a variable exchange rate system. An ILOLR can take two forms: a direct lending between central banks, or through an intermediary like the IMF. Below the latter case is assumed- although in a stylized way. The intermediary (which we call “IMF”) takes a loan from central bank A to obtain A-currency and provides this as credit to central bank B. The loan is assumed here to exactly close the gap of missing foreign reserves to stem the outflow due to current and capital account deficits. System of financial accounts in fixed exchange rate system, with CAP + CUR > F. A bank B bank Loans to corp A (2D+B) Deposits 2D +CA+CU Loans to corp B (2D+B)-F Deposits 2D -CA-CU Loans to corp B F CB credit B -max(0,-F+CA+CU) CB credit B –F +CA+CU Deposits CB B max(0,F–CA-CU) CB A CB B Credit to banks B-max(0,-F+CA+CU) Banknotes B Credit to banks B – F +CA+CU Banknotes B Credit to IMF max(0,-F+CA+CU) Foreign reserves max(0,F–CA-CU) IMF Credit max(0,-F+CA+CU)

IMF Credit to CB B max(0,-F+CA+CU) Credit from CB A max(0,-F+CA+CU)

The intermediary could also have initially, when founded as an international institution, have created a sufficient balance sheet to accommodate such loans. This is displayed in the following diagram where the IMF balance sheet is initially based on paid in capital. This paid in capital is “invested” by the IMF in the form of deposits with the central banks.

CB A CB B Credit banks B-max(0,-F+CA+CU) Banknotes B Credit banks B – F +CA+CU Banknotes B Paid in Capital IMF IMFC/2 IMF dep. IMFC/2 Paid in Capital IMF IMFC/2 IMF dep. IMFC/2 -max(0,-F+CA+CU) IMF credit max(0,-F+CA+CU) Foreign res. max(0,F–CA-CU) IMF Currency reserves IMFC - max(0,-F+CA+CU) Paid in capital IMFC Credit to CB B max(0,-F+CA+CU)

Another example of fixed exchange rate / ILOLR frameworks is ERM II, the new voluntary European Exchange Rate Mechanism introduced for EU countries not (yet) participating to the euro (e.g. ECB, 2006). Participation to ERM II for two years is a criterion for accession to the euro area. In 2008, Denmark, Estonia, Latvia, Lithuania and Slovakia were all in the ERM II, the last four all joined the euro area leaving Denmark as the only ERM II participant. The other EU members never joined the mechanism. Key provisions of ERM II as defined in ECB (2006) may be summarized as follows, whereby the first point covers the setting of “fixed” rates and intervention obligations and the second point covers the ILOLR element. • Bilateral central rates between the euro and participating currencies; +/- 15% intervention bands around the central rates (Art. 1); Interventions at the margin are “in principle” automatic and unlimited (Art. 3); all parties can initiate a procedure aimed at reconsidering central rates (Art. 17). • A “very short term financing facility” (VSTFF) is available to support the central bank that needs to support its currency by providing to the market the other currency (Art 7.1). This central bank shall however make appropriate use of its foreign reserve holdings prior to drawing on it (7.2). The ECB and the other central bank could suspend the VSTFF “if it were in conflict with their primary objective of maintaining price stability” (Art. 7.3). Maturity of the recourse to the VSTFF is 3 months, renewable once (Art 10) and with mutual agreement once more (Art 11). There are ceilings to the access to the VSTFF as set in Annex II of the agreement (ceilings for Denmark: EUR 730 million). The Danish kroner joined ERM II on 1 January 1999, and actually observes a central rate of 7.46038 to the euro with a narrow fluctuation band of ±2.25%. Denmark has never taken recourse to the VSTFF, nor has any of the other countries that have gone through ERM II on their way to euro accession.

7.4 Bretton Woods system Flow of funds representation of the Bretton Woods system The Bretton Woods system was set up in 1944 and included establishing the IMF and the World Bank. The related key convertibility promises were given up in 1971 and 1974, although the role of the IMF as ILOLR has continued until today. The Bretton Woods system was a fixed exchange rate system in which the US committed to fix the value of the USD in gold (and to ensure convertibility), while the others promised to fix the price of their currency in USD (and to defend these fixed exchange rates). Therefore, in principle the US Fed needed gold as reserve asset, while the other members needed USD. The following accounts capture this situation, focusing on the case of flows handled through the foreign reserves of a non-USA country. If country B is Germany, then in the Bretton Woods era (CA+CU)<0, i.e. Germany had balance of payment surpluses and the Bundesbank accumulated foreign reserves.

Household USA Household B Deposits A-bank D+G +CU Equity E Deposits A-bank D + CA Equity E Deposits B-bank D – CA -CU Banknotes B Banknotes B Real assets (E-D-B-G) - CU Real assets (E-D-B) + CU Corporate USA Corporate B Real assets (D+B) Loans (D+B) Real assets (D+B) Loans from Bank B (D+B) -F Loans from Bank A F Bank USA Bank B Loans to corp A (D+B) Deposits 2D +G +CA+CU Loans to corp B (D+B) – F Deposits D -CA-CU Loans to corp B F CB credit B -G CB credit B – F +CA+CU Deposits CB B F –CA-CU CB USA CB B Credit to bank B -G Banknotes B Credit to bank B +CA+CU Banknotes B Gold G Foreign reserves F -CA-CU

Bordo (1993 50-51) notes that a fixed rate system of the Bretton Woods type was subject to the following three problems: Adjustment: According to Bordo, “Under the classical gold standard, balance of payments adjustment worked automatically through the price specie flow mechanism, aided by short- term capital flows. …. Under Bretton Woods, concern over the unemployment consequences of wage rigidity delayed the deflationary adjustment required by a deficit country and, together with the use of short-term capital controls, considerably muted the automatic mechanism. …. The adjustment problem concerned the burden of adjustment between deficit and surplus countries and the choice of policy tools.” Liquidity: “The perceived liquidity problem in the Bretton Woods system was that the various sources of liquidity were not adequate or reliable enough to finance the growth of output and trade. The world’s monetary gold stock was insufficient by the late 1950s, IMF unconditional drawing rights were meagre, and the supply of U.S. dollars depended on the U.S. balance of payments, which in turn was related to the vagaries of government policy and the confidence problem.” The confidence problem: “as in the interwar period, involved a portfolio shift between dollars and gold. As outstanding dollar liabilities held by the rest of the world monetary authorities increased relative to the U.S. monetary gold stock, the likelihood of a run on the “bank” increased. The probability of all dollar holders being able to convert their dollars into gold at the fixed price declined.” This led surplus countries prefer also hoarding gold instead of holding USD, i.e. a run on the US Fed’s gold holdings took place by the surplus countries who made use of the US Fed’s obligation to convert the surplus countries’ USD into gold.

Another way to describe the problem of Bretton Woods is the Triffin-Dilemma (Triffin, 1960). As described by the IMF (Money Matters: An IMF Exhibit -- The Importance of Global Cooperation, IMF Webpage): “If the United States stopped running balance of payments deficits, the international community would lose its largest source of additions to reserves. The resulting shortage of liquidity could pull the world economy into a contractionary spiral, leading to instability. If U.S. deficits continued, a steady stream of dollars would continue to fuel world economic growth. However, excessive U.S. deficits (dollar glut) would erode confidence in the value of the U.S. dollar. Without confidence in the dollar, it would no longer be accepted as the world's reserve currency. The fixed exchange rate system could break down, leading to instability.” In principle, the solution of this problem was supposed to be the “SDR”, the special drawing rights of the IMF which created additional international liquidity.

If the other central banks prefer gold to USD – flow of funds The accounts above were suggesting that the US Fed’s balance sheet is not really affected by the balance of payment deficits of the US – however this was not true as highlighted in the Triffin Dilemma: CB B can also exchange its foreign reserves against gold, and this is indeed what central banks of balance of payment surplus countries, like Deutsche Bundesbank, tended to do. For example, the Bundesbank/Bank Deutscher Länder had no gold reserves in 1949, but more than 3000 tons at the end of the Bretton Woods area in 1974. The desire of e.g. the Bundesbank to not only accumulate US dollar, but also gold, is understandable, since holding US dollar exposes it to risks that the US eventually devalues – which it actually did. A bank run logic applies. The mechanics of the self-fulfilling prophecy works as follows in this case: if surplus countries start to doubt on the ability or willingness of the US to defend the peg of the USD to gold, they are incentivized to start hoarding gold instead of USD, and eventually this leads the US to run out of gold reserves, forcing it to devalue the USD against gold, validating the fears of the surplus countries who “ran” on the USD. Of course, at an early stage, the US could have tried to defend the peg by restrictive monetary policies which would have triggered capital inflows exceeding the negative effects of current accounts. But this would have had significant economic costs in the view of the responsible policy makers, and therefore did not take place. In the financial accounts, the tendency of surplus countries to hoard gold is reflected as follows, assuming the case that country B (Germany) has surpluses and converts these completely into gold.

USA bank German bank Loans to corp A (D+B) Deposits D +G -CA-CU Loans to corp B (D+B) - F Deposits D -CA-CU Loans to corp B F CB credit B -G +CA+CU CB credit B - F +CA+CU Deposits CB B F USA CB German CB Credit to banks B -G +CA+CU Banknotes B Credit to banks B -CA-CU Banknotes B Foreign reserves F Gold G -CA-CU Gold CA+CU

In retrospect, it appears that the Bretton Woods system could have worked relatively well if: • balance of payment imbalances would have been limited; • commitment of countries to defend the pegs would have been very strong, even if it implies domestic adjustment costs (in view of its exposed role, the related commitment and credibility of the US was of overwhelming importance); • the other countries accept to mirror US monetary policies, including e.g. to import inflation, such as to not build up appreciation pressures, such as the DM constantly did. • the other countries accept to accumulate USD as foreign exchange reserves, i.e. do not insist on accumulating gold, in case the US runs balance of payment deficits; • An ILOLR (like the IMF) would have given enough confidence to non-US countries about available liquidity in case of need, such that central banks are not tempted to build up excessively large foreign exchange reserves and thereby contribute to international imbalances. Eventually, there were several changes of pegs before the eventual dismissal of the Bretton Woods pegs in 1974 (see also Bordo, 1993), as shown in the next section.

7.5 Exchange rate adjustments This section provides briefly an overview of the evolution of gold value of currencies and exchange rates since the early days of the gold standard, revealing that exchange rates were only fixed for limited periods, once the 40 years of the stable gold standard ended in 1914. Major milestones for the USD, DM, GBP, FF and CHF were as follows:

United States - [1800] to 1914: 20.67 USD = one ounce of gold (31.1 gram) - 192[5]-1933 Restored gold standard at pre-war parity - 1933 gold standard largely suspended; in 1935 limited convertibility at 35 USD / ounce - 1946 with the launch of the Bretton Woods system, the USD was fixed at 35 USD per ounce gold - 1971 suspension of USD-gold convertibility. In December 1971, the USD was devalued to 38 USD per ounce gold, but anyway full convertibility was not restored. - February 1973, further devaluation of USD by 10% to 42.22 USD per ounce gold (again without actual convertibility). (In 2017 the price of one ounce of gold was approximately 1200 USD)

United Kingdom - [1700] to 1914 (with non- convertibility between 1797 and 1821): 4.248 GBP per ounce of gold - [1925]-1931 return to pre-war gold standard, i.e. 4.248 GBP per ounce of gold or 0.2053 GBP/USD - 1946 devaluation to 0.248 GBP/USD; 1949 to 0.357 GBP/USD; 1967 to 0.417 GBP/USD

Germany: - 1873-1914: 86.814 Mark per ounce of gold - 1924-1931: gold convertibility restored with 86,814 Reichsmark per ounce of gold (1RM=10^9M) - 1931-1945: gold parity in theory unchanged, but no actual convertibility - 1948: 3.33 DM/USD (=> 116,6 DM/ounce gold) - 1949: Devaluation to 4.20 DM/USD - 1961: appreciation to 4.00 DM/USD; 1969: 3.66; 1971: 3.20; 1974: on average 2.59. - 2017: DM/EUR exchange rate of 1.96 in 1999; 1.10 USD /EUR, => DM to be around 1.75 DM/USD

France - 1803-1914: Napoleon introduces the bimetallic standard whereby one French Franc is equal to 0.290032 gram golds, or 95 French Francs buy one ounce of gold (5 Francs = 1 USD). Maintained until 1914 with interruption of convertibility in 1871. - 1928-1933: Gold convertibility restored, gold content of [0.2] grams gold/Franc, i.e. 7.5 FF/USD - 1945: 119 Francs/USD; 1949: 350 Francs/USD; 1958: 494 Francs/USD (1 franc = 0.0018 grams gold). - in 1960, a new FF was introduced with 1 new FF = 100 old FF

Switzerland - 1850 Swiss Francs introduced 1 CHF = 1 FF (0.290032 gram golds) - 1936: devaluation by 30%, i.e. 1 CHF = [0.203125] gram. - 1949-1971: 4.375 CHF per USD (1 CHF = 0.203125 grams of gold) - 1971-1974: Several appreciations of the CHF relative to the USD.

Devaluation in the financial accounts, and settlement of the implied negative central bank capital How to reflect a revaluation in the financial accounts when the exchange rate changes? Assume that currency B appreciates by 100%. The effect on central bank balance sheets are as follows.

CB A (in currency A) CB B (in currency B) Credit to bank B Banknotes B Credit to bank B-F Banknotes B Gold G Foreign reserves 0.5F Negative equity 0.5F

(If currency B depreciates, typically no profits (nor losses) occur for central bank B as most likely this scenario occurs when CB B has exhausted its foreign reserves. If it would have foreign currency reserves,

then it would book a profit, or, if it would be prudent and conservative, it would book instead a revaluation reserves on its liability side.)

The accounts of country A are not affected, but the country B central bank now books a loss and negative equity. If the Government of country B wants to repair this negative equity (e.g. to restore the credibility of the central bank), then it may issue additional debt (or impose extra taxes on households) and recapitalize the central bank. At the end, the devaluation is, unsurprisingly, at the expense of the wealth of the household. One could say that the household sold real assets to A-land, but was only paid for half of the value – retroactively because of the devaluation of A-currency (USD). This is what partially happened in some way to Germany during the Bretton Woods area (although the effect was moderated by the accumulation of gold, instead of only USD).

7.6 Foreign reserves Recent developments of the size of foreign reserves of central banks Although in principle the time of universal fixed exchange rates ended in the early 1970s, many central banks continue to manage their exchange rate by letting their foreign reserves fluctuate accordingly. The following chart illustrate this. Actually, reserve accumulation between 2000 and 2013 is unprecedented in history.

While in 2000, Japan was still the largest holder of foreign reserves and the euro area second, China then rapidly overtook everyone and exceeded in 2013 Japan by a factor of 3. The euro area was also surpassed by a number of other emerging economies. Switzerland is unique in terms of being a small advanced economy and nevertheless ranking third at the end of the captured time period – reflecting its combat against appreciation in view of save haven flows in the context of the euro area sovereign debt crisis. How can one explain this rapid build-up of unprecedented foreign reserves in the years until 2014? In IMF (2011, 9), responses to a questionnaire are reported which provide the following most frequent answers to the question about the reasons for holding (high) reserves: 80%: “Buffer for liquidity needs”; 60%: “Smoothing of exchange rate volatility”; 30%: “Management of exchange rate level”. One might speculate that the frequency of answers also reflect how potentially controversial different explanations are. In reality, the management of the exchange rate level, i.e. preventing appreciation, has likely been the most important reason for the very steep trend of reserve accumulation, which goes beyond needed liquidity buffers.

What do foreign reserves consist in? First, as the following table from the IMF annual report for 2014 (appendix, page 1) reveals, foreign reserves at end 2014 consisted to a very large extent of foreign currency (86%), while gold came second (10%) and IMF related reserves (including SDRs) were third (4%). Second, the currency composition of foreign reserves at end 2013 was (according to IMF annual report for 2014) dominated by USD holdings (66%), followed by the EUR (24%) and the GBP (6%). Third, one can ask in which form the foreign exchange reserves were held: securities, deposits, etc. McCauley and Fung (2003) (see also Borio et al. 2008) review the composition of global foreign exchange reserves in USD and provide the following data on instrument composition (for the year 2000). Securities constitute in total 75%. Deposits and other money market instruments provide the other 25%. It seems noteworthy that the majority of deposits is offshore, i.e. not deposits in USD with US banks, but with banks located in other jurisdictions (mainly London, or other global foreign exchange centres).

Securities: Treasury securities 58% Agency bonds 8% Corporate bonds 1% Equity 8% Deposits / money market instruments Deposits with US banks 3% Money market paper in the US 9% Offshore deposits 12%

How are foreign reserves built up? Essentially one can imagine four ways: (i) Accumulated balance of payment surpluses under a fixed exchange rate system (or managed float). (ii) Creation of mutual foreign reserves through a currency swap, possibly including the involvement of an international organisation like the IMF. This neither requires BoP surpluses, nor will it put pressure on the exchange rate. (iii) Obtaining the foreign claims without counter-flow through e.g. a war indemnity, or a grant. (iv) Acquire reserves in the market without corresponding balance of payment surpluses, assuming variable exchange rates. Banks will, with some elasticity, finance this, i.e. will accept the corresponding capital account flows. But the exchange rate of the country accumulating in this way foreign reserves will likely decline to some extent. Eventually BoP surpluses may then kick in as a consequence of the devaluation.

7.7 Flexible exchange rate system Flexible exchange rates without any central bank intervention require that the FX market has to square itself at any moment in time. Banks can however act as market makers in the foreign exchange market and buffer out short-term random flows stemming from the other sectors – which means that they offer an elastic net foreign position that will provide the balance. We now denominate the accounts of A land in A-currency, and the ones of B-land in B-currency, and reflect that the exchange rate is likely different from 1 by introducing the exchange rate α, i.e α units of B-currency are worth one unit of A-currency. Now, the central bank balance sheet is no longer available for counterbalancing balance of payment flows. Instead, the market has to equilibrate the balance of payment on its own. Below, this takes place by letting banks create between each other cross border claims and liabilities, such that eventually the foreign exchange market is in equilibrium. We call CAHH the capital created by the household and CAB the one contributed by the banking system. The full Capital account is now the sum of the household’s and the banks’ capital account contribution, CAHH + CAB, which is necessarily equal to CU, i.e. such that the capital account exactly balances the current account. If banks are not there to provide some elasticity, then the adjustment of the exchange rate to imbalances of payment flows stemming from households

will be more violent. In other words, the readiness of the financial system to look through short term fluctuations of payment flows and to take temporarily FX exposures is essential in this system to obtain stability (in conjunction of course with adequate central bank and fiscal polies of the public authorities in both countries). In the following financial accounts, the capital account contribution of the banking system, CAB, is shown as CAHH+CU – as indeed it is bound to take this value. Bank A has accepted to export capital into B land by depositing foreign currency in B-bank, while Bank B has accepted to import capital by getting indebted towards bank A. One of the bank also has to hold foreign exchange exposure (depending on whether the deposits is denominated in A or B currency – this is not explicit in the system of accounts below).

The net foreign position of the countries has evolved as it would have done under any other international monetary regime – according to the current account imbalance. The net foreign position is also impacted by the exchange rate. If the claim is denominated in B-currency, then a devaluation of currency B implied that the net foreign position of country A (in A currency) declines (while it did not change for country B), etc.. The central bank foreign reserves do not change, i.e. it is neither involved in current account nor capital account flows. Taking the perspective of country B: (i) Cross border claims of the Household B increase by αX; (ii) The gross cross border liabilities of B bank increase by αX+αCU. The net cross border financial liabilities of country B to country A increase by αCU (all expressed in B- currency). This exactly corresponds to the real goods export from country A to country B.

Household A Household B

Deposits A-bank D + CU Equity E Deposits A-bank αCAHH Equity E Banknotes B Deposits B-bank D – αCAHH-αCU Real assets (E-D-B) - CU Banknotes B Real assets (E-D-B) + αCU Corporate A Corporate B Real assets (D+B) Loans (D+B) Real (D+B) Loans from Bank (D+B)

A bank B bank

Loans to corp (D+B) Deposits D + CAHH +CU Loans to corp (D+B) Deposits D -αCAHH -αCU CB credit B CB credit B Dep. with B bank CAHH+CU Dep. of A bank αCAHH+αCU CB A CB B Credit to bank B Banknotes B Credit to bank B Banknotes B

7.8 European monetary union A monetary union like the euro area can be interpreted as a fixed exchange rate system in which the automatic creation of intra-central bank claims and liabilities plays the role of gold/foreign reserves/IMF loans in standard fixed exchange rate systems. The intra-central bank claims and liabilities are in the case of the euro area the so-called TARGET2 balances, which have found quite some attention starting in 2011 (e.g. Sinn and Wollmershäuser, 2011, Bindseil and König, 2011, Buiter and Rabhari 2012, Cour- Thimann, 2013). The capital flow mechanics in the years up to 2012 are reviewed in more detail in Lane (2013). The mechanics of TARGET2 balances are briefly presented below – in three phases, capturing their history since 1999. Before entering these three phases, consider briefly the general mechanics in a system of financial accounts similar to the one applied to other fixed exchange rate international monetary arrangements. We assume country A is a balance of payment surplus country and country B a balance of payment deficit country. Once more, we assume that both current account and capital account imbalances originate from the household. The two households contribute to capital flight to the same extent by shifting bank deposits from country B to country A. The payment matching the current

100

account transaction is assumed to be from the account of household B with bank B to the account of household A with bank A.

Household A Household B Deposits A-bank D/2+CA/2+CU Equity E Deposits A-bank D/2+CA/2 Equity E Deposits B-bank D/2-CA/2 Deposits B-bank D/2-CA/2-CU Banknotes B Banknotes B Real assets (E-D-B) -CU Real assets (E-D-B) +CU Corporate / Gvt A Corporate and Gvt B Real assets B+D Bank credit B+D Real assets B+D Bank credit B+D Bank A Bank B Loans corp/Gvt D+B Deposits HH D+CA+CU Loans corp/Gvt D+B Deposits HH D-CA-CU

Deposits CB Max(0,-(B-CA-CU)) Credit CB Max(0,(B-CA-CU)) Dep. CB Max(0,-(B+CA+CU)) Credit CB Max(0,B+CA+CU) NCB A NCB B Credit banks max(0,(B-CA-CU)) Banknotes B Credit to bank Max(0,B+CA+CU) Banknotes B Dep. banks max(0,-(B-CA-CU)) Dep. banks max(0,-(B+CA+CU)) T2 claims max(0,CA+CU) T2 liabs max(0, -(CA+CU)) T2 claims max(0,-(CA+CU)) T2 liabs max(0,(CA+CU)) ECB T2 claims CA+ CU T2 Liabilities CA+ CU Consolidated Eurosystem balance sheet = ECB + NCB A + NCB B Eurosystem credit 2B + max(0,-(B-CA-CU)) + max(0,-(B+CA+CU)) Banknotes 2B Deposits of banks max(0,-(B-CA-CU)) + max(0,-(B+CA+CU))

In contrast to foreign reserves, T2 balances are not limited. However, one constraint is the central bank collateral framework and to what extent banks can close the funding gap created by the BoP deficits through additional central bank credit. This is why Hans Werner Sinn and other ECB critics have identified the ECB collateral framework as one culprit for having allowed the Eurosystem contributing to overcome the balance of payment crisis associated with the sovereign debt crisis of 2009-2012. In the accounts above, it is assumed that the banking system of country A is still in recourse to central bank credit: once the cumulated BoP deficits exceed the initial level of banknotes circulating in country A, the banking system would be in a liquidity surplus and the Eurosystem consolidated balance sheet would start to lengthen one-to-one with further BoP surpluses of A-country.

Phase I: Current account deficits and capital account surpluses of the “periphery” Between 1999 to 2007, Target2 balances were rather flat – despite large imbalances of sub-accounts of the balance of payment. Indeed, these years were characterized by current account surpluses of the “core” towards the “periphery”. The following table shows the cumulative current account balances in this period, relative to the GDP of the relevant country in 2007 (although not distinguishing between intra-euro area imbalances and those towards the rest of the world). In addition, we show the average growth rates and savings ratios in these countries. Finally, we also show, for illustrative purposes, the US and China.

Portugal, Greece and Spain have large cumulative current account imbalances. While both Spain and Greece grew by close to 4% annually, Portugal grew by less than 2%, which could have suggested that the current account deficits of Portugal could have been more problematic as it could have been more “consumptive”. Portugal also had the lowest savings ratios. Target balances in contrast changed only marginally during this period, suggesting that current account deficits were financed by capital inflows. Special inflows, such as e.g. structural EU funds, were also non-negligible. It seems that creditors believed that these countries had better growth prospects than Germany, and therefore deserved capital inflows essentially to finance investments into a growing and more dynamic economy.

101

1999-2007 (9 years): Balance of payment Balance of payment items – all in % or 2007 GDP Information items Cumulative Cumulative Change of Change of Residual Average real Average Current Capital T2 balances foreign GDP growth savings account account reserves rate ratio (broad) (+ = claim ↑ (+ = surplus) (+ = deficit) or liability ↓

Germany 19,3 20,5 2,1 -1,0 -2,2 1,7 23,2 Spain -36,4 -28,0 -3,2 -3,1 -2,1 3,9 22,5 Portugal -58,7 -50,7 -3,7 -4,5 0,3 1,8 16,0

Greece -40,2 -37,2 -1,2 -1,7 -0,1 4,1 15,7 Italy -3,4 -10,1 3,0 -0,2 3,9 1,5 20,7 US -34,4 -33,2 0,0 -0,1 -1,1 2,9 18,7 China 25,9 -13,3 0,0 40,1 -0,9 10,2 42,8

In the financial accounts below, we broadly capture this first phase. To simplify, current account and capital flows are assumed to exactly compensate each other (CA = CU), i.e. the countries with current account deficits saw corresponding capital inflows, etc. We capture the Phase I current account by C, and assume to be exactly balanced by the capital account, also expressed by C.

Phase IIa: sentiment towards periphery turns negative, triggering capital flight (2008-2012). Now, the sentiment towards the soundness of e.g. the Spanish growth model suddenly deteriorates: what seemed economically convincing before is suddenly regarded as a bubble and as unsustainable. Probably, the positive beliefs pre-crisis were exaggerated, but so were as well the sudden fears that e.g. the Spanish banking, corporate and government sectors would be insolvent. However, if investors believe so, then they should “run” on Spain, i.e. withdraw credit and do not roll over debt exposure (as explained in the stylized bank run model of chapter 6). The debt haircut on Greece shocked international investors and made them fear similar decisions for other euro area countries.

The Phase IIa capital flight and the role of the central bank in preventing defaults in view of the illiquidity of the real asset stock that was financed through the inflows is captured in the system of accounts below as CA. The central bank basically acts as lender of last resort (chapter 7) of the banking system in country B. This should be based on the assumption that the banking system of country B is solvent. As one may recall, the solvency of the Spanish banking system was also supported by an EU/IMF program specifically dedicated to its recapitalization. The combination of this capital support program (by Governments) and the LOLR (by the Eurosystem) was considered instrumental in turning around Spain and putting it back on a growth path as of 2014 – without any losses for the official sector lenders. The following table presents the period 2008-2012 (5 years). This period in fact encompasses rather heterogeneous sub- periods: for Italy and Spain, the large-scale capital outflow took place in 2011/2012, while for Greece it started in 2009. In 2008, flows were still more limited.

In assessing the capital outflows, it should be noted that they contain for GR and PT large adjustment program funds. Starting from the (still simplistic formula): “Private capital flows + Public sector capital flows + Change in T2 balances + change of foreign reserves = Current accounts”, one can now match the figures to identify exactly the private capital flight. This would have been 82% of GDP for Greece, 6% of

102

GDP for Spain, and 5% for Portugal. As current account imbalances initially remained high in some of the crisis countries, it seems that this period contained still some of the factors that had driven Phase I, namely current account imbalances. However, what is very different in this phase is the role taken by Target 2 balances (in the case of PT and GR apparently complemented by a shrinking of foreign reserves of central banks) to replace parts of the previous capital inflows.

2008-2012 (5 years) BoP Balance of payment items – all in % or 2012 GDP Information items Cumulative Cumulative Of Change Change of Residual Average real Average Current account Capital which of T2 foreign GDP growth savings account public balances reserves rate ratio (broad) sector loans Germany 30,4 3,5 -x 19,6 0,6 6,7 0,7 25,8 Spain -24,1 1,1 +3% -29,0 1,8 2,0 -1,2 19,8 Portugal -46,3 -5,0 +13% -28,0 -12,8 -0,6 -1,3 11,8

Greece -61,9 -18,0 +95% -33,0 -11,4 0,4 -5,3 7,0 Italy -12,4 1,3 -x -17,2 1,0 2,5 -1,4 17,7 US -15,0 -15,1 0,0 0,5 -0,4 0,7 15,6 China 14,6 -8,9 0,0 21,4 2,1 9,4 50,9

Phase IIb: partial reversal of capital flight (2012-2014). In this phase, the mood lightened again, the crisis abated, and the effects of Phase IIa are partially reversed, i.e. capital flows back partially into the periphery.

Phase III: renewed rise of Target2 balances because of the ECB’s purchase programme (2015-2016) Target2 balances started again to increase in 2015 with the launch of the ECB’s large scale asset purchase programme APP, reflecting a general concentration of accounts of clients to hold excess liquidity in the “core” countries, and in particular Germany (see Eisenschmidt et al, 2017). This was no longer linked to acute fears and acute capital flight. Maybe still a remote risk perception may have contributed to the general tendency of excess liquidity to accumulate in the core. Also, generally the euro securities settlement accounts of foreign bond holders and of the international banks that dominate the global bond market tend to be in “core” countries such as Germany, Netherlands or Luxembourg. We capture this case in the accounts below, in which we postulate that bondholders are all located (at least in terms of their bank accounts) in country A, while the securities purchases are conducted equally by the two NCBs. As the accounts illustrate, it does not matter whether the securities are issued by an issuer in country A (“Federal Republic of Germany”) or country B (“Republic of Italy”).

Of course, one could also translate these flows into capital flows, i.e. into the perspective of the balance of payment. The actual capital account imbalance results from the fact that before the securities purchase programme of the central bank, the household A had a cross border claim towards say the Government of country B (an investment fund with his accounts at a bank in Germany held Italian government bonds), while after the purchase programme, this cross-border claim has shrunk.

103

Household A (Germany+International investors) Household B (“Italy”) Deposits A-bank D - S + 2APP Equity E Deposits B-bank D Equity E Securities S - 2APP Banknotes B Banknotes B Real assets (E-D-B) Real assets (E-D-B) Euro area corporate and Government sector Real assets 2D+2B Credit from banks 2D+2B -S Securities issued S A bank B bank Corp loans (D+B-S) Deposits D-S+2APP Corp loans (D+B) Deposits D CB deposits Max(0,-(B-2APP)) CB credit Max(0,B-2APP) CB credit B NCB A NCB B CB credit to bank A Max(0,B-2APP) Banknotes B CB credit to bank B B Banknotes B Asset purchase program APP Dep. A-bank Max(0,-(B-2APP)) Asset purchase program APP T2 Liabilities APP T2 claims APP ECB T2 claims APP T2 Liabilities APP Consolidated Eurosystem balance sheet = ECB + NCB A + NCB B Eurosystem credit B + Max(0,B-2APP) Banknotes 2B Asset purchase program 2APP Deposits of banks Max(0,-(B-2APP))

The following chart shows the evolution of T2 balances of euro selected area central banks, with the Target2 phases marked with vertical bars.

Some (German) authors have requested that euro area Target II balances would be “settled” to reduce risk taking of core countries from the large-scale Target II claims and provided as example the case of the US “Intra-district settlement account” (ISA) which would foresee such settlement. However, as e.g. Klose and Weigert (2012) note, this argumentation is flawed. Generally, it should be remarked that a true “settlement” of cross border net financial claims can be achieved only through a few specific ways, as already noted by Keynes (1919) in the context of German war reparation, such as (i) debtor countries cumulating current account surpluses; (ii) transfer of truly external assets, such as gold, foreign real assets, or foreign reserves; (iii) novation of claims and liabilities involving third countries. What the ISA accounts do in the US is not more than replacing one cross federal reserve district financial claim by another one. E.g. if the Fed New York would have net liabilities toward the other federal reserve banks, and New York would decide to quit the US, then none of the various types of intra-Federal reserve district bank claims would be necessarily enforceable, i.e. correspond to assets to which the other Federal reserve banks could really have unconditional recourse to. As Klose and Weigert (2012, 246) note, the ISA settlement does not involve the actual transfer of securities, but only creates a claim on a share of the overall portfolio.

104

In case of the euro area, the following considerations apply to the settlement idea. The amounts of “foreign” assets (gold, foreign reserves) owned by the Target 2 liability NCBs are limited and would only allow a small partial settlement of balances. This is not a sign of weakness of the NCBs, but relates to the fact that NCBs were not asked to maintain or build up such reserves, and that absent a business case, foreign reserves tend to add financial risks without purpose. Generally, the idea to link the scope for capital flows to the availability of foreign reserves seems to be associated with the case of a fixed exchange rate system, in which the central bank defending a peg can do so for as long as it can inject its foreign reserves into the market, but then eventually may have to give up the peg and return to flexible exchange rates. This is in principle the framework of the European Exchange Rate Mechanism that prevailed between 1979 and 1999. The decision to move to EMU was about overcoming the weaknesses of this system (including the danger of speculative attacks / run on a peg, anticipating that the central bank may exhaust its foreign reserves and then will be forced to devalue).

“Settlement” through domestic financial assets is of more limited value from the perspective of risk protection, as it just replaces one cross border financial claim by another one of similar limited enforceability in a scenario of break-down of international collaboration. For example, one could settle the T2 claims by (i) transferring Government bonds or other financial claims like credits to banks from NCB to NCB balance sheet; or (ii) bundling all Eurosystem security holdings at the ECB (as they are bundled in the US at the Fed NY) and to settle the claims and liabilities through changing the NCB claims towards the portfolio held by the ECB. Total securities holdings of the Eurosystem at end February 2017 were around 2.1 trillion euro and claims to credit institutions were around EUR 0.8 trillion. At the same time, total target liabilities were around EUR 1.2 trillion. Therefore, it would seem that Target 2 liabilities could indeed be “settled” in this sense through transfers of securities or claims to centralised securities portfolios. In fact, by definition, Target 2 liabilities of an NCBs can never exceed the assets of the NCB, and in this sense “settlement” would always be possible. The problem remains that this type of settlement changes little to the nature of the net financial cross-border claims of central banks. Only if one believes that owning a Government bond or having a collateralised claim against a bank in the other country is more secure than having a T2 claim, it could be considered as a relevant change. A truly

105

revolutionary and potentially irrational Government can expropriate foreign creditors regardless of the form of their claim (including whether or not it is collateralised with domestic assets). This is what has happened in the Soviet Union, China or some Latin American countries in the course of the 20th century. This being said, it is also true that the form of cross border claims is not irrelevant for the costs for a government to repudiate them – in particular if the government is only moderately revolutionary, i.e. does not want to completely re-invent the financial system and the economic organisation of the economy.

List of References Abraham, M., T. Adrian, R.K. Crumb, E. Moench and R. Yud (2016), Decomposing real and nominal yield curves, Journal of Monetary Economics, 84, 182-200. Aglietta, M. and B. Mojon (2012), “Central Banking”, in: The Oxford Handbook of Banking, edited by A. N. Berger, P. Molyneux, and J. O. S. Wilson, Oxford University Press. Akerlof, G. (1970), “The market for lemons: Quality uncertainty and the market mechanism”, Quarterly Journal of Economics, 84, 488–500. Ayuso, J. and Repullo, R. (2003), “A model of the open market operations of the European Central Bank”, Economic Journal, 113, 889–902. Bagehot, W. (1873), Lombard Street: A Description of the Money Market, H.S. King Ball, Laurence (2014), “The Case for a Long-Run Inflation Target of Four Percent”, IMF Working Paper, WP/14/92. Bech, M.L. and A. Malkhozov (2016), “How have central banks implemented negative policy rates?”, BIS Quarterly Review, March 2016. Benes, J. and M. Kumhof (2012), The Chicago Plan Revisited, IMF Working Paper, WP/12/202. Bindseil, U., K. Nyborg, and I. A. Strebulaev (2009), “Repo auctions and the market for liquidity”, Journal of Money, Credit and Banking, 41, 1391–421. Bindseil, U. and P. König, (2011), “The economics of TARGET2 balances”, SFB 649 Discussion Paper 2011-035. Bindseil, U. (2013), “Central bank collateral, asset fire sales, regulation and liquidity”, Working Paper Series, No 1610, European Central Bank. Bindseil, U. and J. Jablecki (2013), ‘Central bank liquidity provision, risk taking and economic efficiency’, ECB Working Paper, No. 1542. Bindseil, U. (2014), Monetary policy operations and the financial system, Oxford University Press. Bindseil, U. C. Domnick and J. Zeuner (2015), “Critique of accommodating central bank policies and the ‘expropriation of the saver’ - a review”, No. 161. Bindseil, U. (2018), Pre-1800 central bank operations and the origins of central banking, 15 January 2018, WP. Bindseil, U. (2018a), “Some pre-1800 French and German central bank charters and regulations,” available in SSRN. Bindseil, U., M. Corsi, B. Sahel, A. Visser (2017), The ECB collateral framework explained, ECB Occasional Paper No. 189. Bolton, P. and X. Freixas (2006), “Corporate finance and the monetary transmission mechanism”, Review of Financial Studies, 19, 829–70. Bordo, M. (1993), “The Bretton Woods international monetary system: A historical overview.” (Eds.) M.D. Bordo and B. Eichengreen, A Retrospective on the Bretton Woods System, University of Chicago Press, Chicago. Borio, C., G. Galati, A. Heath (2008), FX reserve management: trends and challenges, BIS Papers No 40, Monetary and Economic Department. Brunnermeier, M., A. Crockett, C. Goodhart, A. D. Persaud, and H. Shin (2009), ‘The fundamental principles of financial regulation’, Geneva Reports on the World Economy 11, International Center for Monetary and Banking Studies. Buiter, W. H. (2009), “Negative nominal interest rates: three ways to overcome the zero-lower bound”, NBER Working Paper 15118. Buiter, W. and E. Rabhari (2012), “The European Central Bank as lender of last resort for sovereigns in the eurozone”, Journal of Common Market Studies, 50, 6-35. Campbell, J.R., C.L. Evans, J.D.M. Fisher and A. Justiniano (2012), “Macroeconomic effects of Federal Reserve forward guidance,” FRB of Chicago Working Paper, WP 2012-03.

106

Chailloux, A., Gray, S. and McCaughrin, R. (2008), “Central Bank Collateral Frameworks: Principles and Policies”, Working Paper, 08/222, International Monetary Fund. Coase, Ronald (1937), “The nature of the firm”, Economica, 4, 386–405. Cochrane, J. (2011), “Determinacy and Identification with Taylor Rules”, Journal of Political Economy, Vol. 119, No 3, pp. 565- 615. CGFS/MC (2015), Regulatory change and monetary policy, Report submitted by a Working Group chaired by U Bindseil and W. R. Nelson (Federal Reserve Board) established by the Committee on the Global Financial System and the Markets Committee. CPMI/MC (2018), Central Bank Digital Currencies, BIS Report, March 2018. Crosbie, P. and J. Bohn (2003), “Modelling Default Risk”, Moody’s KMV. Cour-Thimann, P. (2013), “Target balances and the crisis in the euro area”, CESIFO, Forum, Volume 14, Special Issue D’Amico, S. and T.B. King (2010), “Flow and stock effects of large-scale Treasury purchases”, Finance and Economics Discussion Series, Federal Reserve Board, Washington, D.C.2010-52 Davydenko, S. A., I. A. Strebulaev, and X. Zhao (2012), “A market-based study of the cost of default”, Review of Financial Studies, 25, 2959–99. Diamond, D. W. and P. H. Dybvig (1983), “Bank Runs, Deposit Insurance and Liquidity”, JPE, 91, 401–19. Dyson, B. and G. Hodgson (2016), Digital Cash, Why Central Banks Should Start Issuing Electronic Money, Working paper, positivemoney.org. Eichengreen B. (1992), Golden fetters: The gold standard and the great depression, 1919-1939, New York: OUP. Eisenschmidt, J., D. Kedan, M. Schmitz, R. Adalid, P. Papsdorf (2017), “The Eurosystem’s asset purchase programme and Target balances”, ECB Occasional Paper, No. 196. European Central Bank (2004), ‘Risk mitigation methods in Eurosystem credit operations’, Monthly Bulletin of the ECB, May 2004, 71–9. European Central Bank (2006), Agreement of 16 March 2006 between the European Central Bank and the national central banks of the Member States outside the euro area laying down the operating procedures for an exchange rate mechanism in stage three of Economic and Monetary Union, Official Journal of the European Union, 25.3.2006 Ewerhart, C. and Tapking, J. (2008), “Repo markets, counterparty risk and the 2007/2008 liquidity crisis”, ECB Working Paper, No. 909. Filardo, A. and B. Hofmann (2014), “Forward guidance at the zero lower bound”, BIS Quarterly Review, March 2014, 37-54. Flannery, M. J. (1996), “Financial crises, payment system problems, and discount window lending”, Journal of Money, Credit and Banking, 28, 804–24. Fudenberg, D. and J. Tirole (1991), Game Theory, MIT Press. Gali, J. (2008), Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Framework, Princeton University Press. Glover, B. (2011), “The expected costs of default”, working paper, Carnegie Mellon University. Goodhart, C. A. E. (1999), ‘Myths about the lender of last resort’, International Finance, 2, 339–60. Goodhart, C. and G. Illing (eds.) (2002), Financial Crises, Contagion, and the Lender of Last Resort—A Reader, Oxford University Press Gray, D. and S. Malone (2008), Macrofinancial Risk Analysis, Wiley Finance. Hamilton, A. (1790), Second report on the further provision necessary for establishing public credit (report on a National Bank). Treasury Department. Holmström, B. and J. Tirole (1997), “Financial intermediation, loanable funds, and the real sector”, Quarterly Journal of Economics, 112, 663–91. Huber, Joseph (1999), “Plain Money. A Proposal for Supplying the Nations with the necessary Means in a modern Monetary System”, Revised version October 1999. Hull, John C. (2018), Risk Management and Financial Institutions, 5th Edition, Wiley Finance. IMF (2014), IMF Annual Report 2014: From stabilization to sustainable growth, Washington. Keynes, J.M. (1919), The economic consequences of the peace, Harcourt, Brace and Howe. Kindleberger, C.P. (1984), A Financial History of Western Europe, London, George Allen & Unwin. Kindleberger, C. P. and R. Aliber (2011), Manias, Panics and Crashes, 6th edition, Wiley. Klose, J. and B. Weigert (2012), "Das Verrechnungssystem der Federal Reserve und seine Übertragbarkeit auf den Euroraum," Wirtschaftsdienst, Springer; German National Library of Economics, vol. 92(4), pages 243-250, April.

107

KPMG Island (2016), Alternative monetary systems, A report commissioned by the Icelandic Prime Minister’s Office. Kyle, Albert S. (1985), ‘Continuous auctions and insider trading’, Econometrica, 53, 1315–36. Lane, Philip R. (2013), Capital flows in the euro area, European Commission, Economic Papers, No. 497. Laubach, T., and J. C. Williams (2003). “Measuring the Natural Rate of Interest,” Review of Economics and Statistics, 85, 1063– 1070. Laubach, T. and J.C. Williams (2015), “Measuring the Natural Rate of Interest Redux”, FRB of San Francisco WP 2015-16. Law, J. (1705/1750), Money and Trade consider’d with a proposal for supplying the nation with money, First published in Edinburgh in 1705, Glasgow: A. Fulis. McCauley, R. and B. Fung (2003): “Choosing instruments in managing dollar foreign exchange reserves”, BIS Quarterly Review, March, pp 39–46. Moody’s (2008), ‘Corporate default and recovery rates, 1920–2007’, Special comment, Moody’s Investors Service. Mundell, R. (1961), “A Theory of Optimum Currency Areas”, American Economic Review, Vol. 51, 657-665. Nyborg, K. (2016), Collateral frameworks: the open secret of central banks, Cambridge: Cambridge University Press. Obstfeld M. and K. Rogoff (1995), Foundations of International Macroeconomics, MIT Press. Rochet, J.-C. and X. Vives (2004), “Coordination failures and the lender of last resort: was Bagehot right after all?”, Journal of the European Economic Association, 2, 1116–47. Rogoff, K. (2016), The curse of cash, Princeton University Press. Richter, R. (1990), Money: Lectures on the Basis of General Equilibrium Theory and the Economics of Institutions, Berlin: Springer Verlag. Schacht, H. (1928), Die Stabilisierung der Mark, Deutsche Verlags-Anstalt, Stuttgart. Schacht, H. (1953), 76 Jahre meines Lebens, Kindler und Schiermeyer Verlag, Bad Wörishofen. Schmitt-Grohe, S. and M. Uribe (2010), The optimal rate of inflation, NBER Working Paper 16054 Sinn, H.-W. and T. Wollmershäuser (2011), ‘Target loans, current account balances and capital flows: the ECB’s rescue facility’, NBER Working Paper, No. 17626. Standard & Poors (2013), “Annual global corporate default study and rating transitions”, Ratings direct Stiglitz, J. E. and Weiss, A. (1981), “Credit rationing in markets with imperfect information”, American Economic Review, 71, 393– 410. Sveriges Riksbank (2017): The Riksbank’s e-krona project – Report 1, September. Swanson, E.T. (2017), “Measuring the effects of Federal Reserve forward guidance and asset purchases on financial markets”, NBER WP N. 23311. Thornton, H. (1802), An Inquiry into the nature and effects of the paper credit of Great Britain, Edited with an introduction by F.A. Hayek, 1962, New York: Augustus M. Kelley. Tobin, J. (1963), “Commercial banks as creators of money”, in Dean Carson (ed.), Banking and Monetary Studies, 408–19, for the Comptroller of the Currency, U.S. Treasury, Richard D. Irwin. Triffin, R. (1960): “Gold and the dollar crisis”, Yale University Press, New Haven. Ugolini, S. (2017), The evolution of central banking: theory and history, Palgrave-Macmillan. United Nations (2008), “Updated System of National Accounts (SNA): Chapter 4: Institutional units and sectors”. Välimäki, T. (2003), Central Bank Tenders: Three Essays on Money Market Liquidity Auctions, BOF Studies E:26. Werner, R. (2014), “Can banks individually create money out of nothing? — The theories and the empirical evidence”, International Review of Financial Analysis, 36, 1–19. Werner, R. (2015) “A lost century in economics: three theories of banking and the conclusive evidence”. International Review of Financial Analysis, 1-19. Wicksell, K. (1898/1936), Interest and Prices: A Study of the Causes Regulating the Value of Money, Macmillan, 1936. Translation of: Geldzins und Güterpreise, Eine Studie ueber die den Tauschwert des Geldes bestimmende Ursachen, Gustav Fischer, 1898. Williamson, O.E. (1984), The economic institutions of capitalism, The Free Press. Woodford, M. (2003), Interest and prices, Princeton University Press. Ynesta, I. (2008), “Households’ Wealth Composition Across OECD Countries and Financial Risks Borne by Households”, OECD Journal: Financial Market Trends, vol. 2008, issue 2, 1-25

108