1 Financial sector development and inequality – theoretical model for USA Hanna Szymborska University of Leeds 1 Abstract This paper argues that analyses of inequality based on existing theories of distribution do not adequately account for growing wealth disparities. This is because the division into capitalists and workers traditionally envisaged in the Post Keynesian wage share models has been altered by financialisation, making these categories more heterogeneous. Financial deregulation and securitisation have contributed to the falling wage share of national income. The rich accumulate high-yielding assets while the middle/low-income groups suffer from high leverage due to unsustainable debt accumulation. Rising indebtedness, linked to stagnating wage growth and validated by the growing demand for asset-backed securities among financial investors, has led to massive wealth disparities. Recent contributions to the stock flow consistent modelling literature incorporate some wealth considerations into the Post Keynesian stock flow consistent models by distinguishing between rentiers, non-managerial and managerial workers as well as by allowing for indebtedness of non-supervisory workers and consumption emulation. This paper aims to complement these contributions by focusing on how financialisation has altered the structures of householdsǯ balance sheets, and affected their stability. In particular, the implications of these changes for income distribution are examined in a stock flow consistent model of a US economy with three classes of households and a complex financial sector. The simulation results reveal that balance sheet heterogeneity among households has an important impact on inequality levels. WORK IN PROGRESS – DO NOT QUOTE OR CIRCULATE WITHOUT PERMISSION November 2016 Note The author wishes to thank Yannis Dafermos, Gary Dymski, Giuseppe Fontana, Antoine Godin, Marc Lavoie, Maria Nikolaidi, Ozlem Onaran, Cem Oyvat and Peter Phelps for comments on an earlier draft of the paper. 1 Contact e-mail: [email protected]
Transcript
1
Financial sector development and inequality – theoretical model for USA
Hanna Szymborska
University of Leeds1
Abstract
This paper argues that analyses of inequality based on existing theories of
distribution do not adequately account for growing wealth disparities. This is
because the division into capitalists and workers traditionally envisaged in the
Post Keynesian wage share models has been altered by financialisation, making
these categories more heterogeneous. Financial deregulation and securitisation
have contributed to the falling wage share of national income. The rich
accumulate high-yielding assets while the middle/low-income groups suffer
from high leverage due to unsustainable debt accumulation. Rising indebtedness,
linked to stagnating wage growth and validated by the growing demand for
asset-backed securities among financial investors, has led to massive wealth
disparities. Recent contributions to the stock flow consistent modelling literature
incorporate some wealth considerations into the Post Keynesian stock flow
consistent models by distinguishing between rentiers, non-managerial and
managerial workers as well as by allowing for indebtedness of non-supervisory
workers and consumption emulation. This paper aims to complement these
contributions by focusing on how financialisation has altered the structures of households balance sheets, and affected their stability. In particular, the
implications of these changes for income distribution are examined in a stock
flow consistent model of a US economy with three classes of households and a
complex financial sector. The simulation results reveal that balance sheet
heterogeneity among households has an important impact on inequality levels.
WORK IN PROGRESS – DO NOT QUOTE OR CIRCULATE WITHOUT
PERMISSION
November 2016
Note
The author wishes to thank Yannis Dafermos, Gary Dymski, Giuseppe Fontana, Antoine
Godin, Marc Lavoie, Maria Nikolaidi, Ozlem Onaran, Cem Oyvat and Peter Phelps for
I. Introduction ......................................................................................................................... 2 II. Theories of inequality – an assessment ................................................................. 10 III. Wealth and inequality in stock-flow consistent models .................................. 14 IV. Model specification ........................................................................................................ 15
The household sector .............................................................................................................................. 19
V. Results ................................................................................................................................. 31 VI. Sensitivity analysis ......................................................................................................... 39 VII. Conclusion and future work ....................................................................................... 44 References ....................................................................................................................................... 47
List of Figures
Figure 1. Percentage change in homeownership rate by percentile............................ 4
Figure 2. The top 1% income share, USA 1980-2013 ........................................................ 5
Figure 3. Mean and median net worth, USA 1983-2013................................................... 6
Figure 4. Financial fragility measures by percentile, USA 2010 .................................... 7
Figure 5. Median net worth annual growth rate by decile, USA 1989-2013 ............ 8
Figure 6. Household portfolio composition, USA 2014 ..................................................... 9
Figure 7. Change in the top 10%-bottom 40% and the top 10%-middle 50%
median income ratios, USA 1992-2013 ................................................................................ 22
Figure 8. Simulation results – full model ............................................................................. 33
Figure 9. Simulation results – pure capitalists specification .................................... 36
Figure 10. Simulation results – pure capitalist specification, no rentier debt ... 37
Table 4. Government transfers as a percentage of pre-tax income, USA 2011 ..... 23
2
I. Introduction
The main goal of this paper is to examine the dynamics of income and wealth
inequality in high-income countries and the implications for the stability of
household financial positions across the distribution in the light of financial
sector transformation since 1980s. A theoretical stock flow consistent model is
proposed, aiming to explain the concentration of income and wealth at the top of
the distribution and the diffusion of financial fragility to the rest of the society.
The innovation of the model lies in its interpretation of inequality as balance
sheet structure disparities, based on a reinterpretation of the working and
rentier class and a new conceptualisation of the middle class in Post Keynesian
analysis. Three-class specification of the household sector is developed,
accounting for indebtedness, financial fragility and wage inequality – processes
strongly associated with the impact of financial sector transformation on
inequality.
Financial sector transformation, often described by the umbrella term financialisation , is an extremely complex process occurring at a variety of dimensions. It finds its roots in the persistently high inflation and high interest
rates in the late 1960s, which induced non-financial companies to turn to
financial markets rather than banks for funding investment. This realigned firms objectives away from long-term investment towards short-term profitability,
making them more involved in financial activities (such as issuing shares), which
raised the importance of financial over real profits and contributed to the
growing share of the financial, insurance and real estate sector (FIRE) in the
economy at the expense of manufacturing (Palley 2007:18).
The processes of financialisation gained steam in the 1980s under policies
promoting market liberalisation and retrenchment of the state from public
service provision associated with the leadership of Reagan in USA and Thatcher
in UK (Sawyer 2013:13). Firstly, labour market liberalisation and the associated
rolling back of minimum wage, unemployment protection and union-oriented
policies resulted in gradually declining wage income growth. Simultaneously,
provision of pensions, housing and public goods such as education and
3
healthcare was increasingly delegated from the state to the private sector. With
stagnant wages and diminishing state provision, households found themselves in
need of additional financing through borrowing.
Rising credit demand was paralleled by the massive proliferation of
financial instruments and the development of structured finance. The
aforementioned turn of non-financial companies towards financial markets
resulting from high borrowing costs in 1960s and 70s led financial
intermediaries to seek revenue in the household sector and through innovation
of new financial products (Dymski 2009:157). An increasing volume of financial
obligations — primarily consumer debt and mortgages — was transformed into
securities in a process labelled securitisation, forming collateralised debt
obligations (CDOs), which combined financial instruments of varying risk and
return characteristics (Pollin/Heintz 2013:113). The establishment of credit
default swaps (CDS) and derivatives on existing products allowed investors to
bet against the default of any financial instrument, leading to the transformation
of traditional lending relations based on intermediation towards an originate and redistribute" model, where default risk became originated" by creditors and then spread across the financial system through securitisation. The actors of this
new lending model were not only registered banks, transformed into highly consolidated megabanks as a result of intense merger activity, but also non-
bank intermediaries, which played a role similar to that of formal banks but were outside central bank s jurisdiction in obtaining liquidity ibid.:115). This whole
process was validated by increasing financial deregulation measures such as the
Gramm-Leach-Bliley Act in 1999 in USA, which allowed commercial banks to
engage in financial investment activities.
The combination of demand factors (stagnant earnings, privatisation of
public services) and supply factors (securitisation, deregulation) led households
in high-income economies to become more involved in financial markets,
although to a varying extent in different countries depending on the degree of
liberalisation and deregulation introduced. On the supply side, financial
intermediaries were eager to include more households in their services partly to
compensate for diminishing deposits from non-financial firms (banks) and partly
4
to generate more underlying assets for CDOs so as to keep pace with the rapidly
growing demand for securitised instruments among financial investors (bank
and non-bank intermediaries) (cf. Goda/Lysandrou 2013). In the process, many
non-bank intermediaries took advantage of lax financial regulation and engaged in predatory lending practices by offering subprime" mortgages at extremely harsh conditions to social groups previously excluded from access to credit, such
as the young, women and racial minorities (cf. Dymski et al. 2013). Those
subprime mortgages formed a lion share of securitised assets demanded by
investors. In result, growth in homeownership rates among households at the
bottom of the distribution spiked (Fig.1). Securitisation and tranching of
subprime loans and other instruments into CDOs created an unequal hierarchy
of monetary claims, giving priority to the interests of senior (and wealthy)
financial investors and diminishing possibilities of debt renegotiation and
forgiveness in case of financial distress for the low-income borrowers (cf. Mian
and Sufi 2013). In the wake of the crisis, this resulted in a wave of foreclosures,
evictions and unsustainable indebtedness for the subprime borrowers,
spreading the burden of the crisis unequally between different race and gender
groups (cf. Young 2010).
Figure 1. Percentage change in homeownership rate by percentile,
USA 1989-2007 (source: Survey of Consumer Finances)
These mutually validating processes associated with financial sector
transformation set in motion institutional forces exerting direct impact on the
dynamics of income and wealth distribution in advanced countries. Data show
that various measures of inequality have dramatically increased in high-income
0
5
10
15
20
25
30
Pe
rce
nt
5
countries. In USA, where the trends are the most extreme, Gini coefficient for
income rose from 0.48 in 1982 to 0.57 in 2006 (Wolff 2014:27). Furthermore,
the share of national income held by the richest 1% (excluding capital gains) in
USA increased by 131% between 1980 and 2012, peaking at 18.3% in 2007
(Alvaredo et al., fig.2).
Figure 2. The top 1% income share, USA 1980-2013 (source: Alvaredo et al.)
The growth in inequality at the top tail of the distribution was driven by
financial sector, with financial services sector employees accounting for 15%-
27% of the top 0.1% of the income distribution in USA (and non-financial sector
top executives representing only around 6%, cf. Kaplan/Rauh 2009).
Simultaneously, due to wage growth lagging behind productivity growth, the
share of worker compensation in GDP declined steadily from 62% in 1980 to
56% in 2013 in USA (AMECO Database), suggesting redistribution of national
income towards profits, specifically financial profits (Krippner 2005).
In terms of wealth, the rise in wealth Gini in USA has been less dynamic
than that of income but its level has been persistently higher, reaching 0.87 in
2010 (Wolff 2014). Deepening wealth inequality is further highlighted by the
Pe
rce
nt
0
5
10
15
20
25
Top 1% income share
Top 1% income share inc. capital gains
6
growing gap between mean and median net worth (defined as marketable assets
less current debt) — in USA, the mean-median ratio increased from 3.9 in 1983
to 7 in 2013 (Survey of Consumer Finances, fig.3). Similarly to income, finance
has been strongly associated with rising wealth inequality. Almost a third of
wealth of the Forbes 400 listed rich derives from finance, compared with around
10% from manufacturing or technology (Foster/Holleman 2010).
Figure 3. Mean and median net worth (left axis) and the mean-median ratio
(right axis), USA 1983-2013 (source: Survey of Consumer Finances)
These worrying trends in inequality were only briefly reversed during the
2007 recession. The top 1% income share in USA declined from 18.3% in 2007 to
16.7% in 2009, but it quickly recovered to 18.9% in 2012. Importantly, fall in the
top 1% share of national income was redistributed within the top quintile, as the
share of the top 10% decreased by far less than the top 1% share between
2007-2011 (Dufour/Orhangazi 2016:165). Real wages were temporarily on the
rise and despite growing unemployment, low and middle income households
suffered smaller income losses than the top 1%. The latter saw they capital
income diminished in result of falling asset and property prices
0
1
2
3
4
5
6
7
0
100
200
300
400
500
600
700
1983 1989 1992 1995 1998 2001 2004 2007 2010 2013
Median Net Worth Mean Net Worth Mean-Median Ratio
00
0s,
20
13
US
D
7
3,5
60,6
18,9 21
127
41,2
71,5
134,5
51,3
0
20
40
60
80
100
120
140
Debt / equity ratio Debt / income ratio Principal residencedebt / house value
Top 1%
All HHs
Middle 3
quintiles
(Dufour/Orhangazi 2016:165). The overall Gini coefficient for income fell from
0.57 to 0.55 (Wolff 2014:27). Nevertheless, there are reasons to believe that the
relative income gains for the working class are likely to be short lived as positive
growth of real wages in recent years has been driven primarily by low inflation
(caused mainly by falling commodity prices, which are known to be highly
volatile) rather than rising nominal wages (Gould 2016).
In contrast, while falling asset prices slightly diminished the stocks of
wealth of the rich, the Gini coefficient for wealth increased by 0.035 Gini points
in the post-crisis period (Wolff 2014:32). In fact, while median net wealth fell by
21.2% from 2007 to 2010, mean net wealth saw only 6.5% decline, suggesting an
uneven burden of the crisis across the society (ibid.:24). The increase in wealth
inequality during the crisis was due to different degrees of leverage across the
population (ibid.:32). The ratios of debt to assets and income were unsustainable
for the middle and bottom part of the distribution and amplified the asset price
losses (Fig.4). Consequently, wealth gains experienced by these income groups in
the 1990s and early 2000s relied primarily on asset price inflation and
increasing indebtedness, turning to be illusory as the recession unfolded (Fig.5).
Figure 4. Financial fragility measures by percentile, USA 2010 (source: Wolff 2014)
Pe
rce
nt
8
Figure 5. Median net worth average annual growth rate by decile,
USA 1989-2013 (source: Survey of Consumer Finances)
The key argument of this paper is that these differences in the dynamics
of wealth and inequality are related to balance sheet composition of households
along the distribution (Fig.6). Middle- and low-income households rely more
heavily on primary residence and high homeownership rates (67% share of total
assets compared to 31% for all households) and greater relative indebtedness
(debt-equity and debt-income ratio at 72% and 135% respectively compared to
21% and 127% for the whole sample, see fig.5) driven by mortgage debt, making
their balance sheets more vulnerable to financial shocks (ibid.:22). As was
mentioned before, asset price movements and housing market collapse shortly
before the Great Recession generated a massive drop in median wealth, while
mean net worth suffered less and grew at a faster rate than the median in the
whole period, indicating deepening inequality. The fact that top quintiles
directed most of their wealth into financial assets meant that annual rates of
return were comparatively higher for these wealth groups (ibid.:30-31).
Crucially, these dynamics of household balance sheet structures were directly
related to the political economy of securitisation and household indebtedness
outlined above. Consequently, a powerful case of the impact of financialisation
Pe
rce
nt
-1,78
0,46
-7,84
0,74
3,66
-4,83
2,45
4,23
-1,63
-10,0
-8,0
-6,0
-4,0
-2,0
0,0
2,0
4,0
6,0
1989-2013 1989-2007 2007-2013
Bottom 40% 40th-90th percentile Top 10%
9
on inequality emerges from wealth distribution, household balance sheet
structures and leverage.
Figure 6. Household portfolio composition, USA 2014 (source: Wolff 2014)
Overall, the above analysis of the data reveals that in the context of
financial sector transformation an important aspect of inequality emerges from
the distribution of wealth. The growing need for borrowing arising from
retrenchment of the state and labour market liberalisation policies was matched
by rising demand of wealthy financial investors for securitised assets derived
from loans to households. This led to an emergence of a new class of
homeowners forming the new middle class. Their wealth gains were driven by
the real and financial housing bubble and were largely eroded during the Great
Recession. Coupled with stagnating incomes, the new home owning middle class
lost out the most due to financial sector transformation. It is argued below that
the existing theoretical approaches to inequality do not fully account for this
heterogeneity of wealth among households its impact on inequality. The
proposed theoretical model aims to incorporate these considerations.
9,4
66,6
31,3
5,5
5,9
6,2
7,8
14,2
15,3
25,4
3,1
15,7
50,3
8,9
29,8
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Top 1%
Middle 3 quintiles
All HHs
Principal residence
Liquid assets
Pension accounts
Corporate stock, financial assets, trusts and funds
Unincorp. business equity, other real estate
Other
10
II. Theories of inequality – an assessment
Although aspects of wealth have been increasingly incorporated into
distributional theories, heterogeneity of households financial positions has not
been taken into full consideration explicitly.
Theory putting the largest emphasis on the importance of wealth for
inequality is found in the seminal work of Piketty (2014). The main premise of his Capital in the Twenty-First Century is that inequality is driven by accumulation of persistently higher returns to wealth (r) relative to the growth
of income (g) (historically averaging at 5% and 1% respectively). Compounding
of the returns to wealth overtime generates higher income flows for the wealth
holders and their inheritors (identified with the top 0.1-1%) than for the rest of
the society. Higher capital income in turn allows for greater saving, facilitating
further wealth generation and perpetuating inequality. In other work
(Piketty/Zucman 2014) it is emphasised that due to its high concentration and
the aforementioned accumulation dynamics, inequality of wealth is more
important for the overall structure of inequality in the 21st century than in the
post-war era. Importantly, saving and consumption propensities are not enough
to predict wealth-income levels in advanced countries (higher wealth-income
ratios suggesting large economic power of asset holders and deepening
inequality). This is because capital gains (often driven by housing wealth) are
found to account for around 40% of increase in national wealth to income ratios
between 1970 and 2010 (Piketty/Zucman 2014:1288). Piketty s insight regarding the interplay between income and wealth dynamics and its impact on inequality is particularly relevant in the age of
financialisation. As highlighted in the introduction, financial innovation and
securitisation influenced inequality by generating differential rates of return and
degrees of volatility across the distribution. Large wealth holdings of the rich
allowed them to invest in high-yielding financial instruments (often requiring
large initial payments, which can only be afforded at high levels of net worth),
generating sizeable capital income. Moreover, they were able to use their
economic power to secure higher wages, particularly when employed in financial
sector.
11
Despite the importance of its general conclusions, Piketty s Capital in the Twenty-First century suffers from several drawbacks. The most relevant criticisms for our analysis concern the weakness of Piketty s theoretical explanation and insufficient emphasis on household debt in contributing to
inequality.
While his empirical work is to be applauded, theoretical explanation for
inequality based on r greater than g relies on the expectation that these trends observed in the past would continue into the future (Pressman 2016:159).
Hence, Piketty does not provide any explicit theoretical explanation why returns
to wealth should always exceed the growth of income. Consequently, despite the
relevance of his conclusions, there is no formal link between inequality and
financial sector transformation in Piketty s framework. The alternative body of theoretical literature identified with the Post
Keynesian functional distribution explicitly takes into account the link between
financialisation and distribution. It focuses on the macroeconomic impact of
increasingly unequal functional distribution of national income between two
factors of production – capital and labour – which are associated with higher
propensity to save and consume respectively (cf. Kalecki 1971). The distributive
force of financialisation is seen as the maximisation of shareholder value,
proxied by a higher rentier (i.e. capitalist) income share, related to the
increasingly short-term orientation in firm operations and preference for
financial rather than real investment, which increased the corporate governance
power towards shareholders (cf. Hein 2009, 2015; Hein/Van Treeck 2010; Palley
2012, 2013; Van Treeck 2009). These models often draw from Bhaduri/Marglin
(1990) argument that the macroeconomic effects of income transfers between
wage and profit earners hinge on whether the economy is wage- or profit-led.
Onaran et al. (2011) establish that the majority of advanced economies are
wage-led, which in the Bhaduri/Marglin framework signifies that lower wage
share resulting from financial sector transformation has a negative impact on
aggregate demand and growth by undercutting effective investment demand
because resources are taken away from those who are more likely to spend them
to those who are more likely to hoard them.
12
However, what this theoretical approach has not yet done is to examine
how the transformation in the nature of financial intermediation has
complexified the division of society into two distinct categories. Both groups of workers and capitalists have become heterogeneous, which complicates their
analytical usefulness. In the course of financialisation workers became the
recipients of capital income through homeownership and private pension
schemes, while capitalists became the recipients of the highest wages in the
economy. In fact, the top 10% of earners were the only income group with
above-average income growth between 1979-2012 (Bivens et al. 2014), with the
top 5% of wage earners experiencing a wage increase during the Great Recession
(Table 1). Clearly, not only are there large disparities in the aggregate
characteristics of households within each category but also the boundaries
between the two have become less clear.
Table 1. Average annual growth rates of average hourly wages, USA 1979-2012
(own calculation based on Bivens et al. 2014)
Moreover, Keynesian models are traditionally focused on investment as
the variable most important for macroeconomic growth, treating savings and
consumption as residual and passive (Setterfield/Kim 2013:2). However, since
1980s consumption has become much more volatile and thus more important as
an independent source of aggregate demand. This is largely due to development
of financial sectors and massive expansion of credit to households, leading
household spending to become increasingly disconnected from income.
1979-2012 1979-2007 2007-2012
All households +0.7% +0.8% +0.02%
0th-40th percentile –0.14% +0.05% –1.2%
40th-80th percentile +0.13% +0.3% –0.6%
80th-95th percentile +0.8% +1.02% –0.3%
Top 5% +2.9% +3.4% +0.4%
13
Similar drawback can be identified in Piketty. This is because his
argument relies on comparing average growth rates of wealth and income.
However, there is a substantial variability in income and wealth trends across
the distribution, which is particularly important when it comes to understanding
the impact of financialisation on inequality. As suggested in the introduction, the
top 10% experienced the most rapid and above average wage income and net
wealth growth over the past decades. In contrast, income and wealth gains to the
middle and lower class were illusory as they were underpinned by a housing
price bubble and large household debt holdings relative to income and assets.
Consequently, differential degrees of leverage across the population turned to be
an important driver of inequality, particularly during the 2007 recession. It is not
only the access to financial resources but also the stability of that access
overtime across the population that has implications for inequality. For instance,
financial investors owning a diversified portfolio of securitised assets with
return guaranteed by the seniorage of their claims (due to tranching) are better
able to bear financial losses associated with risky financial instruments than
households whose portfolios are based on housing equity withdrawal (HEW). In
the latter case, price deflation of collateralised assets prevents further
withdrawal of equity to cover outstanding loan repayments, generating higher volatility of household s balance sheet position relative to the former case. Since
interest rates differ for the bottom/middle and high-income households, there is a disproportionate impact of borrowing on financial stability of households balance sheets (Pressman/Scott 2009). When interest payments are considered,
smaller portion of income is available of consumption and hence inequality is
deepened.
Examination of household balance sheets structures remains relevant
after the Great Recession. Scott/Pressman (2015) show that households have
not deleveraged their massive debt levels after the 2008 crisis. Using data from
US Survey of Consumer Finances (SCF) they show that the decrease in total
median monthly debt payments and debt payments to income ratio has been
illusory and reflected low interest rates rather than real reduction in debt. In
fact, mortgage debt levels have not fallen much since the recession. Moreover,
the share of households filing for bankruptcy has been rising since 2010. Because
14
households have not deleveraged properly after the Great Recession, there have
been no increases in consumption and saving allowing for more equitable
growth of the economy.
Consequently, there is a gap in the existing literature on inequality. On the
one hand, Piketty s insight on the interplay between wealth and income is not fully developed on a theoretical level. On the other hand, the Post Keynesian
theoretical literature does not take into sufficiently explore the role of wealth
distribution, household balance sheet structures and leverage for overall
inequality dynamics. This provides an opportunity to complement the existing
literature with a theoretical model incorporating wealth into the analysis of
inequality. We propose a three-class theoretical model aiming to explain the
observed trends in inequality, accounting for disparate wage growth, unequal
returns to wealth and leverage across the population and the role played by the
middle class.
III. Wealth and inequality in stock-flow consistent models
To maintain dialogue with the existing literature on financialisation and
distribution described above, we adopt the method well established among the
Post Keynesians, namely the stock flow consistent modelling (thereby SFCM).
Originating in Copeland (1949) and the works of Tobin and Godley in 1980s, the
framework has recently been formalised by Godley/Lavoie (2007). It is a
macroeconomic tool integrating stocks and flows across real and financial
sectors in the economy in a consistent fashion, relying on the quadruple-entry
system, which necessitates that every inflow has a corresponding outflow in the
system (Caverzasi/Godin 2013).
A number of recent contributions in the SFCM literature take into account
some aspects of household wealth into the analyses of growth and
In contrast to the existing Post Keynesian approaches to distribution, social
groups in our analysis are defined not by the type of employment or ownership
of the means of production but by their balance sheet characteristics. As argued
previously, this is a more suitable method to understanding inequality in the age
of financial sector transformation and massive indebtedness of the society.
Moreover, it links with the theory developed by Piketty, which highlights the
importance of wealth in contributing to overall inequality.
The working class
The working class includes non-supervisory production/ blue collar workers. )n line with the Kaleckian approach, this group has the highest propensity to
consume. Critically, they are the most leveraged group. It is identified with the
bottom 40% of US population, which experience net wealth losses over the past
three decades (see fig.5 above).
One of the phenomena associated with financial sector transformation
has been the massive extension of credit to those previously excluded from
access to it based on their low incomes and low or non-existent wealth. As was argued before, this credit expansion wasn t accidental as household loans, primarily mortgages and consumer credit, constituted the basis for asset-backed
securities. Consequently, there were strong incentives in the financial sector to
generate as many household loans as possible to satisfy the growing demands of
financial investors for securitised instruments. For these reasons, analysis of the
household sector in the model accounting for financial sector transformation
calls for consideration of credit among the lowest income groups. In the present
model, the working class households are seen as subprime borrowers. We
assume that they do not carry enough wealth and income that would allow them
to take out mortgages and hence that all working class households rent houses.
Consequently, it is assumed that credit to the working class households consists
of unsecured short-term consumer credit and payday loans. This has been
particularly relevant in recent years as unsecured debt and payday borrowing
have been on the rise after the crisis (cf. The Pew Charitable Trust 2012; PWC
2015).
20
Working class households rely primarily on wage income (Bivens et al.
2014:6). In our model, real disposable income of the working class consists of
wages and interest earned on deposits, less interest paid on loans and house
rental payments to rentiers. Households consume part c1 of their disposable
income as well as proportion c3 of their wealth, and store the remaining savings
as bank deposits. We assume that the propensity to consume of this income
group is the highest among all households. Furthermore, at this stage of the
model we assume constant propensity to consume out of wealth c3 across all
household groups.
Assuming simple adaptive expectations, borrowing by the working class
is determined by their past consumption level, adjusted by parameter β. β captures household borrowing norms as well as lending norms in the financial
sector (Setterfield/Kim 2013:10). In this way, we are able to indirectly account
for borrowing constraints for workers, reflecting commercial banks attitude towards creditworthiness of borrowers. We can think of β as high during the housing bubble, when lending norms were lax due to the perceived minimisation
of credit risk by securitisation. In times of recessions, β can be thought of as low as lenders are more concerned about creditworthiness and lending norms are
strict. Because workers are constrained in their access to credit, their demand
for loans also includes the debt burden ratio, capturing the repayment capacity
of past loans.
Net wealth of the working class is accumulated entirely in deposits, less
loans. Rental payments on housing are defined as a proportion � of the value of
houses owned by rentiers. � depends positively on the change in rentier demand
for housing. At this stage of the analysis it is not endogenously explained why
households in each group chose to rent or own their house, although the earlier
discussion in this paper explains how financial innovation had the middle class
households turn into homeowners and low-income households rely on
unsecured debt.
Because differential degrees of leverage and unequal ability to cope with
financial fragility along the distribution are important contributors to inequality
21
in a financialised economy (as discussed above), one of the most innovative tasks
of our model is to examine the exact dynamics of household leverage and
inequality. Since measurement of financial distress is a complex task (cf.
DeVaney/Lytton 1995, Boushey/Weller 2008, Ampudia et al. 2014), we include
three different measures of leverage to account for financial fragility in the most
complete way possible at the present stage given our choice of SFCM as
modelling technique. Firstly, the ratio of debt to assets is provided, capturing the
value of loans relative to the value of gross wealth. Secondly, debt to disposable
income ratio constitutes a measure of the stock of loans to the flow of disposable
income in each period. Finally, debt servicing to income ratio shows how much of
gross income is directed towards debt repayments in each period. We assume
that for the working class all of these measures are relatively high.
The middle class
Definition of the middle class is complex. In monetary terms it is defined,
according to the relative size of income, as the middle 60% of the population,
with incomes ranging from 75% to 125% of median income as the standard,
although some studies have extended the upper limit to as much as 300% of
median income (this is because with 125% as the cut-off a disproportionately
large portion of the population in certain countries falls into the upper class
category, cf. Pressman 2007). Atkinson/Brandolini (2011) develop a wealth
criterion to qualify the income definition of the middle class. Based on various
studies, the rich can be classified as having net wealth at least 30 times larger
than mean income. As for the lower cut-off point, members of the middle class
should have enough real and financial assets to be clear from the risk of falling
into poverty for a certain period of time, e.g. 6 months, if income suddenly falls.
Atkinson/Brandolini argue that asset-poor individuals may need to be excluded
from the middle class even if their income exceeds the poverty threshold.
In contrast, definition of the middle class in the present model is centred
on the stylised facts on balance sheet composition and income trends found in
the income and wealth data for USA. Middle class is defined as a group whose
balance sheets depend on housing. Their wealth was rising in the 1990s and
2000s due to increasing house prices, allowing them to refinance their
22
mortgages by taking on more credit and engage in home equity withdrawal, a
strategy which was only feasible in house price bubble. When the price trends
reversed during 2006 and 2007, these households saw their wealth gains largely
wipe out. For these reason, the middle class is assumed to have high leverage
ratios.
Separation of this group from the working class is important as the
evidence shows that in USA inequality growth has been the most striking
between the middle and upper parts of the population rather than between the
top and the bottom (cf. Wolff 2014). Because of the differential rates of return on
wealth of the upper and middle income groups as well as stagnant income for the
latter, the ratio of median income held by the top 10% to median income of
households in 40th to 90th percentile increased by an average 1.5% a year
between 1989 and 2013, compared to 0.9% growth in the ratio of the top 10% to
bottom 40% median income (Figure 7). This difference in the inequality trends
between the middle and bottom population groups is also related to higher
government support and greater contribution of social security to income of the
latter (Table 4).
Figure 7. Change in the top 10%-bottom 40% and the top 10%-middle 50%
median income ratios, USA 1992-2013 (source: Survey of Consumer Finances)
-6%
-3%
0%
3%
6%
9% T10/B40
T10/M50
T10/B40AverageT10/M50Average
23
Table 4. Government transfers as a percentage of pre-tax income, USA 2011
(source: Congressional Budget Office)
Note: Government transfers include local, state and federal level in-kind benefits and cash payments from social insurance and government assistance programs.
This definition of the middle class encompasses the portion of the
population between the 40th and the 90th percentile and thus includes the
median household. The lower cut-off has been chosen as households below the
40th percentile saw negative wage and net worth growth between 1989-2013
(see table 1 and fig.5). In contrast, the upper cut-off has been chosen as only
households above the 90th percentile experienced above average income growth
(Bivens et al. 2014).
Because the middle class is assumed to account for 50% of population in
our analysis, issues associated with heterogeneity of this group need to be
acknowledged. Currently, the middle class in our model includes both subprime
mortgage borrowers, whose incomes resemble more the income of the working
class, and the middle-managers in the 80th-90th percentile, whose incomes and
wealth are closer to the rentier households.
We argue that heterogeneity issues cannot be avoided in analysing the
household sector. Three class division adopted here is superior to the two-class
conceptualisation of households in the literature because it allows for a more
intricate examination of household balance sheets, leverage and incomes in the
age of financialisation, which altered the traditionally envisaged economic
relationships. There is a possibility of extending the division of households even
further, which has been done by Dafermos/Papatheodorou (2015). Such detailed
division is not necessary in the present model for two reasons. Firstly, it would
introduce a considerable degree of complexity to an already elaborate model of
Bottom quintile 36.9%
Second quintile 34.6%
Middle quintile 24.8%
Fourth quintile 14.4%
Highest quintile 4.4%
24
heterogeneous households and financial institutions. Secondly, in an aggregate
model that SFCM is, it would be difficult to meaningfully break down the social
classes into upper/lower groups and introduce a drastically different picture of
balance sheets than already provided in the three class model. This is because at
the aggregate level the most important distinctions between the different types
of debt and wealth accumulation possibilities are already made.
Real disposable income of the middle class consists of wage income and
interest earned on deposits less interest payments on loans. A fraction of
disposable income and wealth is consumed. Residual income is saved as
deposits, including realised capital gains on housing.
Borrowing of the middle class depends on their target consumption and
their debt burden. Target consumption incorporates past consumption (due to
simple adaptive expectations) and relative consumption concerns, which depend
on rentier consumption adjusted by an emulation parameter η. η is exogenously
defined as the Ravina emulation parameter (Ravina 2007). Consumption
emulation has recently emerged as a potentially important driver of borrowing
(cf. Cynamon/Fazzari 2008, Pressman/Scott 2009), leading to lower levels of
consumption than income inequality (cf. Krueger/Perri 2006). However, while in
existing SFCM studies emulation is applied to low-income workers (see above
and Kapeller/Schuetz 2015; Detzer 2016), we restrict relative consumption to
the middle class. This approach is more reflective of reality as emulation motives
are more likely to be relevant among the more affluent households belonging to
the middle class, who can afford necessities such as owning their house. In
contrast, working class households are more concerned with maintaining their
living standards in the light of rising living costs (rent payments). Their demand
for loans is thus more likely to be driven by necessitous borrowing concerns (cf.
Pollin 1988) rather than their desire to follow the celebrity lifestyle of the rich. It
would be possible to introduce emulation of the middle class consumption by the
working class, in line with the expenditure cascades theory where each group
emulates consumption of the one just above it in the distribution (Frank et al.
2014). However, we believe that in the age of financial sector transformation,
due to falling median incomes and increases in the prices of housing, rising
25
demand of low-income households for unsecured credit such as payday loans is
motivated primarily by sustaining a constant standard of living rather than
achievement of social status.
Net wealth of the middle class is composed of deposits and housing, less
loans. We therefore assume that middle class households are owner-occupiers of their houses and hence that they don t rent out their property) and that loans to
the middle class consist exclusively mortgages. Demand for houses by the middle
class depends positively on their income and change in the provision of
mortgages and negatively on their consumption and debt-to-income ratio,
adjusted by the price of housing. As in the case of the working class, different
measures of financial fragility for the middle class are presented, including the
debt-to-asset ratio, debt-to-income ratio and the debt-service-to-income ratio.
Rentier class
Households in this group are defined as the top 10% of the population. In
contrast to the other household groups, they saw income growth equal or above
the average since 1980s (Bivens et al. 2014). Moreover, their balance sheets are
relying primarily on financial wealth rather than housing or wages, which
differentiates this group from the middle and the working class respectively (see
fig.6).
Existing studies accounting for distributional heterogeneity often adopt
social classification from the times of Marx and treat the rich as pure rentiers,
deriving their income purely from capital ownership. This is also envisaged by
Piketty – as wealth becomes inherited and compounding returns to wealth
exceed income growth overtime, the rich abandon work as they are able to live
off the returns to their wealth. While this was true in the pre-Fordist era and
seems like a plausible scenario for the future in light of the deepening wealth concentration, it doesn t describe the realities seen since the post-war period.
Data for USA show that inheritance accounts for a small portion of existing
wealth for the rich (Keister/Lee 2014:20). In turn, much of the income of the top
10% derives not only from large returns to capital but also from extremely high
salaries, particularly for financial sector executives (cf. Kaplan/Rauh 2010). To
26
account for growing wage inequality we can describe the rentier class in our
model as working rentiers . This complements the traditional Post Keynesian
view of the capitalist class as owners of capital earning no wage income.
Importantly, the rentier class engages in work not because of necessity (as is in
the case of the working and the middle class) but because institutional
conditions made employment an alternative investment strategy for the rich along the ownership of capital, as they are able to use their financial power to
influence their earnings.
Furthermore, in contrast to the majority of SFCM studies including debt,
we allow for indebtedness of the rich. This is because the analysis of household
survey data reveals that the top decile undertakes sizeable debt and constitutes
the most indebted income group in terms of both participation and the amount
of debt. Consequently, in our model it is assumed that rentiers borrow from
banks to consume and invest in excess of their wage and capital income. Rentier
borrowing depends positively on their wealth, which serves as a collateral. What
is different about indebtedness of the rich is their leverage. In contrast to other
income groups, debt of the top decile constitutes a small portion of their assets. Rentiers disposable income consists of wages, interest on deposits, part
of the profits of firms, commercial banks and institutional investors, return on
equity, institutional investors shares as well as housing rent payments by the
working class households, less interest paid on loans. As other household groups,
rentiers consume a fraction of their income and wealth. In line with Kalecki,
rentiers are assumed to have the lowest propensity to consume among all
household groups. Deposits of rentiers consist of residual saving as well as
realised capital gains on housing and equity.
Borrowing of rentiers depends on their past consumption and debt
burden ratio and does not include relative consumption concerns. It should be
mentioned, however, that since growth in the national income share of the top
10% is driven by the top 1%, and the growth of the top 1% share is driven by the
top 0.1% (cf. Piketty 2014), relative consumption motives are bound to be
especially strong among the richest 10%, who engage in luxury goods consumption and aim to attain the highest status and the associated celebrity
27
lifestyle . (owever, high aggregation of SFCM and the elaborate character of the
current model prevent us from modelling the precise consumption behaviour of
different income groups within the top 10%.
It is assumed that the allocation of rentiers wealth between houses, equities, institutional investors shares and deposits (treated as a buffer stock)
follows a Tobinesque portfolio principle and depends on the relative rates of
return offered on these assets (Caverzasi/Godin 2015:16). Business equity
accounts for an important part of wealth for the richest 10% and thus rentiers in
our model are assumed to own all firm equity. Return on housing considered by
the rentiers is given by the ratio of rent payments by the working class and
capital gains on housing to the value of housing in the previous period.
Firms
To rein in the complexity arising from the three class composition of the
household sector, firm analysis remains simple. Profits are residual and the
profit share is determined as a mark-up over unit labour costs. It is assumed that
firms invest in housing and produce a single capital good on demand so that
capital inventories are not taken into account. Furthermore, we assume that
firms retain part of their profits and distribute the rest to rentiers.
Output of the modelled economy is given by consumption spending of
households as well as investment in productive capital and housing. Wage bill
follows from a bargaining process and is defined according to an exogenously
given wage share of output. Wage rates of the working and the middle class
depend on the share of each group (Nw and Nm respectively) in total population.
Importantly, wages paid to rentiers are linked to a variable remuneration dependent on firms profits. The rentier wage premium is given by a premium
mw > 1 over the workers wage rate, the profit sharing element ℎ and exogenous
parameter ∈ , reflecting the relative importance of profit remuneration in
the wage rate determination (Dafermos/Papatheodorou 2015:13).
Investment is defined simply as the growth rate of capital stock. A fraction x
of investment spending is financed by equity issue.
28
Apart from productive capital, firms invest in housing, which depends on
the difference between housing demanded by rentiers and the middle class and
the available housing supply in the previous period. In every period, a stock of
houses remains unsold, depending on the change in the supply and demand for
housing among the middle class (note that the Tobinesque portfolio equation
implies that all houses demanded by rentiers are sold). Change in the price of
housing is given by the difference between the change in the demand for housing
by rentiers and the middle class and the change in supply of housing by firms.
Commercial banks
Since the aim of our model is to account for inequality determination in the age
of financialisation, commercial banks are envisaged as active profit-seeking
entities rather than passive intermediaries between debtors and creditors.
Profits of commercial banks are generated by charging higher interest rates on
loans than offered on deposits. A constant interest rate on deposits is assumed
for all households, defined as an exogenous premium �1 over a given central
bank interest rate. The interest rate on loans is set by charging an exogenous
premium �2 over the deposit rate. Commercial bank profits are thus derived as a
sum of interest payments on non-securitised mortgages of the middle class,
consumer loans of the working class and loans to rentiers, less interest payments
on deposits to households. All profits are transferred to rentier households, who
are the owners of all financial institutions.
Commercial banks accept deposits from the household sector. However,
each household group faces a different rate of interest depending on the
perception of their creditworthiness by banks. Interest on loans to the working
class is higher than the rate charged to the middle class and rentiers. This risk
premium depends on exogenous parameters 0 and 1, capturing institutional
conditions in financial markets, the debt to income ratio of the working class, and
their debt service ratio.
Importantly, part of mortgages taken out by the middle class are
securitised and sold to underwriters and their SPVs. The share of securitised
loans depends on an exogenous parameter s0 (capturing institutional conditions
29
such as the degree of financial regulation) and the target yield on mortgage-
based securities (MBS) (given by the past yield under the assumption of simple
adaptive expectations), adjusted by parameter s1.
Middle class loans are subject to a mortgage rate, defined as a spread over
the commercial bank lending rate. The mortgage spread depends positively on
parameter 0, the debt service ratio and the debt to income ratio of the middle
class adjusted by parameter 2, and negatively on the rate of return on MBS
adjusted by parameter 3.
SPVs/underwriters
The main role of the sector of SPVs and underwriters is to transform securitised
mortgages bought from commercial banks into mortgage-backed securities
(MBS). It is assumed that SPVs/underwriters pay no administrative fees to banks
for this transaction.
It is assumed that all MBS are sold to institutional investors without any
fee in the form of coupon payments at a coupon rate determined by an
exogenous spread over the mortgage rate. Consequently, the SPVs/underwriters
sector accumulates no profits. Importantly, MBS issued are assumed to be of the single pass-through type rather than consisting of various pooled MBS cf. Nikolaidi 2015:4).
Institutional investors
The institutional investors sector includes entities such as pension funds, mutual
funds, hedge funds, insurance companies, and investment banks (cf. Davis 2003).
They earn revenue from holding MBS and finance their operations by issuing
shares, which are purchased by rentiers. For simplicity, a constant price of
shares equal to $1 is assumed. Demand for MBS follows the portfolio principle,
where the return on MBS depends on the yield and capital gains on MBS.
Institutional investors accumulate profits equal to the coupon payments
from SPVs/underwriters, which are entirely distributed to rentiers. Return on institutional investors shares is given as the ratio of their profits to shares demanded by rentiers in the previous period.
30
Simulations
The model is calibrated to the US economy. The main objective of the simulation
exercise is to examine the impact of the proposed model on inequality patters.
Specifically, we analyse how changes in household balance sheet composition
and leverage affect quantitative measures of income inequality such as the Gini
index, the Atkinson index (with inequality aversion parameter �=2) and the
squared coefficient of variation. While the Gini and Atkinson indices range
between 0 and 1, squared coefficient of variation ranges from 0 to infinity. In all
indices, higher value indicates higher inequality level. This follows the
benchmark exercise outlined in Dafermos/Papatheodorou (2015) where the
choice of these three inequality measures is motivated by their different
sensitivity to inequality in different moments of the distribution (the middle, the
bottom and the top of the distribution respectively).
In addition, we calculate the Theil T index to capture wealth inequality.
This is because the other measures of income inequality incorporated in our
model cannot be readily adapted to the distribution of wealth due to possible
negative net worth values (cf. Cowell 2009:72). Theil T index is a generalised
entropy measure of inequality, ranging between 0 and infinity, higher value
corresponding to a higher inequality level (World Bank 2005). To compare the
distributions of income and wealth in our model, we also compute the Theil T
index for income.
It is expected that the balance sheet heterogeneity should produce more
acute long-run polarisation of income. This is because the inclusion of wealth in
the model creates forces which pull the upper class even further away from the
rest of the distribution, drowning the middle and working class in debt.
Consideration of the different types of debt, which is reflected in our distinction
between the working and the middle class, could also explain the middle class
meltdown in countries like USA and should reproduce the illusion of short-run
prosperity for the middle class in the run up to the crisis.
Firstly, a full model, which is outlined above, is simulated for 100 periods.
For clarity, simulation results are presented from period 20 onwards to allow for
31
adjustment of the system to a steady state. The steady state is defined as a
situation where all variables in the economy grow at the same rate, given by the
exogenous growth rate of capital gk. Results for the income Gini coefficient, the
Atkinson index and the squared coefficient of variation as well as for the Theil T
index for income and wealth are presented. Additionally, we report the three
measures of leverage for each household group.
Secondly, we compare the above results of the full model with reduced
form specification without the novel features introduced in our model, namely
rentier wage, rentier debt and securitisation.
V. Results
Simulations of the model produce a consistent result of increasing inequality
according to all measures. The Gini index in the model tends towards 0.6, which
is close to the actual 2006 value recorded in USA (see introduction). The
Atkinson index tends towards 0.45 and the squared coefficient variation towards
1.25 (Fig.8, panel A). Furthermore, model results show that wealth inequality is
higher than income inequality, which reproduces the stylised fact outlined in the
introduction (panel B in fig.8). This is measured using Theil T indices for both
income and wealth to maintain comparability.
Interesting results follow from simulating various financial fragility
measures. Looking at the debt-to-asset ratio, the working class is the most
leveraged, with the ratio stabilising at 0.5 (panel D in fig.8). The ratio for the
middle and the rentier class reaches 0.4, with rentiers being slightly less
leveraged than the middle class. This is because of the presence of housing on the
asset side of the middle class balance sheet. However, although the ratio for
rentiers reaches similar values as the middle class, rentiers do not face the
negative consequences of large debt holdings as the middle and the working
class due to high returns to their assets and diverse income sources. This is best
highlighted by examination of the debt service to income ratio (panel C, fig.8).
This measure shows clearly that debt is the most burdensome for the working
class, as debt repayments in each period correspond to 8.7% of their income.
Similarly, despite lower debt-to-asset ratio of the middle class, their debt
repayment ratio of 0.077 puts them closer to the working class in terms of their
32
balance sheet fragility. Conversely, due to multiple income sources and large
high-yielding asset holdings rentiers debt service corresponds to only 3.8% of
their income in each period.
In contrast, an opposite picture emerges from the debt-to-income ratio
analysis (panel E, fig.8). By this measure, the working class is leveraged the least,
with the ratio reaching 0.87. The ratio for the middle class stabilises at 1.3 and
for rentiers at 1.4. This order is surprising and does not corresponds to the debt-
to-income ratios found in the household survey data. Hence, while our model
reproduces the empirical fact that debt of rentiers is large, it either understates
the demand for loans by the working and the middle class or it overstates their
income. This may be either because the part of the wage share accruing to the
working and the middle class is overstated in our model compared to the real
world or because the impact of securitisation on household indebtedness does
not generate enough supply and demand for debt among the lower and middle
income groups. Both of these explanations are related to the aggregate nature of
the SFCM method and the inability to decompose the imposed aggregated
structures. Consequently, in the context of our model it is important to examine
household financial fragility holistically, as each of the commonly used measures provides different information on households capacity to handle financial distress.
33
Figure 8. Simulation results – full model
De
bt
serv
ice
to
in
com
e r
ati
o
Working class Middle class Rentier class
(A)
(B) (C)
(D) (E)
Gini index Atkinson index Squared coefficient of variation
Working class Middle class Rentier class
Working class Middle class Rentier class
Theil wealth Theil income
34
Secondly, we present the simulation results of a reduced form model to
highlight the importance of the novel features presented in our model for
analysing inequality. Figure 9 reports the simulation results of the model with a pure capitalist class, i.e. it is assumed in line with the existing literature that
rentiers earn only capital income and no wages. In this case, the overall trends in
the indicators reported in the full model are replicated. However, all measures of
inequality are understated. The Gini index for income is lower at 0.5, the
Atkinson index decreases to 0.37 and the squared coefficient of variation falls to
0.8 (panel A, fig.9). Similarly, the reported Theil T indices are lower, with values
of 0.024 and 0.013 for wealth and income respectively (panel B). The leverage
indicators remain largely unchanged, although the debt-to-asset ratio of the
rentier class increases slightly to 0.4 (panel D).
Similar results follow from a reduced form specification without neither
wage nor debt holdings for rentiers (fig.10). The Gini index and the Atkinson
index decrease to 0.5 and 0.38 respectively, while the squared coefficient of
variation falls to 0.85 (panel A). The values for the Theil indices for wealth and
income decrease to 0.028 and 0.016 respectively (panel B). Since no rentier debt
is considered, leverage ratios are only reported for the working and the middle
class. The values for both groups remain similar to the full specification, although
the debt service to income ratio for the middle class decreases slightly to 0.074
(panel C).
Finally, we present results from a reduced specification without
securitisation (fig.11). In this case, mortgages are not securitised and commercial
banks are the only financial institutions in the model. The asset side of rentiers balance sheet is reduced as they do not earn profits of institutional investors nor
do they purchase shares of securitised assets. Similarly to previous reduced
specification results, inequality measures are lower than in the full model. The
Gini index settles at 0.54, the Atkinson index falls to 0.41 and the squared
coefficient of variation falls to 0.99 (panel A, fig.11). The Theil T indices for
wealth and income stabilise at 0.026 and 0.013 respectively (panel B). In terms
of leverage measures, the debt service to income ratio falls slightly to 0.083 for
the working class (panel C).
35
The comparison of the reduced specification results with the full model
shows clearly that heterogeneity of household balance sheets along the
distribution matters for inequality. Firstly, it is striking that factors commonly
omitted in the theoretical literature, such as rentier debt and rentier wage, have
an important impact on inequality measures, as is shown by the higher values of
all inequality indicators in the full model than in the reduced specifications.
Secondly, the results reveal that in light of household balance sheet
heterogeneity leverage of different income groups needs to be analysed
holistically. This is because each measure of financial fragility captures a
different aspect of indebtedness and thus does not represent the true capacity of
households to handle financial distress when analysed by itself. Consequently,
the results of our model strongly show that the theory of inequality in 21st
century in the context of financial sector transformation needs to take into
account different balance sheet positions of households and the associated
implications for financial distress.
36
Figure 9. Simulation results – pure capitalists specification (A)
(B) (C)
(D) (E)
Theil wealth Theil income Working class
Middle class Rentier class
Working class Middle class Rentier class
Working class Middle class Rentier class
Gini index Atkinson index Squared coefficient of variation
37
Figure 10. Simulation results – pure capitalist specification with no rentier debt
Gini index Atkinson index
(A)
(B) (C)
(D) (E)
Working class Middle class
Working class Middle class
Working class Middle class
Theil wealth Theil income
Gini index Atkinson index Squared coefficient of variation
38
Figure 11. Simulation results – reduced specification without securitisation
(A)
(B) (C)
(D)
Theil wealth Theil income
Working class Middle class Rentier class
Working class Middle class Rentier class
Working class Middle class Rentier class
(E)
Gini index Atkinson index Squared coefficient of variation
39
VI. Sensitivity analysis
In order to test the robustness of our finding that greater household balance
heterogeneity contributes to inequality, a range of sensitivity test is performed to
examine the volatility of the proposed model to specific parameter values. 12
parameters are identified as crucial to model results, reflecting the underlying
assumptions about economic behaviour. Two types of sensitivity analysis are
conducted – a univariate test, where the full model scenario is re-run changing
only one parameter at a time while leaving the others constant, and a
multivariate test, where variation in full model result is assessed by changing all
parameter values simultaneously. The model outcome is then seen as robust if
the values of key variables do not change significantly despite substantial
variation in parameter calibration.
Univariate sensitivity test
The choice of the sensitivity test values is motivated by changes in the actual
economic environment in the US after the 2007 crisis. All parameter values are
subsequently shocked in period 50.
One of the key distributional variables in our model is the central bank
interest rate rcb, as it constitutes the baseline for the interest rates on loans and
deposits set by commercial banks. In the sensitivity analysis, central bank
interests rate is shocked to increase from 0.25% to 0.5%. This corresponds to the
actual change in the interest rate level adopted by the Fed at its December 2015
meeting. Thus, apart from assessing the robustness of the model result, this
exercise also allows us to examine the impact of monetary policy on inequality
levels in the modelled economy.
Another parameter relevant for the interest rate level is 0, reflecting
institutional conditions in the lending market. Higher level of 0 indicates stricter
lending standards among the intermediaries, contributing to a larger transfer of
income from low- and middle-income households to rentiers via the banking
sector. In the sensitivity analysis, the value of 0 is increased from 0.03 to 0.04.
40
Furthermore, the exogenously given share of wages is important for
distribution as it determines the portion of national income going to wage
payments. The wage share parameter sw is decreased from 57% to 50%,
additionally allowing us to analyse the impact of falling wage share on inequality
levels.
Another parameter crucial to the model dynamics is s0, which captures
the institutional conditions in financial markets in the equation defining the
share of securitised mortgages. The greater the proportion of securitised
mortgages the higher the transfer of the middle class wealth to rentiers via
securitisation. To examine volatility of the model outcome to the value of this
parameter, s0 is decreased from 0.6 to 0.4, reflecting slowdown in the mortgage
securitisation market after the crisis.
Further parameter influencing the distribution of income in our model is
the firm profit retention rate sf. Higher value of this parameter is likely to prevail
in recessionary periods as firms are more credit constrained. The value of sf
increases to 0.5, which corresponds to the actual post-crisis value.
Additionally, the sensitivity analysis assesses model robustness by
decreasing the rentier portfolio equation parameter λ30 from 0.33 to 0.25.
Importantly, due to the adding-up constraint requiring λ10, λ20 and λ30 to sum up
to unity (cf. Godley/Lavoie 2007), fall in λ30 necessitates a simultaneous rise in
one of the remaining two values. It is assumed that λ10 increases to 0.5. Rise in
the value of λ10 indicates greater preference for firm equities among rentiers and
hence smaller demand for securitised assets among institutional investors.
Choice of these parameters is once again motivated by the fall in demand for
mortgage-backed securities after the 2007 crisis.
In addition to the above parameters directly affecting the distribution of
income and wealth in our model we consider five parameters important for the
overall model dynamics. Firstly, we test model sensitivity to parameter β,
capturing household borrowing and lending norms. The value of β is decreased
from 0.1 to 0.05, reflecting more stringent lending conditions after the 2007
crisis. Secondly, propensity to consume out of wealth c3 is increased from 0.1 to
0.2, maintaining the assumption that each household group consumes the same
41
proportion of its wealth. Thirdly, parameters h1 and h3 are decreased from 0.5 to
0.1, indicating a slowdown of housing supply by firms and a brake on the house
price growth respectively. Finally, parameter �10 in the institutional investors portfolio equation is decreased from 0.3 to 0.1, suggesting fallings demand of
institutional investors for MBS.
Overall, the univariate sensitivity analysis shows that our model results
are robust to changes in most of the key parameters. When the values of rcb, s0
and �10 are shocked in period 50, the model outcome exhibits no variation from
the baseline full model specification. Similarly, following a shock to the values of
h1, h3, 0, λ20 and λ30 model results do not change their long-term steady state
values, experiencing only very slight variations in the short-run.
However, the model outcome is sensitive to the values of parameters sw, sf,
β and c3. Fall in the wage share sw leads to higher steady-state levels of inequality,
with largest increases in the squared coefficient of variation for income (value of
1.6) and the Theil T index for wealth (0.033). The Gini index, the Atkinson index
and the Theil T index for income rise to 0.7, 0.53 and 0.02 respectively. Leverage
measures increase for the working and the middle class in the short run,
returning to their pre-shock levels in the long run. Conversely, leverage
indicators for rentiers decrease slightly following the shock, retaining their
original value in the long run.
In contrast, fall in the value of sf is associated with lower levels of
inequality but similar leverage levels in the model. Squared coefficient of
variation decreases the most to 0.73, with smaller fall in the Gini coefficient to
0.46 and the Theil T index for income to 0.01. The Atkinson index decreases to
0.4 and the Theil T index for wealth settles at 0.024. This occurs as the lower
level of distributed profits reduces the rentier income. Thus the decrease in
inequality is driven by redistribution at the top. The long run values of leverage
measures remain close to the baseline specification, with the middle class debt-
service to income ratio falling from 7.7% to 7.3%.
Furthermore, fall in parameter β has little influence on the levels of
inequality but it does result in lower values of the leverage measures. The debt
service to income ratio decreases to 0.051, 0.05 and 0.025 for the working, the
42
middle and the rentier class respectively. The debt to asset ratio falls to 0.43,
0.31 and 0.29 respectively for the working, the middle and the rentier class. The
debt to income ratio settles at a lower level of 0.66, 0.91 and 0.93 for the working
class, the middle class and rentiers.
Finally, following the change in c3 all reported inequality and leverage
measures experience an initial decrease, followed by a rise and a subsequent fall.
The Gini index, Atkinson index and the squared coefficient of variation settle at a
slightly higher level of 0.63, 0.46 and 1.3 respectively. Similarly, the Theil T
indices for income and wealth rise to 0.017 and 0.029 respectively. The debt
service to income ratio increased only for rentiers, rising from 3.8% to 3.9%. The
debt to asset ratio increases from 0.5 to 0.64 for the working class as well as
from 0.4 to 0.57 and 0.55 for the middle and rentier class respectively. Finally,
debt to income ratio rises across households, from 0.87 to 0.89 for the working
class, from 1.3 to 1.38 for the middle class and from 1.4 to 1.45 for rentiers.
Multivariate sensitivity test
Having examined the sensitivity of the model result to changes in individual
parameters, we proceed to analyse its variation to changes in all chosen
parameters simultaneously. Since at the present stage of the analysis the choice
of sensitivity values for different parameters is not random, one multivariate
scenario corresponding to the post-crisis conditions in USA is considered,
maintaining consistency across parameter changes.
Introducing shocks to parameter values in period 50, the model is able to
reproduce the overall trends in economic dynamics experienced by USA after the
2007 crisis. Firstly, the change in parameter values is associated with a recession
in the model as following an initial acceleration from 2.5% to 16.5%, the steady
state growth rate of output falls to –6.5% and gradually returns to its pre-shock
level after around 20 periods.
Secondly, three out of four income inequality measures indicate falling
income inequality in the periods following the shock, settling at a lower steady
state level after approximately 10 periods. Squared coefficient of variation
experiences the largest fall from 1.25 to 0.88. The Gini index decreases by a
43
smaller amount, from 0.6 to 0.49. Similarly, the Theil T index for income falls
from 0.016 to 0.01. In contrast, after the shock the Atkinson index for income
increases from 0.45 to 0.49. This suggest that changes in inequality occurring in
USA after the crisis have been different for various income groups. Sensitivity of
the squared coefficient of variation and the Gini index to changes at the top and
in the middle of the distribution respectively indicates that the fall in income
inequality post-2007 has been driven by its decrease among the top income
group, and less so by the middle. Conversely, sensitivity of the Atkinson index to
changes at the bottom of the distribution suggests that income inequality has
increased as the lowest income groups experienced a larger fall in incomes than
the rest. Furthermore, the model reproduces the fact that, unlike income, wealth
inequality as measured by the Theil T index increased after the crisis from 0.026
to 0.028, peaking at 0.033 in the period immediately after the shock.
Thirdly, among the leverage measures, debt service to income ratio and
debt to income ratio indicate overall deleveraging of households, while debt to
asset ratio indicates rising financial distress across household groups.
Specifically, the debt service to income ratio falls the most for the working class,
settling at a lower level of 5.2% after an initial increase from 8.5% to 15%. The
new steady state level of this ratio is significantly lower that the 8.7% result in
the baseline full model specification. The trend in the debt service to income
ratio is similar for the middle class, albeit of smaller magnitude. After a peak of
10% after the shock, the ratio settles at 5.3%, down from 7.7% in the baseline.
Similarly, the ratio for rentiers falls to 3%, after an initial increase to 5%, which
is also lower than the baseline result of 3.8%.
Furthermore, the debt to income ratio dips slightly, peaks and
subsequently falls across all household groups. This variation is the highest for
the rentiers, with the ratio rising to 1.75 before settling at a new value 0.98,
lower compared to the baseline result 1.4. The debt to income ratio for the
middle class peaks at 1.45 and subsequently decreases to 0.89 (1.3 in the full
model specification). Similarly, the ratio for the working class rises to 0.9 before
falling to 0.65, down from 0.87 in the baseline specification. Consequently, the
multivariate sensitivity analysis preserves the counterintuitive result regarding
44
the order of magnitude of the debt to income ratio for different household
groups encountered in the full model simulation.
In contrast, the debt to asset ratio increases across all household groups
relative to the baseline simulation result. Following the shock, the ratio initially
peaks at approximately 0.6 for all households, before settling at a new level equal
to 0.55, 0.47 and 0.45 for the working, middle and rentier class respectively.
These values are higher than the baseline result of 0.5 for the working class and
0.4 for the middle and rentier class. Higher values of the debt to asset ratio
suggest that following a simultaneous variation of all key parameter values, loans
are falling less rapidly than the value of assets across all households.
Overall, the sensitivity analysis shows that the model results presented in
the previous section are robust to changes in most of the key parameters,
particularly in the long run. Sensitivity of the model outcome to changes in the
wage share and the profit retention ratio as well as household lending norms and
the marginal propensity to consume out of wealth suggest that not only income
but also wealth channels are important for inequality determination. The
proposed model setout, putting emphasis on balance sheet heterogeneity of
different household groups and a more active financial sector does well in
explaining trends in inequality in USA before and after the 2007 crisis.
VII. Conclusion and future work
Summary
The model outline presented here constitutes a first attempt of the author to
develop a theoretical model of inequality in the age of financialisation. SFCM is
adopted to account for the interactions between the financial and real sector and
their impact on the distribution of income and wealth in a financialised economy.
Unlike the existing functional distribution literature, in the current model
inequality is understood in terms of differential balance sheet and net wealth
structures among various income groups in the society. It is argued that this is a
more suitable approach to analysing inequality in times of financial sector transformation as the traditionally envisaged groups of workers and capitalists in the Post Keynesian literature became more heterogeneous since
45
1980s. While low- and middle-income households became actively involved in
financial markets through securitisation, the rich captured an increasing share of
income and economic power due to high returns to their wealth in result of
financial innovation and deregulation as well as high incomes received in the
financial sector. Thus, the innovation of our model is to reinterpret the groups of
workers and rentiers as well as to reconceptualise the middle class and its role in
inequality trends since 1980s.
The main distributional channels in our model emerge through credit
provision to the working and the middle class (firstly, because the interest
payments by the latter are ultimately received by the rentiers, and secondly,
because loans to the working and middle class are transformed into derivative
instruments held by rentiers); the housing sector (directly through rent
payments by the working class households to rentiers and indirectly through
interest payments on mortgages); and inequality is also reflected in the relative
consumption undertaken by the middle class.
Future work
At this early stage, the model is necessarily simplistic. In the near future, I aim to
extend the model so as to account for important processes influencing
distribution in the age of financial sector transformation, which could not be
considered at present due to their novelty and complexity.
The most innovative aspect which will be considered in the model is the
addition of more complex microeconomic behaviour using agent-based
modelling (ABM) techniques (cf. Gaffeo et al. 2007, Delli Gatti et al. 2011, Caiani
et al. 2016). The present SFCM representation is too aggregate to study changes
in the shape of the wealth and income distribution in detail. This is because its
macroeconomic character imposes a top-down structure of behaviour in the
model. This macroeconomic rigour is certainly important as shown by
Dafermos/Papatheodorou (2015) since aggregate mechanisms provide
important feedback mechanisms into the distribution of income, which could
give misleading outlook on the dynamics of inequality overtime if omitted.
However, it may not be a suitable starting point for the analysis of inequality
based on understanding what determines portfolio decisions of households and
46
hence their balance sheet structures in the times of financialisation. Agent-based
dynamics could inform what drives household behaviour when interacting with
different social groups, employers and the financial sector. It could also help to
correct the puzzle regarding the opposite than expected order of the debt-to-
income ratios in the present model.
Furthermore, I will analyse the influence of specific balance sheet
structures on income shares of different household groups. Decomposition
technique could be adopted to reveal which aspect of balance sheet inequality
has the biggest impact on distribution. This issue remains ambiguous in the
literature. While the Post Keynesian theories of inequality reviewed earlier
suggest that it is debt which exacerbates the distribution of income away from
workers, empirical studies often find that it is the asset side of the balance sheets
that contributes more to inequality (cf. Fredriksen 2012). Similar conclusion can
be drawn from Piketty, according to whom high capital income from assets held
by the top 1% drives economic inequality. The unique setout of our model would
be capable of testing these competing claims.
Further extension to the present model will concern the inclusion of
social transfers. This is particularly important in the recent years, as due to
stagnating incomes and worsening working conditions (due to globalisation,
privatisation and labour market liberalisation), many especially low income
households (corresponding to the working class in our model) rely increasingly
on social security in their income. Furthermore, it would shed light on how
different taxation policies influence inequality.
47
References
Alvaredo, F., Atkinson, A.B., Piketty, T. and Saez, E., The World Top Incomes
3/08/2015]; Ampudia M., van Vlokhoven, (., and Zochowski, D. Financial fragility of euro area households , Working paper , European Central Bank; Atkinson, A.B. and Brandolini, A. On the identification of the middle class , EC)NEQ Working Paper -217, Bivens, J., Gould, E., Mishel, L. and Shierholz, (. Raising America s Pay. Why )t s Our Central Economic Policy Challenge , Economics Policy )nstitute Briefing
Paper #378; Bhaduri, A., and Marglin, S. Unemployment and the real wage: the economic basis for contesting political ideologies , Cambridge Journal of Economics, , pp. 375-393; Boushey, (., and Weller, C. (as Growing )nequality Contributed to Rising (ousehold Economic Distress? , Review of Political Economy, 20(1), pp. 1-22;
Caiani, A., Godin, A., Caverzasi, E., Gallegati, M., Kinsella, S., Stiglitz, J.E. (2016) Agent based-stock flow consistent macroeconomics: Towards a benchmark model , Journal of Economic Dynamics and Control, 69, pp. 375-408; Caverzasi, E. and Godin, A. Stock-flow Consistent Modeling through the
Ages , The Levy Economics )nstitute Working Paper No. ; Caverzasi, E. and Godin, A. Financialization and the sub-prime crisis: a
Stock-Flow Consistent model , Working paper; Congressional Budget Office (2014), The Distribution of Household Income and
Federal Taxes, 2011. Available at: https://www.cbo.gov/publication/49440
Copeland, M. A. ( Social Accounting for Money Flows , The Accounting
Review, 24(3), pp. 254–264;
48
Cynamon, B.Z. and Fazzari, S.M. (ousehold Debt in the Consumer Age: Source of Growth—Risk of Collapse , Capitalism and Society, 3(2), pp. 1-30;
Dafermos, Y. and Papatheodorou, C. Linking functional with personal income distribution: a stock-flow consistent approach , )nternational Review of Applied Economics, 29(6), pp. 1-29; Davis, E.P. )nstitutional )nvestors, Financial Market Efficiency, and Financial Stability , E)B Papers, , pp. -107;
Delli Gatti D., Desiderio, S., Gaffeo, E., Cirillo, P. and Gallegati, M. (eds) (2011)
Macroeconomics from the Bottom-up, Milan: Springer; Detzer, D. Financialisation, Debt and )nequality – Scenarios Based on a Stock Flow Consistent Model , )MK Working Paper No. / ; DeVaney, S.A. and Lytton, R.(. (ousehold insolvency: A review of household debt repayment, delinquency, and bankruptcy , Financial Services Review,
4(2), pp. 137-156;
Dos Santos, C.(. and Macedo e Silva, A.C. Revisiting and Connecting Marglin-Bhaduri and Minsky: An SFC Look at Financialization and Profit-led Growth , The Levy Economics Institute Working Paper No. 567;
Duesenberry, J.S. (1949) Income, Saving and the Theory of Consumption
Behaviour, Cambridge, MA: Harvard University Press;
Dufour, M. and Orhangazi, O. Growth and distribution after the – US financial crisis: who shouldered the burden of the crisis? , Review of Keynesian
Economics, 4(2), pp. 151-174; Dymski, G.A. Racial Exclusion and the Political Economy of the Subprime Crisis , Historical Materialism, 17, pp.149-179; Dymski, G.A., (ernandez, J. and Mohanty, L. Race, Gender, Power, and the US Subprime Mortgage and Foreclosure Crisis: A Meso Analysis", Feminist Economics,
19(3), pp.124-151;
Foster, J.B. and Holleman, (. The Financial Power Elite , The Monthly
Review, 62(1);
49
Frank, R.(., Levine, A.S. and Dijk, O. Expenditure Cascades ; Review of
Behavioral Economics, 1(1–2), pp 55-73; Fredriksen, K.B. Less )ncome )nequality and More Growth — Are they
Compatible? Part 6. The Distribution of Wealth , OECD Economics Department Working Papers No. 929;
Gaffeo, E., Catalano, M., Clementi, F., Delli Gatti, D., Gallegati, M. and Russo, A. Reflections on modern macroeconomics: Can we travel along a safer road? , Physica A, 382, pp. 89-97;
Goda, T., and Lysandrou, P. The Contribution of Wealth Concentration to the Subprime Crisis: A Quantitative Estimation , C)BS Working Paper No. ; Godley, W. and Lavoie, M. (2007) Monetary Economics. An Integrated Approach to
Credit, Money, Income, Production and Wealth, 1st Edition, Basingstoke: Palgrave
Macmillan;
Gould, E. (2016) Wage inequality continued its 35-year rise in 2015 , Economic Policy Institute Briefing Paper #421; (ein, E. Financialisation in a comparative static, stock-flow consistent
Post-Kaleckian distribution and growth model , Macroeconomic Policy )nstitute )MK Studies 21/2008; (ein, E. Finance-dominated capitalism and re-distribution of income: a Kaleckian perspective , Cambridge Journal of Economics, , pp. 907-934; (ein, E. and Van Treeck, T. Financialisation and Rising Shareholder Power in Kaleckian/Post-Kaleckian Models of Distribution and Growth , Review of Political Economy, 22(2), pp. 205-233;
Kalecki, M. (1971) Selected Essays on the Dynamics of the Capitalist Economy,
Cambridge: CUP; Kapeller, J. and Schuetz, B. Conspicuous Consumption, )nequality and Debt: The Nature of Consumption-driven Profit-led regimes , Metroeconomica, 66:1, pp.
51-70;
50
Kaplan, S.N. and Rauh, J. (2010) Wall Street and Main Street: What Contributes to the Rise in the (ighest )ncomes? , The Review of Financial Studies, , pp. -
1050;
Keister, L.A. and Lee, (.Y. The One Percent: Top Incomes and Wealth in
Sociological Research , Social Currents, 1(1), pp. 13-24; Krippner, G. The financialization of the American economy , Socio-
Economic Review, 3, pp.173-208; Krueger, D. and Perri, F. Does )ncome )nequality Lead to Consumption )nequality? Evidence and Theory , Review of Economic Studies, 73, pp. 163-193;
Minsky, H.P. (1986) Stabilizing an Unstable Economy, Yale: Yale University Press; Nikolaidi, M. Securitisation, wage stagnation and financial fragility: a stock-flow consistent perspective , Greenwich Papers in Political Economy; Onaran, O., Stockhammer, E. and Grafl, L. Financialisation, income distribution and aggregate demand in the USA , Cambridge Journal of Economics, , pp. 637-661; Palley, T.). Financialization: What )t )s and Why )t Matters , The Levy
Economics Institute Working Paper No. 525; Palley, T.). Wealth and wealth distribution in the neo-Kaleckian growth model , Journal of Post Keynesian Economics, , pp. -474; Palley, T.). Enriching the neo-Kaleckian growth model: Nonlinearities, political economy, and financial factors , PER) Working Paper No. ; Pew Charitable Trust Payday Lending in America: Who Borrows, Where They Borrow, and Why , Report; Pew Research Center The American Middle Class )s Losing Ground ,
Report;
Piketty, T. (2014) Capital in the Twenty-First Century, Harvard: HUP;
Piketty, T. and Zucman, G. Capital is back: wealth-income ratios in rich
countries 1700– , The Quarterly Journal of Economics, pp. 1255-1310;
51
Pollin, R. and (eintz, J, Study of U.S. Financial System", FESSUD Studies in Financial Systems No. 10;
Pressman, S. (2007) The Decline of the Middle Class: An )nternational Perspective , Journal of Economic Issues, 41(1), pp. 181-200;
Pressman, S. (2016) Understanding Piketty’s Capital in the Twenty-First Century,
Abingdon: Routledge;
Pressman, S. and Scott, R. Consumer Debt and the Measurement of Poverty and )nequality in the US , Review of Social Economy, 67(2), pp. 127-148;
PWC , Precious Plastic. How Britons fell back in love with borrowing , Report; Ravina, E. (abit formation and keeping up with the Joneses: evidence from micro data . Available at:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=928248 Sawyer, M. What is Financialisation? , International Journal of Political
Economy, 42(4), pp. 5-18; Sawyer, M. and Passarella Veronese, M. The Monetary Circuit in the age of Financialisation , University of Leeds Working Paper;
Scott, R. and Pressman, S. )nadequate (ousehold Deleveraging: )ncome, Debt and Social Provisioning , Journal of Economic Issues, 49(2), pp. 483-492; Setterfield, M. and Kim, Y.K. Debt Servicing, Aggregate Consumption, and Growth , Trinity College Department of Economics Working Paper -16;
Survey of Consumer Finances, 1989-2013, Federal Reserve. Available at:
http://www.federalreserve.gov/econresdata/scf/scfindex.htm Van Treeck, T. A synthetic stock-flow consistent macroeconomic model of financialisation , Cambridge Journal of Economics, , pp. -493;
Veblen, T. (1899) The Theory of the Leisure Class, New York: Macmillan; Wolff, E.N. (ousehold wealth trends in the United States, 1983- , Oxford Review of Economic Policy, 30(1), pp. 21-43. World Bank )nequality Measures , Chapter , Poverty Manual;
Young, B. The Gendered Dimension of Money, Finance and the Subprime Crisis . )n: Bauhardt, C. and Caglar, G. eds) Gender and Economics. Feministische Kritik
der politischen Oekonomie, Wiesbaden: VS Verlag für Sozialwissenschaften; Zezza, G. U.S. growth, the housing market, and the distribution of income , Journal of Post Keynesian Economics, , pp. -401.
