The demand for safe assets in France : a long time perspective

(1)

University of Amsterdam

Faculty of Economics and Business

Master of Science in Economics – Specialization: Monetary Policy, Banking and

Regulation

The Demand for Safe Assets in France: A Long

Time Perspective

Ernest Lecomte

Student ID: 11646985

Master’s thesis under the supervision of professor Dr. Enrico Perotti

Spring semester 2018

(2)

Abstract

Safe assets – information-insensitive assets that can be used to transact without fear of adverse selection – play a key role in the economy. Their use as collateral in financial transactions, as safe store of value through time, as well as for meeting prudential regulation, make them socially valuable. Two key characteristics make safe assets intriguing: their demand as a share of total assets has remained surprisingly stable over the last 60 years in the United States, and several studies have found that the financial sector actively responds to a scarcity of safe government debt by issuing liquid, safe liabilities against its own risky, illiquid assets. While the growing literature on the subject has mainly focused on the US, the present study seeks to assess whether the two aforementioned characteristics of safe assets hold true for France.

Key words: Safe Assets, Sovereign debt, Monetary Economics, Privately-issued Safe Assets

Statement of Originality

This document is written by Student Lecomte Ernest who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Acknowledgements

This master’s thesis would not have been possible without the help of Dr. Enrico Perotti who carefully guided me throughout the entire research process and always answered my questions with great care. You have set an example of excellence as a researcher, mentor, instructor, and role model.

I would like to thank Dr. Massimo Giuliodori for giving me some last minute, yet insightful advices on my work.

I would also like to thank my amazing classmates for the time we have spent together, the intellectually challenging discussion we have had, and your contribution to my thesis through various discussions and through the moral support you have provided me with. A special thanks to Simone Fumagalli and Cole Pearce.

Lastly, I would like to thank my family and Flora Wierzbicki for their unconditional moral support during my master program and my thesis in particular.

(4)

I. Introduction

1.1. Background and Significance

Safe assets have in recent years been under scrutiny by policy-makers and academics alike, as reflected by the burgeoning literature on the nature of safe assets and on their role in the global economy. Safe assets are assets that can be used to transact without fear of adverse selection; that is, there are no concerns that the counterparty privately knows more about the value of the asset (Gorton, 2016). They can easily be used to exchange for goods or services or to exchange for another asset, which explain why they are referred to as “money-like” (Krishnamurthy et al., 2012). Furthermore, safe assets can also be used to store value through time or be used as collateral in transactions between financial entities (Trésor Direction Générale, 2013). Gorton et al. (2012) notes that there is a substantial demand for collateral for purposes of mitigating counterparty risk in derivatives and settlement systems. The collateral in these transactions is required to be extremely safe, which also drives the demand for a safety attribute (Gorton et al., 2012). Safe assets therefore perform a critical role in the global financial system and have implications for transactions and savings efficiency, financial crises, and monetary policy (Gorton, 2016).

A key characteristic of the global economy in recent years has been the apparent shortage of government safe assets. Massive financial flows from emerging markets in search of safety – including central banks – have flooded the US and Europe for the past two decades, with the consequence that interest rates were continuously pushed down while asset price skyrocketed, thereby playing a significant role in fueling the credit boom that led to the 2008 financial crisis (Caballero and Krishnamurthy, 2009). It has been documented that in respond to such shortage of government safe debt, financial intermediaries have actively created “quasi-safe” assets in the form of short-term liabilities, effectively carving out “safe assets” from risky ones. In the case of the US, this process has relied heavily on securitization (ibid). With some of these assets losing their risk-free status in the wake of the recent financial crisis, and with demand for safe assets expected to exhibit structural growth, particularly as a result of prudential reforms, some analysts are worried that risk-free assets may become increasingly scarce (Trésor Direction Générale, 2013).

(6)

Safe assets are safe debt instruments – i.e. liabilities – that are mainly issued by two entities: the government – which includes, using a wide concept, the central government, the central bank as well as governmental agencies and administrations – and the financial sector. High-quality, short-term debt issued non-financial firms – in the form of high-grade commercial paper and AAA corporate bonds – can also be considered safe. Government-issued and government-guaranteed safe assets are the safest – with Treasuries and central bank liabilities being the quintessential form of safe assets – while certain financial sector liabilities also carry a safety attribute. Several studies have indeed provided evidence for the existence of a safety premium on privately issued assets – provided that their maturity is short enough (Krishnamurthy and Vissing-Jorgensen, 2012; 2015, Pérignon et al., 2017). In that sense, private and government safe assets are substitutes, albeit imperfect ones given the lower default risk of sovereign debt.

The moneyness and safety of those assets translates into a convenience yield – referred to as the safety premium – which captures the lower interest rate that investors are ready to accept for the non-pecuniary qualities those safe assets offer. This conceptualization of the transactional qualities of money finds its origin in the “money in the utility function” approach, which many contemporary economic models hinge upon (Sidrauski, 1967, in Perotti and Golec, 2016). Through price and quantity adjustment mechanisms, the demand for safe assets implies a lower yield on the latter relative to assets of similar liquidity but lower safety, or similar safety but lower liquidity. Two major characteristics of safe assets have attracted attention from economics researchers and justified the present research. The first one is the apparent stable demand for safe assets. Several studies have found that the demand for the latter in the US – as a proportion of wealth (Gorton et al. 2010) or of GDP (Gourinchas, 2012) – has been stable over the last half century. In other words, economic agents – households, firms, and financial-corporations – desire to hold a relatively constant share of their portfolio in the form of safe assets. This would imply the existence of a distinct market for safe debt secluded from speculative investment markets (Perotti and Golec, 2016). Furthermore, it would also indicate that firms, households and financial corporations exhibit stable preferences for safety and liquidity, in order to meet their value-storing requirements, and to hold liquid and safe assets to be used for exchange and as collateral for certain transactions.

The second major finding about safe assets is that the financial sector actively responds to a shortage of safe-public debt – as reflected by a drop in the safety premium – by issuing liquid

(7)

and safe short-term debt against its own risky and illiquid assets (Krishnamurthy, 2015). Krishnamurthy and Vissing-Jorgensen (2015) show that investors have a large demand for safe and liquid investments, and that short-term bank debt satisfies this demand. Investors’ demand translates into low yields on short-term debt that is both safe and liquid, as mentioned above. The financial sector supplies such debt by investing in risky assets (securities, loans etc.) that is funded by short-term debt (Krishnamurthy and Vissing-Jorgensen, 2015). Fluctuations in the supply of government debt translate into variations in the safety premium, which compels banks to adjust their own supply of short-term liquid debt – demand deposits – to take advantage of the spread between the return on risky assets to the non-financial sector – loans to businesses and households – and between the interest rate on the deposits they take. In a context of scarcity of safe government debt, a crowding in of financial sector’s lending funded by short-term debt is expected to occur, as a reduced supply of treasuries lowers the yields on the former, thereby increasing the yield spread between illiquid/risky loans and safe/liquid assets (i.e. Treasuries), making it more attractive for banks to invest in riskier assets using deposits. Conversely, an increase in the supply of Treasury bonds should make the latter more attractive to the financial sector through an upward pressure on their yields, resulting in a substitution of Treasury securities away from loans and corporate securities, assuming that the interest rate on the latter stay constant, or at least that some frictions result in a subdued increase in interest rates on loans in a context of increasing yields on Treasuries. Krishnamurthy and Vissing-Jorgensen call this mechanism the bank portfolio substitution channel.

The second channel through which government debt is expected to crowd out short-term debt issued by the financial sector is linked to the portfolio rebalancing process that households perform: in a context of constant demand for safety, a low supply of government debt causes yields on treasury securities to drop further, making alternative “quasi-safe” assets issued by the financial sector more attractive. On the contrary, households are expected to substitute government bonds for short-term claims on banks – i.e. deposits – when Treasury supply increases. The two effects are complementary in to the extent that, assuming government debt securities and short-term bank debt are good substitutes, a lower supply of Treasuries makes households willing to accept even lower interest rates on their deposits in banks – or on any other short-term claims they would hold

(8)

1.2. Scope and Delimitation

Empirical studies on safe assets on countries other than the US are lacking, hence the need for further research in this domain. In the present study, we intend to test two hypotheses about France, with the ultimate goal of shedding light on the mechanisms by which private safe assets can substitute themselves for government-supplied safe assets, under which circumstance they do so, and on the share of safe assets as a fraction of total assets (i.e. wealth) over time. The time frame that is used is 1870-2017. One advantage of using long time series, historical data, is the enhanced likelihood of incorporating periods of time where the supply of government-issued safe assets shoots up suddenly, such as wars. The two following hypotheses will be tested:

Hypothesis 1: There is an inelastic demand for safe assets on behalf of resident households, firms and financial-corporations, as a proportion of total assets (i.e. wealth). The share of safe assets in the portfolio of the latter three categories has remained relatively constant over time. Both government liabilities and certain financial liabilities carry safety and moneyness attributes, and as such, are substitutes.

Hypothesis 2: Government debt crowds out short-term debt issued by the financial sector, through a bank substitution portfolio effect and a household debt substitution effect. The financial sector actively responds to variations in the safety-premium – arising from the portfolio demand for safe and liquid assets – by substituting Treasuries for risky/illiquid assets when the supply of Treasuries increases.

While the first hypothesis focuses on the demand side, the second hypothesis is concerned with the supply side of safe assets provision. The two hypotheses complement each other and, if confirmed or proven plausible, would provide useful insights into the functioning of safe assets markets. Furthermore, the first hypothesis serves as a basis for underpinning the second hypothesis, as the latter presupposes that there is a relatively stable and inelastic demand for safe assets. If our results suggest that there is indeed a constant safe-assets-to-wealth ratio, the corollary would be that we could treat demand as being essentially fixed. Lastly, as will be explain in the methodology section, the first hypothesis encompasses a large array of instruments and assets, issued on one hand both by the financial and non-financial private sector, and issued or guaranteed

(9)

by the government on the other. By contrast, the second hypothesis focuses on the dynamics of safe assets provision by the financial sector as a function of the outstanding supply of sovereign debt instruments – Treasury bills, notes and bonds – thereby offering two perspectives on the matter.

1.3. Outline

In the next chapter, we present a literature review in order to introduce the main relevant findings of previous research pertaining to safe assets. Thereafter, the methodology chapter will announce the conceptual and empirical framework to be used, followed by a discussion on the

shortcomings and caveats of the present study. Then, the data description section will provide more details on the variables we have constructed and their characteristics. Finally, the results and discussion chapter contains the results of our study and their economic interpretation. A conclusion will summarize the main findings of the study and discuss the significance of the results for economic research and policy making.

II. Literature review

Several studies have documented the existence of a safety premium on certain types of assets. Krishnamurthy and Vissing-Jorgensen (2012) find that there exists a safety premium on US Treasury debt, distinct from a liquidity premium. Safety can be thought of as an absolute security of nominal repayment, whereas liquidity refers to “moneyness”, or the ease with which an asset can be traded. As such, an asset is qualified as “safe” if it guarantees an unconditional financial promise of nominal repayment with no credit risk (Perotti, 2016). Any debt issued or guaranteed by a “safe” government – with its own central bank, and a stable currency – can be regarded as a safe asset.

(10)

liquid debt in the economy and offer different perspectives on the role of safe assets in relation to credit cycles and prudential regulation.

Krishnamurthy and Vissing-Jorgensen (2012) confirm the presence of a price-sensitivity of safe assets to supply and demand fluctuations. In particular, they find an inverse correlation between corporate and treasury bonds yield spread and the share of privately owned government debt to GDP in the US, from 1926 to 2008. By comparing pairs of assets with different safety but similar liquidity – lower and higher rated commercial paper with three months maturity – and conversely, pairs of assets with similar safety but different liquidity – 6-months insured certificate of deposit and Treasury-bills – they conclude that Treasuries carry both a liquidity and a safety attribute. Furthermore, the supply of Treasuries drives both the equilibrium price of safety and that of liquidity. They find that the safety premium is high at times of scarcity of US debt. They estimate the convenience yield of long-term Treasuries – capturing both liquidity and safety attributes – to be 73 bp for the period 1926-2008, of which at most 46 relates to liquidity and at least 27 relates to safety.

In a model of financial intermediation based on a fundamental demand for safety, Perotti and Ahnert show that intermediaries emerge to ensure an efficient provision of safety when self- storage is too inefficient for some agents (Perotti and Ahnert, 2017). As no private asset is without risk, safe claims must be carved out of risky assets. The issue is that an increase in safety needs – as was the case in the US in the run up to the 2008 crisis – systematically induces more fragility to the financial system, as intermediaries invest in risky assets to satisfy the demand for safety. Indeed in equilibrium, in a setting where agents have different endowments of wealth level and safe storage, only agents with a good access to safe storage take up the role of intermediaries. The situation is such because agents with high access to safety are risk-tolerant, as vice-versa, which creates an agency problem in which intermediaries have a lower risk-aversion than their investors. Gorton et al. (2012) finds evidence of a historically stable demand for safe assets in US households’ portfolio. In fact, the so-called “safe assets share” – defined as the sum of government debt and the safe component of private financial debt – has remained constant at around 33% in the US since 1952 (Gorton et al., 2012). In the same study, he also shows that government debt and privately produced safe assets in the United States are strongly negatively correlated.

(11)

In a later study, Krishnamurthy and Vissing-Jorgensen find an inverse relationship between supply of government-issued safe assets and privately-issued ones (Krishnamurthy and Vissing-Jorgensen, 2015). They estimate that a one dollar decrease in Treasury debt increases short-term financial debt by 0.5 dollar. Furthermore, they also find that financial intermediaries match one for one short term debt issuance with long term lending. Indeed, a decline in government debt appears to boost the private production of – quasi – safe assets in the form of short-term liabilities – repos and commercial paper – which in turn generates a credit expansion and long-term investments by financial intermediaries. As a consequence, a decline in government debt increases credit supply and maturity mismatch. Those findings carry a lot of significance, as they suggest two things. First, that privately-issued safe assets can – albeit imperfectly – substitute themselves for safe government debt, as they share similar attributes. Second, that the production of safe assets by the private sector positively responds to an increase in demand for safety, in a context of scarcity of safe public debt.

With respect to France, Kacperczyk, Pérignon, and Vuillemey (2017) have documented the existence of a safety premium which correlates with the supply of government debt. By analyzing the supply of Treasury Bills on a daily frequency from 2008 to 2014 in France, and contrasting it with privately-issued safe assets – commercial paper (CP) and certificate of deposits (CD) – it appears that short-term private assets can have a safety attribute similar to government safe assets. Furthermore, the issuance of the short-term CDs – but not CPs – increases at times when issuance of Treasury bills goes down (Pérignon et al., 2017), thereby supporting Krishnamurthy and Vissing-Jorgensen’s findings. Nonetheless, two elements warrant further research. First, the time frame is particularly short, as their study spans from 2008 until 2014. Second, they restrict their definition of public and private safe assets to Treasury bills and CPs and CDs respectively, thereby reducing the scope of the maturity spectrum to its short-end.

(12)

III. Methodology

In order test the two hypotheses, we need to construct a model of safe assets holdings that is consistent across the entire time frame, which is 1870-2010 for the first hypothesis and 1951-1995 for the second one. We intend to analyze the evolution of the demand for safe assets in France by using low-frequency, annual data. One issue lies with the disappearance of certain debt instruments and the emergence of new ones over time. We will include every instrument that fits our concept of safe asset, namely low default-risk and information-insensitivity.

3.1. Hypothesis 1

Regarding the first hypothesis, the entities we are interested in are households, firms (i.e. non-financial corporations) and non-financial-corporations. The aim is to quantify the amount of safe assets held by these three sectors across time, as a proportion of all the assets they hold – i.e. their wealth. Yearly data on residents’ holdings of government securities (Treasury bills, bonds, notes, banknotes and coins, regulated savings accounts, as well as central bank liabilities) and safe financial sector liabilities (sight deposits in commercial banks and short-term money market instruments) will be collected. They will be divided by net national wealth, which measures the total amount of a country’s financial assets, land, currency, and real estate. The wealth estimate that will be used comes from Piketty and Zucman (2014)1_{, who collected data on wealth and} income for the years 1700-2010 in several countries including France. We will only include the holding of these aforementioned assets by residents, which is why exclude sovereign debt instruments held by non-residents. I will not test whether the said safe assets really are safe, but instead postulate that certain assets are safe. While the near-absolute safety of French sovereign debt is unequivocal, assets issued by the private agents cannot be perfectly safe. The data description section down below will discuss the safety and information-insensitivity of various types of suspected safe assets, in order to motivate our selection of assets to include.

1_{Piketty, Thomas., Zucman, Gabriel. (2014). “Capital is Back: Wealth-Income Ratios in Rich Countries,} 1700-2010”. The Quarterly Journal of Economics (2014), 1255–1310.

(13)

An OLS regression will be performed to assess whether resident households, financial-corporations and firms substitute government-issued safe debt for privately issued debt and vice versa, in an attempt to maintain a balanced portfolio with a roughly constant share of safe assets. We expect a negative relationship between government debt and privately-issued debt holdings, which would indicate that a reduction in the availability of safe government liabilities is matched by an increase in demand deposits or other safe, short-term debt originating from the private – mostly financial – sector. We include two control variables, namely a measure of the real short-term and real long-short-term interest rate. The reason for their inclusion is that we are interested in finding whether the demand for safe assets is inelastic to changes in interest rates and wealth levels. Indeed, if the inclusion of the real short-term and long-term interest rate in the equation does not change the results by a large margin, it would suggest that, contrary to standard macroeconomic literature, the demand for certain assets that exhibit safety attributes is not affected by variations in interest rates. The following equation will be used to shed light on the degree of substitutability between private and government-issued safe debt that we are interested in.

𝑺𝑻_𝒕 = 𝜷_𝟎+ 𝜷_𝟏𝑮_𝒕+ 𝜷_𝟐𝒍𝒓_𝒕+ 𝜷_𝟑𝒔𝒓_𝒕+ 𝜷_𝟒𝒓𝑮𝑫𝑷 + 𝒖_𝒕

Where:

STt = Safe Assets issued by the private sector held by residents / wealth Gt = Government-issued safe asset held by residents / wealth

𝑙𝑟₆ = Real long-term interest rate 𝑠𝑟₆ = Real short-term interest rate rGDP = Real GDP (index=100 in 2005)

For the first hypothesis, our definition of Government-issued safe assets (which we use interchangeably with Government supply) is the following:

(14)

= (Treasury debt supply available to residents + sum of currency + savings on regulated savings accounts + liabilities of the Bank of France towards its residents and deposits at the Treasury) / wealth

Our definition of Safe Assets issued by the private sector is the following:

= Safe Assets issued by the private sector / wealth

= (Sight deposits in commercial banks + commercial paper + certificates of deposits + repurchase agreements + Money Market Mutual Funds shares) / wealth

The data description section will provide more details on the characteristics of the above variables.

3.2. Hypothesis 2:

The second hypothesis will be assessed using Krishnamurthy and Vissing-Jorgensen (2015) methodology. By constructing a balance sheet for the financial sector in the United States for the years 1875-2014, and observing the impact of variations in government debt, they looked for evidence for the previously described crowding out effect, in which the supply of treasury liabilities crowds out financial sector lending financed by short-term debt. The model they present involves a financial sector, a household sector, and a government. The financial sector can invest in loans and treasuries, which are financed by short term debt and equity, while households can invest in short term debt vis-à-vis the financial sector – i.e. deposits – as well as treasuries, as displayed in the graph below. For data availability and reliability issues, we will focus on the period 1951-1995. Our intention is to reproduce their model vis-à-vis France, for which we will need to construct an estimate of net short-term debt issued by the financial sector. We will thus borrow the methodology of Krishnamurthy and Vissing-Jorgensen, while making the necessary adjustments to fit our conceptual framework.

(15)

Source: Arvind Krishnamurthy and Annette Vissing-Jorgensen (2015)

3.2.1. Modelling Net Short-Term Assets

In order to test the prediction that increases in government bond supply crowds out short-term debt backed by risky long-term assets, one would need to isolate net short-term liabilities.

In their model, the balance sheet of banks only contains loans to the economy and government securities on the asset side, and short-term debt and equity on the liability side. In reality, banks and other financial institutions also hold short term assets and are partly financed with long-term debt. As a consequence, we need to subtract short-term assets from short term liabilities. In addition, as we are only interested in risky/illiquid assets that are financed with short-term debt, we need to remove the financial sector’s holding of government debt.

The variable that we need to obtain thus is the following:

• Net short-term debt = short-term liabilities of the financial sector minus short term assets of the financial sector minus financial sector’s holding of treasuries. We denote it 𝑵𝑺𝑻_𝒕

(16)

3.2.2. Definition of the Financial Sector

The financial sector is an incredibly complex system of interwoven and overlapping entities, with a plethora of actors playing different roles, providing different services, and relying on different sources of funding. This renders modelling the financial system a difficult task, and it would be simplistic to limit our analysis to banks. As our goal is to analyze financial actors that invest mainly in risky, illiquid loans to the economy, and fund these in large parts with short-term liabilities, we will include the following categories: all the deposit taking institutions – private and national –, finance companies, money market mutual funds, and the Banque

Française du commerce extérieur. In addition, la Banque de France – the French central bank – will be included until the year 1978. The reason for its inclusion in the financial sector category in the second hypothesis, as opposed to the government sector, is that it directly provided a significant amount of loans to the French economy from its creation in 1808 until the late 1970s, and took a substantial amount of deposits. In 1971, the Bank of France supplied 9.16% of total credit to the economy outside of those financed by the Treasury and its specialized agencies (Conseil National du Crédit, 1977). That share declined to an insignificant amount by the late 1970s, reaching less than 0.1% of total credit by 1990. We will therefore exclude the Bank of France’s credit to the economy and supply of short-term liabilities from 1979. Our results are however not materially affected by the inclusion or exclusion of the Bank of France from our estimation.

3.2.3. Accounting for Cross Holdings of Financial Instruments among the Financial Sector

One major issue with the process of quantifying the supply of financial sector liabilities is that the latter sector is characterized by large degrees of cross-holdings. Indeed, banks and other financial institutions write claims on each other, with the consequence that certain financial instruments constitute both an asset and a liability. Money market fund shares, as well as certain money market instruments such a commercial paper and certificates of deposits constitute a prime example of these persistent cross-holdings. Most financial institutions are credited with a large amount of Money market fund shares, which also appear on the liabilities side of their balance sheet. The task consists in obtaining a net estimate for each instrument, by measuring liabilities minus assets

(17)

for those that are on average a net liability for the entire time period. Similarly to Krishnamurthy and Vissing-Jorgensen (2015), our estimate for financial sector short-term liabilities will be lower than its outstanding counterpart, as cross-holdings are netted out.

3.2.4. Treasury Supply

To test our second hypothesis, we will only focus on French sovereign debt instruments – i.e. Treasury bonds, bills and notes – as opposed to the approach used for testing the first hypothesis, which includes all government-issued and government-insured assets. Indeed, the provision of government-insured savings account is not fixed and determined – although capped at a ceiling level of 22,950 € for the “livret A” as of 2018 – as opposed to available treasury debt (Services-publique, 2017). In other words, we cannot take deposits in government-insured savings banks as an independent and exogenous variable, as their level is a function of demand and is always below the maximum amount that could potentially be deposited if every holder of these regulated savings account reached the ceiling level. On the contrary, government debt is more or less exogenous, and its supply is relatively inelastic to demand, as sovereign debt reflects the financing need of the state. Similarly, considering that coins and banknotes in circulation offer no interest rates, they cannot be included as an independent variable for assessing whether government liabilities crowds out privately-issued short-term debt. Moreover, although being considered as safe assets, coins and banknotes cannot satisfy the demand for store of value for large volumes of money, considering the costs associated with holding large amounts of cash. Only treasury debt securities constitute the kind of exogenous supply that is of interest for us here, in a context where we are concerned with the crowding out impact of tradable and liquid sovereign debt on short-term financial liabilities. Lastly, in agreement with Krishnamurthy and Vissing-Jorgensen, we will include the entire supply of government securities, including those held by the central bank, as opposed to the approach used in hypothesis one. However, while they removed foreign official holdings from their Treasury supply estimate, no such reliable data exists for France. The Agence France-Trésor – the agency that provides for the government's cash needs so that the State can meet its financial commitments at all times – indicates on their website that such information is not available as the identity of investors is to be maintained

(18)

3.2.5. Financial Sector Short-Term Liabilities

For the second hypothesis, our estimate of financial sector safe assets includes all net short-term liabilities – with a maturity of less than three months and a year for deposits and securities respectively – towards the non-financial sector. In line with Krishnamurthy and

Vissing-Jorgensen, we then distinguish between checkable and non-checkable deposits, in order to assess whether those categories are subject to different levels of crowding out by government debt. Importantly, within the second hypothesis, we include deposits in regulated savings account in the financial sector category. Indeed, although being guaranteed by the state, the latter are issued by banks – both private and national – and therefore constitute financial sector liabilities. We include three control variables, namely the real interest rate – short-term and long-term – and the the real level of GDP, in order to control for the business cycle.

The following time series equation will be used to test hypothesis two:

𝑵𝑺𝑻𝒕 = 𝜷𝟎+𝜷𝟏 𝑻𝑺𝒕+ 𝜷𝟐𝒍𝒓𝒕+ 𝜷𝟑𝒔𝒓𝒕+ 𝜷𝟒𝒓𝑮𝑫𝑷 + 𝒖𝒕

Where:

𝑁𝑆𝑇6 = Financial Sector’s Net Supply of Short-Term Liabilities / GDP

𝑇𝑆₆ = Treasury supply / GDP

𝑙𝑟₆ = Real long-term interest rate /GDP 𝑠𝑟6 = Real short-term interest rate / GDP

rGDP = Real GDP (index=100 in 2005)

Notice that we divide the above quantities by GDP to scale them by the size of the French economy. In hypothesis one, our variables are scaled to wealth because the supposition to be tested is that residents hold a constant share of safe assets in their portfolio – i.e. their wealth, and not in proportion to the annual national output.

(19)

3.3. Limits of the study

A major concern with our crowding out result (hypothesis two) is that it might not be driven by safety demand but instead by the standards crowding out mechanism by which government spending and government debt reduces private capital formation by raising the interest rate. We include a measure of the short-term and long-term real interest rate in our regressions and show that the crowding out effect is robust to including these control variables.

Another issue concerns the fact that our results might not be causal but instead driven by an omitted variable bias or a reverse causality bias. For instance, the supply of French sovereign debt is a function of war and of the business cycle, both of which also affecting the financial sector’s lending to the economy and its use of short-term debt, independent of Treasury supply. Similarly, a banking crisis – translating into a drastic reduction in the financial sector’s lending and supply of short-term deposits – would have the unintended consequence of increasing

government debt, as the government would increase its spending to bail-out the banks and would see its tax revenues decline and social transfers expenditures increase in the face of the ensuing recession. This would entail an inverse causality in which lower financial sector supply of short-term debt would drive up Treasury supply. Nonetheless, our time frame for the second

hypothesis (1951 until 1995) exclude the two major financial crises of 1929 and 2008, and the 1945-1971 period – sometimes referred to as the Golden age of capitalism – is virtually free of financial crises. Although France has been involved in several conflict during the time period under scrutiny (Indochina war 1946-1954, Algerian war 1954-1962, Chadian–Libyan conflict 1978-1987, Gulf war 1990-1991) all of these conflicts took place outside of mainland France, and thus probably did not have much of an effect on the financial sector’s activities.

Furthermore, we show that our crowding out results are unaffected by the inclusion of real GDP growth as a control variable, thus controlling for the business cycle. Regarding the first

hypothesis, we exclude the period 1915-1947 due to the effects of the two world wars and the great depression, and due to the lack of reliable data of the supply of sovereign debt available to resident firms, households and banks, as will be explained in the next section. Furthermore, considering that we are using of low-frequency data over such an extended period of time, the

(20)

categorically rule out the existence of omitted variables or reverse causality bias. Nonetheless, the precautions we take through the use of control variables and the low-frequency of our data should allow us to infer certain patterns from our results.

Importantly, if Treasury Supply/GDP and Net short-term debt are stationary, our approach is adequate. While the Dickey-Fuller test we run cannot reject that Net short-term debt/GDP and Treasury Supply/GDP are integrated of order one (i.e. have a unit root), our results bring insightful information about safe assets provision and about the substitutability between safe government debt and short-term claims on the financial sector. Further research could examine whether net short-term liabilities issued by the financial sector and Treasury supply are cointegrated, and explore the speed of adjustment of safe assets provision by the financial sector as a function of variations in Treasury supply. An Error Correction Model could be used for that purpose. We leave those alternative research paths to be investigated by other empirical economists. Lastly, we need to emphasize that some of our variables might have been estimated with error, as historical data on sovereign debt and bank deposits going back to the 19th century is likely to contain some degrees of approximation. While we have endeavored to

maintain a rigorous nomenclature for our variables of interest, the data had to be collected and compiled from different sources, and the classification and measurement methodology for various elements is subject to change throughout time. While the time frame for the first

hypothesis is extensive enough to unveil rich and reliable information about the safe assets share, the time period for the second hypothesis is relatively narrow, as it contains 45 observations. Although qualifying for a time series analysis in light of the rule of thumb of 30 observations minimum, further research with a longer time frame would be desirable from the perspective of establishing a more reliable and robust relation between the two key variables of interest.

(21)

IV. Data Description

4.1. Classification of safe assets

We distinguish two categories of safe assets, namely government-issued (or government-insured) safe assets and privately issued safe assets. The former corresponds to sovereign debt instruments – government and central bank liabilities – and any asset guaranteed by the state – such as insured deposits – while the latter corresponds to financial and corporate sector net short-term debt. Aside from (some) government debt which is virtually risk-free – France last defaulted on its debt in 1812 and its sovereign debt is widely considered to be a safe store of value (Allen and Blitz, 2017) – other assets cannot unequivocally be categorized as safe. For example, most of the securitized products and mortgages issued in the US during the 2000s were regarded as safe by investors, as reflected by the AAA-rating that most of them had and the low risk-weight assigned to mortgages on banks’ balance sheet. However, when the crisis unfolded and contagion from toxic securitized US assets spread throughout the world, an entire market worth US$ 893 billion annually collapsed, with the consequence that these assets were no longer safe (Krishnamurthy and Caballero, 2009). The next two sections present an overview of our classification of safe assets.

4.2. Hypothesis 1

4.2.1. Government-issued safe assets

Any liabilities of the French government can be regarded as safe, provided that inflation is low. However, we are only interested in liabilities that can be regarded as assets, enjoying both the “moneyness” or “liquidity”, and the “safety” attributes. As such, the so called “diverse commitments” category of French debt, such as pensions of public servants, as well as dividends owed to former shareholders of nationalized companies, will not be taken into account, because they are not tradable and are not necessarily the product of deliberate investments or acquisitions. Similarly, the deposits from the “Correspondants du Trésor” at the Treasury will not be included.

(22)

These organisms include municipalities, local communities and public-sector establishments (Vallet, 2003). The centralization of their deposits at the Treasury amounts to around 60 billion euros every year, thereby representing a significant source of financing. However, as these deposits essentially are intergovernmental loans, we will discard them.

All debt securities issued by the French Republic will be considered, in addition to the sum of currency, certain savings accounts guaranteed by the State, as well as the liabilities of the central bank – La Banque de France – available to residents. This section provides details about the instruments that are included.

a) Sovereign debt securities

The nature and characteristics of French Sovereign debt has evolved over time, which compels us to establish and define a clear methodology in order to maintain a consistent estimate for sovereign debt instruments. Several obstacles are making this task difficult.

First, in order to assess the volume of government debt available to resident firms, households and financial institutions, we need to consolidate sovereign debt in order to isolate only the share that is available to the latter categories, by removing the share bought by other actors such as non-residents and, most importantly, by other governmental agencies and administrations, and by the central bank. Indeed, until 1994, the Banque de France was allowed to directly purchase debt instruments from the Treasury, and it maintained a credit line available to the latter, thereby effectively allowing the central bank to monetize government debt. The Bank of France and l’Agence France-Trésor disclose some information about the residency of the government’s creditors since 1997 only. However, in its Annuaire Statistique de la France, l’INSEE provides a distinction between interior and exterior debt for most years for the period 1875-1997, and also provides details about the volume of government bonds subscribed by international organizations. The amount of sovereign debt held by non-residents, by public administrations and government agencies, as well as by the central bank will therefore be deducted whenever possible, as it is effectively not available to resident households, firms and financial-corporations. However, for the period 1914-1945, the amount of Treasury bonds, bills and notes held by the central bank remains unclear. Therefore, our estimate for the volume of government safe assets available to the public during that period is to be interpreted cautiously. On the other hand, the Bank of France

(23)

only became nationalized in 1945, and it essentially performed all the tasks that a commercial bank does – except that it already had an exclusive monopoly on money creation – which makes the lack of data on central bank’s holding of government securities less problematic.

For the 1870-1995 period, we will include all short-term, medium term and long-term French sovereign debt in the forms of tradable securities issued by the State and its agencies and local administrations2_{. We will deduct these securities held by the central bank, other government} agencies, and international organizations in order to only keep those instruments that are available to domestic firms, households and financial institutions. For the 1995-2017 period, we use the total debt of public administrations in line with the Maastricht classification, from INSEE database. Only debt instruments will be considered (BTF, BTAN and OAT), while short-term and long-term credits, as well as deposits by the correspondants du trésor will be discarded – as they correspond mostly to public administrations and agencies depositing money at the treasury. The advantage of using Maastricht debt is that it is consolidated, meaning that intergovernmental flows and stocks of debt and credit are netted out. As such, the values used do not incorporate debt between governmental agencies and administrations.

b) Regulated savings accounts

In addition to sovereign debt, we need to consider all assets guaranteed by the state. In the case of France, these include certain regulated savings accounts (Livrets réglementés), such as the “Livret A”, which are offered by savings institutions called “Caisses Épargne” - i.e. savings banks. The first Caisses d’Épargnes emerged in 1818 in France as a philanthropic endeavour and were officially deemed as private institutions of recognized public utility by the law the 5th_{of June 1835} (Christen-Lécuyer, 2004). In 1882, the State created its own National Caisses Épargne, thereby introducing some degree of competition. The State regulate these savings accounts by determining

2_{The 1870-1964, the following instruments are used: Debt redeemable by annuity installments and short-term debt,} War damages (in the forms of Treasury bonds and bills), Amortizing long-term loan, Short-term maturity loans,

(24)

their interest rate, as well as by setting the ceiling on the amount each individual is allowed to deposit. All savings all fully guaranteed by the French government, on top of the insurance on deposits which currently stands at 100.000 euros (Service-public, 2017). Furthermore, three out of the four categories of regulated savings accounts are exempt from taxes.

c) Deposits at the Treasury, the Central Bank, and Postal Checking Accounts

Since its creation and until the 4th_{of August 1993, individual citizens and firms were allowed to} open a checking account at the Banque de France, in order to deposit money. Similarly, the decree of June 16, 1806, authorized individual citizens to deposits their money at the Treasury (Vallet, 2003). On the 2nd_{of February 2001, this decree was repealed. Deposits both at the Treasury and at} the Banque de France will be categorized as public safe assets, as they were guaranteed by the State, backed by the taxing power of the government and the money creation monopoly of the central bank. Moreover, Postal checking accounts will be included in the estimation of public safe assets, as the postal service is public and the savings it holds are guaranteed by the state.

Lastly, coins, banknotes, and paper money (metal-backed banknotes, convertible in gold and silver) will be counted as safe assets. For much of history, coins made of precious metals were thought of as safe assets (Vilar in Gorton, 2016). Indeed, coins and metal-backed banknotes carry all the attributes of safe assets. Their “moneyness” means that they are liquid and can be traded easily and accepted as payment if they have legal tender. They are information-insensitive too, as it does not cost money to obtain information about their value: economic agents simply accept French francs and euros without striving to obtain information about their value, at least within the time frame considered in this study. Lastly, metallic currency is fairly safe as they can store value through time. Fiat currency, which is only a recent innovation, can be regarded as safe upon the condition that inflation is low. Even without being backed by metal supply, fiat money – which appears on the liabilities side of the central bank, as does metal currency – has not lost its safety as opposed to metal currency. It remains a claim on production and wealth, and as long as there is confidence in the sovereign, as well as a reasonably low-inflation environment, fiat money can be treated as a safe asset.

(25)

d) Liabilities of the Bank of France

While Krishnamurthy includes total Treasury supply in his estimate, and discards central bank liabilities, we removed the central bank’s holding of government securities but included some of the Bank of France’s liabilities towards its residents. Krishnamurthy argues that the central bank does not add any net safe assets, to the extent that it backs its liabilities with treasuries (Krishnamurthy, 2015). As his aim is to analyze the mechanisms through which the non-financial sector balances its portfolio with financial sector liabilities and government liabilities to meet a demand for safety, and given that central bank reserves are not available to households and firms – except before 1993 under certain condition – his approach does not need to include central bank liabilities. However, the question we are asking for the first hypothesis requires a different approach, as we are interested in resident households, firms, and financial-corporations’ holdings of safe assets, and the latter (i.e. banks) do have access to central bank reserves. Indeed, the Bank of France liabilities towards domestic banks – in the form of reserves – constitute a safe asset, arguably the safest form of it (Gorton, 2016). Furthermore, the central bank is backed not only by the treasuries it holds – in addition to other safe assets – but also by the taxing power of the central government. The central bank can issue more safe “claims” than it effectively holds in the form of treasuries, for the reason that it is protected by an implicit public backstop. Indeed, there are reasons to think that the government would recapitalize the central bank if it ever was to go bankrupt. As a consequence, all the liabilities of the central bank – even those in excess of the treasuries it holds – should be regarded as safe.

4.2.2. Privately-issued safe assets: financial and non-financial institutions

The financial sector can create quasi-safe assets by offering short-term liabilities to investors. Gorton argues that there are two ways to produce safe debt: by backing it to the government’s taxing power, or by using collateral (Gorton, 2016). The financial sector can offer safe debt to economic agents in the form of sight deposits, payable on demand, which are typically invested in

(26)

in the form of deposits offers a convenience yield – measured as the spread between the interest rate on such deposits and on other assets of similar safety but without the liquidity or moneyness attribute specific to bank debt. As debt is senior to equity, its value has some immunity to the arrival of information, which makes bank debt relatively resilient to information over the long-run (Gorton, 2016). Depositors can always withdraw their cash if they lose confidence in the solvability or liquidity of a financial institution, although doing so fosters the development of a self-fulfilling prophecy by triggering a bank run. Bank debt is by no means always safe, but the convenience it offers – by its moneyness and convertibility – and the fact that it is collateralized by assets which themselves are sometimes collateralized – mortgages are backed by the underlying property – make it relatively safe. Furthermore, Gorton argues that demand deposits are over-collateralized to the extent that banks have equity, which serves as a loss-absorbing buffer (Gorton, 2016).

In the present study, for the majority of the time frame to be analyzed, short-debt deposits (sight deposits) will constitute the main privately-issued safe asset. Latter development, such as the rise of the money market and the emergence of securitization, brought new instruments into play. Hence, short and medium-term negotiable debt securities, known as TCNs in French (Titres de Créance Négociables) will be incorporated. Most of these short-term liabilities have an immediate or overnight maturity – sight deposits, whether checkable or non-checkable, can be withdrawn on demand. Note that, although “regulated savings accounts” are now offered by almost all banks, not only the so called national “Caisses d’épargne” (savings banks), they will not be included in the financial sector category, but instead in the government supply category, as they are guaranteed by the State. This choice is rather arbitrary, as one could argue that these liabilities are now effectively offered predominantly by private banks, as opposed to national savings banks. However, for the sake of the first hypothesis, we are mainly concerned about the overall share of safe assets vis-à-vis total assets, whether private or public. As a consequence, the rather crude, dichotomous framework that we use should not be problematic, even if it does not capture the nuances of the present financial products that are characteristic of French financial institutions. The following section describes the elements that will be included.

(27)

a) Safe Assets issued by the financial sector

The following data is collected from the online database of La Banque de France (Webstat) and l’INSEE3_.

• Certificate of deposits (CDs), which have a maturity comprised between one day and one year. France started issuing CDs in 1985.

• Sight deposits (also called demand deposits, or call deposits), which constitutes the largest category of safe debt issued by the financial sector.

• Money market mutual fund shares (called ”organisme de placements collectif en valeurs mobilières” in French, or OPCVM). As they only invest in debt securities with short maturity and low credit-risk, they can be considered as safe.

• Repurchase agreement stocks.

• Short-term securitized notes (although of questionable safety, they constitute an insignificant share of total safe assets).

b) Issued by non-financial institutions:

Commercial Paper – known as billet de trésorie in the French money market – will be included. They typically have a maturity of less than 270 days, and no more than 365 days. Although not being collateralized, commercial paper is very safe.

4.3. Hypothesis 2

The second hypothesis requires some modification of our nomenclature. The data itself is similar to that used to test hypothesis one. The present section introduces the elements that differ from hypothesis one.

(28)

4.3.1. Government-issued safe assets

As stated in the methodology section, only Treasury bonds, bills and notes will be included in the Treasury supply estimate for the second hypothesis, as opposed to all government-insured and government-issued assets.

4.3.2. Safe assets issued by the financial sector

All short-term liabilities of the financial sector – as described in the methodology section – towards the non-financial sector will be included. Again, the so called regulated savings accounts offered by the Caisses d’Épargnes will be included in this category.

(29)

V. Results and Discussion

5.1. The demand for safe assets

As displayed in figure 1 below, French residents’ holding of safe assets – both government-issued and privately-issued – is remarkably constant. The first hypothesis appears to hold for the years 1875-1914 and 1948-2010. The safe asset share in those years averages 21.38% (table 1 in the appendix). Overall, if we exclude the years 1914-1947, a slight upward trend in the safe assets share stands out, as the latter increases from 17.87%, to 22.85% and 24.4% for 1875-1914, 1948-1980 and 1981-2010 respectively. Three time-segments thus stand out – if we excluded the war years – characterized by low variance in the safe asset share. However, from the beginning of the first World War until the late 1940s, the share of safe assets over total assets held by residents in France (i.e. wealth) increases and fluctuates dramatically. The years 1914-1945 are characterized by a large volume of sovereign debt instruments available to residents, in a context of increased public spending to finance the two world conflicts as well as the aftermaths of the great recession. In 1941, the safe-asset-to-wealth ratio even exceeds 1, standing at 112.1%. It may seem surprising that the safe-assets-share exceeds 100%, as it implies that there are more safe assets than total assets. In reality, government debt is nothing else than a promissory note, a payment that the State has committed to honor in the future. Nothing prevents the State from issuing more debt than it can immediately pay, or in the present case, more than debt than the entire country’s national wealth. The relatively long maturity of sovereign debt, and the years of sustained inflation and economic growth that followed the end of WWII made it possible for France to reduce its debt-to-GDP to a fraction of what it was, reaching around 35% by the mid 1950s. Moreover, one problem pertains to the fact that during the two world wars, a significant share of government debt was held by the Banque de France and the Caisse des Dépôts et Consignations. The data used for the figure one is net of exterior debt, but it contains debt held by the central bank for the 1915-1945, in the form of treasury bonds, bills and notes. As a consequence, our data overestimate the true safe asset share during those years.

(30)

exceptional, considering the profound and disruptive effects armed conflicts generate on the economy. One interesting question is whether the sustained increase in the safe assets share originates from an expanded demand for safety, amidst a period of uncertainty and economic contraction, or whether it mostly originates from a surge in the supply of safe assets. Considering the sharp increase in government debt between the two world wars, it appears obvious that the surge in the safe-assets-share mostly originate from a supply shock. Households, firms and banks were also urged to participate in the war effort during the first world war by purchasing treasury bonds. However, on the demand side, it seems plausible that economic agents desire to hold more safe assets in times of economic distress, as they become more risk averse.

(31)

Table 2 in the appendix contains the results from the first equation, in which we test the degree of substitutability between government-issued and privately-issued safe assets. All the coefficients are negative and significant at the 1% level. Without control variables, the coefficient is -0.632. The inclusion of the real short-term and long-term interest rate separately only slightly increases the coefficient (-0.629 and -0.612 respectively), while their simultaneous inclusion decreases the coefficient to -0.716. An increase of one euro of residents’ holding of government-issued safe debt is matched by a decrease of 0.716 euro in privately-government-issued safe debt holdings, controlling for real interest rates. The fact that the coefficient does not vary by a large margin when the real short-term and long-term interest rates are included implies that there is a persistent demand for safe and liquid assets that is inelastic and insensitive to variations in interest rates. In fact, table 3 confirms this characteristic of safe assets, as a regression of safe-assets-over-wealth on the real short term and real long-term interest rates fails to find a statistically significant correlation. This finding differs from what standard macroeconomic models would predict, as fluctuations in interest rates are expected to impact investments and savings. In particular, an increase in interest rates should increase savings through the intertemporal substitution channel between consumption and saving, by increasing the opportunity cost of consumption in the present period. However, the assets under scrutiny appear to follow a different pattern, and suggest that firms, households and financial corporations desire to hold a relatively constant proportion of their total assets in the form of safe debt, irrespective of the return on the said debt instruments. The substitutability between the two categories of assets seems robust and persistent, thereby confirming the validity of the first hypothesis.

Interestingly, figure 2 indicates that the share of government-issued safe assets in total safe assets has experienced a sustained decline almost throughout the entire time frame, with a reversal in 1992 and 2008 caused by the Early 1990s recession and the Great Recession respectively. During the latter, the French money market contracted in the face of the financial crisis, translating into a drop in the issuance of certificates of deposits, commercial paper, and repurchase agreements, paralleling a surge in government debt (Banque de France, 2013). While a reverse causality is obviously responsible for the inverse relationship between privately-issued and government-issued safe debt during that period – as described in the “limits of the study” section

(32)

between private and government safe assets is likely to break down in times of market distress. Figure 5 displays the composition of the safe asset share for the years 1870-2010. While coins, banknotes, and sovereign debt instruments constituted the majority of safe assets in the late 19th century, bank deposits, deposits in regulated savings account and in postal accounts, as well as central bank liabilities progressively acquired a preponderant share of total outstanding safe assets. By the 2000s, the dominant instruments were money market instruments – repos, certificates of deposits, commercial paper etc. – deposits in banks and well as sovereign debt. Noticed that while the nominal amount of government debt has continuously increased since 1973, the ratio of public debt available to residents over wealth and over total safe assets has been diminishing from 1997 until 2007. The reason for that paradox stems from the sustained increase in the share of government debt held by non-residents, rising from 17.4% in 1997 to 66.2% in 2009 (ECB database, 2018). The consequence of such acquisitions by non-residents is that these assets are withdrawn from the domestic market, making them unavailable to residents. Figure 2 and 3 illustrate the relationship between GDP, wealth, and the outstanding amount of safe assets, and attest of the stable relationship between safe assets and wealth. No stable relationship is present between safe assets and GDP. Lastly, it is worth noticing that Gorton et al. (2012) similarly finds a stable safe asset share in the US for the period 1952-2010, although his figure is sensibly higher. He finds a 30-33% ratio, which exceeds our estimates by about a third. One crucial difference between his methodology and ours relates to the nature of our denominator. While Gorton measure the safe asset share by dividing total safe assets by total assets – which he approximates as the combined liabilities of the government, the financial sector and the household sector, in addition to total equity – we have used net national wealth, which we borrowed from Piketty and Zucman (2014). Net national wealth is not restricted to financial assets as measured by Gorton, but also includes land and real estate. As such, the lower estimate we find could originate from the fact that we have a larger denominator. Nonetheless, in light of the question we are interest in, this should not pose any issue, as it merely reflects a slightly different conceptual framework whereby we are interested in safe assets holding as a proportion to total assets.

(33)

5.2. The Crowding out effect of Government debt

Figure 2 below displays the relationship between Treasury supply and Net short-term debt – short-term debt issued by the financial sector minus short-term assets held by the financial sector minus the financial sector’s holding of treasuries. Table 1 below also displays the main statistical results pertaining to the second hypothesis, while the appendix contains further regression outputs underpinning our findings. As suspected, there is a strong negative correlation between Net short-term debt and Treasury supply significant at the 1% level for every specification, except for the one controlling for the real short-term and real long-term interest rate

simultaneously, which is significant at the 5% level. Without control variables, the coefficient is 0.5, indicating that a one euro increase in Treasury supply reduces the supply of Net short-term debt by 0.5 euros. Controlling for the real interest rates makes the coefficient larger in absolute terms, while the inclusion of all the control variables simultaneously reduces it to 0.368. The crowing out effect of government debt seems very plausible in light of these results.

(34)

Table 1: Regression of Net Short-Term Debt on Treasury Supply, with control variables (1951-1995)

(1) Only Treasury Supply

(2) Treasury Supply + real short-term interest rate

(3) Treasury Supply + real long-term interest rate

(4) Treasury Supply + real GDP

(5) Treasury Supply + real short-term interest rate + real long-term interest rate

(6) Treasury Supply + real short-term interest rate + real long-term interest rate + real GDP

rSTir = Real short-term interest rate rLTir = Real long-term interest rate

In line with the predictions of the bank portfolio substitution effect, there is statistically significant positive correlation between Treasury supply and the financial sector’s holding of Treasuries (table 4 of the appendix). Without resorting to control variable, a one euro increase in government debt supply increases the financial sector’s holding of government debt securities by 0.367 euros. After having controlled for the real short-term and long-term interest rate, as well as the real GDP level, the coefficient increases to 0.456 while remaining highly significant. Within the framework of the bank portfolio substitution effect, this relationship makes sense. An

* p<0.05, ** p<0.01, *** p<0.001 t statistics in parentheses N 45 45 45 45 45 45 (12.94) (-1.26) (-0.57) (12.94) (-0.50) (2.51) _cons 0.399*** -0.456 -0.237 0.399*** -0.197 0.603* (9.24) rGDP 0.00396*** (1.55) (-1.59) (0.94) rLTir 0.633 -1.626 0.597 (2.37) (2.39) (-1.62) rSTir 0.854* 2.223* -1.037 (-3.04) (-3.58) (-3.42) (-3.04) (-2.68) (-3.75) Treasury -0.500** -0.568*** -0.585** -0.500** -0.458* -0.368*** NetSTdebt NetSTdebt NetSTdebt NetSTdebt NetSTdebt NetSTdebt (1) (2) (3) (4) (5) (6)

(35)

increase in government debt should increase the yield on government securities, which – provided that the interest rate banks charge on their loans remains constant or does not fully match the increase in interest rates on government debt – makes it more profitable for banks to substitute Treasuries for loans and mortgages to the non-financial sector. While Arvind and Vissing-Jorgensen show that Net long-term assets financed by short-term debt from the financial sector are crowded out by Treasury supply, we could not construct a series for long-term assets. However, considering that our results seem to confirm the crowding out mechanism, and given that our measure of Net short-term debt corresponds to the proportion of long-term assets funded by short-term liabilities, we can expect Treasury supply to crowds out long-term loans to the non-financial sector.

Table 5 and 6 of the appendix bring further insight into the crowding out effects of government debt on short-term debt, with a distinction between checkable and non-checkable deposits. Our analysis point to a larger crowding out effect on non-checkable deposits (-0.495) than on checkable deposits (-0.156), both significant at the 0.1% level. The inclusion of control variables does not affect our results by a large margin, though the coefficient on non-checkable deposits decreases in absolute terms after controlling for the short-term and long-term real interest rate and the real GDP level. Arvind and Vissing-Jorgensen offer an explanation for such disparities in the degree of crowding out. They suggest that Treasuries are important for backing checkable deposits, more so than non-checkable deposits (Arvind and Vissing-Jorgensen 2015), such that a reduction in Treasury supply allows banks to issue more non-checkable deposits but not more checkable ones, as the required Treasuries are scarce, and their yield is low. It would therefore be a reflection of the bank portfolio substitution effect. Another explanation that comes to mind is that in many countries including France, interest rates on checking deposits were forbidden. This ban was lifted on the 23rd_{of February 2005 in France, following a decision of the} European Justice Court to repeal the sanctions imposed on Caixa by the French Banking

Commission for having paid interests on checking deposits in its French branch (LeMonde, 2005). This ban was justified by the desire to channel savings towards medium and long-term investments. As a consequence, one would expect the demand for checking deposits to be