Transcript Political mechanism

What determines income
distribution and how income
distribution might affect growth
Branko Milanovic
World Bank Training Poverty and Inequality Analysis
Course
March 3, 2011
A. From Kuznets to Piketty:
determinants of income
distribution
1. Relationship between income
and inequality: the rise and fall of
the Kuznets hypothesis
Kuznets inverted U-shaped curve (defined in 1955)
• As income increases, inequality at
first goes up and then declines
• “It seems plausible to assume that in the process of growth, the
earlier periods are characterized by…forces that may have
widened the inequality…for a while because of the rapid growth
of the non-A [non-agricultural] sector and wider inequality
within it.
• It is even more plausible to argue that the recent narrowing in
income inequality observed in the developed countries was due
to a combination of • the narrowing inter-sectoral inequalities in
product per worker, • the decline in the share of property
incomes in total incomes of households, and • the institutional
changes that reflect decisions concerning social security and full
employment."
Kuznets curve: history
• Evidence for Kuznets curve in cross-sectional
data analyzed in the 1970s, 1980s (Paukert,
Lecaillon, Koeble & Thomas)
• More than 90% of pooled time-series and crosssectional Gini variability is due to differences
between countries => factors that determine
country inequality are stable (Li, Squire, Zhou)
• Elusive evidence in time-series (Oshima)
• Strong historical evidence for Western Europe
before and during Industrial revolution (Lindert
and Willianson, van Zanden, Prados)
• Modifications of the Kuznets curve: “strong” and
“weak” formulations
The same idea; Tocqueville 120
years earlier
• If one looks closely at what has happened to the
world since the beginning of society, it is easy to
see that equality is prevalent only at the
historical poles of civilization. Savages are equal
because they are equally weak and ignorant.
Very civilized men can all become equal
because they all have at their disposal similar
means of attaining comfort and happiness.
Between these two extremes is found inequality
of condition, wealth, knowledge-the power of the
few, the poverty, ignorance, and weakness of all
the rest. (Memoir on pauperism, 1835).
• General formulation (used by Ahluwalia
1976)
Giniit  o   1 ln Yit   2(ln Yit)   kZit  eit
2
k
• We expect β1>0 and β2<0
• Control variables include socialist dummy,
government transfers, share of state
sector employment, openness, age
structure of population (Milanovic 1994;
Williamson and Higgins 1999)
60
Relationship between Gini and GDP per capita;
(about 1100 Ginis between 1970 and 2005)
40
30
20
Gini
50
Kuznets
curve
6
7
8
9
10
ln of GDP per capita in international dollars
11
twoway (scatter Giniall lngdpppp if Giniall<65) (qfit Giniall lngdpppp, yline(20 60, lpattern(dash))), legend(off)
xtitle(ln of GDP per capita in international dollars) ytitle(Gini)
From global_new2.dta
• No controls; a weak inverted U
relationship (more than 1300 Gini obs)
• Huge variability in inequality; R2 only 0.11
• The upward sloping part of the curve
generally hard to discern
• Turning point quite unstable; here about
$PPP 4,000 (level of Sri Lanka or
Paraguay in 2008)
• Some disenchantment with the
hypothesis: hard to see inverted U in timeseries for a single country
40
35
30
30
35
40
Gini from my allGini file
45
45
No downward portion plotted against time or
income: example of the USA
1950
1960
1970
1980
1990
year when the survey was conducted
2000
twoway scatter Giniall year if contcod=="USA", connect(l) ylabel(30(5)45)
From allginis.dta.
25000
30000
35000
40000
constant 2005 ppp, based on icp05
45000
twoway scatter Giniall gdpppp if contcod=="USA", connect(l) ylabel(30(5)45)
From global_new2
Example of China
Against income, 1970-2004
40
35
30
30
35
40
45
Gini from my allGini file
45
Against time, 1950-2004
1950
1960
1970
1980
1990
year when the survey was conducted
twoway scatter Giniall year if contcod=="CHN" &
year<2005, connect(l) ylabel(30(5)45)
Based on giniall.dta
2000
0
1000
2000
3000
constant 2005 ppp, based on icp05
4000
twoway scatter Giniall gdpppp if contcod=="CHN" & year<2005, connect(l)
ylabel(30(5)45)
From global_new2.dta
2. Credit market imperfections
theory : “pull yourself by your
bootstraps”
Credit market imperfections
• Poor households do not invest in human K even
if the returns are high; they invest in
subsistence-related types of investment
• Indivisibilities: minimum threshold of K needed
for investment; convex returns
• Societies with these problems both more
unequal and wasteful in terms of human and
capital resources
• Example of win-win strategy (inequality&growth)
• Solutions: asset redistribution, no school fees,
deeper capital markets, micro finance
Credit constraint, education,
democracy (Li, Squire & Zhou)
pooled
IV formulation
Schooling 1960
-4.6**
-4.4**
Democracy
1.6**
1.5**
Land Gini 60
0.16**
0.15**
Financial depth
(M2/GDP)
-7.7**
-10.1**
R2
0.62
No. of obs.
166
166
3. Political theory of income
distribution
Methodologically, move from
household survey data to fiscal
data
Long-run studies using income and inheritance tax
data (Picketty et al.): France 1901-98
• Secular decline in inequality
• Due to the declining share of top 1%
• Due to the decreasing importance of large
capital income
• Due to progressive taxation and
progressive (and high) inheritance taxes
• Produces no effect on average K stock but
truncates large K holdings (lower
concentration of capital income)
Story for the US (Piketty & Sanz)
• Top K incomes decreased during the
Depression and WW2 and never recovered (top
estates still lower in real terms than around
1900)
• Total K income did not decrease; its
concentration did
• Change in factoral income composition among
the top 1%; no longer mostly capitalists but
salaried workers. Δ more pronounced in the US
than in France
• Conclusion: No spontaneous decline in
inequality. Role of depression, wars and
progressive taxation. Policy and politics matter
the most. A political theory of income distribution
Explanation by E. Saez in “Striking it richer”
The labor market has been creating much more
inequality over the last thirty years, with the very top
earners capturing a large fraction of macroeconomic
productivity gains. A number of factors may help
explain this increase in inequality, not only
•underlying technological changes but also the
retreat of institutions developed during the New
Deal and World War II - such as • progressive tax
policies, • powerful unions, • corporate provision of
health and retirement benefits, and • changing
social norms regarding pay inequality.
Recent findings
• A number of similar studies for developed
countries reaches the same conclusion: a
U-shaped inequality in the 20th century in
English-speaking countries (UK: Atkinson
03; Netherlands: Atkinson & Salvedra 03;
Italy: Brandolini)
• But also for India: Banerji and Piketty 2005
• Long L shaped curve for the rest of
developed countries
A. Top 0.1% incomne share in English Speaking Countries
12%
United States
United Kingdom
Canada
Income Share
10%
8%
6%
4%
2%
Source: Piketty and Saez (2006)
1998
1993
1988
1983
1978
1973
1968
1963
1958
1953
1948
1943
1938
1933
1928
1923
1918
1913
0%
Fig 5. Top 0.1% income share in Germany and Japan
12%
Japan
Germany
Income Share
10%
8%
6%
4%
2%
1885
1890
1895
1900
1905
1910
1915
1920
1925
1930
1935
1940
1945
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
0%
Source: Piketty and Saez (2006)
Business Income
Capital Income
1951
Salaries
4.5%
1926
Fig 3: Share and Composition of top 0.01% in the US
Capital Gains
4.0%
3.5%
3.0%
2.5%
2.0%
1.5%
1.0%
0.5%
Source: Piketty and Saez (2006)
1996
1991
1986
1981
1976
1971
1966
1961
1956
1946
1941
1936
1931
1921
1916
0.0%
But this finding crucially depends on strong ρ btw.
Gini and top income share; while true among
recent data, not true historically!
30
Kenya
Byzantium
Chile
Kenya
20
China
Nueva España
Peru
Japan
Netherlands
France
Roman Empire
India-Moghul
India-British
Florence
Kingdom of Naples
Hol1732
10
Eng1290
Bihar
Java1924
Java1880
BrazilEng1759
Eng1688
Levant
Siam
Eng1801
Maghreb
Old Castiille
0
South Serbia
20
30
40
50
60
70
gini2
twoway scatter top_percent gini2 if Dancient==1, msize(vlarge) mlabel( country) xlabel(20(10)70)
Source: Milanovic, Williamson and
Lindert (2009)
B. How inequality might affect
growth
Channel 1: The median voter
hypothesis (Meltzer-Richard)
Political mechanism
• Greater inequality in
factor income=>
• Relatively poor μ
voter=>
• Chooses relatively
high tax rate
Economic mechanism
• High redistribution
and distorsionary
effect of taxes =>
• Lower growth rate
Extent of redistribution =
= fct (inequality in market income)
• Hypothesis 1. More market-unequal
countries redistribute more (using two
definitions of market income, without and
with government pensions)
• Hypothesis 2. An increase in market share
of a given decile is associated with a lower
sharegain
• Question. If countries do redistribute more,
is the mechanism through which it
happens, the median voter hypothesis?
4
2
0
sharegain in percent
6
8
Redistribution is greater if market income
share of the poor is less
-1
From figure.do based on data_voter_checked.dta
0
1
2
share of poorest market decile in percent
3
Source: Milanovic 2000, 2009
It holds for all deciles: if a decile is better-off in terms of marketP
income distribution, it loses more through the redistribution
2
3
sharegain in percent
4
0
0
1
2
sharegain in percent
6
4
5
Second decile
8
Bottom (first) decile
-1
0
1
2
share of poorest market decile in percent
3
1
5
Second richest decile
-1
-2
sharegain in percent
-4
-6
20
25
30
share of poorest market decile in percent
35
-3
-8
sharegain in percent
-2
0
0
Richest (top) decile
2
3
4
share of poorest market decile in percent
15
16
17
share of poorest market decile in percent
18
More market unequal states of the world associated with
greater Gini reduction through redistribution
Gini of marketP income
Without controls
With controls
+0.438**
(8.9)
-0.010**
(-5.3)
0.47
+0.473**
(8.2)
+0.002
(0.9)
-0.004
(-0.6)
-0.070
(0.3)
0.54
110
100
Openness
GDP per capita (in logs)
Constant
R2 (within)
Number of observations
Dependent variable: Gini reduction through redistribution.
Country fixed effects regression. Source: Milanovic (2009)
-4
-6
-8
gain of the middle
-2
0
But we cannot show that the middle deciles
gains more if market inequality high
36
Source: Milanovic 2000
38
40
42
44
share of the middle in market income
46
8
10
Sharegain of the very poor, 1973-2005
(using market income)
6
Germany
2
4
USA
1970
1980
1990
year
2000
2010
twoway (scatter gain3 year if contcod=="DEU" & decile==1, connect(l)) (scatter gain3 year if contcod=="USA" & decile==1,
connect(l)), legend(off) text(4 2000 "USA") text(7 2000 "Germany") ytitle(Distributional gain of the bottom decile)
Based on data_voter_checked.dta
Channel 2: Inequality and property
rights
Political mechanism
• Greater inequality
creates cleavages =>
• They are particularly
strong if coincide with
ethnic differences
(high horizontal
inequality)=>
• Insecure property
rights
Economic mechanism
• Insecure property
rights =>
• Lower growth rate
Inequality and property rights
(Keefer & Knack)
Dependent variable:
protection of property rights
Ln GDP per capita 1985
Cross section
7.61**
Ethnic tensions
-0.933**
Income Gini circa 1985
-0.196**
Land Gini 1985
-0.097**
R2
No. of obs.
0.80
64
Dependent variable: Property rights: ICRG measure 1986-95. Ranges
from 0 to 50.
Excursus: the reverse link and the reverse
sign: greater protection of property rights
increases inequality
Dependent
Gini
Property rights
0.929**
Gini (time-dummies
included on the RHS)
0.709*
Financial development
(M2/GDI)
-0.064**
-0.07**
Education
0.026
-0.016
Land inequality
-0.016
-0.02
Democracy
0.438**
0.323**
Prop. Rights x
Democracy
-0.056**
-0.046**
R2 within (N)
0.26 (203)
0.35 (203)
• Greater protection of property rights increases
inequality
• The rich elite is also politically powerful and
protects its economic assets
• The effect is mitigated by the introduction of
democracy
• => The negative effect of property rights
protection is particularly strong in lowdemocracy environments
• But the regression does not include an income
term
(results based on Savoia and Easaw, 2007; World Development, Feb.
2010)
Channel 3. Inequality caused by
“morally irrelevant” characteristics
• Inequalities which are independent of individual
effort, entrepreneurship or luck
• “Wasteful” (vs. instrumental or “useful”)
inequalities
• Examples: education, health, opportunity to
better oneself economically, to have a political
voice
• Horizontal inequalities between ethnic/religious
groups, education levels, socio-economic
categories, geographical areas
Assumed ρ’s for different parts of the world
Base case
Optrimistic
(high mobility)
Pessimistic
(low
mobility)
Average Gini
(year 2002)
Nordic
0.2
0.15
0.3
27.5
Rest
WENAO
E. Europe
0.4
0.3
0.5
33.7
0.4
0.3
0.5
30.6
Asia
0.5
0.4
0.6
37.6
LAC
0.66
0.5
0.9
53.8
Africa
0.66
0.5
0.9
42.6
Also a super-optimistic: ρ=0.2 for all; and super-pessimistic: ρ=0.9 for all.
ρ’s based on literature review.
How one’s income depends on circumstances:
(dependent variable: own household per capita income, in $PPP, logs)
4 (Base)
5 (Optimistic)
6 (Pessimistic)
Mean per capita
country income (in ln)
0.986
(0.00)
0.987
(0)
0.991
(0)
Gini index (in %)
-0.019
(0.00)
-0.019
(0.00)
-0.019
(0.00)
Parents’ estimated
income class (ventile)
0.105
(0.00)
0.100
(0.00)
0.109
(0.00)
Constant
-0.513
(0.00)
-0.462
(0.00)
-0.582
(0.00)
Number of
observations
232,000
232,000
232,000
R2 adjusted
0.81
0.80
0.83
Number of countries
116
116
116
Eq.
• Circumstances at one’s birth (country +
parents’ income class) explain between 83
percent (if world is fairly income-mobile
within countries) and 85 percent (if there is
less social mobility) of variability in income
globally
• => thus, only a very small portion of
global income differences can be due to
effort
• Coefficient on country mean income
remains 1; coefficient on parental income
0.1 (each notch is worth 10% increase in
children’s income); coeff. slightly higher if
there is less social mobility
• As a proxy, WDR06 looks at the contribution of
horizontal inequalities to total inequality, or total
“feasible between- inequality” (total Y of
country=given; number and sizes of
groups=given; ‘pecking order’ by group mean
incomes= given; => find new group mean
incomes that maximize the between component)
• Up to 40-45% of “feasible between inequality”
explained by education differences
• Inequality traps and the interaction between
political and economic power