1
University of Groningen
Faculty of Economics and Business
Investigating the link between financialisation and debt-led
consumption booms: A case study of EU member states
M.Sc. International Economics and Business Master Thesis
June 2017
2 The research work disclosed in this publication is partially funded by the
Endeavour Scholarship Scheme (Malta). Scholarships are part-financed by the European Union - European Social Fund (ESF) - Operational Programme II – Cohesion Policy 2014-2020
“Investing in human capital to create more opportunities and promote the well-being of society”.
European Union – European Structural and Investment Funds Operational Programme II – Cohesion Policy
2014-2020
“Investing in human capital to create more opportunities and promote the well-being of society”
Scholarships are part-financed by the European Union - European Social Funds (ESF)
3
Abstract
4
Table of Contents
1. Introduction ... 5
2. Literature Review... 7
2.1 Defining Financialisation ... 7
2.2 Financialisation and its impacts ... 8
3. Data and Methodology ... 11
3.1 Data ... 11
3.2 Hypotheses ... 14
3.3 Measures ... 16
3.4 Methodology ... 19
4. Results and analysis ... 23
4.1 Robustness Checks ... 29
4.2 Hypotheses Evaluation ... 35
5. Conclusion ... 36
5.1 Limitations of the paper ... 37
6. References ... 38
5
1. Introduction
The end of the 1970s and beginning of the 1980s saw the ushering in of a new era for developed nations, many economists believe – the era of financialisation. This process has been defined by various economists in different ways (discussed further in the literate review), however in general it refers to the rapid proliferation of financial services and instruments, and the growing reliance of a given nation on this sector in order to generate economic activity and value added – while that of the real sector of the economy (industry, manufacturing, agriculture, etc.) steadily declines (Krippner, 2005).
T Palley (2013) believes that this financial neoliberalism was driven by the economic conditions of the 1970s, and it resulted in over thirty years of credit expansion and uncontrolled asset price inflation – eventually culminating itself into the recent financial crisis of 2008. This process of financialisation is a fairly new phenomena, and there is a divide in the current literature, with some economists warning that this is a dangerous process which, if not managed properly, can cause many ails to future economic development, (Aalbers 2008, Arcand 2012) while others argue that it is the next logical step that ought to be taken in rich nations that have successfully undergone industrialisation.
This thesis aims to investigate whether financialisation of a given economy does in fact lead to a debt-led consumption boom as explained by authors Hein and Dodig (2014) in a recently published research paper. The two authors outlined this channel as one of the four routes by which financialisation of the economy impacts the macroeconomy. The intuition behind this relationship relates to the increasing availability and affordability of financial instruments to households, changing financial norms and deteriorating creditworthiness standards (Hein & Dodig, 2014).
6 Since the specific topic of financialisation is relatively in its infancy, there is a lack of econometric models and attempts to measure and quantify the degree of this process and its subsequent effects. This thesis therefore adds to the literature as it serves as a simplistic attempt to make use of a measure of financialisation and apply it to an econometric model and identify and analyse its correlation with certain macroeconomic variables.
7
2. Literature Review
One of the most fundamental developments that occurred over the last three to four decades is a process that many researchers and economists have begun to refer to as ‘financialisation’. The precise definition of the term is still debated to some extent, but in general it refers to the surge in size of the financial sector of a certain economy, as well as an increase in the reliance of that economy on the financial sector in order to generate and maintain profits, employment and economic activity. The following subsection discusses the variations found in related literature with regard to the definition of the financialisation process. This is followed by a section focusing on the empirical evidence on whether this process is occurring, and if so, what the economic implications are.
2.1 Defining Financialisation
Despite the different means chosen to describe financialisation, the fundamental underlying process that the literature is attempting to analyse is the same. Thomas L. Palley (2013) describes financialisation as a process that radically changes the way an economy functions at the macro as well as the micro level. The author defines the process as the three following main effects; 1) the increased importance of the financial sector in comparison to real sector; 2) a redistribution of income from the real sector of the economy to the financial sector; and 3) higher income inequality and wage stagnation. The latter income effect stems from an increase in capital’s share at the expense of that of labour.
8 There are authors however that have a different view regarding what should be focused on to best encapsulate financialisation. Malcolm Sawyer (2015) for example notes that certain authors have attempted to describe this process in terms of shareholder value, practices and power, attributing‘the financialisation of the corporation to the emergence of shareholder value as the main guiding principle of corporate behaviour (cf. Rappaport 1986).’ Aalbers (2008) disagrees with this train of thought, and by pooling from numerous empirical sources, he explains that financialisation is not simply encompassed by increasing shareholder value in numerous sectors of the economy, but rather it is also the process by which the financial industry has transformed itself from an industry that facilitates growth for the economy and other firms, into a growth industry in and of itself. This is therefore more consistent with the views held by Krippner regarding this topic.
Aalbers also emphasises the point that financialisation is not restricted solely to financial markets, but also involves the real sector of the economy, such that non-financial firms become increasingly participative in capital and money markets. He also stresses the ‘inverted relation between the financial and the real’, drawing on research by Sweezy (1995). This inverse relationship is also supported by authors such as Stockhammer (2004). Another point of concern for Aalbers is the increasing dependence that capitalism has on finance growth in order to raise capital, as a result of the dwindling overall growth rate and stagnating real economy (drawing from his analyses of the USA).
2.2 Financialisation and its impacts
Since financialisation is a phenomena that has been brought to attention relatively recently, this research field is still in its infancy. Despite this, there is sufficient literature and empirical research in existence to prove that financialisation is indeed occurring, and to draw upon it when making arguments that provide insight into its effects.
9 These four channels are all important to observe in order to obtain a full picture of the changes brought about by such a process, however one channel – that of household debt and consumption - could arguably be the most central factor to the recent financial crisis. Apart from the clear explosion in national debt for numerous EU countries during and leading up to the crisis, household debt specifically played a role in the upheaval. According to A Chmelar (2013), household debt in Europe was a vital driver of economic growth until the mid-2000s, yet the debt overhang that prevailed, coupled with the contraction of demand, were the main obstacles to a return of growth. Both lenders and borrowers had not adequately accounted for the likelihood of a downward turn in the business cycle, which ultimately led to borrowing and lending based on ill-informed or irrational income expectations, and hence to over-indebted households and a highly leveraged population. It is for this reason that the household debt and consumption channel were chosen for analysis in this paper.
The empirical evidence so far has not been entirely clear-cut, with some studies concluding that financialisation and financial development exert a positive effect on economic growth, while other research has found that when the financial industry reaches a certain size, it reduces growth and can even be a drain on growth, causing the economy to contract. With regard to the former conclusion, Arestis, Chortareas and Magkonis (2015) conduct a meta-analysis of the existing empirical data on the effects of financial development on growth. They argue that there is no strong evidence of causality between financial development and output in the short run. In the long run, however, financial development and output exhibit a clear positive relationship, provided that the financial deepening is supplemented with trade openness.
As for evidence for the latter relationship (i.e. a large financial industry being counterproductive to growth), a 2012 IMF working paper (Arcand et al, 2012) uses ‘different datasets and empirical approaches to show that there can indeed be “too much” finance.’ The main findings of the paper are that there is indeed a positive relationship between a growing financial industry and economic growth, however, when credit to the private sector stands between 80% and 100% of GDP, then the incremental effect of financial deepening reduces growth.
The study also rules out the chances that this effect is being driven by output volatility, banking crises, low institutional quality or differences in bank regulation. The author posits that the most plausible cause of negative returns is that rapid expansion of credit to a large part of the economy can induce more macroeconomic volatility or cause a banking crisis.
10 More specific to the EU and the Eurozone, there is a recent body of research that has found that those countries that were part of the common currency area suffered more during the 2008 financial crisis than their non-eurozone counterparts. Carrasco and Ferreiro (2016) show precisely this phenomena in a FESSUD working paper, stating that ‘membership to the European Monetary Union has implied a significant differential effect, and, thus, the impact of the crisis has been larger inside the Eurozone than in the non-euro EU countries or in the other developed economies.’
When measuring the impact of the GIIPS countries however, they find that the impact of the crisis had been more significant in the latter group. Thus when comparing the core Economic and Monetary Union (EMU) countries to the non EMU countries, the effects of the recession were found to be similar. However, a separate FESSUD paper by Ferreiro et al. (2016) supports the hypothesis that countries within the common currency area – in particular those who joined after 1999 - performed worse during the economic crisis than those EU members that did not adopt the euro. The authors show a growing divergence in the macroeconomic performance of euro-countries.
There is also empirical research that deals with means of preventing the negative effects of financialisation from materialising and devastating the real sector of an economy. A robust regulatory environment has been found to be a crucial part of countering the undesired influence of financialisation. Michal Jurek (2014) is one author who has argued the latter point, stating that public authorities ought to be more engaged in reconciling the interests of the financial institutions, those of the real sector and general public interests. He further suggests the regulation and supervision of all financial institutions while simultaneously developing favourable conditions for the growth of financial institutions that are not private-owned or profit-oriented.
Further supporting the claims made by Jurek regarding regulation, a research paper by Özgür Orhangazi (2014) concludes that financial deregulation played a crucial role in creating the conditions in which the 2007-08 financial crisis was able to take hold and spread. ‘In the run up to the crisis, deregulation created an environment in which mortgage lending expanded and speculation in other financial markets were heightened, even though riskiness was steadily increasing’ (Orhangazi, 2014).
11
3. Data and Methodology
3.1 Data
The regression analysis undertaken below makes use of data on 22 different countries; 21 of which are member states of the European Union and the other being Norway. Although the UK is set to leave the EU, by all practical measures it is still a member of the union; and was a member for the entirety of the period under observation in this paper, and was thus included. Norway has been included because despite not being a member of the EU, it is still part of the European Economic Area (EEA), in the context of being a European Free Trade Association (EFTA) member. The country is also a regional neighbour and one of the most important trade partners of the EU, with the European Commission reporting that 75% of Norway’s import and export market is EU oriented.
The data utilised is secondary in nature, as it was taken primarily from the publicly available Organisation for Economic Co-operation and Development (OECD) online database, and the period under observation ranges from 1995 to 2016. Figure 1 below illustrates both the household debt rate as a percentage of GDP (blue) and household consumption to GDP (red) for all the countries used to construct the panel dataset for years 1995 to 2016. The vertical axis is not identical for each diagram, therefore it only illustrates variation between zero and the relevant maximum for each country. It is clear that for the majority of these countries, household debt proportional to GDP has followed a significant upward trend during the period under observation.
It is also clear that with the exception of seven countries (Czech Republic, Estonia, Germany, Hungary, Ireland, Spain and the UK), household debt continued to rise or remained constant despite the financial crisis of 2008. This is especially concerning, as the high level of indebtedness was one of the key contributors to the crisis.
12 Figure 1: Household Debt (blue) and Consumption (red) (% of GDP) for 21 EU countries and Norway; 1995-2016
13 By generating a rudimentary scatter plot of FIRE sector contribution to GDP, household debt and household consumption (pooling from all countries included in the dataset) one is able to observe some basic trends. The vertical axis runs from 0 to 1.5, which represent a range of percentage of GDP for both debt and consumption of households, running from 0% to 150%. The horizontal axis measures the percentage contribution to GDP of the FIRE sector, spanning from 0% to 35%.
It can be noted that for the majority of countries, the FIRE sector contribution clusters between 10% and 15% of total Gross Domestic Product. Also, taking the data as a whole, there is a clear upward trend of household debt that that coincides with a growing FIRE sector size. As for consumption, the upward trend is much less pronounced and it is dubious whether there is indeed and upward slope in the data. There is also a clear group of outliers at the far end of the horizontal axis, in which despite the very high levels of FIRE sector contribution to GDP, household debt and consumption (to GDP) values are no higher than the average of the clustered data points – around 50% of GDP.
By analysing a time series generated using the same panel data (figure 2, right panel), which takes the mean value of all countries at each given year for both debt (mean) and consumption (mean2) and plots it against time, the above analysis is further supported. Household debt on average has experience a significant upward trend in the 22 countries under observation while consumption remained rather constant.
Figure 2: Scatter plot against FIRE (left) and time series plot (right) for Household Debt and Consumption
14 3.2 Hypotheses
As explained before, the primary purpose of the paper is to determine whether financialisation leads to a debt-led consumption boom. Given the intuition and economic reasoning laid out in the Hein (2012) and Hein and Dodig (2014) papers, taken together with the Krippner-inspired (2005) definition of financialisation (discussed further in section 3.3), one is able to formulate a number of hypotheses.
As the FIRE sector grows in terms of importance to GDP contribution proportionate to the total, new financial instruments are created, such as credit card debt and collateralised debt obligations, which facilitate the undertaking of debt for households. Financial norms within the economy also change, as credit is demanded in order to maintain a consumption level in line with relative preferences (‘keeping up with the Joneses’). A further change is that credit worthiness standards fall, stemming from debt securitisation and originate and distribute strategies undertaken by the banks. This rise in debt as a result of the aforementioned factors allows consumption to rise faster than median income. These arguments are supported by Hein and Dodig (2014). Thus, we hypothesise that an increase in financialisation ought to cause a rise in debt as well as consumption:
H1: FIRE sector size is positively correlated with household debt to GDP
H2: FIRE sector size is positively correlated with household consumption to GDP
If the above two hypotheses are accepted, and FIRE sector size is also correlated with the growth rate of household debt and consumption to GDP, then further tests can be conducted to test whether a third hypothesis holds. That hypothesis is whether or not FIRE sector size correlates positively with a debt-led consumption boom. If however, there is no significant relationship between this sector size and the growth rates of debt and consumption, then hypothesis 3 must be automatically rejected, as there will be no grounds for evidence of an acceleration in the growth rates (boom) for either debt or consumption:
H3: FIRE sector size is positively correlated with a debt-led consumption boom
Another set of hypotheses can be formulated that relates to the Eurozone dummy variable. Membership in the EMU brought with it a common currency for participating members as well as coordinated economic policies. The goal was to create a single harmonised market which made use of a single currency with common regulations throughout, in order to further facilitate cross-border trade, entrepreneurship and competition.
15 Countries with underdeveloped or relatively less developed credit markets are expected to attract firms from more financially advanced nations that are able to provide the necessary service and turn a profit. According to A Chmelar (2013) ‘Economic growth, the establishment of the single market in financial services, the expansion of credit products and falling interest rates have facilitated access to all types of credit across Europe’. This author also reiterates how the single market facilitated the recognition of banking firms across the union, and thus significantly increased the degree of competition. The aforementioned changes occurred in a time of global financial deregulation and thus had an impact on European legislation, in particular with reference to the European mortgage markets and their subsequent liberalisation. Unless the large, established banks and credit providers from western Europe act in an anti-competitive manner and divide the market or collude with one another, avoiding competing with one another (which is strongly monitored for and prohibited in the EU due to their strict competition policy (De Haan, 2012) and therefore unlikely), then it is expected that the cost of credit in the more financially developed parts of Europe will also decline, as competition intensifies and the larger firms of a newly integrated market compete for market share and dominance. Households within these countries would therefore have more opportunities to finance their consumption through borrowing as it becomes more available and more affordable. These changes culminated in an altered trend for EU households, in which their credit behaviour began to closely resemble that observed in the US. The following hypotheses are thus formulated:
H4: Eurozone membership is positively correlated with household debt to GDP
16 3.3 Measures
The objective of this paper is to test whether the process of financialisation leads to a debt-led consumption boom. If it is found that there is indeed a significant correlation, this would be consistent with the assessments of Hein (2012) and Hein and Dodig (2014), i.e. that a debt-led consumption boom is one of the four channels through which financialisation affects the macro economy at large.
Therefore, the two dependent variables to be used in separate models are household debt and household consumption. These two variables are measured in terms of a ratio of a given country’s respective household debt and household consumption in terms of that country’s GDP (%), in order to facilitate meaningful comparison among the nations in the dataset. Household debt refers to the amount of outstanding consumer credit and mortgages owed by a household to a financial institution. Household consumption is defined as that portion of income that is not retained in savings, and also includes consumption expenditure financed through borrowing.
As for the main independent variable of this paper, a proxy for financialisation was constructed. Economists and researchers have defined financialisation in a multitude of different ways, as has been already discussed in the literature review section. One of the most pragmatic definitions was given by Greta R Krippner (2005), defining it as a ‘pattern of accumulation in which profit-making occurs increasingly through financial channels rather than through trade and commodity production’. This refers not only to an overall decline in the importance of non-financial firms and an increase in that of non-financial firms with regard to profit generation and economic activity, but also refers to non-financial firms’ increasing reliance on portfolio investment and financial instruments in order to turn a profit.
17 Financialisation defined in this way must also take into consideration the declining relative contribution of the Real sector of the economy to total GDP, while the FIRE sector simultaneously becomes more important. For this reason, a second variable was considered that would gauge the size of the Real sector in a similar way. This variable was to include the agriculture, forestry, fishing, industry (including energy), industry (including manufacturing), construction, distributive trade, transport, repairs, information and communication, accommodation and food services – as given by the OECD database.
However, this variable would cause problems in the regression analysis due to its identity relation with the FIRE sector GDP contribution, i.e. as the FIRE sector output grows in terms of percent of total GDP, it automatically implies a fall in the contribution of the aforementioned sector that make up the Real sector. Therefore, it was decided that the FIRE variable (Ft) would
be sufficient to gauge the level of financialisation within the economy. This variable will be used as the main independent variable throughout the analysis, as is explained in the next section.
The remaining explanatory variables have been included in the regression because according to the literature and existing empirical research, they are known to have an observable and significant effect on both consumption and household debt. Variable ‘it’ is the short term
interest rate, defined by the OECD as the rates at which short-term borrowings are effected between financial institutions or the rate at which short-term government paper is issued or traded in the market. The intuition of the interest rate’s bearing on household debt is addressed by Guy Debelle (2004). A fall in the short term interest rate lowers the level of borrowing costs incurred by households, thus increasing the maximum amount of debt that a household is willing to undertake, as well as raising the amount a financial institution would lend to a given household.
The second control variable, ‘pt’ represents the inflation rate within and economy, specifically
the total growth rate (%) of the Consumer Price Index (CPI). Guy Debelle (2004) also explains how a significant part of the spike in household borrowing during the 1980s and 1990s in developed nations stems from a falling inflation rate that coincided with declining interest rates. A high inflation rate implies a rapid loss of value of debt for the lender, thus lending constraints are tightened and the upfront payments on say, a mortgage, made by the borrower are a higher proportion of income than would be under low inflation – thus reducing the incentive to borrow for consumption.
18 As for savings, there are two opposing views with regard to its impact on consumer credit. A higher propensity to save enables higher growth rates (more income channelled into investment) and thus works counter to the need for consumer credit. The opposing view is that higher savings induce more lax financial constraints, thus fuelling credit growth for consumption.
In place of savings, household wealth (Wt) is used as an explanatory variable as it has been
shown by a number of econometric studies that household wealth is a significant determinant of consumption (Hein, Dodig 2014; Ludvigson, Steindel 1998). These studies have focused particularly on the United States, and the significance and prevalence of wealth-based/financed consumption. The measure of household wealth used in this paper is household financial assets, defined by the OECD as ‘currency and deposits; securities other than shares; loans; shares and other equity; net equity of households in life insurance reserves; net equity of households in pension funds; prepayments of premiums and reserves against outstanding claims; and other accounts receivable’.
19 3.4 Methodology
The aim of this paper is to analyse the relationship between financialisation and a debt-led consumption boom using a panel-data series, therefore two separate regressions are considered and tested; one that relates to consumption and one relating to debt. These two main equations are illustrated below.
Dt = 𝛽0 + 𝛽1Ft-1 + 𝛽2it + 𝛽3it-1 + 𝛽4it-2 + 𝛽5pt + 𝛽6yt + 𝛽7yt-1 + 𝛽8yt-2 + 𝛽9St + 𝛽10Ez + εt (1)
Ct = 𝛽0 + 𝛽1Ft-1 + 𝛽2it + 𝛽3it-1 + 𝛽4it-2 + 𝛽5pt + 𝛽6yt + 𝛽7yt-1 + 𝛽8yt-2 + 𝛽9lnWt + 𝛽10Ez + εt (2)
The dependent variables, ‘Dt’ and ‘Ct’ represent the household debt to GDP ratio and the
household consumption to GDP ratio respectively – at time ‘t’. Data on these two variables was collected from the aforementioned public OECD online databases. They are initially measured at nominal percentage levels, yet are later also measured in the form of growth rates, in order to test for the robustness of the initial results. If it is found that there is a positive and robust relationship between both Dt and Ft-1 as well as Ct and Ft-1, then a further regression
would be run to test for correlation between debt (at time t-1) and consumption (at time t). The first independent variable ‘Ft-1’ refers to a lagged variant of the FIRE variable described in the previous section. This variable is the central focus of this regression analysis, with the resulting coefficients bearing insight on whether or not a relationship can indeed be established between financialisation defined in such a way and a debt-led consumption boom as described by Hein (2012). The lag is included to take into account any delay that is expected to occur between financialisation and the proliferation in debt and consumption.
The interest rate (it) has also been included in the model as a lagged variable; by both one year and two years. The reason for this is due to the inherent delay that occurs between the moment the interest rate is lowered/raised and the subsequent translation to increased/decreased borrowing costs for consumers and households. One source of this lag relates to fixed interest rates for borrowing and saving. Consider a household with a fixed rate mortgage, with fixed interest payments for say 2 years. Until this time has elapsed, homeowners will not have to face a higher interest rate. Another source of delay is that of imperfect knowledge. The average consumer does not regularly check for changes in the short term interest rate, and so it may take time until they become aware of the rate hikes or cuts and adjust their borrowing accordingly.
20 Table 1: Summary Statistics
Variable Obs Mean Std. Dev. Min Max
Years 475 2005.627 6.284239 1995 2016 Country 474 11.44726 6.39258 1 22 Debt 436 0.545045 0.295867 0.042922 1.476941 Consumption 453 0.534762 0.080608 0.328461 0.741446 FIRE 475 0.134759 0.045751 0.080577 0.34509 Inflation 472 0.026163 0.028448 -0.04478 0.283055 Interest 460 0.035429 0.036477 -0.007 0.320417 Income 413 0.01797 0.028637 -0.10561 0.222068 Savings 428 0.049288 0.051392 -0.19301 0.173731 Eurozone 475 0.498947 0.500526 0 1 L1_Interest 437 0.037306 0.036449 -0.0028 0.320417 L1_Income 408 0.018001 0.028686 -0.10561 0.222068 L2_Interest 415 0.039160 0.036457 0.00209 0.320417 L2_Income 387 0.017705 0.029276 -0.10561 0.222068 lnWealth 435 10.62595 0.792829 8.19149 11.98816 Dgrowth 412 0.015941 0.030987 -0.093 0.119868 Cgrowth 431 -0.00263 0.023452 -0.43643 0.062056 L1_FIRE 452 0.134472 0.045605 0.08058 0.345090
Given the numerous lagged variables included in the model, as well as the well documented relationship between interest rates and inflation, it is expected that the variables included in the main regressions suffer from problems of multicollinearity. According to Frederic S Mishkin (1992), ‘Empirical evidence finds no support for a short-run Fisher effect in which a change in expected inflation is associated with a change in interest rates, but supports the existence of a long-run Fisher effect in which inflation and interest rates have a common stochastic trend when they exhibit trends’. It is also expected that an increase in the cost of credit will be reflected in price levels throughout the economy.
21 Mean VIF 1.74 Savings 1.09 0.920351 L1_FIRE 1.33 0.751794 Eurozone 1.42 0.703760 L2_Income 1.49 0.672015 Income 1.54 0.647255 L1_Income 1.90 0.525761 L2_Interest 2.54 0.393422 Interest 2.59 0.386044 Variable VIF 1/VIF
Therefore, the inflation variable was removed from the model and the new regression was executed, followed by another vif test and another variance covariance matrix. The vif test once again revealed that multicollinearity persisted even in the adjusted model, and the matrix indicated that the source of this problem was the interest rate variable lagged by one year – with a high correlation with both other interest rate variables. Thus, the interest rate lagged by a factor of one year (Interest_F1) was removed from the model and the tests were repeated.
Mean VIF 3.27 Savings 1.13 0.887651 L1_FIRE 1.35 0.738851 Eurozone 1.45 0.691522 L2_Income 1.54 0.649466 Income 1.59 0.628761 L1_Income 1.91 0.522683 Inflation 2.03 0.491942 L2_Interest 5.35 0.186793 Interest 6.31 0.158411 L1_Interest 9.99 0.100083 Variable VIF 1/VIF
Table 2: Variance inflation factor (vif) test for equation (1)
Table 3: Variance covariance matrix for independent variables for equation (1)
22 Mean VIF 1.89 Eurozone 1.36 0.737271 L2_Income 1.49 0.671472 L1_FIRE 1.53 0.652749 Income 1.54 0.650753 lnWealth 1.85 0.541022 L1_Income 1.91 0.524321 Interest 2.64 0.378979 L2_Interest 2.80 0.357770 Variable VIF 1/VIF
_cons 0.0833 -0.1335 -0.3494 -0.9346 -0.0898 -0.0340 -0.0037 -0.0398 1.0000 Eurozone -0.1174 0.1188 0.1548 -0.0233 0.1316 0.0191 0.0294 1.0000 L2_Income 0.1452 -0.0516 0.0068 -0.0604 -0.0170 -0.4282 1.0000 L1_Income 0.0876 -0.1298 0.0936 -0.0046 -0.4268 1.0000 Income 0.1218 -0.0027 0.0069 0.0168 1.0000 lnWealth -0.4086 0.1108 0.2954 1.0000 L2_Interest -0.0400 -0.6452 1.0000 Interest -0.0342 1.0000 L1_FIRE 1.0000 e(V) L1_FIRE Interest L2_Int~t lnWealth Income L1_Inc~e L2_Inc~e Eurozone _cons
Mean VIF 1.82 L1_FIRE 1.28 0.780084 Eurozone 1.36 0.735601 L2_Income 1.48 0.674290 Income 1.53 0.654605 L1_Income 1.90 0.526156 L2_Interest 2.55 0.391715 Interest 2.61 0.383302 Variable VIF 1/VIF
The latter vif test (table 4) revealed that the problems of multicollinearity among the explanatory variables had been greatly reduced following the removal of the two aforementioned variables. The same process was repeated for the equation with consumption as the dependent variable (equation 2), and very similar results were obtained. However, after removing the variables causing high multicollinearity problems, there remained an issue with the lnWealth variable, showing a high correlation with the FIRE variable (-0.41). Therefore, this variable was also removed from the model.
Table 5: Variance inflation factor (vif) test for equation (2) excluding Inflation
Table 7: Vif test for equation (2) excluding Inflation, Interest_F1 and lnWealth
23 The two original models shown previously were therefore modified before executing the regressions; shown below.
Dt = 𝛽0 + 𝛽1Ft-1 + 𝛽2it + 𝛽3it-2 + 𝛽4yt + 𝛽5yt-1 + 𝛽6yt-2 + 𝛽7St + 𝛽8Ez + εt (3)
Ct = 𝛽0 + 𝛽1Ft-1 + 𝛽2it + 𝛽3it-2 + 𝛽4yt + 𝛽5yt-1 + 𝛽6yt-2 + 𝛽8Ez + εt (4)
4. Results and analysis
The tables below hold the results of the two regressions run using cross-sectional time series data of the chosen 21 EU Member states and Norway – for years 1995 to 2016. Table 8 is the outcome of regressing Household Debt to GDP (dependent variable) over the main FIRE variable, as well as the control variables and a dummy indicator. Prior to executing the following regressions, a hausman test was conducted in order to determine whether a fixed effects model would be more appropriate than a random effects model for this data, or vice versa. The null hypothesis stating that random effects are suitable for this model was rejected for both equation (3) and (4); i.e. fixed effects ought to be applied to the regression.
However, due to the nature of the model, fixed effects would imply ignoring heterogeneity among countries in the panel dataset. Therefore, random effects were chosen to be applied to both models. This did not significantly alter the results when compared to the fixed effects model for equation (4). The model was also tested for heteroskedasticity, as it was expected that the dataset did in fact contain heteroskedastic errors. This was carried out by conducting a Breusch-Pagan/Cook Weisberg test for heteroskedasticity. The null hypothesis of homoskedasticity was soundly rejected for equation (3), accepting instead the alternate hypothesis which states that the model indeed contains problem of heteroskedasticity. Therefore robust standard errors need to be applied to the model in order to correct for this. The second model however (equation (4)) had no issues of heteroskedasticity, yet robust errors were applied to the model anyway, as is common practice in many econometric studies. All test results can be found in the appendices section.
24 It must be pointed out however, that there are certain issues with implied causality where this measure is concerned. Given that mortgages comprise the bulk of household debt, whenever a mortgage is issued and fees are paid to the financial institutions, both household debt and FIRE sector income/output rise simultaneously. This identity relationship may be a source of upward bias in the regression results, and thus undermines the causal relationship that is expected between the two variables.
Table 8: Household Debt to GDP as dependent variable
(1) (2) (3) (4) (5) (6) (7) (8) L1_FIRE 5.157*** 3.534*** 3.361** 2.804* 2.730* 3.124** 3.787*** 3.856*** (4.43) (3.57) (3.23) (2.44) (2.24) (2.66) (3.4) (3.56) Interest -2.407*** -1.279** -1.057** -1.044** -1.246** -0.999** -0.933* (-5.13) (-2.58) (-2.63) (-2.61) (-3.04) (-2.74) (-2.46) L2_Interest -1.419*** -1.479*** -1.488*** -1.430*** -1.718*** -1.656*** (-5.17) (-5.59) (-6.01) (-5.67) (-5.56) (-4.91) Income -1.179*** -1.154*** -1.052*** -1.076*** -1.036*** (-4.04) (-4.50) (-5.11) (-5.37) (-5.66) L1_Income -0.059 -0.197 -0.204 -0.192 (-0.22) (-1.07) (-1.03) (-1.01) L2_Income 0.428 0.365 0.384 (1.84) (1.49) (1.48) Savings 0.765* 0.738* (2.17) (2.03) Eurozone 0.0281 (0.97) _cons -0.138 0.163 0.212 0.332* 0.342* 0.292 0.18 0.151 (-1.04) (1.22) (1.51) (2.09) (2.03) (1.82) (1.15) (1.01) N 417 414 390 370 369 355 354 354
RE Yes Yes Yes Yes Yes Yes Yes Yes
Within R2 0.2302 0.4223 0.4582 0.4882 0.4845 0.4855 0.5135 0.5171
Between R2 0.0181 0.0505 0.0475 0.2248 0.2278 0.225 0.1823 0.1683
Overall R2 0.0335 0.1110 0.105 0.2504 0.2507 0.2442 0.2179 0.2050
25 The coefficients of the interest rate variable and the interest rate lagged by a factor of two years are both negative and highly significant – maintaining their high significance (mostly at the 1% level) throughout the analysis. The negative correlation between household debt to GDP and the interest rate is in line with expectations based on economic literature. A lower interest rate reduces borrowing costs for households and thus ought to facilitate rising household debt levels (Debelle 2004). Adding the interest variable and its respective lagged values had the strongest negative effect on the FIRE coefficient, causing a reduction from 5.16 to 3.36 – thus suggesting that the interest rate is a more important determinant of household debt rather than size of the FIRE sector.
As expected and discussed in the previous section, the coefficient of the income variable is negative and significant, thus supporting the argument that a reduction in income (or a fall in the growth rate of income) would incentivise households to borrow in order to maintain consumption at the levels prior to the fall in income (Barba & Pivetti, 2008). It was also expected that the lagged values of income growth ought to serve as better predictors of household debt due to the time taken by households to adjust their debt levels accordingly following changes to income. However, from table 8 it is clear that the coefficients are insignificant. Also, since the income coefficient has a value of around -1, and remains so throughout the table, together with the large increase of between and overall R2 when this variable is added, this implies an identity correlation. This arises from the fact that, for a given level of consumption, falling income will translate into rising household debt.
The savings coefficient is positive and significant at the 10% level, thus consistent with the intuition that rising household savings result in looser credit constraints and thus facilitate the undertaking of debt by individuals. The Eurozone coefficient is also consistent with the argument that membership in the EMU should exacerbate the growth of debt due to more closely integrated financial markets of individual member states.
26 Table 9: Household Debt to GDP as dependent variable
(1) (2) (3) (4) (5) (6) L1_FIRE 5.157*** 3.534*** 3.361** 2.804* 3.518** 3.635*** (4.43) (3.57) (3.23) (2.44) (3.23) (3.49) Interest -2.407*** -1.279** -1.057** -0.842* -0.717 (-5.13) (-2.58) (-2.63) (-2.24) (-1.95) L2_Interest -1.419*** -1.479*** -1.773*** -1.650*** (-5.17) (-5.59) (-5.62) (-4.82) Income -1.179*** -1.227*** -1.136*** (-4.04) (-4.08) (-4.34) Savings 0.771 0.764 (1.90) (1.84) Eurozone 0.0535 (1.60) _cons -0.138 0.163 0.212 0.332* 0.213 0.159 (-1.04) (1.22) (1.51) (2.09) (1.38) (1.12) N 417 414 390 370 369 369
RE Yes Yes Yes Yes Yes Yes
Within R2 0.2302 0.4223 0.4582 0.4882 0.5148 0.5265
Between R2 0.0181 0.0505 0.0475 0.2248 0.1786 0.1492
Overall R2 0.0335 0.1110 0.1050 0.2504 0.2225 0.1964
The dependent variable is the household debt as a percentage of total GDP. Definitions and descriptive statistics of all variables are in Table 1 above. T-statistics are in parentheses, *** p<0.01, ** p<0.05, * p<0.1.
The main results of running this more optimised model is that the coefficients of FIRE and the two interest rate variables are slightly smaller, while those of Income and Savings are slightly larger. Also, savings is no longer significant and the non-lagged interest rate variable loses on average a degree of significance than in the previous regression. The FIRE variable however, once more retains significance at the 1% and 5% levels throughout (with the exception of column (4)). Judging from these results, one may suggest that there is indeed a significant correlation between the size of the FIRE sector and the level of household debt in relation to national GDP. A 1% increase in the FIRE sector contribution to total GDP is associated with a 3.63% percent higher household debt to GDP ratio.
27 It is by no means conclusive evidence, particularly due to the identity relationship between mortgages and FIRE sector size as well as that of income and debt as discussed above, however some relationship between the two variables does appear to exist. Despite optimising the model, the Eurozone variable remained insignificant, thus the condition of being a nation within the euro common currency area appears to have had no additional correlation with the growth rate of household debt to GDP as the FIRE sector grows accordingly.
As for the link between the FIRE sector variable and household consumption, the results of executing equation (4) as a regression were tabulated in Table 10 below. Household consumption as a percent of total GDP is the endogenous variable, and was regressed over the FIRE variable and the remaining control variables. From the first row it can be noted that the FIRE sector coefficient is positive but insignificant for columns (1) – (3), and gains significance once the income variables have been added, after which the coefficient rises and maintains its sign and 1% significance for columns (5) to (7). Despite the income growth variable’s lack of significance, it is one of the main determinants of consumption in economic literature, and due to its theoretical importance should not be omitted from the model.
The interest rate variable is significant (with the lagged variant causing the original to lose its significance), however the resulting coefficient is not consistent with conventional economic intuition which states that interest rates and consumption are negatively correlated – because as interest rates rise they will raise the monthly cost of mortgage repayments for many households, thus leaving them with less disposable income for consumer spending. The borrowing costs on newly issued loans will also rise, thus dissuading consumer credit. One possible explanation for the positive coefficient of the interest rate shown in table 10 below is that a rising interest rate is often soon after followed by a corresponding rise in inflation. Such changes can actually induce consumers to increase their consumption, if it is believed that further inflation will follow and their purchasing power will continue to be eroded in the near future.
28 The R2 of this model is slightly higher than the previous model, explaining around 25% of the average within and between variation in household consumption once all control variables have been added. By contrast, the previous two models (table 8 and table 9) scored an R2 of around 0.2 (20%), thus proving to be less accurate.
Despite these issues, the FIRE sector variable was shown to be positively correlated with household consumption, such that a 1% increase in the contribution of the FIRE sector to GDP is associated with a 0.78% increase in household consumption as a portion of GDP. Once again, these results are far from conclusive, yet there does appear to be a significant indication that the two variables are positively correlated with one another.
Table 10: Household Consumption to GDP as dependent variable
(1) (2) (3) (4) (5) (6) (7) L1_FIRE 0.0699 0.235 0.415 0.615** 0.684*** 0.767*** 0.784*** (0.25) (0.84) (1.53) (3.07) (3.44) (4.02) (4.08) Interest 0.217** 0.123 0.165* 0.160* 0.149 0.161* (2.81) (1.34) (2.00) (2.08) (1.75) (2.06) L2_Interest 0.158** 0.136** 0.138** 0.190*** 0.199*** (3.20) (2.80) (3.09) (4.61) (4.60) Income -0.0484 -0.0603 -0.0688 -0.0627 (-0.85) (-1.24) (-1.37) (-1.38) L1_Income 0.0273 -0.00710 -0.00537 (0.69) (-0.21) (-0.16) L2_Income 0.0594 0.0615 (1.20) (1.28) Eurozone 0.00458 (0.74) _cons 0.525*** 0.495*** 0.467*** 0.451*** 0.442*** 0.430*** 0.424*** (16.47) (15.69) (15.35) (15.55) (14.65) (13.76) (13.87) N 431 420 393 375 373 358 358
RE Yes Yes Yes Yes Yes Yes Yes
Within R2 0.0023 0.0822 0.1466 0.2053 0.2231 0.2588 0.2610 Between R2 0.0219 0.0058 0.0110 0.2006 0.1905 0.1923 0.1998 Overall R2 0.0157 0.0003 0.0001 0.2259 0.2286 0.2332 0.2448
29 4.1 Robustness Checks
For each of the models above, a number of robustness checks were conducted in order to strengthen the findings and conclusions that could be drawn. As can be seen in the tables below, four models have been tabulated for each of the dependent variables, i.e. four models for household debt to GDP and four for household consumption. The first model (column 1 in tables 11 and 12) is simply the final column of tables 8 and 10 above, respectively, for the sake of comparison with the new models checking for robustness. The models in column 2 in the following two tables are essentially the same as those in column 1 with the difference that rather than using the standard percentage rate of lagged FIRE sector contribution to GDP, the first difference (D2_FIRE) of this lagged variable is taken so as to correct in part for the implied causality between a rise in household debt to GDP and the value added of the debt providers (FIRE sector).
The models in columns 3 and 4 have incorporated two dummy variables that are important, according to the literature (A Chmelar, 2013), with regard to debt and consumption growth within the EU. These variables are used as interaction terms with the FIRE variable, and they are ‘Precrisis’ (indicating the pre-crisis era i.e. prior to 2008) and ‘Periph’ (which takes the value of 1 if the country is considered to be in the periphery group rather than the core group of EU countries). A Chmelar (2013) reports that the impact of the new accessibility of credit products to households in the past two decades, coinciding with the unprecedented uptick in homeownership in those countries classified as the European periphery (see Appendix 2) has been enormous, and has particularly effected household leverage. Prior to joining the EU, peripheral countries benefitted from positive externalities emanating from European integration, most importantly through the establishment of the single market. The drop in interest rates enjoyed by new entrants into the EU, together with the low household debt level (in relation to income) of peripheral countries provided a ‘scope for catch-up’ and economic conditions conducive to providing it. We thus expect a positive and significant correlation of Periph with household debt.
30 Table 11: Household Debt to GDP as dependent variable
(1) (2) (3) (4) L1_FIRE 3.856*** 2.683* 2.380 (3.56) (2.51) (1.60) Interest -0.933* -1.266** -0.236 -0.949* (-2.46) (-3.09) (-0.94) (-2.55) L2_Interest -1.656*** -1.781*** -1.374*** -1.744*** (-4.91) (-6.51) (-4.83) (-5.69) Income -1.036*** -1.273*** -0.536* -1.107*** (-5.66) (-8.10) (-2.39) (-5.41) L1_Income -0.192 -0.380* -0.131 -0.0841 (-1.01) (-2.02) (-0.89) (-0.47) L2_Income 0.384 -0.0823 0.528* 0.395 (1.48) (-0.27) (2.15) (1.56) Savings 0.738* 0.458 0.343 0.760* (2.03) (1.33) (1.15) (2.01) Eurozone 0.0281 0.0102 (0.97) (0.31) D2_FIRE 3.814*** (3.45) Precrisis -0.128 (-1.36) 1.Precrisi#L1_FIRE 0.152 (0.19) Periphery -0.144 (-0.67) 1.Periph#L1_FIRE 2.166 (1.28) _cons 0.151 0.688*** 0.350* 0.268 (1.01) (10.39) (2.15) (1.16) N 354 354 354 335
RE Yes Yes Yes Yes
Within R2 0.5171 0.4785 0.6022 0.5176
Between R2 0.1683 0.1072 0.209 0.3096
Overall R2 0.205 0.1669 0.2475 0.3267
31 Table 12: Household Consumption to GDP as dependent variable
(1) (2) (3) (4) L1_FIRE 0.784*** 0.827** 0.621* (4.08) (4.17) (2.29) Interest 0.161* 0.111 0.0908 0.131 (2.06) (1.62) (1.23) (1.44) L2_Interest 0.199*** 0.160** 0.171*** 0.175*** (4.60) (2.73) (3.89) (4.21) Income -0.0627 -0.134** -0.0994* -0.0860 (-1.38) (-2.91) (-2.32) (-1.71) L1_Income -0.00537 -0.0672* -0.0108 -0.0000261 (-0.16) (-2.07) (-0.34) (-0.00) L2_Income 0.0615 -0.0273 0.0504 0.0565 (1.28) (-0.38) (1.02) (1.04) Eurozone 0.00458 0.000290 (0.74) (0.04) D2_FIRE 0.115 (0.50) Precrisis 0.0226 (1.05) 1.Precrisi#L1_FIRE -0.125 (-0.82) Periphery 0.00160 (0.03) 1.Periph#L1_FIRE 0.165 (0.51) _cons 0.424*** 0.532*** 0.421*** 0.444*** (13.87) (26.91) (12.71) (10.60) N 358 358 358 339
RE Yes Yes Yes Yes
Within R2 0.2610 0.1436 0.2843 0.2842
Between R2 0.1998 0.1999 0.1992 0.1466
Overall R2 0.2448 0.1028 0.2400 0.1983
32 In table 11 above, it is clear that replacing the lagged FIRE variable with the first difference of that same variable (D2_FIRE) does not significantly change the results, and all coefficient signs remain unaltered. One noticeable change is the decrease in the size of the coefficient of D2_FIRE (compared to L1_FIRE in model 1) and a slight increase in the coefficient size of the remaining significant variables. Also, the Income variable lagged by 1 year becomes significant at the 5% level. In table 12, the same conclusions cannot be drawn by comparing the first two columns. D2_FIRE is no longer significant in contrast to its non-differenced counterpart. The interest rate is also no longer significant, yet income and its 1year lag counterpart become significant at the 5% and 10% level respectively.
In columns 3 and 4 in table 11, despite some gains in significance for some variables and losses in significance for others, as well as slight changes to coefficient size, the precrisis variable and the periphery variable were shown to have no significant correlation with household debt to GDP – and neither did the interaction terms. The same is true for table 12, in which the dependent variable is household consumption rather than household debt. The FIRE variable remained significant for the majority of the above robustness checks. However, the reason for this could be due to an implied correlation of the FIRE variable with household debt and consumption by definition.
33 Table 13: Growth rate of Household Debt to GDP (Dgrowth) as dependent variable
(1) (2) (3) (4) L1_FIRE -0.0757 -0.220 0.213 (-1.32) (-1.63) (1.54) Interest 0.265** 0.269** 0.139 0.234** (2.95) (2.78) (1.77) (2.65) L2_Interest -0.0839 -0.0855 -0.103* -0.0788 (-1.57) (-1.49) (-2.46) (-1.63) Income 0.295* 0.315* 0.192 0.319* (2.15) (2.24) (1.55) (2.33) L1_Income 0.275** 0.293*** 0.244** 0.185*** (3.24) (3.61) (3.02) (3.34) L2_Income 0.194 0.193 0.164 0.289** (1.76) (1.85) (1.48) (2.81) Savings -0.180*** -0.170*** -0.177*** -0.229*** (-3.41) (-3.65) (-3.51) (-3.60) Eurozone 0.0106* 0.00936 (2.18) (1.93) D2_FIRE 0.494 (1.52) Precrisis -0.0298 (-1.20) 1.Precrisi#L1_FIRE 0.337 (1.80) Periphery 0.0494 (1.83) 1.Periph#L1_FIRE -0.438* (-2.13) _cons 0.0103 -0.000195 0.0349* -0.0134 (1.15) (-0.03) (1.98) (-0.78) N 337 337 337 318
RE Yes Yes Yes Yes
Within R2 0.3662 0.3665 0.3860 0.4265
Between R2 0.3141 0.3303 0.2932 0.1861
Overall R2 0.3316 0.3367 0.3540 0.3591
34 Table 14: Growth rate of Household Consumption to GDP (Cgrowth) as dependent variable
(1) (2) (3) (4) L1_FIRE 0.0465 0.0701 0.0784 (1.48) (1.19) (1.95) Interest 0.155 0.154 0.129 0.119 (1.45) (1.44) (1.48) (1.17) L2_Interest -0.0124 -0.0153 -0.0386 -0.0316 (-0.43) (-0.53) (-1.52) (-1.15) Income -0.0227 -0.0326 -0.0424 -0.0382 (-0.73) (-1.04) (-1.02) (-1.10) L1_Income 0.0549 0.0472 0.0524 0.0403 (1.60) (1.21) (1.39) (0.94) L2_Income -0.0363 -0.0386 -0.0401 -0.0182 (-0.95) (-1.01) (-1.08) (-0.47) Eurozone 0.00777 0.00824 (1.73) (1.73) D2_FIRE -0.138 (-0.64) Precrisis 0.000235 (0.02) 1.Precrisi#L1_FIRE -0.00161 (-0.02) Periphery 0.00158 (0.24) 1.Periph#L1_FIRE -0.00660 (-0.13) _cons -0.0174* -0.0112 -0.0138 -0.0155 (-1.96) (-1.63) (-1.25) (-1.86) N 337 337 337 319
RE Yes Yes Yes Yes
Within R2 0.0163 0.0168 0.0113 0.0099
Between R2 0.2769 0.2609 0.0657 0.0672
Overall R2 0.0313 0.0305 0.0147 0.0134
35 4.2 Hypotheses Evaluation
Using the data and results obtained in the previous section, one can now evaluate the hypotheses laid out in section 3.4 on this paper. Based on these tables, hypothesis 1 (H1) is accepted as it is clear in tables 8 and 9 that the FIRE sector variable coefficient is positive and highly significant, thus serving as evidence that the two variables are indeed positively associated with one another. This variable and its differenced counterpart (D1_FIRE) remained significant in the robustness checks in table 11.
As for the second hypothesis (H2), despite the initial lack of significance of the FIRE sector variable, it remained positively correlated with household consumption throughout the model in table 10, as well as the robustness checks of table 12. This hypothesis therefore was also accepted in light of the results.
H1: FIRE sector size is positively correlated with household debt to GDP
H2: FIRE sector size is positively correlated with household consumption to GDP
The final robustness checks for both endogenous variables however (tables 13 and 14), failed to produce significant results, therefore undermining the relationship between household debt growth and FIRE sector size as well as that of household consumption growth and FIRE sector size. Since there was no significant correlation between the two growth rates and the FIRE measure, it is not expected that there would be significant correlation between a term measuring growth acceleration (in order to measure the debt-led consumption boom) either. One must therefore reject the claim of a debt-led consumption boom (H3) as a result of financialisation, according to the data and methods used in this paper.
H3: FIRE sector size is positively correlated with a debt-led consumption boom
The successive hypotheses, (H4 and H5) can be confidently rejected, as it is evident that the Eurozone dummy variable is consistently insignificant in all the presented tables in the previous section. Membership in the Eurozone therefore seems to have made no difference to the levels of household debt and consumption that emerged over the period under observation.
H4: Eurozone membership is positively correlated with household debt to GDP
36
5. Conclusion
The main objective of this thesis was to test for a positive relationship between financialisation of the chosen 21 European Union member states’ (and Norway’s) economies on the one hand, and household debt and consumption on the other – between the years 1995 and 2016. This channel was identified by Hein and Dodig (2014) as one of the four ways through which financialisation of the economy impacted the macroeconomy at large. The size of the contribution of the financial services, insurance and real estate sector to the given nation’s gross domestic product was used as a proxy for financialisation, for lack of a more intricate measure. Two main models were constructed in order to carry out this analysis; one with household debt as a percentage of GDP as the endogenous variable, and the other with consumption as a percentage of GDP as the dependant. The main determinants of household debt and consumption, according to the literate, were added to the respective models as control variables; namely the short term interest rate, income growth and savings as a proportion of GDP.
A dummy variable indicating country membership in the Eurozone was also plugged into the model, in order to determine whether adopting the common euro currency exacerbated the rise of household debt and consumption relative to the other EU nations. The FIRE sector variable serving as a financialisation measure was the explanatory variable this research was most concerned with. The relevant data was collected and a dataset was created, which was used for regression analysis in the form of a panel model. Four hypotheses were subsequently formulated in order to test for the relationships being examined.
A significant and positive relationship was initially found between the FIRE sector size in proportion to GDP, and both household debt as well as household consumption to GDP. However, robustness checks revealed that the correlation between the FIRE measure and growth rates in both household consumption and household debt was not as strong as suspected. Therefore both hypotheses 1 and 2 were accepted, yet hypothesis 3 was rejected as since the growth rates of the two endogenous variables did not positively correlate with the FIRE variable, neither will there be correlation with acceleration in these growth rates and the FIRE measure. Thus, the claim of a debt-led consumption boom correlating with financialisation in these EU countries was rejected. More detailed analyses using different measures of financialisation and different models ought to be used in order to add on to the evidence relating to this particular query, (discussed further in the limitations section below).
37 5.1 Limitations of the paper
This research paper had a number of limitations which, to some degree, diminish the significance of the end result. The main limitation is the use of a very basic measure of the FIRE sector, relying only on its size in proportion to GDP as a proxy for financialisation. This is a rather simplistic approach, which overlooks a number of key elements of financialisation as outlined by G Krippner (2005), such as the nature of the transformation undergone by the private non-financial sector, by which they become more reliant on portfolio income for revenue streams rather than the non-financial aspect of their business. Implicit in this measure is also the relationship between the FIRE sector and debt which exists by definition, as the value added of debt providers increases (FIRE sector), then debt is also expected to increase. A more sophisticated approach ought to be employed in order to bypass this biasing effect on the results. The identity relation between consumption and income is also an issue that is likely causing biased results in this paper, and thus alternative control variables for consumption ought to be explored.
Collecting the necessary data on the largest non-financial firms in the EU and the relevant breakdown of their income sources proved to be beyond the scope of this research paper. It is therefore recommended that further research on the topic is carried out using different models and varied measures of financialisation that more closely align with the methods used in the Krippner paper (2005).
38
6. References
Aalbers, M. B. (2008). The financialization of home and the mortgage market crisis. competition & change, 12(2), 148-166.
Alvaredo, F., Atkinson, A. B., Piketty, T., & Saez, E. (2013). The top 1 percent in international and historical perspective. The Journal of Economic Perspectives, 27(3), 3-20.
Arcand, J. L., Berkes, E., & Panizza, U. (2012). Too much finance?
Assa, J. (2012). Financialization and its consequences: The OECD experience. Finance Research, 1(1), 35-39.
Barba, A., & Pivetti, M. (2009). Rising household debt: Its causes and macroeconomic implications—a long-period analysis. Cambridge Journal of Economics, 33(1), 113-137.
Brown, A., Passarella, M. V., & Spencer, D. (2015). The nature and variegation of financialisation: a cross-country comparison (No. 127). FESSUD Working Paper Series.
Carrasco, C. A., Ferreiro, J., Gálvez, C., Gomez, C., & González, A. (2016). 13. The impact of the financial and economic crises on European Union member states. Financialisation and the Financial and Economic Crises: Country Studies, 299.
Cecchetti, S. G., & Kharroubi, E. (2012). Reassessing the impact of finance on growth.
Chmelar, A. (2013). Household debt and the European crisis. European Credit Research Institute (ECRI) and the Centre for European Policy Studies (CEPS), Brussels.
Chortareas, G., Magkonis, G., Moschos, D., & Panagiotidis, T. (2015). Financial development and economic activity in advanced and developing open economies: Evidence from panel cointegration. Review of Development Economics, 19(1), 163-177.
Cynamon, B. Z., & Fazzari, S. M. (2008). Household debt in the consumer age: source of growth–risk of collapse. Capitalism and Society, 3(2), 1-30.
De Haan, J., Oosterloo, S., & Schoenmaker, D. (2012). Financial markets and institutions: A European perspective. Cambridge University Press.
39 Epstein, G. A. (Ed.). (2005). Financialization and the world economy. Edward Elgar Publishing.
Ferreiro, J., Galvez, C., Gomez, C., & Gonzalez, A. (2016). The impact of the Great Recession on the European Union countries (No. wpaper150).
Fligstein, N., & Calder, R. (2001). Architecture of markets. Emerging Trends in the Social and Behavioral Sciences: An Interdisciplinary, Searchable, and Linkable Resource.
Fontana, G., & Sawyer, M. (2015). The Macroeconomics and Financial System Requirements for a Sustainable Future. In Finance and the Macroeconomics of Environmental Policies (pp. 74-110). Palgrave Macmillan UK.
Guttmann, R., & Plihon, D. (2008). Consumer debt at the centre of finance-led capitalism. Paris: CEPN, Jan.
Hein, E., & Dodig, N. (2014). Financialisation, distribution, growth and crises: Long-run tendencies (No. 35/2014). Working Paper, Institute for International Political Economy Berlin.
Hein, E., & Detzer, D. (2015). Finance-dominated capitalism and income distribution: a Kaleckian perspective on the case of Germany. Italian Economic Journal, 1(2), 171-191.
Hein, E., & van Treeck, T. (2008). 'Financialisation'in Post-Keynesian models of distribution and growth-a systematic review (No. 10/2008). IMK Working Paper.
Jurek, M. (2013). Report on the structure of ownership in the financial sector across the EU (No. wpaper16).
Jurek, M. (2014). Role and impact of different types of financial institutions on economic performance and stability of the real sector in selected EU member states (No. wpaper36).
Krippner, G. R. (2005). The financialization of the American economy. Socio-economic review, 3(2), 173-208.
Lin, K. H., & Tomaskovic-Devey, D. (2013). Financialization and US Income Inequality, 1970–2008 1. American Journal of Sociology, 118(5), 1284-1329.
40 Mishkin, F. S. (1992). Is the Fisher effect for real?: A reexamination of the relationship between inflation and interest rates. Journal of Monetary economics, 30(2), 195-215.
OECD, n.d. National Accounts Statistics Database. Web. Retrieved on 3 Mar. 2017, from: http://stats.oecd.org/Index.aspx?DatasetCode=SNA_TABLE1
OECD, n.d. National Accounts Statistics Database. Web. Retrieved on 3 Mar. 2017, from: https://data.oecd.org/hha/household-net-worth.htm
Orhangazi, Ö. (2015). Financial Deregulation and the 2007–08 US Financial Crisis. The Demise of Finance-dominated Capitalism: Explaining the Financial and Economic Crises, 289.
Palley, T. I. (2013). Financialization: what it is and why it matters. In Financialization (pp. 17-40). Palgrave Macmillan UK.
Passarella, M. V., & Sawyer, M. (2014). Financialisation in the Circuit (No. wpaper18).
Raza, H., Gudmundsson, B., Kinsella, S., & Zoega, G. (2015). Experiencing financialisation in small open economies: An empirical investigation of Ireland and Iceland (No. wpaper84).
Sawyer, M. (2015). Financialisation, financial structures, economic performance and employment (No. wpaper93).
Settlements, Bank for International and Debelle, Guy, Household Debt and the Macroeconomy (March 1, 2004). BIS Quarterly Review, March 2004.
Stockhammer, E. (2004). Financialisation and the slowdown of accumulation. Cambridge Journal of Economics, 28(5), 719-741.
