• No results found

Is credit creation for unproductive purposes a burden to the real economy? Evidence from the United Kingdom.

N/A
N/A
Protected

Academic year: 2021

Share "Is credit creation for unproductive purposes a burden to the real economy? Evidence from the United Kingdom."

Copied!
57
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Is credit creation for unproductive purposes a burden to the real economy?

Evidence from the United Kingdom.

MSc Thesis Economics (EBM877A20)

Rijksuniversiteit Groningen

Faculty of Economics and Business

2017-2018

Abstract:

This thesis, building on the frameworks of Bezemer (2014) and Werner (2012), proposes a theoretical distinction between credit creation for productive and unproductive purposes. Using a

Vector Error Correction Model, evidence is provided for the proposition that credit creation for unproductive purposes negatively impacts real economic activity, as measured by Assa’s (2017) FGDP measure. Furthermore, this thesis develops two channels through which credit creation for

unproductive purposes could crowd out productive purposes, however, no statistical evidence is found to support these channels.

Written by Thomas van der Donk (s2508508)

Supervisor : Prof. Dr. Jan Marc Berk

Co-assessor: Dr. Sebastiaan Pool

JEL Codes: B50, E44, E51, G21, P44,

(2)

2

Introduction

In the tradition of the credit theory of money, banks create money when they grant a loan,

ex nihilo, by simultaneously creating a deposit for the borrower. This implies that,

contrary to the view held by many economists, such as Krugman (2015), banks are not mere intermediaries between deficit and surplus units, but actively create purchasing power by granting loans to their customers. When this loan is paid back, the created purchasing power disappears from the economy entirely.

(3)

3

increase in flows to the financial sector, according to these proposed channels, should crowd out positive Net Present Value (NPV) projects by negatively impacting the

earnings before interest and taxes (household-indebtedness channel) and increasing the amount of capital expenditures necessary (production asset- inflation channel).

The propositions that that credit creation for unproductive purposes negatively affects credit creation for productive purposes is tested using a Vector Error Correction Model, and the proposition that rising real estate prices negatively affect credit creation for non-financial firms is tested using a cointegration analysis, utilizing quarterly data for the United Kingdom over the 1995-2014 period.

This paper is novel in the sense that it uses the FGDP index, developed by Assa (2016) as an index for economic activity and aims at combining insights from the

hierarchy of money, credit theory of money and the literature on the relation

between finance and economic growth. It adds a conceptual distinction between credit creation for productive and unproductive purposes and introduces two explicit theoretical channels through which credit creation for unproductive purposes could crowd out credit creation for productive purposes. This paper focusses on the United Kingdom, as it is the fifth biggest economy in terms of GDP1, and the United Kingdom has seen a major shift from manufacturing (27% as

fraction of GDP in 1970 to 10% in 2017)2 towards a (financial) service oriented economy (from 69% as fraction of GDP in 1970 to 79% in 2017)3 over the past decades.

Assa’s measure for economic activity, FGDP, in contrast to GDP4, regards fees paid to the FIRE (Financial, Insurance and Real Estate) sector as input costs of the economy, and not as value adding activities. Assa (2017) argues that this measure outperforms traditional GDP measures in various respects. It regards the output of the FIRE sectors as

intermediary consumption of other sectors, as its end product, money or finance, does not have direct use value. That is, money cannot be directly consumed. This measure will

1 Data from the IMF World Economic Outlook database 2 Data from The House of Commons library.

3 Data from The House of Commons library.

(4)

4

be discussed in chapter two and contrasted to the GDP measure, which is established in the 1993 U.N. System of National Accounts.

In order to provide the reader with the theoretical framework of this thesis, I will discuss the credit theory of money and the hierarchy of money in chapter one. Chapter two will review theoretical literature and empirical evidence on the effect of credit creation on economic growth. Furthermore, in this chapter Assa’s proposed measure for economic growth is introduced and contrasted to the traditional measurement for economic growth, GDP. In chapter three, the two proposed channels that could explain a possible crowding out of credit creation for productive purposes by credit creation for

(5)

5

Chapter one: The nature of credit creation in a modern economy

The view that banks are more than just financial intermediaries between surplus and deficit units, and that loans create deposits is, over time, becoming more widely

accepted in both academic and professional economic circles. Examples include, but are not limited to, important institutions such as the Bank of England. The implication of this proposition is that the money supply is endogenously determined and demand driven. According to Werner (2014), the idea that individual banks create money when they issue credit originates back to Macleod, who in 1855 published the book The Theory and

Practice of Banking. This idea was adopted by various important thinkers such as

Schumpeter, Keynes and Wicksell, but was over time dominated by the financial

intermediary theory of banking and the fractional reserve theory of banking. These

theories remain very prevalent within orthodox economics, although many prominent monetary economists and prominent institutions, have come out in favour of the idea that loans create deposits. In order to explain the nature of credit creation in a modern monetary economy, it is necessary to discuss and delineate the concepts of money and credit. It is necessary to touch upon these concepts as there exists various rival

definitions of what exactly constitutes money, which can have a profound impact on an argument related to the creation of money.

Money, credit and the hierarchy of money

According to Bell (2001), there has been an extensive discussion for centuries on what exactly is and constitutes money. Similarly, there exist extensive amounts of literature on the origin of money. I will abstain from discussing the origin of money, as it is beyond the scope of this thesis. Instead I will introduce several definitions of money and discuss the hierarchy of money in a modern monetary economy.

(6)

6

of account, for it provides the terms in which prices of goods and debt are recorded and denominated. Lastly, money serves as a medium of exchange, that is, it can be used to sell and purchase goods and hence makes indirect transactions possible.

While these purposes are commonly accepted, the definition of money remains

controversial. In a much discussed paper, published in 1913, Innes introduces a different definition of money. Innes argues that credit, and in his words, only credit, is money. According to Bell (2001), building on Innes’ definition, money represents a debt-relation between human beings, in the form of a promise or in the form of an obligation, that cannot be regarded in a vacuum, but can only be identified in its institutional usage. Similarly, Kaldor (1970) defines money as “those forms of financial claims which are

commonly used as means of clearing debts.” (Kaldor, 1970, p. 7). It is Kaldor’s definition

that, in the rest of this thesis, will be used to define money.

Kaldor’s definition implies that money simultaneously exists as an asset, in the form of credit, and as an liability, in the form of debt. Minsky (1986), argued along similar lines, for he stated that everyone can create money. The problem, however, is not creating money, but getting this money accepted by others. Bell notes, importantly, that as money is a debt relation, and hence always a balance sheet operation, it is not money until it actually has become accepted by others. This is, because without acceptance it cannot serve as a debt relation between individuals that can clear other outstanding debt. When money is defined in terms related to debt and credit, it is for clarity important to note that these words are interchangeable. Both represent the different viewpoints of an IOU. To use Innes’ words: “What A owes to B is A's debt to B and B's

credit on A. A is B's debtor and B is A's creditor. The words "credit" and "debt" express a legal relationship between two parties, and they express the same legal relationship seen from two opposite sides” (Innes, 1913, p. 392).

(7)

7

(8)

8

Money creation in a modern monetary economy

It has been argued by various authors, such as Palley (1991), Lavoie (2014), and Werner (2014), that when banks create a loan, they simultaneously create a deposit for the borrower. Banks when granting a loan, thus effectively create money, provided that other banks accept their deposits. Banks can do this, according to Werner (2014), because they are the only entity that can legally put deposits from surplus units on their own balance sheets. This function gives them the ability to create commercial bank money, as it is a priori

impossible to distinguish between a deposit derived from loan creation, or one derived from surplus unit saving. Once both deposits are on the balance sheet of the bank, they both are (second tier) money, as both the surplus unit deposit and the deposit originating from a debt contract, represent financial claims on the bank. The result is that both deposits, in practice, can be exchanged for legal tender (government money) issued by a central bank under the jurisdiction of a (national) government. So despite the central bank being the only entity that can create money out of nothing, for instance by issuing money and using it to buy bonds or other assets, regular commercial banks can create second tier money for their borrowers by granting credit (lines) to their creditworthy customers.

Furthermore, it has been argued that banks first grant a loan and look for the funding later. This view is also stated by the vice president of the ECB, Constancio, in a speech in 2011, This implies that the causality runs from loans to commercial bank reserves at the central bank, rather than is often proposed, from commercial bank reserves at the central bank to loans, which is a necessary condition for banks to grant loans ex nihilo. According to Lavoie (2014), banks do have to be solvent in order to create these loans ex nihilo. Otherwise, a central bank (in general) or the interbank market, will not provide a liquidity backstop to an illiquid bank, as it is now insolvent. One could argue that this would result in this illiquid bank losing its second tier place in the hierarchy of money, as the financial claims on the bank are now no longer commonly accepted as means than can be used to settle

outstanding other debt. Thus, banks will extend credit only to those that can provide evidence, mostly in the form of collateral, to the bank that they will not forfeit on the repayment of their loans.

(9)

9

determined by the central bank via a corridor system. The size of the mark-up, at which banks are willing to create credit for their customers, depends on the risk and illiquidity premium associated with the project, and is determined by the bank (Lavoie, 2014). In case of a change of the short run rate by the central bank, the adjustment to the new interest rates is not instantaneous (Gambacorta, 2008). Over time, however, short run interest rate pass-through is one for one.

The amount of bank loans that a bank can create is still up to debate. Various authors, such as Moore (1998), claim that loan supply is horizontal. Moore thus argues that the interest rate that banks charge their customers is independent of the amounts of loans created by the bank, but only dependent on the short term interest rate, the mark up and the individual characteristics of the project (i.e. liquidity, risk), mandatory reserve

(10)

10

Chapter two: the effects of credit creation on economic growth and the

distinction between credit created for productive- and unproductive

purposes.

The relation between credit creation and economic growth

Schumpeter, in 1934, already argued that for additional production above the current maximum capacity to take place, the creation of extra liquidity is a necessity. Based on these observations, Graziani (2003) introduced the circuit theory of production, in which the production process starts with the creation of money in the form of a loan being granted by a bank, and ends with the repayment of this debt after wages have been paid and

consumption has taken place. In the case that a consumer decide to save, both banks and firms will compete for these savings. A bank uses these deposits to be able to provide short term liquidity and conversion of their liabilities in to government money, whereas the firm will try to get the cheapest way to provide themselves with extra funding. The moment a bank loan is repaid, the money is effectively destroyed. So, this theory argues that expansion of a production process starts with the creation of bank credit, and only when savings

happen, financial markets become relevant. It is important to note that these savings can only exist if another unit is in debt. Productive debt, thus, is the driver of economic activity and more productive credit extension should lead to higher economic growth levels. Such a view is also supported by Lavoie (1984), who argues that for firms to increase their production above a stationary level, they need to attract financing, for instance in the form of bank credit lines. Stationarity, in this respect, is defined as the level where firms are liquid enough to finance their expenditures. This bank credit line, when used, is an asset on the bank’s balance sheet. The income (i.e. wage income) generated by the bank financed activities flows back to the liability side of the bank balance sheet in the form of deposits, once the factors of production have been paid their respective shares. Empirically, there is a broad array of literature on the relation between financial development and economic growth. For instance, De Gregorio & Guidotti (1995) and Calderon & Liu (2003) indeed find positive evidence for the existence of this relation.

(11)

11

performance (i.e. by influencing the amount of transaction and monitoring costs). Changes in effectiveness often originate from financial innovation. It is argued, for instance by Litan (2010), that financial innovation has drastically increased the access of both the general public and private corporations to finance, which positively affects GDP, as households are able to further smoothen their consumption over their lifetime. The effect of financial innovation on economic growth depends per innovation. Examples of positive financial innovations include ATMs and debit cards. There are ample negative financial innovations as well, such as the poorly designed SIVs and CDOs. I will however not further touch upon this subject, as this is beyond the scope of this thesis.

A quadratic relationship between credit-to-GDP and economic growth?

Bezemer (2014) observes that at a certain ratio of debt-to-GDP, the relation between debt creation and economic growth ceases to hold and even turns negative. Bezemer finds an explanation in the ideas of Schumpeter, who argued that loans can be created for productive and unproductive purposes. Werner (2012) argues that productive loans are loans that can generate value adding income flows that are large enough to pay interest payments and repay the principle. Werner (2012) also argues that credit creation for unproductive purposes, implies credit creation for non-value adding transactions. As the economy does not expand from these transactions, this credit creation for unproductive purposes will only result in asset price inflation. Bezemer (2014), basing his argument on Schumpeter’s analysis of a monetary economy, argues that there are two waves of credit creation. The first wave of credit creation will consist mainly of productive credit and result in growth and

innovation. The second wave, following endogenously from financial development and economic growth, will predominantly consist of unproductive credit. It will follow

endogenously from the first wave, as agents expect that the observed rates of change will continue endlessly, and start speculating on that outcome.

(12)

12

stock in combination with the size of the stock that is relevant, rather than the stock of debt in general.

Werner (2012) argues that credit creation for unproductive purpose, at the aggregate level,

is an unsustainable zero-sum game. This credit creation is unsustainable as these

unproductive projects are reliant on capital gains and do not generate new income

necessary to service and repay the debt. Furthermore, in order to realise these capital gains, the owners of these unproductive assets are reliant on further credit creation for

(13)

13

renders the bank insolvent, and thus results in bank defaults. So, while not resulting in real economic growth, these loans for unproductive purposes do result in financial fragility. Bezemer & Hudson (2016) introduce a table, table 1, that conceptually distinguishes between the several types of loans. It can be seen that consumer loans, according to Bezemer & Hudson, results in high levels of GDP growth. This is, because these loans create liquidity that is used in the real economy. However, they come with a downside compared to a non-financial business loan. This downside is that there is a separation of expenditure and income flows. That is, while the (real) expenditures do generate income on a

macroeconomic level, they (in general) do not generate income for the debtor herself. This separation can result in macroeconomic instability.

Table 1 : Reprinted from Bezemer & Hudson (2016).

A conceptual distinction between credit creation for unproductive and productive purposes

I am aware that it is a priori impossible to establish whether credit creation for productive purposes will be actually successful in adding value to the economy, as this can only a

posteriori established after consumption has taken place. It is a priori impossible to establish

whether a project will add value to the economy, as value added is measured based on final consumption, thus only after consumption has taken place, one could inspect whether a project has added value to the economy. There are plethora examples of projects being undertaken that never generated the income flows necessary to service the debt.

(14)

14

ownership of this house will change hands, but this transaction does not add value to the economy (in other words, the economy does not expand from these transactions). In contrast, a credit flow for productive purposes could be deemed as such when it can be a

priori established that this could, on a macroeconomic level, generate income equal to the

loan and the interest payments. Leaving regulatory capture, moral hazard, fraud, and corruption aside, it is hard to imagine a situation in which an economic agent takes out a loan with the intention of not repaying this loan, and using this in a non-productive manner. The other party involved in the credit creation process, the commercial bank, is assumed, to act in a profit maximizing manner and a potential loss on a non-value generating project would have to be covered by the bank’s own funds. Hence, the bank is assumed to provide credit only to projects that a priori are expected, by the bank, to have a positive net present value. The fact that a posteriori a project might not be able to service the debt is irrelevant for this discussion, as it is inherent to the notion of uncertainty in an economy. Hence we are left with three cases, unproductive credit for productive purposes, productive credit for productive purposes and credit for unproductive purposes5. In order to provide the reader with some clarity surrounding these concepts, I will provide the reader with several

examples in table 2.

Similar to credit creation for real production, credit creation for finance, is not necessary unproductive. Minsky (1992) argued that only hedge finance is sustainable, as this form of finance is able to generate the necessary income flows to service their debt. Speculative finance, however, is unable to repay the principal amount, but is able to service the interest payments. Hence there is the constant need to roll-over the speculative debt. Minsky also describes Ponzi finance, which can neither repay the principal nor pay the interest

payments, and hence needs to attract additional credit to pay interest payments, resulting in a Ponzi scheme. Speculative and Ponzi finance can be considered as credit created for

unproductive purposes. Minsky (1992) argued that in prolonged booms, the financial structure of capitalist economies endogenously moves from being dominated by hedge finance towards speculative and Ponzi finance. Lavoie (2014) describes these Minskyan dynamics using the concept of the paradox of tranquillity. He argues that in a world full of

(15)

15

uncertainty, and with imperfect knowledge about the economic fundamentals, several successful years tend to induce a state in which agents tend to forget the hardships

experienced in the past. Instead, economic agents start to take more risk. Furthermore, he claims that the more prolonged a boom is, the more computed risk models, such as the Value-at-Risk model, will show that risk is decreasing, as the previous recession is just a mere exception in a sequence of successful years. Hence, these dynamics will induce economic agents to expose themselves to more risk. The paradox of tranquillity then, is that a stable state, itself, is destabilizing (Lavoie, 2014).

Table 2: The three different types of credit creation.

Type A priori

expectations

Examples A posteriori

result

Unproductive credit for productive purposes

Productive (value adding)

Unfinished highways in Southern Italy, unused airports in Spain.

Unproductive

Productive credit for productive purposes

Productive (value adding)

Bank loans to successful (macroeconomic) value adding projects, where successful is defined as being able to service the debt.

Productive

Credit for unproductive purposes

Unproductive (non-value adding)

Mortgages for existing real

(16)

16

Alternative explanations for a diminishing effect of credit growth on GDP

Alternative explanations as to why the relation between credit-to-GDP turns negative at a certain ratio are discussed by Arcand, Berkes and Pannizza (2015). These authors include the reasoning that as economies develop credit markets lose importance for the financing of operations, and that financial markets become more dominant in providing this role. Another line of reasoning that is provided by Arcand, Berkes and Pannizza (2015) is that as financial sectors grow, they drain intellectual capacity from the real economy. Kneer (2013) provides evidence of such a brain drain, by arguing that financial liberalization has resulted in a decline in the growth rate of skilled workers in skill-intensive industries. I argue

however, that the previously described effects can co-exist with Bezemers argument, as over the past decades most developed economies have gone through a process of financialization (Assa, 2017). Financialization is defined by Krippner (2005) as the process where profit-making in increasing levels occurs via financial channels, rather than via the production of goods or trade. Krippner defines these financial channels as referring to activities related to the provision and/or transfer of liquidity in the expectancy of future capital gains, dividends or interest. These financial channels can, following the discussion on the previous pages, be regarded as unproductive channels.

I argue that as there is an overextension of initial liquidity, resulting from credit extension for unproductive purposes, used to purchase financial or existing real assets, these transactions have to result in a capital gain somewhere else in the system. This capital gain, however, is financed completely with increased indebtedness of the household and financial sector. This capital gain will have either to be saved, invested or consumed. In the case of the second, this will result in a growth of financial markets, as the unit that made a capital gain will search for the most profitable manner to reinvest this gain. Bezemer (2014) argues that the money saved by a household has to be reinvested6, and that this is predominantly done in financial assets and in real estate rather than in fixed capital formation, and hence will drive up the price of the asset they are invested in. This implies that despite the high achieved returns on the investments, it does not generate more income on the macroeconomic level, as income can only be derived from production. In contrast to income, money can be

created with the stroke of a pen. As discussed before, the increased capital gains eventually

(17)

17

have to be repaid from future productive income. The increased amounts of saving (that also, as discussed before, represent a claim on another unit) resulting from the capital gains made, could give rise to an increase in the amount and size of financial intermediaries in the system, which are market based. When the brain drain argument is regarded an externality of this increase in size (and the overall increased complexity of the financial system), Kneer’s brain drain argument seems likely and compatible with Bezemers theory.

Existing empirical evidence on the effects of different types of credit creation on GDP Several authors have tried to empirically establish the effects that the different types of credit have on economic performance. In this section I discuss several of these empirical papers. However, as the focus of research is slowly shifting towards the relation of private debt composition and GDP, rather than solely focusing on the total private debt stock, empirical evidence remains relatively scare compared to the research on the relation between the Debt-to-GDP ratio and the economic performance of a country.

Jordà, Schularick, and Taylor (2016) find, using disaggregated credit data, that mortgage loans have increased from one third of the total banking business in the beginning of the 20th century, to almost two thirds today. They also show using sophisticated econometric techniques that financial crisis risk has shifted towards mortgage lending booms, and that these are associated with lower growth rate in the post-crisis recovery. Lastly, they show that the modern business cycle is increasingly shaped by the extension of mortgage credit, and that the effect of non-mortgage lending on the contemporary business cycle is minor. Angeles (2015) finds, using a GMM estimator, that credit to households is negatively related to GDP growth. When Angeles controls for financial crises, which he presumes are mostly the result of credit to the household sector, it becomes apparent that there is a negative to a negligible effect on GDP growth, and that only credit to firms actually results in GDP growth. Sahay et al (2015), in an IMF staff paper, also find that credit to firms has bigger growth effects than credit extended to households. In line with other findings, Beck et al (2009) find using a panel data model using 45 countries that the effect of household debt on GDP growth is negligible. Credit to non-financial firms however, in line with the previous

(18)

18

household credit growth weakly negatively affects GDP growth. Enterprise credit however, is statistically strongly related to GDP growth.

A new leading indicator? Final Gross Domestic Product vis-à-vis Gross Domestic Product

Gross Domestic Product (GDP), since the introduction of the 1993 edition of the System of National Accounts (SNA), includes the financial fees paid to financial institutions as value adding activities. GDP can be approximated via three approaches. The value added

approach, or also known as the production approach, estimates the gross value of output of each sector and subtracts intermediate consumption. The expenditure approach is a

summation of all domestic consumption expenditures, gross fixed capital formation, government expenditure and net exports. The income approach is to add up all income earned by both firms and households (that is, labour income, rental income, interest income and profits).

As previously discussed, since the 1993 edition of the SNA, finance has become a productive industry in terms of GDP. For instance, the 2008 SNA discusses explicit fees, such as the buy-sell spread on derivative trades, which are essentially fees paid to market makers, and regards these as value-adding activities7. The capital gains/losses resulting from these trades however, are regarded as transfers and are thus not regarded as value adding activities. The 2008 SNA also states that not all fees are explicitly priced but that financial services are added in to the price of loans and deposits. In the case of a bank deposit, the SNA argues that the interest rate is lower than the reference rate (often taken to be the interbank interest rate). Similarly, it states that the interest rate on loans is also higher than the reference rate, due to implicit service charges introduced into the lending rate. This implies that these Financial Intermediary Services Implicitly Measured (FISIM) are calculated according to the following function; 𝐹𝐼𝑆𝐼𝑀 = (𝑅𝑙− 𝑅𝑟)𝐿 + (𝑅𝑟− 𝑅𝑑)𝐷, where 𝑅𝑙 is the bank lending interest rate, 𝑅𝑟 is the reference rate, 𝑅𝑑 is the bank deposit rate, 𝐿 is the amount of outstanding loans, and 𝐷 is the amount of bank deposits. The reference rate is often defined as the interbank rate, although there exist rival definitions. While FISIM and explicit fees are subtracted as intermediate costs of corporations and governments, for households they are regarded as final consumption.

(19)

19

Assa (2017), however, argues that financial fees should not be part of GDP. To the contrary, he argues that these fees should always be subtracted, regardless of whether they are implicit or explicit. He claims that all these financial fees are intermediary inputs for production, and that the end product of financial intermediaries, credit or money, has no direct use value. Assa states that money can be used to purchase consumption or production goods, but cannot directly be consumed.

In an Indigo Prize winning essay, Coyle and Mitra-Kahn (2017), state similar views, arguing that as the banks’ business models have changed from traditional business banking away to investments in financial capital, they should be netted of GDP. As an alternative to GDP, Assa provides a measure, Final GDP (FDGP), which effectively is GDP minus the value added of the FIRE sector, and minus the output of the FIRE sector. Both value added and output are subtracted, as subtracting only value added would result in an economy without finance, and would not make it an intermediary input of cost to all the sectors that produce output, that has direct use value.

Assa argues that it is important to reconcile the FGDP measure with the income and expenditure approaches, as all are based on accounting identities. To reconcile with the expenditure approach, Assa, using input-output tables, calculates the contribution of each sector to the four expenditure categories8. Then, Assa reduces each of these four categories proportionally to the share of financial value coming from these categories. Reconciling with the income approach is more complicated. Assa argues here that as finance represents a cost to the economy, it’s income should be treated in similar terms. This implies that he proposes to reclassify the financial fees to taxes or costs. Doing so, he proportionally subtracts the share of finance from each of the three income categories9.

Despite measurement difficulties, Assa’s (2017) index does include sectors such as the government, health care and education, as these sectors produce “goods” with final use-value. Assa relates the inclusion of these sectors to the labour intensity of the work, as he argues that if more medical check-ups are performed, more doctors are employed. For

8 As discussed before, these are final consumption expenditure, gross fixed capital formation, government expenditure and net exports.

(20)

20

finance, Assa claims that this does not hold, as he finds a negative correlation between finance’s share in total GDP and its share in total employment.

He argues that, besides his conceptual argument, there are three (other) main reasons to support his FDGP measure over GDP in economic analysis. Assa claims that GDP is failing as a leading indicator, that it is diverging from employment measures and that it has become decoupled from median income. The FGDP measure, he argues, is more effective in measuring real economic growth, and outperforms traditional GDP in several respects. For instance, it outperforms traditional GDP in an out-of-sample forecast of changes in employment, and is capable of explaining the Great Moderation. Assa explains this

(21)

21

Chapter Three: The quantity theory of credit and two non-productive credit

channels

In this section I provide two channels via which credit creation for unproductive purposes could crowd out credit creation for productive purposes. I will do so by building on the arguments by Werner (2012) and Bezemer (2014). In a 2012 paper, Werner introduces the quantity theory of credit. This theory is based on Fisher’s equation, 𝑀𝑉 = 𝑃𝑄, with 𝑀 being the active money supply, 𝑉 the velocity of money, 𝑃 represents the price level and 𝑄 is defined as the quantity of the transactions in a specified time span. Often, as 𝑄 is, or used to be before central banks became the clearing houses of commercial bank transactions, hard to measure, 𝑄 is substituted for 𝑌, which represents income.

Werner criticizes contemporary use of Fisher’s equation for monetary analysis, as it does not adequately describe money in a modern economy and does not acknowledge that Y, proxied by GDP, is a mere subset of all transactions that take place in an economy. As discussed before, money is a commonly accepted financial obligation that can be used to settle debt and can be created with the stroke of a pen, and its measurement, M3 or M4, measures the outstanding stock of all deposits, regardless of what these financial obligations are used for. Credit, in contrast, always has a directly purpose, namely to be used for a certain type of transaction. Werner hence argues to substitute 𝑀 for 𝐶, where 𝐶 implies credit creation. This gives rise to the following function; 𝐶𝑉 = 𝑃𝑄. Rather than using just 𝑌 to approximate 𝑄, Werner proposes to separate the right hand side in two subsets, 𝑄𝑓 and 𝑄𝑟, where the subscript denotes respectively the transactions in the financial and real sector. 𝑄𝑟 is proxied by Y. This also implies that both sectors have their own price deflator, 𝑃𝑓 and 𝑃𝑟. Similarly, Werner decomposes 𝐶 into 𝐶𝑓 and 𝐶𝑟.

This yields a new quantity equation that encompasses all transactions in a monetary economy; 𝐶𝑓𝑉𝑓+ 𝐶𝑟𝑉𝑟 = 𝑃𝑓𝑄𝑓+ 𝑃𝑟𝑌, for which holds that 𝐶𝑓𝑉𝑓 = 𝑃𝑓𝑄𝑓 and 𝐶𝑟𝑉𝑟 = 𝑃𝑟𝑌. This also implies that Δ(𝐶𝑟𝑉𝑟) = Δ(𝑃𝑟𝑌).

(22)

22

The new quantity equation can be used to analyse what happens when credit is granted for unproductive purposes. Suppose, a household wants to buy an already existing house, and applies for a mortgage at a commercial bank. The commercial bank grants the household the loan. After this credit for unproductive purposes is granted to a household to buy this asset, 𝐶𝑓 increases with the amount of that loan. From the 𝐶𝑓𝑉𝑓 = 𝑃𝑓𝑄𝑓 equation, it follows that 𝑃𝑓𝑄𝑓

𝑉𝑓 must increase equiproportional with the increase 𝐶𝑓. As 𝑉𝑓 is presumed to be constant,

and unaffected from this transaction, it follows that there is an increase in the price of these non-GDP assets. That is, if the capital gains made from this transaction by the selling unit are reinvested in another non-GDP asset. While this assumption is very restricting in practice, it does show how money created for non-GDP transactions can drive up non-GDP asset prices. As previously discussed, Werner argues that this credit creation for non-productive purposes must be a Ponzi scheme, as on the macroeconomic level these loans do not generate

income. This implies that in order to keep generating capital gains on the individual level, the economy has to indebt itself at increasing pace, and keep investing in these financial assets. Such a dynamic is not unrealistic, as with higher asset prices the debt-to-equity ratio of bank, with a balance sheet marked-to-market, decreases. This gives the bank the incentive to increase their credit creation for unproductive purposes, resulting in higher FIRE asset prices, which in turn can result in an upwards financial asset price - leverage spiral.

As discussed before, this dynamic is related to banking crises and asset bubbles, both of which can be very costly to the real sector. I argue that there are two other mechanisms through which credit creation for unproductive purposes can crowd out credit creation for productive purposes. In the next section I will introduce these two mechanisms, where one is a demand related mechanism, and the other is a supply related mechanism.

The household indebtedness channel

The demand effect is what I refer to as the household indebtedness channel. This channel is an explication and extension of the argument that Bezemer (2014) makes. Bezemer (2014) argues that liabilities resulting from credit creation for unproductive purposes will have to be paid out of (future) real income. However, as discussed in the previous chapter, credit

(23)

23

flows sufficient to service the debt, hence less future income can be used to purchase goods and services.

The household indebtedness channel can be described in several steps, and is dependent on several assumptions. As discussed before, Bezemer (2014) showed that there exists a strong relation between credit extended to the FIRE sector and real estate prices, this implies, following 𝐶𝑓𝑉𝑓 = 𝑃𝑓𝑄𝑓, that there is relatively little leakage from the non-GDP sector to the real economy. As there is little leakage, the income from capital gains does not enter the real economy, although there could be second round effects, such as increased household

spending via a net- wealth effect. Given that the central bank in an inflation targeting country adjusts the interest rate not on changes in 𝑃𝑓 but only on changes in 𝑃𝑟, the asset inflation resulting from increases in 𝐶𝑟, resulting from credit creation for mortgages or other financial assets, will not automatically be controlled counter-cyclically via changes in the short term rate by the central bank10.

This mortgage induced indebtedness is unrelated to interest payments associated with the indebtedness, as the lending rates are determined as the short run rates with a risk and illiquidity premium and do not react to the amount of outstanding mortgage loans. This can clearly be observed from the sub-prime crisis in the United States where mortgage rates did not increase as more mortgages were granted. Hence, the principal payments, on the sector level, and the interest payments associated with these loans, have to be repaid are higher vis-à-vis a situation where this credit had not been extended.

As (future) income is lower, these implies that for future investment projects dependent on this –now forgone- demand, there are lower earnings before interest and taxes (EBIT), as there is going to be less demand for the goods produced by the (would-be) entrepreneur. From this it follows that the discounted free cash flows originating from the investment project will be lower than they would be vis-à-vis a situation without these credit flows and the resulting asset inflation. While on the individual level many projects will still be able to get funding, the reduction in future aggregate demand on the macroeconomic level will reduce the total EBIT of projects, on the macroeconomic level, and hence result in lower

(24)

24

credit creation for productive purposes, as on aggregate there are less projects that fulfil the positive NPV requirement.11

The production asset inflation channel

The second channel that could explain this crowd out is a supply side argument. It can be argued that within urban areas, households and firms compete for the same scarce good, land (i.e. stores, cafés, commercial offices, industrial production locations, ateliers or pubs versus apartments and houses). While urban areas can expand over time, I assume land supply within the city is fixed in the short and medium run, as exploiting new land and making it commercially viable is time-consuming. In most urban areas there are zoning laws which dictate for what purpose the land can be used (i.e. commercial or residential). Hence, the markets for commercial land and residential land are directly (and physically) separated from one another by regulation.

As established earlier, when bank credit flows to the FIRE sector (including household mortgages), this results in higher real estate prices. While an increase in credit to non-financial corporations also presumably drives up commercial estate prices, such increased costs would presumably be transferred to customers via increased prices, thus resulting in inflationary pressure in the market for consumer goods. Hence, one can argue that there are countercyclical (interest rate) policies in place by an inflation targeting Central Bank, to prevent excessive increases in commercial estate prices. As previously discussed, while the short run rate decided upon by the Central Bank affects real estate inflation, it is not part of the Central Bank reaction function, unless the capital gains from selling real estate assets are used for consumption purposes12.

Due to price discrepancies between the different land, and land uses, individual land owners, speculators and developers can profit majorly from changes in the zoning plan. Hence, a growing discrepancy between real estate prices and commercial estate, resulting from the amounts of credit created for these purposes, could result in pressure on local authorities to convert commercial areas into residential areas. Similarly, a revenue maximizing local

11 Another negative effect on the firm’s EBIT could be that as firms (with a downward sloping marginal cost curve) face less future demand, they will have to reduce their production, resulting in higher costs per unit, thus reducing overall economic performance.

(25)

25

government could choose to sell of publicly owned land to developers, in order to maximize their revenue in the short run.

According to a Greater London Authority (GLA) economics report (2016), there is indeed pressure on the municipality to convert commercial areas into residential areas. In the same report it is argued that over the past 15 years, industrial land in London fell from 8,282 hectare in 2001 to 6,976 hectare in 2015. This is however, not due to a lack of demand for industrial locations by entrepreneurs in London, as becomes clear from the core strategy of the Camden local board; “The Camden Employment Land Review 2008 found that the cost of

industrial locations in Camden is high, indicating that supply does not meet demand. However, there has been pressure to redevelop the borough’s stock of land used for employment purposes, particularly manufacturing and industry, for higher value uses, principally housing. Once employment land in the borough has been developed for an alternative use it is very unlikely it will ever be returned to industrial use.” (Camden Local

Plan, 2010, p. 60).

(26)

26

speculative, or even Ponzi, motive to holding onto (land destined for) real estate13. Similarly, a speculative household could decide that the expected return on their asset, the house, outweighs the total costs (including time) of the increased commuting distance.

As both channels work simultaneously and in the same direction, these effects combined could make the entrepreneur, who without these effects would have been creditworthy, non-creditworthy, as with the inflated asset prices, their lower free discounted cash flows could be insufficient to service their now higher debt. Hence, on the macroeconomic level projects that would have been feasible, are not able to get funding, because of the credit creation for unproductive purposes and the resulting asset inflation.

(27)

27

Chapter four: Methodology and empirical model

In this chapter I will discuss the methodology and the empirical model that is used to test the hypothesis that credit creation for productive purposes drives real economic activity, the hypothesis that credit creation for unproductive purposes negatively affects real economic activity, and to test the household indebtedness channel and the asset price inflation channel. The main claims of this thesis can be summarized in four hypotheses, namely:

H (1): In the long run, an increase in credit creation for productive purposes results in an increase in economic performance.

H (2): In the long run, there exists a negative relation between credit creation for unproductive purposes and economic performance.

H (3): In the long run, increasing average housing prices result in a reduction in credit creation for non-financial firms.

H(4): In the long run, an increase in credit creation for unproductive purposes results in a reduction of credit creation for productive purposes.

Hypothesis (2) reflects the overall expected negative effect of credit creation for

unproductive purposes on economic performance, while hypothesis (3) and (4) could be regarded as simplified versions of the respective productive asset price inflation – channel and the household indebtedness -channel.

(28)

28

model. Next, a Johansen cointegration test is executed, to see whether there exists a cointegration vector. If there exists a cointegration vector, then a VEC model is estimated.

(Non) Stationarity and cointegration

For proper statistical inference of time series, it is of paramount importance that all the statistical variables, or a linear combination of the variables, are of a stationary nature. Applying econometric techniques to non-stationary time series can result in spurious regressions. There are ample examples of published spurious regressions that yield a very high 𝑅2, yet also have an extremely low Durbin-Watson statistic (Granger and Newbold, 1974). The concept of stationarity can very well be explained by introducing a time series following an autoregressive process, modelled by the equation 𝑥𝑡 = 𝛼 + 𝜃𝑥𝑡−1+ 𝜖𝑡. This series is stationary if the coefficient |𝜃| < 1, as a shock to the time series will result in the variable 𝑥𝑡 moving back to its mean 𝑥 over time. Suppose now that |𝜃| > 1, then 𝑥𝑡 will explode over time in the case of a shock, and the series can be regarded as non-stationary. If 𝜃 = 1, then the variable is said to contain a unit root and is regarded non stationary. In that case shocks will not fade over time, and are thus permanent and the variance will increase as (t) → ∞. A stationary variable is integrated of order 0, which is denoted as 𝐼(0).

(29)

29

Vector Error Correction Model

A Vector Error Correction (VEC) Model is a special form of a vector autoregressive (VAR) model. The VAR model is given by 𝑌𝑖,𝑡 = 𝛼𝑖 + 𝜃1𝑌𝑖,𝑡−1+ ⋯ + 𝜃𝑝𝑌𝑖,𝑡−𝑝+ 𝜖𝑖,𝑡, where Y is a vector of 𝑖 variables of interest, and 𝑡 denotes the time period. Given the structure of this VAR model, all variables 𝑖 are regarded as endogenous variables, although exogenous variables could be added. If the variables 𝑖 are all 𝐼(1), and they are cointegrated, then the VAR model can be rewritten as error correction model for the first differenced series14. The error correction terms (ECT) will then capture the extent of the deviations from the long run equilibrium.

(30)

30

Chapter five: Data

In this chapter, the data that is used for the testing of the proposed channels will be introduced and the results of the relevant statistical tests will be discussed. As discussed before, the focus of this thesis is on the United Kingdom. Due to data availability limitations, the data that will be used to test the earlier stated hypotheses covers the period 1995 until 2014. The data on the amounts of net outstanding credit and the changes in net lending are retrieved from the Bank of England, the central bank of the United Kingdom. These data are in millions Pound Sterling and thus are nominal. Data on the quarterly national accounts is retrieved from the Organisation for Economic Co-Operation and Development (OECD) statistics database, and is in millions current price Pound Sterling. Data on the average UK house prices are retrieved from the Office of National Statistics, and is in Pound Sterling. Furthermore, all data is transformed into their natural logarithms.

Data on credit creation

As a proxy for credit creation for mortgage purposes, the amounts of outstanding sterling M4 lending secured on dwellings is used as stock measure for the last quarter of 199415, then for each subsequent quarter, the net M4 lending secured on dwellings in that period is added16. So, the manually calculated stock of lending secured on dwellings for the first quarter of 1995 is given by the amount outstanding in the last quarter of 1994 plus the net lending in the first quarter of 1995. For the next quarters, the net lending in this period is added to this previous stock. Net lending is chosen, as this reflects net money creation in a monetary economy, for repayments of loans imply the destruction of money and new loans imply the creation of money, as they represent claims on the bank. The reason to use this computation rather than just using the amounts of outstanding M4 lending secured on dwellings provided by the Bank of England is that this measure, includes revaluations and write-offs. Write-offs, represent the writing off by banks on the firm and consumer debt they hold. These are tier 3 and 4 debts in the hierarchy of money, and following the application of Kaldor’s definition of money, this does not represent money destruction, as these debts are not defined as money in the first place. Their counterpart liability, the banks own funds, is its own capital, which does not represent a direct claim on the bank. A reduction of the banks

(31)

31

own funds resulting from the write-off of debt does thus not reduce the amount of money in the economy. This also holds for the other proxies. While this measurement of mortgage lending does not encompass credit created for the real estate sector, as is done in for instance in the industrial analysis of monetary financial institutions lending to UK residents series, it encompasses the sterling lending of all monetary financial institutions, whereas the more detailed industrial analysis of monetary financial institutions lending to UK residents includes only UK resident banks. Credit created for financial purposes is proxied by the amount outstanding Sterling M4 lending to other financial corporations in the last quarter of 199417, and in each subsequent period the change in net lending from the last quarter of 1994 to that period is added18. The sum of these two proxy series will proxy the total stock of unproductive credit in the United Kingdom in a given quarter.

As proxy for credit creation for production purposes, the amount of outstanding Sterling M4 lending to non-financial corporations is used for the last quarter of 199419, and in each subsequent period the net change in lending from the last quarter of 1994 to that period is added20. Furthermore, the amount of outstanding Sterling M4 lending to unincorporated business and non-profit institutions serving households is used for the last quarter of 199421 and in each subsequent period the net change in lending from the last quarter is added22. The sum of these two stocks represents the outstanding amount of credit used for

production purposes. Similarly, as proxy for credit creation for consumption purposes, the amount of outstanding Sterling M4 consumer credit lending to individuals, excluding securitisations is used for the last quarter of 199423, and in each subsequent period the net change in lending from the last quarter of 1994 to that period is added24. The sum of these three series will proxy the total stock of credit created for productive purposes in the United Kingdom in a given quarter.

In order to ascertain that there is no double counting of credit flows, the sum of total outstanding lending to other financial institutions plus the sum of outstanding lending of

(32)

32

lending to non-financial corporations plus the sum of outstanding lending to the household sector is subtracted from total outstanding lending to the private sector. If there is a case of double counting, then the value of this calculation should be smaller than zero. However, despite some minor statistical discrepancies, this value is effectively equal to zero. Hence, there is no double counting of credit flows.

House prices

Rather than using an index for housing prices, I have chosen to use the simple average of the house prices in the United Kingdom. This includes the average price of both new houses and existing houses. This data originates from the Office of National Statistics, can be found in the House price simple averages dataset, and is in Pound Sterling.

FGDP measure

As a measure for real economic performance, the measure developed by Assa (2017), FGDP, will be used. This measure is constructed as displayed in Figure 1. It subtracts from Gross Value Added, the value added of the FIRE (Financial, Insurance and Real Estate) sectors to arrive at an economic measure where the financial sector is neutral to economic output. Then, following Assa (2017), the FIRE sector is regarded as an input cost/tax on other

sectors, hence its output is subtracted. Then, this is multiplied with the GDP/GVA ratio, as to prevent overinflating the FGDP measure and to remain consistent with the two other

approaches used for the calculation of the Gross Domestic Product.

The data for the disaggregated GDP(O) measure is retrieved from the OECD database. This data is in current prices Pound Sterling, and seasonally adjusted. The reason to choose nominal values, is that while theoretically, credit creation for productive purposes could be deflated using for instance CPI, there does not exist such a deflator for non-GDP

(33)

33

order to calculate FIRE output, the quarterly value added of the FIRE sector is multiplied with the interpolated output-to-value added ratio. A schematic derivation of the FGDP measure can be found in figure 1.

Figure 1: The derivation of the FGDP measure

Stationarity and summary statistics

In order to test for stationarity, the Philips-Perron (PP) test and DF-GLS test are utilized. The Phillips-Perron test, as proposed by Phillips & Perron (1988), is used, as it allows for

heteroscedasticity and autocorrelation, by using a non-parametric distribution. In order to ascertain the validity of these results, a Dickey Fuller–General Least Squares (DF-GLS) test as proposed by Elliott, Rothenberg and Stock (1996) is executed. As the latter has higher power, in case of conflicting outcomes, the DF-GLS result is to be preferred.

As credit is a nominal value, and thus loses purchasing power over time, an upward trend is to be expected and an intercept is expected. An upward trend is also expected for the FGDP per capita measure, as over time economies should become more efficient and an intercept is expected. Hence, the PP and DF-GLS tests are estimated using a trend and intercept. The spectral estimation method used in the PP test is the Barlett kernel, and bandwidth is

(34)

34

𝐼(2) series at the 10 % significance level by the PP test, and that all the variables are 𝐼(1) at the 10% significance level according to the DF-GLS test. As the latter has higher power, all the series are regarded stationary at the 10% level. This is unsurprising, as most economic time series follow a linear trend. The summary statistics are shown in table 3. Furthermore, the data is shown in a graphical form in graph 1 and 2.

LNNFGDP LNPROD LNUNPROD LNHP LNBUS

Mean 12.15467 13.14737 13.92186 11.93432 12.81434 Median 12.20583 13.24051 13.94848 12.14081 12.83294 Maximum 12.50143 13.62154 14.58015 12.53444 13.31091 Minimum 11.72748 12.31549 13.04290 11.04018 12.07387 Std. Dev. 0.215112 0.428237 0.530102 0.487407 0.410182 Skewness -0.352407 -0.475459 -0.166574 -0.465041 -0.290954 Kurtosis 2.107912 1.750289 1.524880 1.662153 1.569909 N 80 80 80 80 80

(35)

35 Graph 1 & 2 : The development of the time series described in table 3 over time.

Cointegration

Two, or more, 𝐼(1) variables are said to be integrated if there exists a linear combination of these variables that is 𝐼(0). In order to proceed to testing hypotheses (1), (2) and (4) that have been stated in chapter 3, and to prevent the results from being spurious, it is necessary that there exists one, or more cointegration vectors between the nominal FGDP measure, outstanding credit for unproductive purposes and credit for productive purposes. This relation will be hereafter referred to as the FUP relation.

Similarly, in order to test hypothesis (3) it is necessary that there exist one or more

cointegration vectors between the nominal FGDP measure, the outstanding credit for non-financial corporations and the nominal average housing prices. This relation will hereafter be referred to as the FPH relation.

(36)

36

Lag length

In order to test these hypotheses, the test developed by Johansen (1991) is employed. In order to select the lag length, a vector autoregressive (VAR) model is estimated and the lag length is based on the AIC. Using the Akaike information criterion, it is determined that the for the FUP relation the Johansen test should be executed with two lags. That is, if the errors for the estimated VAR model are not serially correlated. Using a LM multiplier test, shown in table A2, it is shown that there is no evidence of autocorrelation in the VAR model with two lags. Hence, the Johansen cointegration test is estimated for the FUP relation using two lags For the FPH relation, the VAR estimation lag length selection yields, following the AIC, that the Johansen test should be estimated with 6 lags. As shown in table A4, there is no evidence of autocorrelation at the 5 % level for this relation.

The Johansen (1991) test

The Johansen test can be estimated for different forms of cointegration25. The five assumptions that are considered by Johansen are the following:

1) The data do not follow a trend, and the cointegration equations do not have intercepts

2) The data do not follow a trend, and the cointegration equations have intercepts 3) The data follow a linear trend, and the cointegration equations have intercepts 4) Both the data and the cointegration equations have linear trends

5) The data have quadratic trends, and the cointegration equations have linear trends

For both relations, although imperfect due to the financial crisis of 2007-2008, the trend in the data does appear to follow an upward sloping linear trend. Hence, for both relations, assumption 3 are chosen.

Johansen Test statistics

The Johansen test, when used with three variables, tests whether there exist no, one or two cointegration equations. The test statistics are calculated using the trace statistic and the maximum eigenvalue-statistic. For the FUP relation, I find that the null hypothesis of no

(37)

37

cointegration can be rejected at the 1% level. The null hypothesis of at most 1 cointegration cannot be rejected, at the 5% level. The test results are displayed in the appendix in table A3. This implies that there exists one linear combination of the I(1) variables that is I(0). For the FPH relation, the null hypothesis of no cointegration cannot be rejected at the 5% level. This implies that there does not exist a cointegration vector for the FPH relation, and that hypothesis (3), should be rejected. This test is shown in table A5.

Vector Error Correction Model

(38)

38

Chapter five: Empirical Results

In this chapter, I will discuss the results of estimating respective VEC models for the FUP relations, and relate this to hypothesis (1) and (2). Furthermore, as a VEC Model does not directly address the causality relations, I will discuss the causality relations using a Granger causality test.

The FUP relation

For the FUP relation, that is used to test hypothesis (1) and (2), the VEC model is estimated using the assumption of a linear trend in the data and the existence of an intercept in the cointegration equation, with two lags. The cointegration vector [1, −𝛽1, −𝛽2 , −𝛽3 ]′ and the error correction terms are given in table A6.

The cointegration vector can be interpreted in the following way, and depicts the long run equilibrium. The t-values are given in parenthesis and ** denotes significance at the 1% level.

𝑙𝑛(𝑁𝐹𝐺𝐷𝑃) = 4.817388 + 2.206002 ln(𝑝𝑟𝑜𝑑) – 1.556591 ln(𝑢𝑛𝑝𝑟𝑜𝑑) (1) (−6.42977)∗∗ (5.07016)∗∗

The coefficients of equation (1) can be interpreted in the usual manner for log-log models, that is, the change of one unit (which is in percentages) results in a percentage change in another unit. From table A6 it can be noted that only the adjustment coefficient for the first difference of unproductive credit is significant at the 5% level, however as all values are between 0 and -1, the system is stable. Testing for autocorrelation in the residuals yields that there is no evidence of autocorrelation at the 5% level. Testing for heteroscedasticity yields that there is evidence of heteroscedasticity at the 5% significance level, which could result in biased standard errors. These test results are shown in the appendix in tables A7 and A8. Normality, appears to be a problem for the model, as there is evidence at the 5% of jointly non-normal residuals, as shown in table A9. This indicates, that as the estimators are derived using maximum likelihood estimation the robustness of the estimators cannot be guaranteed. However, Cheung and Lai (1993), argue that the bias imposed by the non-normality of the residuals has relatively minor effects on the Johansen test.

(39)

39

(2) should be accepted. however, a cointegration vector should not be interpreted as a direct causality relation. For causal inferences to be made, it is important to perform predictive causality tests. Hence, I will test the causal relations using a using a Granger causality/block exogeneity test. In order to see whether there exists causality from 𝑋𝑡→ 𝑌𝑡, the Granger causality test tests whether past values of 𝑋𝑡 can increase the predictive power in explaining 𝑌𝑡 compared to using just past values of 𝑌𝑡. The results of the Granger causality tests for predictive causality are shown in table A10. As can be noted from the Granger causality test, there appears to be evidence at the 5% significance level that there is a causal relation from credit for unproductive purposes to nominal FGDP, yet there is no evidence for credit creation for productive purposes. There is also no significant evidence of a reverse causality. Therefore, I advise against interpreting the coefficient of credit creation for productive purposes as empirical evidence for hypothesis (1), yet also do not have enough evidence to reject this hypothesis. Furthermore, there also appears at the 5% significance level, that there is no causal negative channel from credit creation for unproductive

purposes and productive purposes. There does, however, appear to be reverse causality at the 5% significance level. Hence, hypothesis (4) should be rejected.

(40)

40

Chapter six: Discussion

Analysis of the findings

As discussed in the previous section, the hypothesis that credit created for productive purposes is the driver of economic activity in the United Kingdom, cannot statistically be causally inferred from the VEC model. Simultaneously, there also appears not to be evidence of reverse causality, from NFGDP to credit created for productive purposes. A possible factor that could explain these findings, is the aftermath of the credit crunch that took place during the financial crisis in 2007-2008. According to Shin (2009), the beginning of the financial crisis (August 2007) resulted in a withdrawal of credit that hit the entire market. Eventually, this resulted in the bankruptcy of Northern Rock. One could argue that such major shocks to the system, for a profound period of time, alters the behaviour of firms and banks.

Bernanke and Gertler (1995) show that if the credit supply is constrained, then firms could find them shut off by their credit supplier and have to incur search and agency costs to find new credit providers. While Bernanke and Gertler discuss this bank lending channel in the context of changes in monetary policy, I think that their analysis could also be applied to the liquidity provision situation during and after the financial crisis. While credit supply was not constrained by the short term interest rate, as this was decreasing from July 2007 onwards, the uncertainty regarding the future could have impacted forecasts of firm performance. This could result in a credit crunch for these non-financial firms, which could have been deemed more risky due to uncertainty in the market. In the aftermath of such a credit crunch, in order to prevent incurring search costs and the possibility of becoming illiquid if a credit crunch were to present itself again in the near future, non-financial firms might have altered their capital structures towards more leverage.

(41)

41

for pre-crisis measurements the use of credit to non-financial firms is an adequate proxy for credit creation for productive purposes, but that during and after the crisis, the proxy that is used for credit created for productive purposes includes an amount of credit for

unproductive purposes. The empirical result is that the VEC estimates could be biased downwards, the Granger causality tests could be negatively influenced, and possibly explain the non-normality that is present in the VEC Model.

While the normality and heteroscedasticity issues also pose a problem for the acceptance of hypothesis (2), a possibly increased leverage ratio for pure liquidity purposes, resulting from the financial crash, for financial firms does not bias these coefficients. This is because this credit is even under this situation created for non-productive purposes. Therefore, it can indeed be inferred that an increase in credit for unproductive purposes negatively affects real economic performance. This is in line with the theoretical framework provided by Bezemer and Hudson (2016), who argue that extensive credit extended for mortgage purposes has a minor positive effect on GDP growth, and results in financial fragility. It is unsurprising that the coefficient for unproductive credit is considerably larger than in other studies, such as in Angeles (2015) and Sassi & Gasmi (2014). This could be related to the use of the FGDP measure, rather than GDP, as measure of economic performance. Remember that in the FGDP measure, the FIRE sector is regarded as a cost, rather than a value adding activity. This implies that the value added to the real estate sector is negative to this measure. Similarly, as the output of financial firms is regarded as a cost to real economic output, it is of no surprise that as credit extended for unproductive purposes increases, this negatively affects nominal FGDP. It can reasonably be expected that the FIRE output and value added increases as the credit flows to the sector increase. As these are subtracted from FGDP, but are regarded as value adding in the GDP measure, it should be of no surprise that using the FGDP measure can result in considerably higher coefficients relative to the GDP measure.

(42)

42

transform land destined for productive (income generating) purposes into residential land might not best be approximated by a linear relation, as is done using a Johansen

cointegration relation. Instead, it is not unreasonable to assume that this process could best be depicted as a high order relation, such as depicted in graph 3. The reason why such a process might be more appropriate, is that the valuation of the public, which after all

decides via their representatives on the allocation of the urban land, might not coincide with the market valuation. This could be, because (local) businesses can perform more than just economic functions, for instance local businesses can perform social functions. Hence, the constituents and other stakeholders might accept that there is a disparity between the price of residential land and the commercial land price, and thus not demand the revision of the zoning laws, until a certain ratio is achieved. As the valuation of these social functions differs per person and community, it could be that pressure on the local board increases relatively slowly at small to moderate differences between prices for residential and commercial land, but that at a certain threshold ratio, momentum starts to develop for a zoning law revision. Hence, a situation such as depicted in graph 3 could arise. In such a case, using a Johansen test would be inappropriate, as the Johansen tests describes a linear cointegration relation between variables.

Referenties

GERELATEERDE DOCUMENTEN

4–6 While the cyclohexane triamide core (CH, Figure 1), used thus far in this thesis, can form gels with simple amino acid side groups.. 7–9 However, the cyclohexane core is

Verder is het feit dat de uitoefening van het stakingsrecht enkel in de Europese context wordt beperkt door verkeersvrijheden problematisch voor werknemers die voor werkgevers werken

Hoewel Heuth 37 argumenteer dat die eerdtydse status quo van hoërfunksietale (Afrikaans en Engels) teenoor laerfunksietale (die nege inheemse ampstale) sedert

In order to examine dynamic relationships between property prices and bank loans to listed real estate development companies, I use the following variables:

Credit rating announcements published before the financial crisis show large negative cumulative average abnormal returns for downgrades in all event windows except for

The results shown display the tested relationship between corporate social responsibility, firm characteristics and investor ‘value’ multiples – and include period fixed effects

The results in this model indicate that an appreciation of the local currency against the US Dollar has a more significant effect on domestic credit growth

Unfortunately,  these  results  are  not  new:  limited  use  is  a  common  problem  in  PHR  evaluations  [27].  Several  recent  systematic  reviews  focusing