Faculty of Economics and Business
International Economics and Globalisation
Master thesis
Supervision by Boe Thio
The Relative Income Hypothesis and the Financial
Crisis
Rianne Luijendijk
5965942
9 July 2015
Amsterdam
Statement of Originality
This document is written by Rianne Luijendijk 5965942 who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
I. Abstract
The outbreak of the financial crisis has triggered a renewed interest in the relationship between income inequality and household spending. The inequality narrative, based on contributions of Rajan (2010), Kumhof and Ranciere (2010) and Stockhammer (2012), draws upon the Relative Income hypothesis (RIH) of Duesenberry (1949) and connects the rise in top-income shares to increased household indebtedness. In contrast to the Life Cycle Hypothesis(LCH) and the Permanent Income Hypothesis (PIH), the RIH posits that upward looking comparisons determine households’ optimal consumption. The inequality narrative incorporates this mechanism and hypothesizes that the disappearance of households’ liquidity constraints, in combination with a rise in top-incomes, has led to a spending spree throughout the majority of developed economies. However, the validity of the mechanisms underlying the RIH is contested on a macroeconomic level. This thesis contributes to the ongoing debate by assessing the manifestation of the inequality narrative within Europe. By means of a panel data analysis, covering 20 European countries over the period 1995-2008, the relationship between income inequality and household saving is analysed. The results suggest that income inequality has a positive effect on household saving, until the level of private-credit to GDP exceeds the threshold level of approximately 170%. However, in line with previous research, the findings are not robust to other specifications. In addition, it is suggested that the highest income groups hold the largest stock of debt, whereas the RIH posits that especially the lower/middle income groups accumulate debt to keep up with the rich. Thus, it is concluded that the rise in household spending cannot be explained on the basis of the RIH.
4
II. Table of contents
I. Abstract ... 3
II. Table of contents ... 4
1. Introduction ... 7
2. Consumption theory and the income distribution ... 11
2.1 Consumption theory ... 11
2.1.1 Keynes and the Kuznets Paradox ... 11
2.1.2 The Relative Income Hypothesis ... 12
2.1.3 The Life-Cycle Hypothesis and the Permanent Income Hypothesis ... 13
2.1.4 Optimal consumption and income inequality ... 15
2.2 Household consumption and the financial crisis ... 17
2.2.1 A more efficient financial market ... 17
2.2.2 The inequality narrative ... 18
Changes in the relative income distribution ... 19
Enabling excessive debt ... 20
The political economy motivation ... 20
The macroeconomic motivation ... 21
2.3 The inequality narrative in Europe; a cross-sectional description of the income distribution. ... 23
2.3.1 A deterioration of households’ relative position ... 23
2.3.2 Financing consumption through dissaving ... 26
2.3.3 A household’s relative position and its optimal consumption ... 30
The relative position and household debt ... 30
The relative position and household saving ... 32
3. Empirical evidence ... 36
3.1 Private sector credit and income inequality ... 36
3.2 The household saving rate and income inequality... 38
4. Econometric analysis of household saving in Europe:1995-2008 ... 42
4.1 Methodology and data ... 42
4.2 Basic empirical model ... 43
4.2.1 Interaction with financial deregulation ... 44
5. Findings ... 47
5.1 Statistical description of the variables ... 47
5.2 Outcomes of the basic empirical model ... 50
5
5.4 Interaction between income inequality and financial deregulation ... 55
5.5 Discussion ... 59
5.5.1 Limitations and future research ... 60
6. Conclusion ... 63 7. References ... 65 A. List of Figures... 70 B. List of Tables ... 70 C. List of Abbreviations ... 71 D. List of Symbols... 71
Annex I. Growth in HH debt and GDI ... 74
Annex II. Households holding debt ... 76
Annex III. Data description Cross-sectional analysis of European debt distribution ... 77
Annex IV. Data description variables included in the regression ... 79
7
1. Introduction
The European Credit Research Institute (ECRI) reports that the real amount of consumer credit in Europe expanded by approximately 150% in the period 1995-2008. At the same time, the amount of household debt, made up of mortgage loans and credit used for consumption, increased immensely resulting in debt-income ratios exceeding the 200% for several core European economies such as the Netherlands (Chmelar 2013, p.15) . This development coincided with the deregulation of the financial market; from the 1990s onwards, banks and non-bank intermediaries started to rely on each other, facilitating securitization and off-balance sheet activities to reduce banks’ use of regulatory capital and increase investment opportunities. This increasingly complicated the financial system, making it more difficult to monitor and assess risk (Schularick and Taylor 2012).
In 2011 the European Commission (EC) introduced the Macroeconomic Imbalance Procedure (MIP), which contains safety thresholds of 160% for private debt-GDP ratios, and 15% for private credit-GDP ratios (EC 2012). Figure 1 illustrates that, just before the crisis, several European countries did not comply with the MIP threshold of 160% (depicted by the red line). The fact that private sector indebtedness is considered an important indicator of macroeconomic stability in the post crisis environment raises the question what factors, other than financial deregulation and low interest rates, have influenced the rise in household debt.
A recent theoretical framework, henceforth referred to as the inequality narrative, based on contributions of Rajan (2010), Kumhof and Ranciere (2010), Stockhammer (2012) and others (see Goda (2013), and Michell (2015) for an overview) connects income inequality to the growth in household debt. As established by Piketty and Saez (2013), wages at the lower end of the income distribution stagnated throughout the majority of developed economies, whereas those of the upper decile increased rapidly from the 1970s onwards. Consequently, a gap was created between the upper and lower/middle parts of the income distribution. Throughout this thesis, the “upper part” refers to the top 10% of the income distribution, and the “lower/middle parts” to the remaining 90%. This distinction is motivated by the findings of Piketty and Saez (2013). According to the inequality
narrative, this gap triggered low and middle income households to spend more. Consequently,
households have financed consumption through dissaving, which means that they either saved a smaller share of their income, or increased their debt holdings. Thus, the inequality narrative posits that the rise in income inequality can explain why household debt increased (Van Treeck and Sturn 2013).
The theoretical assumptions underlying the inequality narrative are rooted within Consumption theory and draw upon the Relative income hypothesis (RIH) of Duesenberry (1949). However, empirical studies so far remain divided and generally two other consumption models prevail; the life cycle hypothesis (LCH) and the permanent income hypothesis (PIH) (Van Treeck 2012). Nevertheless evidence for the assumptions underlying the RIH is provided by Christen and
8
Morgan (2005), Bowles and Park (2005), Frank, Levine and Dijk (2010), and Bertrand and Morse (2013).
This thesis contributes to the current debate by exploring the validity of the RIH over the period pre-ceding the financial crisis. If evidence is found for the macroeconomic generality of the RIH, it could provide a motivation for the mitigation of top-incomes in order to sustain macroeconomic stability in the future. The focus is on European countries because empirical accounts already exist that validate the inequality narrative for the US in specific, see for instance Frank et al. (2010) and Bertrand and Morse (2013). Accordingly, 20 European countries, all members of the Organisation for Economic Cooperation and Development (OECD), are analysed. Amongst them are non, traditional and new European Union (EU) members. The chosen timeframe is 1995-2008 because of the increase in financial deregulation. The research question reads:
To what extent can the Relative Income Hypothesis explain the rise in household spending over the period preceding the recent financial crisis?
To formulate an answer to this question, several steps are taken. Primary, it is addressed to what extent the RIH differs from other consumption models, and what role is ascribed to income inequality. Accordingly, the mechanisms underlying the RIH are compared to those of the LCH and the PIH. Subsequently, the inequality narrative is discussed and its similarities to the RIH are assessed. Thereafter, the inequality narrative is projected upon the European environment by providing insight into household spending patterns throughout the income distribution. The purpose of this cross-sectional view of the income distribution is to make the mechanism underlying the RIH more tangible and to provide a basis for the interpretation of the econometric results.
The econometric analysis is founded upon the basic empirical model of Bofinger and Scheuermeyer (2014) and captures the relationship between income inequality and household saving on a macroeconomic scale. If European households structurally took on more debt in the run-up to the crisis, this should be depicted in their saving behaviour. Income inequality is measured by the share of income going to the top1% of the income distribution and the Gini coefficient, as published by the SWIID (Solt 2014). If income inequality triggers low/middle class households to structurally spend beyond their means, this causes a downward pressure on the aggregate savings rate. (Barba and Pivetti (2009); Frank et al. (2010); Hein (2011); van Treeck (2013); Bofinger and Scheuermeyer (2014))
Throughout this thesis, the following structure is pursued. The second chapter discusses the Consumption theory and places the inequality narrative under the RIH. The third chapter assesses the empirical evidence of the inequality narrative. The fourth chapter explains the methodology and presents the basic empirical model. Subsequently, the findings are presented in the fifth chapter. Finally, an answer to the research question is formulated in the conclusion.
9
Source: Author with data OECD (2014)
11
2. Consumption theory and the income distribution
The renewed interest in the relationship between income inequality and household consumption is rooted within Consumption theory. In this chapter, three important consumption models are discussed briefly, namely the RIH, the LCH and the PIH. Thereafter it is explained to what extent they differ in the role they ascribe to income inequality. Subsequently, the inequality narrative is connected to the RIH and opposing explanations are connected to the LCH and the PIH. Finally, several propositions reflecting the inequality narrative are explored on the basis of a cross sectional view of household spending throughout Europe.
2.1 Consumption theory
In order to give an overview of consumption theory, it is necessary to set out the absolute income hypothesis of Keynes (1936) and the Kuznets paradox (1946) first. Thereafter the three consumption models that provide a theoretical answer to this paradox are explained. Given that the focus is on the role ascribed to income inequality, the LCH of Modigliani and Brumberg (1954) and the PIH of Friedman (1957) are discussed together. In contrast, the RIH of Duesenberry (1949) is based upon a different perception with respect to the effect of income inequality.
2.1.1 Keynes and the Kuznets Paradox
Keynes (1936) is the founding father of the consumption theory. A simplified description of Keynes’ absolute income hypothesis suggests that aggregate consumption is a stable (not necessarily linear) function of current disposable income:
𝐶𝑡 = 𝑎 + 𝑏𝑌𝑡 (1)
Current income (Yt) is divided between consumption (𝐶𝑡) and saving, thus the share of income that is not consumed is automatically saved1. The Marginal Propensity to Consume (MPC), depicted by b2, indicates how much consumption increases when income rises with one unit. In addition, the Average Propensity to Consume (APC)3 captures the ratio of consumption to current income. Thus, if a>0, then MPC<APC, and people spend a decreasing share of their income (Keynes 1936). In other words, the more a household already has the less utility it gains from consuming one more additional unit. In case a<0, it follows that MPC=APC, which means that households spend a fixed share of their income. (Keynes 1936) If this is the case, households gain the same amount of utility from consuming an additional unit, no matter their initial income.
1 Yt = Ct + St 2
δC/δY
12
Keynes’ early empirical work suggests that the MPC<APC. Hence, the MPC decreases with income. This makes sense on an intuitive basis; if you already have two cars, a third one will not give you the same utility as the first did. Therefore it is suggested that a redistribution of income from the poor to the rich, results in a higher level of aggregate saving (Keynes 1936). This mechanism is referred to as the distribution effect. In contrast, Kuznets (1946) established that the share of income spend on consumption remains stable over time, even as income rises. Accordingly, in the long run, consumption is a proportion of income and aggregate saving remains constant (Kuznets 1946). This empirical contra diction is referred to as the Kuznets paradox.
Three consumption models were developed to provide an explanation to this paradox, the RIH and the LCH/PIH. These are discussed in the following paragraphs.
2.1.2 The Relative Income Hypothesis
The main idea of the RIH, as developed by Duesenberry (1949), is that the determination of household consumption is subject to temporal and social comparisons. These are captured by two complementary effects; the ratchet effect and the demonstration effect.
The ratchet effect posits that consumption depends on previous levels of income, since it is difficult to downwardly adjust standards of living. Therefore a fall in current income is not accommodated by a proportional fall in current consumption (Duesenberry 1949). For instance, if a household’s income decreases, and its mortgage payments cannot be altered because they are backed by long term contracts, a household can choose to save less to compensate for the income loss.
The demonstration effect focuses on social comparisons and claims that households imitate the consumption of their reference group. This group generally comprises the richer households one likes to be associated with (Duesenberry 1949). For example, consider an Afro-American family whose income is generally lower than that of a Caucasian family. When both families live in the same country and have the same income, the RIH predicts that the Caucasian family spends a larger share of its income because its reference households are richer, in comparison to those of the Afro-American family. Thus, the Caucasian family consumes relatively more because its position in relation to its reference group is worse than that of the Afro-American family.
To illustrate the RIH by means of a simple equation4, Keynes’ linear consumption function is supplemented to account for the consumption level of the households’ reference group (𝑅𝑡) and to
account for its previous consumption level (𝐶𝑡−1). The 𝜐 and the σ are unknown parameters which
capture the weight a household attaches to social and temporal comparisons.
𝐶𝑡 = 𝑎 + 𝑏𝑌𝑡 + 𝜐𝑅𝑡 + 𝜎𝐶𝑡−1 (2)
4
It should be stressed that the purpose of this equation is solely to clarify the inclusion of temporal and social comparisons on the basis of the RIH. It is not intended as an official equation to capture the RIH.
13
The implication of the RIH is that relative income, in addition to absolute income, determines how much a household consumes. In other words, when everyone’s income within a community increases with the same percentage, the share of income consumed does not change because the relative position of each household remains the same (Parker 2010). In figure2 this scenario is depicted by an upward shift of the aggregate consumption function which results in a constant APC for both household A and B, as depicted by the dotted line. Note that consumption expenditure of A and B has risen with the same proportion as income, resulting in a constant APC.
Duesenberry (1949) posits that a household’s relative position towards its reference group does not change over time because the returns on economic growth are proportionately divided over the different income groups. Thus, as time passes, the demonstration effect predicts that consumption increases proportionately to the income of the reference group. In case income decreases, consumption does not decline that much due to the ratchet effect. This explains why the share of income is constant in the long run as indicated by the Kuznets paradox (Parker 2010).
2.1.3 The Life-Cycle Hypothesis and the Permanent Income Hypothesis
The central point of the LCH, developed by Modigliani and Brumberg (1954), is the transference of purchasing power from one period to the next, referred to as consumption smoothing. Individuals plan their savings in order to maintain a stable level of consumption throughout their life. Assuming that
Figure 2: The effect of a proportional increase in income
14
the level of income increases with working years, an individual borrows while studying, builds up wealth and repays previous loans while working, and finally lives of accrued savings once retired (Modigliani and Brumberg 1954). See figure 3 for a graphical depiction of this mechanism. (Romer 2011).
The main implication of the LCH is that consumption remains constant among each period of life. Accordingly, saving is perceived as future consumption and the real interest rate represents a trade-off between consumption now (t) and in the future (t+1). Consumers are rational decision makers and strive to maximize lifetime wealth. For simplicity, it is assumed that the marginal utility of consumption is always positive, the discount rate is zero, and that there is no uncertainty. See Romer (2011) for more detail.
𝐶𝑡 =1
𝑇(𝐴0+ ∑ 𝑌
𝑡) 𝑇
𝑡=1 for all t (3)
The consumption function of the LCH indicates that consumption is constant and depends upon the level of initial wealth (𝐴0), the summation of lifetime income (∑𝑇 𝑌𝑡)
𝑡=1 , and the time a person will
live (T).
The PIH, developed by Friedman (1957), builds upon the LCH and incorporates fluctuations in households’ income due to life-cycle effects, business cycles or other factors. A distinction is made between permanent income, which captures expected income each period, and transitory income, which captures deviations from the expected level of income (Friedman 1957). Thus, looking at equation 3, the right hand side captures permanent income. The difference between current and
Figure 3: The Life Cycle Hypothesis (Romer, 2011. P.365)
15
permanent income represents transitory income. See the right hand side of equation 4. In case transitory income exceeds permanent income, the difference is discounted over multiple periods (1𝑇 ) and therefore has little impact on consumption (Romer 2011).
The distinction between permanent and transitory income is made to clarify that households save and borrow to keep their consumption constant over time. When transitory income is high, saving is high, and when transitory income is low, dissaving is high. The latter implies that households decrease their savings, or obtain debt, to finance consumption (Friedman 1957).
𝑆𝑡 = (𝑌𝑡−1𝑇∑𝑇𝑡=1𝑌𝑡)−𝑇1𝐴0 (4)
Equation 4 illustrates that current saving (𝑆𝑡) depends on current income (𝑌𝑡) minus average expected
lifetime income, minus the part of consumption that is based on initial wealth. See Romer (2011) for a more detailed description.
The central point of the PIH comprises that consumption is predominantly determined by changes in permanent income because this changes the level of 𝑌𝑡 for the current, and all future periods. Thus, the MPC is close to one when the change in income is of a permanent nature and below one when it is not. Given that the economy is affected by volatile business cycles in the short run, which trigger transitory income shocks, households (dis)save to maintain a constant level of consumption in the long run (Friedman 1957)..
2.1.4 Optimal consumption and income inequality
The main difference between the RIH and the LCH/PIH lies within the determination of optimal consumption. On the one hand, the LCH and the PIH state that households save to maintain a constant level of consumption over time. This is determined on a rational basis and incorporates a realistic expectation of one’s individual income, now and in the future (Friedman 1957). Thus, the determination of optimal consumption is not influenced by the behaviour of other households. On the other hand, optimal consumption within the RIH is dependent upon intertemporal and interpersonal comparisons. This implies that a household’s relative consumption changes when its position within the income distribution changes (Duesenberry 1949).
A prerequisite for achieving the optimal level of consumption throughout all the consumption models is the absence of liquidity constraints. This implies that households are able to borrow and lend money without limits (Romer 2011). Accordingly, households have the ability to absorb temporary changes in income, as predicted by the LCH/PIH, and to finance additional consumption when its relative position deteriorates, as predicted by the RIH.
Moreover, when assessing the role ascribed to income inequality throughout the consumption models, three outcomes are distinguished. Recall that income inequality refers to a redistribution of
16
income from the poor to the rich. First of all, Keynes (1936), suggests that income inequality has a positive effect on aggregate saving (distribution effect). Second of all, the LCH/PIH does not attribute an effect to income inequality itself, solely to changes in permanent and transitory income components. Thus, if the redistribution of income towards the upper decile is of a transitory nature, the poor/middle dissave to keep consumption constant, and the rich increase their savings. However, if it concerns a permanent change, consumption is adjusted downwards for the poor/middle, and upwards for the rich. Given that the rich save a larger amount of their income, savings increase (Friedman 1957). Third of all, the RIH ascribes a direct effect to income inequality because it triggers a deterioration of a household’s position within the income distribution. Assuming that the rich constitute the reference group, a widened income gap implies that it is more costly for the poor/middle to attain the optimal relative consumption level. Thus, the RIH predicts that income inequality triggers dissaving among household whose income decreased (the ratchet effect) and also among households whose reference group has become richer (the demonstration effect) (Duesenberry 1949).
In short, the RIH argues that income inequality can alter a household’s optimal consumption level, whereas the LCH/PIH does not.
17
2.2 Household consumption and the financial crisis
Throughout the literature, two explanations are put forward for the rise in US household debt preceding the financial crisis. The first explanation draws upon the LCH/PIH and perceives the rise in household debt as the outcome of a better functioning financial market. The second explanation draws upon the RIH and claims that income inequality triggered households to spend more. The prior is discussed first, followed by the latter. This sequence is chosen because the inequality narrative builds upon the first arguments.
2.2.1 A more efficient financial market
Financial deregulation changed the structure of the lending market and facilitated access to consumer credit and mortgages for lower- and middle-income households in the US. Accordingly, the rise in household indebtedness since the 1980 reflects optimal consumption decisions in response to the mitigation of constraints (Debelle 2004). Whether the increase in household debt can be explained by more efficient consumption smoothing, as argued by Debelle (2004), and others (See Michell (2015) for an overview), depends on the nature of the income shocks that trigger the rise in income inequality. If transitory income falls below permanent income, households should dissave to smooth consumption, however if permanent income decreases, consumption should be adjusted and no additional dissaving occurs (Romer 2011).
Krueger and Perri (2006) assess whether income inequality in the US was of a permanent or temporary nature. On the basis of a cross-sectional analysis based on Consumer Expenditure Survey (CES) data, they conclude that the variance of income increased with 21 percent, whereas that of consumption only increased with 5 percent over the period 1980-2003. This suggests that income inequality was not accommodated by consumption inequality. To capture the nature of the income shocks, they distinguish “within-group” and “between-group” inequality. To clarify, the prior captures the growth of income inequality among members of the same group, let’s say Caucasian females with a university degree, and the latter compares the growth of income inequality between this group and another, for instance, Afro-American females with a university degree. The distinction is made by regressing income and consumption on several “fixed” characteristics of the household, such as sex, race, years of education and experience. Since these characteristics cannot be altered, they permanently affect households’ income. In contrast, households can insure themselves against temporary characteristics by borrowing money. Krueger and Perri (2006) find that the divergence between income inequality and consumption inequality was a lot smaller for the “between-group” than it was for the “within-group” Therefore they conclude that the rise in income inequality was of a transitory nature, and that households borrowed to keep consumption constant. Similar results are found by Blundell, Pistaferri and Preston (2008) and Heathcote, Perri and Violante (2010).
18
Van Treeck and Sturn (2013) argue that the method of Krueger and Perri (2006) is problematic because it suggests that all income variance within one group is driven by fixed characteristics. Recently, several explanations for the rise in US income inequality have been put forward that are not based on fixed characteristics (Van Treeck and Sturn 2013). Amongst them are rent-seeking behaviour by top executives in the financial sector (Philippon and Resheff 2009), changes in the tax systems (Piketty and Saez 2007), and changes in labour market institutions which triggered an erosion of the real minimum wage and a decline of trade unions (Levy and Temin 2007).
In addition, later research suggests that income inequality is caused by permanent income shocks. Moffitt and Gottschalk (2008) claim that permanent increases in wage dispersion played a much bigger role than transitory fluctuations from the 1990s onwards. Their analysis is based on the Michigan Panel Study of Income Dynamics for the period 1970-2004. To distinguish between transitory and permanent income components, the classical definition is used; the primer represent an income shock that fades out over time, whereas the latter represents a shock that never goes away. Accordingly, they conclude that the increase in cross-sectional inequality is mainly due to permanent income shocks and should therefore have a larger effect on consumption. Additionally, Kopczuk, Saez and Song (2010) used Social Security Administration longitudinal earnings data since 1937, and conclude that the variance of annual income since the 1970s can be explained by permanent changes only.
In short, Debelle (2004), and Krueger and Perri (2006), Blundell et al (2008) and Heathcote et al. (2010) support the idea that transitory income shocks motivated households to maintain consumption constant, which was enabled by a more efficient financial market. In contrast, Moffitt and Gottschalk (2008) and Kopczuk et al. (2010) contradict this conception and conclude that income inequality was caused by permanent income shocks.
2.2.2 The inequality narrative
If income inequality in the US is caused by permanent income shocks, the LCH/PIH suggests that the rise in debt was not used to smooth consumption, which implicates that households must have spent beyond their means. Accordingly, two questions are addressed within the inequality narrative: why did households structurally spend in excess of their income, and how were they able to do so? In the following paragraphs both questions are addressed. (Bofinger 2012). First, on the basis of the ratchet effect and the demonstration effect it is assessed how income inequality triggered households to increase their spending. Second, the connection between financial deregulation and household spending is discussed.
19
Changes in the relative income distribution
A growth at the top-end of the income distribution, and stagnation at the lower/middle, resulted in increased income inequality in the US. Thus, the rich accounted for a larger share of total income, whereas that of the poor decreased. Accordingly, this resulted in a deterioration of the relative position of lower/middle income households within the income distribution (Michell 2015). To clarify, Piketty and Saez (2013) illustrate that the top decile in the US accounted for 35% of total income in the 1970s, and for 50% of total income just before the crisis. Simultaneously, this implies that the remaining 90% of the income distribution only accounted for 50% of total income in the same year. In addition, hourly wages of the middle/lower income classes stagnated since the mid-1980s (Piketty and Saez 2007). Thus, the gap between the rich and the poor/middle widened.
Barba and Pivetti (2009) argue that more households had to take on debt due to the stagnation of lower/middle class wages, in combination with a sharp reduction in public services. Their analysis indicates that US households started to become more leveraged from the 1980s onwards, especially consumer credit and mortgage loans increased substantially. Figures of the Federal Reserve Board suggest that outstanding consumer credit reached 25% of GDI in 2006 (Barba and Pivetti 2009, p.115). Simultaneously, the private savings rate fell from 10% of DPI in 1980, to 0.5% in 2005 (ibid, p.123). In addition, by comparing the debt-income ratio, debt-asset ratio and the debt-service ratio throughout the income distribution, Debelle (2004) concludes that debt was mainly accumulated by lower/middle income groups. Accordingly, this argument is led back to the ratchet effect; households did not adjust to lower standards, instead they dissaved to maintain their standards of living.
Van Treeck (2014) calls for the renaissance of the RIH because the inequality narrative acknowledges that upward looking comparisons shape households’ spending behaviour. (see van Treeck 2012). The RIH is associated with the “Veblen” or the “keeping up with the Joneses” effect which was first put forward by Thorstein Veblen in 1934. It argues that social comparisons shape consumption patterns instead of rational utility considerations (Adkission 2009, p.609). Bowles and Park (2005) describe the effect in the following words: “Spending is driven by relative status considerations, that is by the desire to be a particular type of person as much as by the desire to enjoy the consumer goods per se” (p. 397). Thus, consumption choices can be explained by class differences in which the less well-off consume to improve their relative position within society. Given the widened income gap in the US, the consumption of the rich became more costly. Accordingly, low/middle income households borrowed more to finance additional consumption. This mechanism can be led back to the demonstration effect because it is suggested that the rich serve as a reference group to the lower/middle income classes (Van Treeck 2012).
In addition, Frank et al. (2010) extend the idea of the rich as the reference group by formulating the concept of expenditure cascades; lower/middle income classes do not only compare themselves to the rich, but every household compares its position to the consumption of the income group above. In other words, consumption is always subject to upward looking comparisons,
20
regardless the level of income. Therefore, stronger income gains at the top cause a spending wave throughout the entire income distribution, which results in reduced saving from the bottom to the top (Frank et al. 2010). On the basis of an analysis of 50 US states, over a time-span of ten years, Frank et al. (2010) show that rapid income growth at the top results in increased spending throughout the entire distribution. In addition, Bertrand and Morse (2013) show that consumption expenditure of the richest 20%, affects that of the remaining 80%. Accordingly, an increase in the consumption of the rich is a significant predictor for spending amongst lower income groups. Hence, Bertrand and Morse (2013) suggest that: “middle income households would have consumed between 2.6 to 3.2 percent less by the mid-2000s, had incomes at the top grown at the same rate as median income.”(ibid, p.24). Thus, the expenditure cascades are able to explain why household dissaving occurred throughout the entire income distribution.
In short, the change in households’ relative income position explains why they were willing to consume excessively in the period preceding the crisis, and thus chose to dissave.
Enabling excessive debt
Given that the RIH predicts that households dissave once their relative position in the income distribution changes, they still need to be able to do so. As discussed before, financial deregulation improved access to consumer credit and mortgages for low- and middle income groups (Schularick and Taylor 2012). Despite the outcome of financial deregulation, there are several theoretical explanations as to why the US government restructured the financial market.
The political economy motivation
Rajan (2010) takes a political stance and argues that the US government stimulated financial deregulation and credit expansion in order to mitigate the political consequences of stagnating wages at the lower/middle end of the income distribution.
“The political response to rising inequality – whether carefully planned or an unpremeditated reaction to constituent demands – was to expand lending to households, especially low-income ones. The benefits – growing consumption and more jobs – were immediate, whereas paying the inevitable bill could be postponed into the future.” (Rajan 2010, p.9).
At the same time, home ownership, a key element of the American dream, was facilitated through government agencies and departments. Whereas, Fannie Mae and Freddie Mac were originally set up by the Federal Housing Administration, these agencies were increasingly used to provide mortgages to low/middle-income US households (Rajan 2010). Thus, lower/middle income groups remained inattentive to their stagnating wealth because they were able to increase consumption and had the ability to finance a house, be it through increased indebtedness. Accordingly, over the period
1999-21
2007, US house prices doubled due to the widespread availability of cheap mortgages. Finally, the downturn in the housing market and the sub-prime mortgage crisis exposed the unsustainability of the credit bubble that was created by these measures (Rajan 2010).
Kumhof and Ranciere (2010) incorporate the reasoning of Rajan (2010) in a DSGE model on the basis of two fundamental assumptions. First of all, two different classes of economic agents are distinguished: the investors, who represent the top 5% of the income distribution, and the workers, who represent the lower 95%. The income of investors consists of returns to capital, whereas income of the workers consists of returns to labour. The idea is that workers are confronted with stagnating wages, and are therefore triggered to interact with financial institutions to borrow money. However, investors benefit from lending money to the lower income groups because this bears an interest which they assume to be above the return to labour. Consequently, the workers lose bargaining power in relation to the investors, because they increasingly rely on the loans facilitated by them (Kumhof and Ranciere 2010). Second of all, workers strive to maintain a minimal level of consumption which is based on their relative position within society. When they see their relative position declining, they turn to credit markets in order to borrow money. Hence a downward spiral is created: as the workers get more indebted to the investors, their relative position worsens, which triggers them to borrow even more in order to maintain position, and so on. This process eventually results in excessive debt accumulation and increases the probability of a crisis (Kumhof and Ranciere 2010).
Thus, Rajan (2010) and Kumhof and Ranciere (2010) take a political stance, and stress that the US government increasingly deregulated the financial market so that lower/middle income groups were kept inattentive to their stagnating wealth. At the same time, this benefited the higher income groups because it gave them the opportunity to reinvest their growing income.
The macroeconomic motivation
Stockhammer (2012) takes an economic stance and stresses the importance of macroeconomic dynamics in order to explain the motivation for financial deregulation. He argues that:
“The economic imbalances that caused the present crisis should be thought of as the outcome of the interaction of the effects of financial deregulation with the macroeconomic effects of rising inequality” (Stockhammer 2012, p.1).
Stockhammer (2012) assumes that a redistribution of income from the poor to the rich results in a downward pressure on aggregate consumption, because the rich have a lower MPC. Accordingly, this creates the risk of reduced output and leads to economic stagnation. Thus, in order to maintain a high level of aggregate demand and sustain economic growth, it is beneficial to extend credit to lower ends of the income distribution (Stockhammer 2012).
22
The argument of Stockhammer (2012) does not specifically focus on the US political climate, and can therefore be extended to the international sphere. Whereas financial deregulation eased credit expansion on a national level, it created the possibility for countries to run larger current account on an international level. Accordingly, he identifies two kinds of international growth models: the debt-led growth model, which thrives on consumption booms and current account deficits, like the US, and an export-led growth model, which is characterised by large export flows and a current account surplus, like China. The interaction of these two regimes creates international financial imbalances.
Thus, Stockhammer (2012) reasons from a more general macroeconomic viewpoint and illustrates how inter-related dynamics create an economic motivation for financial deregulation, both on a national and an international scale, which can explain the rise in household debt in the US.
Despite the fact that Rajan (2010) and Kumhof and Ranciere (2010) focus on political factors, whereas Stockhammer (2012) gives a macroeconomic explanation for financial deregulation, all embrace the RIH as opposed to the LCH/PIH, because the latter sees no link between permanent income inequality and household consumption, and thus no need for the government to maintain a high level of aggregate consumption. Thus, in essence, both arguments suggest that government intervention was necessary to support aggregate demand and that this increased household spending (Van Treeck 2012).
In contrast to the inequality narrative, recent research suggests that the spending behaviour of the rich followed a similar cycle as that of the lower income groups. Bakker and Feldman (2014) show that 55 percent of the consumption growth in the US was financed by the upper 10% of the income distribution, from 1993-2003. In addition, Maki and Palumbo (2001) and Kapur, Macleod and Singh (2005) claim that excessive consumption should be attributed to the income growth of the rich. These authors reason on the basis of wealth effects and suggest that the rich were triggered to save a smaller share of their income because they felt richer. Accordingly, a lower savings rate was applied to a larger income which resulted in lower aggregate savings (Kapur et al. 2005).
In summary, the inequality narrative posits that income inequality altered the relative position of lower/middle income households. Subsequently, these households dissaved more in order to keep up with the increased consumption of the richer reference group. In addition, increased financial deregulation enabled this. However, it remains questionable whether the decrease in household saving can be explained on the basis of the RIH, since other research indicates that the upper income groups also increased their spending prior to the crisis.
23
2.3 The inequality narrative in Europe; a cross-sectional description of the
income distribution.
In order to make the inequality narrative more tangible, and project the assumptions onto the European situation over the period pre-ceding the crisis, thee propositions are developed on the basis of the literature:
Proposition 1: The relative position of the lower/middle income households deteriorated Proposition 2: Lower/middle income households financed consumption through dissaving Proposition 3: The larger the deterioration of a household’s position, the larger the trigger to increase its relative consumption.
The validity of these propositions is assessed by means of a cross-sectional view of the European income distribution. The aim of this approach is to gain insight into the consumption levels of different income groups. Accordingly, a fundament is created on which the findings of the econometric model can be interpreted.
The first and second propositions are discussed first because these form essential prerequisites for the manifestation of the RIH. Subsequently, proposition three is discussed which refers to the demonstration effect of the RIH and argues that proposition 1 and 2 are related to each other.
2.3.1 A deterioration of households’ relative position
Recall that the RIH posits that the optimal level of consumption is subject to the consumption of the richer reference group. As explained before, as the income gap between the upper and the lower parts of the income distribution widens the relative position of the lower/middle classes deteriorates.
On the basis of data provided by SWIID (Solt 2009) a rising trend is perceived in the share of income going to the top one percent of the income distribution, henceforth referred to as share1 income. Accordingly, the income of the remaining 99% of the distribution must have decreased, and thus the relative position of these households deteriorated. However, the growth rates differ substantially throughout the sample. Whereas top incomes grew with 53% in the Czech Republic and with 44.5% in Germany, they decreased in Austria, Hungary and Greece. For the other countries, the growth rates varied from 6% to 36% over the period (See figure 4).
In addition, two groups can be distinguished within the sample. On the one hand, there is a group of countries for which share1 income did not fluctuate a lot and stayed below the 7 percent level. On the other hand, the remaining countries are characterised by a share1 income above the 7 percent level. At the same time, share1 income in the last group varied more and moved within a wider bandwidth. Furthermore, Norway and the UK clearly diverge from the trend. Whereas share1 income
24
in the UK constantly stayed above the 10 percent level, it fluctuated a lot in Norway and reached a peak of almost 17% in 2005 where after it fell to 8% (See figure 5).
Thus, an overall rise in income inequality is perceived throughout the sample and therefore proposition 1 is confirmed: The relative position of the lower/middle income households has deteriorated. However, the levels and growth rates of share1 income vary heavily throughout the sample.
Figure 4 The total growth in share1 income over the period 1995-2008
25
Figure 5: The development of share1 income from 1995-2008
26
2.3.2 Financing consumption through dissaving
Two theoretical explanations for an increase in household spending are put forward. On the one hand, the LCH/PIH argues that households dissave to keep consumption constant when they are hit by transitory income shocks. On the other hand, the RIH claims that households dissave when their reference group becomes richer to maintain relative position within society. Thus, it should be pointed out that a confirmation of proposition 2 does not explain why households increased their spending.
Given that dissaving can be approached by looking at saving rates and debt holdings, a thorough analysis of proposition 2 should consider both. However, due to a lack of data, it is not possible to analyse dissaving of different income groups over time. In addition, given that the saving rate is a flow variable, which implies that it is established anew every year, the analysis of a single year does not give an impression of the way dissaving developed over the period 1995-2008. In contrast, the stock of accumulated debt holdings is able to illustrate this, even if it only one year can be reflected. To clarify, it the household saving rate in year X is 4.5% of GDI, it does not indicate how high it was over previous years. However, if households hold 100 euros of debt, it is implied that this amount was accumulated over the preceding years. Accordingly, if lower/middle income households are characterised by a large debt-income ratio, it can be deducted that they dissaved over the preceding years. For this reason, dissaving is solely explored on the basis of households’ debt holdings throughout the income distribution
Before the lower/middle classes are considered, it is necessary to establish whether households dissaved on an aggregate level throughout Europe over the period 1995-2008. On the basis of the debt-income ratio published by the OECD (2015), the growth in gross household debt can be compared to the growth in household GDI5. It follows that, on average, the annual growth in debt exceeds the annual growth in income for all the European countries. Thus, it is established that households increasingly financed consumption through dissaving. See appendix I for the growth rates per country and a detailed description of the variable.
In order to see whether dissaving specifically occurred among the lower/middle income groups, The Eurosystem Household Finance and Consumption Survey is used as a starting point (ECB 2013). These statistical tables provide insight into the spread of debt over different income groups. Accordingly, the income distribution is divided into income deciles (p10-p90), or into income quintiles (bottom 20% - 80%-100%). These divisions are constituted on the basis of households’ gross income, which is defined as the sum of labour and non-labour income for all household members (ECB 2013, p.85). Furthermore, it should be noted that the survey focuses on a different timeframe. Whereas this thesis analyses the period 1995-2008, the reference years of the survey vary between 2008-2009 and 2010-2011 (ECB 2013). Nevertheless, the sectional view still provides insight into the accumulation of debt over the lower/middle income classes.
27
The debt holdings throughout the income distribution are analysed on the basis of the total stock of debt, the debt-income ratio and the debt-service income ratio. The first is expressed in thousands of euros and refers to the absolute stock of household debt. The other two are relative measures, of which the debt-income ratio relates total liabilities to total household gross income, and the debt service-income ratio gives monthly debt payments to gross monthly income (ECB 2013 p.85). It should be noted that the data is conditional upon households holding debt. Given that a relatively small amount of low income households hold debt, see appendix II for an overview, it should be taken into account that the data gives a somewhat distorted view. Overall, three observations stand out.
First of all, the greatest stock of debt is held by the upper incomes. Figure 6 depicts the sum of debt per decile throughout 11 European countries and illustrates that the lower half of the income distribution barely has any debt holdings. In contrast, the last decile holds approximately eight times as much debt as the 4th decile (ECB 2013). Hence, it appears that the absolute stock of debt rises together with income. Moreover, figure 7 gives a sectional view of the accumulation and depicts that the amount of debt holdings vary substantially per country. For instance, the amount of household debt in the Netherlands is much higher than in other countries (ECB 2013). Nevertheless, in general it is concluded that the rich hold the largest stock of debt.
Second of all, figure 8 gives a sectional view of the debt-income ratio throughout the European countries and illustrates that the spread of relative debt varies a lot. In some countries, the highest debt- income ratios are located at the lower ends of the distribution, such as in the Netherlands and Portugal, in which the ratios far exceed the 160%, whereas in other countries the rich have the largest debt-income ratios, such as in Denmark and France (ECB 2013). Thus, the relative spread of debt varies a lot and depends upon the country.
Third of all, the lower income groups have the highest debt-service income ratio. Figure 9 clearly suggests that the bottom 20% of the distribution bears the highest debt-service income ratio. Furthermore, the ratio decreases rapidly from the 20th until the 40th percentile, and remains more or less constant from the 40th until the 80th percentile, where after it decreases again (ECB 2013). In addition, a cross-sectional view of the countries confirms that the debt-service income ratio is high for the poor, and low for the rich. However, the way the ratio is spread throughout the middle section differs per country (ECB 2013).
In conclusion, the rich hold the largest stock of debt, the division of debt throughout the middle classes is heavily dependent upon the country, and the poor have the largest debt-service income ratio. Accordingly, proposition two is only partly confirmed; low income groups do have high relative debt payments, but not only lower/middle income households have financed additional consumption through dissaving.
28
Source: Author, based on data ECB (2013)
Figure 7 Accumulation of the stock of debt per decile (in EUR 1000s)
Source: Author, based on data ECB (2013)
29
Source: Author, based on data ECB (2013)
Figure 8 The Debt-income ratio per quintile (in % GDI)
Source: Author, based on data ECB (2013)
30
2.3.3 A household’s relative position and its optimal consumption
The demonstration effect predicts that households consume in order to maintain position in relation to their reference group (Duesenberry 1949). Before proposition 3 can be assessed, four things need to be pointed out. First of all, the RIH assumes that the lower/middle income classes use the rich as their reference group, and the expenditure cascade effect of Frank et al. (2014) assumes that the consumption of any group is determined by upward looking comparisons. Second of all, the demonstration effect is magnified by loosened liquidity constraints (Van Treek 2013). Third of all, consumption is approached indirectly on the basis of households’ debt holdings and savings behaviour. Fourth of all, a difficulty lies within the distinction between the mechanisms of the LCH/PIH and the RIH, given that the prior also acknowledges that (dis)saving can change. However, in case of a transitory shock, the increased savings of the rich should still be able to outweigh the dissaving of the poor/middle. Thus, if saving rates are characterized by a downward trend it is assumed that the mechanisms of the RIH are at work.
Recall that proposition one is confirmed for the majority of countries, but that the growth in income inequality differs substantially throughout the sample (Solt 2009). At the same time, the distribution of debt over the various income groups depends heavily upon the country (ECB 2013). In order to assess whether there is a connection between these two observations, the growth in share1 income is compared to the debt holdings of different income groups. In addition, the growth in share1 is also related to the development of the savings rate published by the OECD (2015). The comparison between household debt and the growth in share1 is discussed first, followed by the assessment of the savings rates.
The relative position and household debt
The spread of the debt-income ratio per country is analysed, together with the change in share1 income over the period 1995-2008. Accordingly, the countries are organised on the basis of their growth in share1 income as presented. See figure 10.
The highest growth in share1 income is documented for Finland, approximately 34% over the period 1995-2008 (ECB 2013). At the same time, the debt-income ratio is higher for the middle classes (the 60th -90th percentile), than it is for the upper income groups. Moreover, the three countries that follow Finland (Denmark, Italy, and Slovak Republic) are also characterized by high debt-income ratios among the middle classes. However, it depends on the country whether the highest ratios are documented for the 80-90th percentile (Denmark and Finland), for the 60-80th percentile (Denmark), or for the 40th-60th percentile (Slovak Republic) (OECD 2015). Nevertheless, it is clear that the highest debt-income ratios are not located at the lower end of the distribution, as is the case in Greece, Portugal, the Netherlands and Spain. At the same time, the overall debt-income ratio seems higher for the countries with a low growth in share1 income (below the 10% level), and lower for the countries
31
with a high growth in share1 income (above the 10% level). Austria is a clear exception. From France onwards, the growth in share1 income passes the 10% level (Solt 2009). Thus, no relationship is found between the growth in income inequality and the debt-income ratios.
Low
Growth in share1 income High
Figure 10 Relation debt-income ratio and growth in share1 income
Source: Author, based on data ECB (2013) and Solt (2009)
32
The relative position and household saving
A preliminary view of households’ saving behaviour throughout Europe indicates that overall savings decreased with 1/3 over the 1995-2008 period (OECD 2015). See figure 11.
The saving rates of the countries with the largest growth in share1 income are plotted in figure 12. The selection of the countries is based on a growth in share1 income exceeding the 20% (Solt 2009). Despite the sharp increase in income inequality, there is no trend observable with respect to households’ saving behaviour. On the one hand, a downward tendency is perceived for the United Kingdom and Poland. On the other hand, the saving rates fluctuate in the majority of countries, and do not depict a downward nor upward trend. Inclusion of the remaining countries does not give any additional insight (see figure 13). In general, it appears that the only countries with a clear downward trend are: the United Kingdom, Portugal, Hungary, Poland, Belgium, and the Netherlands. At the same time, these countries are characterised by divergent share1 growth rates.
Figure 11 The development of the aggregate European saving rate
33
Thus, no interaction can be found between the growth in share1 income and the household saving rate. The behaviour of the saving rate varies strongly per county and is subject to business-cycle fluctuations.
Overall, no relationship is found between the growth in share1 income and household spending. Accordingly, the deterioration of a household’s position does not seem to influence its relative consumption level, and thus it is suggested that the demonstration effect cannot be detected on a macroeconomic scale. For this reason, proposition three is rejected.
On the basis of the three propositions it is established that no preliminary evidence can be found for the validation of the inequality narrative within Europe. The relative position of lower/middle income households has deteriorated. At the same time, the lowest income groups are characterised by the highest debt-service income ratio. However, the debt holdings of the rest of the income distribution differ substantially per country, just as the saving rates. Moreover, there does not seem to be a connection between the rise in income inequality and household consumption.
Figure 12 The saving rates for countries with an growth in share1 income exceeding the 20%
34
In summary, the inequality narrative draws on the RIH and posits that the surge in household spending is connected to the rise in income inequality. Rajan (2010) and Kumhof and Ranciere (2010) argue that this effect was able to manifest itself in the US due to the increased financial deregulation of the financial markets. However, as pointed out by the Economist (17/3/2012): “Mr Rajan's story may
work for America's 2008 crisis. It is not an iron law.” A preliminary analysis of the European trends in household spending indeed suggests that the narrative is not self-evident within Europe. Nevertheless, there are several authors have which have empirically assessed the inequality narrative on a macroeconomic scale. Their findings are discussed in the following chapter.
Figure 13 Development of the saving rates for the entire sample (in % of NDI)
36
3. Empirical evidence
The connection between the outbreak of the financial crisis and income inequality is dependent upon two consecutive channels. First of all, it should be established that the rise in income inequality triggers an overall increase in household spending. Second of all, an increase in household spending should lead to financial instability and thus be able to trigger a crisis (Bofinger 2012). In this chapter, the empirical evidence concerning the first channel is set out.
The empirical accounts that capture the macroeconomic effect of income inequality on household spending assess whether the mechanisms of the RIH are large enough to outweigh the distribution effect. The line of reasoning is as follows: if lower/middle income households substantially took on more debt in the run-up to the crisis, a downward pressure on aggregate household saving should be perceived (Bofinger 2012).
However it is difficult to capture causality on the basis of aggregate data because, as pointed out by Bofinger and Scheuermeyer (2014): “The link between saving and the personal distribution of income is ambiguous, as there are various partly opposing effects on a microeconomic level, which could be offsetting each other in the macroeconomic aggregate” (B&S 2014, p.3). Accordingly, the empirical evidence concerning the macroeconomic validity of the inequality narrative remains contra dictionary and seems to depend upon the sample, time-span and the included variables.
An overview is given of the most recent empirical contributions. As mentioned before, dissaving is twofold and refers to the financing of consumption either by increasing debt, or by decreasing savings. Accordingly, the relationship of interest is approached on the basis of two measures throughout the empirical literature:
1) The effect of a rise in top income shares on private sector credit growth 2) The effect of a rise in top income shares on household saving rates
Both approaches are discussed separately to survey the different methods, their drawbacks, the generated outcomes and the meaning of the results. Afterwards, it is determined which approach captures the inequality narrative best.
3.1 Private sector credit and income inequality
Bordo and Meissner (2012) analyse 14 OECD countries over the period 1920-2008. Household credit is approximated by loans given out to the private sector by traditional banks. To measure inequality, share1 income is used. The dependent variable is based on the log of real credit growth. Furthermore, traditional credit determinants are included to assess whether the observed positive correlation between income inequality and credit growth still holds. The hypothesis reads: “[…] credit growth has
37
no relationship to changes in income concentration after conditioning on other factors” (B&M 2012, p.2153). Their results indicate that there is no effect of income inequality on credit growth (B&M 2147).
Gu and Huang (2014) explicitly contest the estimation methods of Bordo and Meissner (2012) and repeat the analysis while accounting for cross-sectional correlation, endogeneity between the dependent and independent variables, and non-linearity. (504). Their results, which are based on a similar time-frame and panel as Bordo and Meissner (2012), do provide robust evidence for the existence of a positive relationship between income inequality and private credit expansion. This relationship is particularly strong for Anglo-Saxon countries which were subject to high levels of financial deregulation. Thus, Gu and Huang (2014) suggest that there is a positive relationship between income inequality and private sector credit.
Perugini et al. (2014) also build upon Bordo and Meissner (2012) and focus on 18 OECD countries over the period 1970-2008. They employ a different dependent variable in order to capture the influence of the shadow banking system; private credit in relation to GDP. In addition, they point out that excessive credit makes households’ vulnerable to macroeconomic shocks and sharp price fluctuations; therefore they also include the level of credit besides its growth rate. Furthermore, a control for financial development is included, in order to take the moderating effect of financial deregulation into account. In line with Bordo and Meissner (2012), proxies for investment and for the monetary policy environment6 are included. To measure inequality, share1 income is used. Their approach accounts for reverse causation and endogeneity through a combination of OLS with Panel Corrected Standard Errors (PCSE), year and country fixed effects, and the system GMM method. The effect of income inequality on private sector indebtedness is highly significant throughout all methods and is robust to other inequality variables (top5 and top10 income share). Therefore Perugini et al. (2014) conclude that there is a “[…] statistically significant, positive relationship between income concentration and private sector indebtedness when controlling for conventional credit determinants.” (p1).
Brazillier and Hericourt (2014) distinguish three major drawbacks in the work of Bordo and Meissner (2012), of which the latter two also hold for Gu and Huang (2014) and Perugini et al. (2014). First of all, reverse causality should have been accounted for because there is reason to believe that financial development also affects the level of inequality (see Brazillier and Hericourt (2014) for an overview of this literature). Additionally, credit growth affects some of the macroeconomic variables included in the regression and therefore results in endogeneity. Second of all, they only assess the impact of top income shares, whereas the stagnation of the lower/middle class wages is also an important dimension of the inequality narrative. Third of all, the use of a gross credit variable, which
