Transcript The Financial Kuznets Curve

1
The Financial Kuznets Curve:
Financial Deepening and
Inequality in the Euro Area
DONATELLA BAIARDI
CLAUDIO MORANA
DEMS – DEPARTMENT OF ECONOMICS, MANAGEMENT AND STATISTICS
UNIVERSITY MILANO-BICOCCA - ITALY
2
Introduction - Kuznets curve
• The Kuznets curve describes the inverse-U-shaped relationship
between income distribution and economic growth. This idea can be
traced back to Kuznets (Amer. Econ. Rev., 1955), in a context where
economic development is the main determinant in the evolution of
income disparities.
Turning Point
Economic
Inequality
Economic
Development
3
Introduction - Kuznets curve
• More specifically, at the early stages of development, when
per capita income is low and the economy is principally based
on agricultural activities, inequality increases because workers
move from the larger agricultural sector to the smaller
industrial one, where wages are higher. Therefore, through
urbanization, the society becomes more unequal.
• Then, when higher standards of living are achieved (post-industrial or service economy), a subsequent fall of income
differentials is verified since wages of lower-income groups
rise, and overall income inequality narrows.
4
Literature review
• More importantly, this inverse-U shaped relationship between
income inequality and the level of economic development is
strictly linked to the shift of the economy from an
unsophisticated to a modern financial system,
characterized by advanced banking activities and stock
markets (Greenwood and Jovanovic, J. Polit. Economy,
1990).
5
Literature review
• In particular, at the early stages of development, economic
inequality increases over time since only the rich initially
benefit from better financial markets, while the poor have to
rely on informal, family connections for funding (Banerjee
and Newman, J. Polit. Economy, 1993);
• On the other hand, once the diffusion of financial
intermediation has sufficiently progressed, this trend reverses
(Greenwood and Jovanovic, 1990; Smith, J. Money, Credit,
Banking, 2003; Deidda, J. Monet. Econ., 2006; Kim and Lin,
J. Compar. Econ., 2011).
6
Literature review
 In the long-run, financial development influences directly
and indirectly economic inequality;
 With regard to the direct influence of financial system on
inequality:
• Li et al. (Econ. J., 1998), Clarke et al. (Southern Econ. J.,
2006), Beck et al. (J. Finan. Econ., 2001; J Econ Growth,
2007) find that financial deepening leads to a more equal
income distribution.
• Kim and Lin (J. Compar. Econ., 2011) demonstrate that a
minimum level of financial development is a necessary
precondition for achieving reduction of income inequality
through financial development.
7
Literature review
 With regard to the indirect influence of financial system on
inequality:
• It is well-known in the empirical literature that financial
development has a positive long-run effect on economic
growth (see, for example, Loayza and Ranciére, J. Money,
Credit, Banking, 2006 and Demirgüç-Kunt and Levine, 2008
for a survey).
• Consequently, if the Kuznets curve exists, an increase of
financial development fosters economic growth with an
additional indirect negative effect in terms of income
inequality, which should be reduced.
8
The aim of this paper
• We propose a new and fresh specification of the traditional
Kuznets curve in order to directly conditioning its turning
point (i.e. where inequality begins to reduce as per capita
income continues to increase) to the level of financial
development
Financial Kuznets Curve
• This specification takes into consideration both the direct and
indirect effects of financial markets on inequality mentioned
above
9
The aim of this paper
We also propose an innovative method to obtain robust
estimates, making use of all information available ex-ante,
instead of compare ex-post different specifications;
We demonstrate that
• Our methodology allows us to obtain more efficient and
robust estimates;
10
The model
• Consider the following data-field specification of the Kuznets
curve
(1)
Ineqi    byi  cyi2   i
where the subscript i refers to the i-th country, i=1, …, N.
• Ineqi is a measure of income inequality, yi is a economic
development indicator (traditionally real per capita GDP), εi is
the iid error term.
• The Kuznets curve hypothesis is verified if Coefficients b and c
obey to the following restrictions:
b>0 and c<0
11
The model
• The turning point of the Kuznets curve is obtained by
maximizing Equation (1) with respect to yi :
yi
b

2c
(2)
• Following Bradford et al. (Economic Analysis and Policy,
2005), we differentiate Equation (1) with respect to time and
after substituting Equation (2), we obtain:
Ineqi
y
 (b  2cyi ) i   ( yi  yi ) gi
t
t
y
(3)
• where   2c  0 and gi  i is per capita income growth
rate in each country. t
12
The model
• Equation (3) suggests that, on average, the instantaneous
change in economic inequality depends on the income growth
rate gi and on the distance of yi from its turning point y*i;
• Moreover, assuming that gi >0, inequality increases when yi  yi*
and decreases when yi  yi*.
• The main advantage of the specification in (3) is that it avoids
conditioning on nonlinear transformations (e.g., squares and
cubes) of potentially nonstationary variables, like per capita
GDP, typically occurring in empirical analyses using the
traditional Kuznets curve specified by Equation (1).
13
The model – a Financial Kuznets Curve
• In order to explicitly capture the impact of financial
development on the turning point of the Kuznets curve, we
assume that
yi*  0  1Fi
(4)
where parameter λ0 is constant and Fi is an (average) measure
of the level of financial development for the i-th country.
• The inverse relationship between the turning point of the
Kuznets curve and financial development, is then given by the
sign of parameter λ1:
λ1<0
14
The model – a Financial Kuznets
Curve
• Equation (4) assumes that a country with more developed
financial markets reaches the turning point of the Kuznets
curve at a relatively lower income level than a country with a
less developed one.
• This is in line with the financial deepening-economic growth
linkage and with the assumption that a threshold level to be
passed before financial development might lead to a reduction
in inequality (Greenwood and Jovanovic, J. Polit. Economy,
1990; Lee, J. Devel. Econ., 1996; Acemoglu and Zilibotti, J.
Polit. Economy, 1997; Smith, 2003; Deidda, J. Monet. Econ.,
2006; Kim and Lin, J. Compar. Econ., 2011).
15
The model – a Financial Kuznets Curve
• We then substitute Equation (4) into Equation (3), in order to
get
Ineq
   yi   0  1Fi  gi
t
(5)
• After some computations (we integrate with respect to time and
we assume that the average values yi, gi and Fi are constant in
time), we obtained the following cross-sectional (long-term)
specification:
Ineqi  i  0  yi gi   1  gi    2  Fi gi    ' z i   i
Kuznets Curve
(6)
Financial Kuznets Curve
16
The model – a Financial Kuznets Curve
• In particular, the coefficients
β0=α<0 as required in order to verify the existence of the
Kuznets curve;
β2=-α λ1 <0 consistent with the hypothesis of an inverse
relationship between financial development and the turning
point of the Kuznets curve
17
The model – a Financial Kuznets Curve
• Moreover, the sign of the parameters λ0 and λ1 can be indirectly
verified as follows:
λ0=-β1/α
and
λ1=-β2/α <0
Financial Kuznets curve
18
Estimation Strategy
Three distinct measures of income inequality are used in the
empirical analysis:
• Income inequality is measured by means of market (GM) and
net income (GN) Gini coefficients, computed by using
household market and disposable income (post-tax, posttransfer), respectively.
• The variables GN and GM are provided by the Standardized
World Income Inequality Database (SWIID), recently updated
by Solt (2014).
• In the analysis we have also employed Gini index data (GW)
from the World Income Inequality Database (WIID).
19
Estimation Strategy
Moreover, three indicators of financial development are used
in the empirical analysis:
• the GDP share of domestic credit to the private sector (C).
• the GDP share of Liquid Liabilities (measured by M3
aggregates) (M),
• the GDP share of a broad market share price index (S).
• The source of the series is the International Financial
Statistics (IFS) database for C and S, while European Central
Bank (ECB) database for M.
20
Estimation Strategy
Equation (6) is estimated by running a single OLS regression
using all data jointly (ex-ante averaging strategy).
• In particular, we construct new dependent and independent
variables (called as Gi and Fi respectively) by stacking the
observations available in order to get that:
Gi= [GNi’ GMi’ GWi’]
and
Fi=[Ci’ Mi’ Si’]
21
Estimation Strategy
• Moreover, we demonstrate that
• Ex-ante and ex-post pooled OLS estimator are equivalent, with
weights determined according to the relative variation of the
candidate regressors;
• Ex-ante OLS estimator is consistent and asymptotically normal
under the same conditions of optimality of the disjoint OLS
estimator;
22
Estimation Results – Data description
 Dataset characteristics
• Our study is focused on the 19 countries (Austria, Belgium, Cyprus,
Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia,
Lithuania, Luxemburg, Malta, the Netherlands, Portugal, Slovakia,
Slovenia and Spain) that have currently adopted the Euro in the
time period 1985-2013.
• Our analysis leads to an accurate assessment of the various
dimensions through which financial deepening and income
inequality are interrelated, since the time spanned covers the most
important events of the Monetary Union history, as well as major
episodes of global economic downturns.
23
Estimation Results – Data description
 Dataset characteristics
• Since we are interest on the long-run relationship between
inequality, financial development and economic growth, a
cross-sectional database related to the 19 selected countries is
constructed.
• Since data available have annual frequency in the time period
1985-2013, cross-sectional data are computed by averaging
annual data.
24
Estimation Results – Data description
Dataset characteristics
• In the case of per capita GDP and of the three proxies of
financial markets, we follow a procedure similar to Bradford
et al (2005) by decomposing observed data into trend and
cycle. Then, we consider the trend level series at mid-sample
(i.e. at year 2000).
• Ex-ante pooled OLS estimator allows us to consider 171
instead of 19 observations;
25
Estimates Results – Linear Specification of Equation (6)
(1)
OLS
(2)
OLS
(3)
OLS
(4)
OLS
(5)
IV
SPREADi
-0.122**
(0.054)
-0.001
(0.056)
-0.395***
(0.080)
0.648***
(0.054)
-0.526***
(0.074)
-
TRADEi
-
-0.193***
(0.065)
-0.022
(0.060)
-0.327***
(0.098)
0.546***
(0.050)
-0.540***
(0.071)
-0.214***
(0.063)
-
URBANi
-
-
-0.327***
(0.105)
-0.114
(0.086)
-0.298***
(0.102)
0.573***
(0.056)
-0.571***
(0.068)
-0.233***
(0.062)
0.179*
(0.106)
-
Constant
0.000
(0.052)
0.000
(0.050)
0.000
(0.049)
-0.329***
(0.092)
-0.220***
(0.084)
-0.337***
(0.097)
0.576***
(0.056)
-0.533***
(0.065)
-0.337***
(0.074)
0.309***
(0.100)
-0.233***
(0.071)
0.000
(0.048)
-0.297*
(0.163)
-0.203**
(0.103)
-0.386*
(0.224)
0.584***
(0.066)
-0.544***
(0.078)
-0.332***
(0.081)
0.302***
(0.100)
-0.238***
(0.073)
0.000
(0.048)
R-squared
0.560
0.592
0.600
0.626
0.626
4.634 [0.000]
0.062 [0.96]
3.398 [0.065]
171
7.896 [0.000]
3.080 [0.214]
2.508 [0.113]
171
7.525 [0.000]
3.641 [0.162]
2.778 [0.096]
171
7.365 [0.000]
5.269 [0.072]
5.946 [0.015]
171
6.924 [0.000]
4.541 [0.103]
5.946 [0.015]
171
yigi
gi
Figi
DEPi
PEi
Diagnostic Statistics
Hetero
Normality
Endogeneity
Obs
26
Estimates Results – Logarithm Specification of Equation (6)
(1)
OLS
(2)
OLS
(3)
OLS
(4)
OLS
(5)
IV
SPREADi
-0.226***
(0.039)
-0.054
(0.051)
-0.289***
(0.066)
0.620***
(0.054)
-0.444***
(0.066)
-
TRADEi
-
-0.274***
(0.042)
-0.070
(0.052)
-0.243***
(0.070)
0.501***
(0.045)
-0.460***
(0.061)
-0.237***
(0.058)
-
-
-
Constant
-0.000
(0.050)
-0.000
(0.047)
-0.000
(0.047)
-0.254***
(0.042)
-0.106
(0.084)
-0.252***
(0.069)
0.492***
(0.050)
-0.439***
(0.061)
-0.289***
(0.073)
0.022
(0.077)
-0.102
(0.072)
-0.000
(0.047)
-0.211***
(0.077)
-0.032
(0.043)
-0.379***
(0.116)
0.527***
(0.039)
-0.508***
(0.059)
-0.222***
(0.061)
-
URBANi
-0.269***
(0.044)
-0.057
(0.073)
-0.244***
(0.071)
0.499***
(0.047)
-0.457***
(0.059)
-0.236***
(0.058)
-0.019
(0.070)
-
R-squared
0.593
0.634
0.634
0.639
0.626
4.605 [0.000]
0.977 [0.610]
1.271 [0.259]
171
8.182 [0.000]
2.725 [0.256]
1.857 [0.173]
171
8.000 [0.000]
2.179 [0.336]
1.790 [0.181]
171
12.219 [0.000]
2.537 [0.281]
1.482 [0.223]
171
7.103 [0.000]
5.296 [0.071]
1.857 [0.173]
171
yigi
gi
Figi
DEPi
PEi
-0.000
(0.048)
Diagnostic Statistics
Hetero
Normality
Endogeneity
Obs
27
Estimation Results
We find that:
• Coefficient β0 (see yigi) is negative and highly statistical
significant
Kuznets curve hypothesis
• Coefficient β2 (see Figi) is negative and highly statistical
significant
Financial Kuznets curve hypothesis
• Therefore, given the existence of the Kuznets curve, countries
with more developed financial markets reach their turning
points at a lower income level than less developed ones
28
Estimation Results
• Furthermore, given the strict relationship between financial
deepening, economic growth and consequently inequality, a
set of four dummies variables indicating whether the
country’s legal system is based on French, English, German
and Scandinavian traditions (see, for details, La Porta et al., J.
Polit. Economy, 1997, Levine et al., J. Monet. Econ., 2000 and
Laeven et al., J. Finan. Intermediation, 2015). Moreover, four
additional series are generated by the interaction between
these four dummies with the variable gi.
• Endogeneity test generally confirms that the direction of
causality comes from financial markets to economic growth.
29
Estimation Results - Robustness Checks
• Equation (6) has also been estimated in different contexts:
1) Three proxies of income inequality and three proxies of
financial development are alternatively considered.
Consequently, the dependent variable is alternatively GNi,
GMi and GWi, while the variable Figi is equal to Cigi, Migi
and Sigi (19 obs);
2) A unique measure of inequality Gi is used as dependent
variable, while three distinct proxies of financial
development (Cigi, Migi and Sigi) are considered separately
(57 obs);
3) Three distinct Gini indexes are used as dependent variables
(GNi, GMi and GWi) and a unique measure of financial
development Fi is considered (57 obs).
30
Estimation Results - Robustness Checks - KC
0
GM-F
GW-C
-0.05
GM-C
G-C
-0.1
GW-M
GM-S
-0.15
β0
GW-F
G-S
-0.2
-0.25
Linear
specification
GN-C
GN-F
G-F
GN-S
-0.3
-0.35
0.5
0.52
0.54
0.56
0.58
0.6
0.62
0.64
R-squared
Our results are
confirmed
0
-0.05
G-M
-0.1
β0
-0.15
-0.2
GN-M
GM-C
GW-C
GN-C GN-S
Logarithm
specification
GM-F
GM-S
G-C
GW-M
-0.25
-0.3
GN-F
G-F
GW-F
-0.35
-0.4
-0.45
0.59
0.61
0.63
0.65
R-squared
0.67
0.69
31
Estimation Results - Robustness Checks - FKC
0.05
-0.05
GM-S
GN-S
-0.25
β2
Linear
specification
GW-S
-0.15
G-S
GM-C
-0.35
G-F
GN-C
GW-F
-0.45
G-C
-0.55
GM-F
GN-F
GW-M
G-M
GW-C
-0.65
-0.75
0.45
0.5
0.55
0.6
GN-M
0.65
0.7
0.75
Our results are
confirmed
R-squared
0
-0.1
GM-S
GN-S
-0.2
Logarithm
specification
β2
G-F
GN-C
-0.3
G-C
GM-C
GM-F
GN-F
-0.4
GW-C
-0.5
G-M
GW-M
GN-M
-0.6
0.6
0.62
0.64
0.66
R-squared
0.68
0.7
32
Future research
Given the existence of the Kuznets curve in the Euro Area, we
would proceed as follows:
• Starting from the estimates derived by the logarithm
specification of Equation (6), parameters λ0 and λ1 can be
indirectly estimated.
• Then, by means of Equation (4) and of the definition of
elasticity, we would compute the turning point of the Kuznets
curve for each selected countries. In this way, we expect that
those countries which play a central role in the Euro Area
exhibit a lower level of income at correspondence of their
turning points with respect to the most peripheral nations.
33
Future research
Given the existence of the Kuznets curve in the Euro Area, we
would proceed as follows:
• Finally, we would like to deeply test the Financial Kuznets
Curve hypothesis by means of panel data instead of crosssection
• This can allow us to distinguish any possible short-run and
long-run effects of financial deepening on inequality.
• Many papers indeed note that, in the short-run, speculation
and volatility characterizing financial markets can increase
inequality (Rodrigues-Pose & Tselios, J Regional Sci 2009;
Roine et al., J Public Econ, 2009; Jauch & Watzka, CESIFO
Working Papers, 2012).