Chart 1. The long term evolution of inequalities in Hungary: Gini coefficients, 1962-2005

1
WP5 Political and cultural impacts Draft of
Discussion paper 5.2.3. Prepared to the Y1
meeting in Milan 3-5 February 2011
István György Tóth Tamás Keller Income
distributions, inequality perceptions and
redistributive claims in European societies
2
Outline of the paper

• Introduction
• Research questions
• Data and definitions
• Inequalities, their perceptions and
  redistributive attitudes across countries (macro
  perspectives)
• Micro- and socio-economic correlates
  (multivariate analysis, individual and contextual
  effects)
• Summary and conclusions

3
Broad frame of understanding Inequality
voting redistribution
The proposition by Meltzer Richard (1981)
Country U is more unequal than country E.
Therefore, it redistributes more (tU gt tE)
No of persons
BUT Empirically, this is not really the case.
The evidence is rather mixed!
tE
tU
incomes
mean
Median (U
Median (E)
4
An even broader frame of understanding
Redistribution
Translation mechanisms (from policies to modified
inequalities) Tax-transfer shemes Regulation,
etc
Translation mechanisms (2) from demand for
redistribution to policies Macro (political
system) Actors (parties, bureaucracies,
etc) Electoral rules (majoritarian,
proportional etc)
Translation mechanisms (1) socio-economics to
redistributive attitudes Micro
(motivations) Perceptions Interests Attitudes
Inequality
5
Theoretical framework
A list of factors why empirics might deviate from
MR predictions

• People base their opinions/judgements on an
  assessment of their relative positions what if
  they misjudge their positions?
• Their motivation depends on self interest what
  about alternative motivations (public values,
  altruism, convictions about good, caring
  society etc)
• Self interest taken at direct money terms
• What about expectations (of their mobility, of
  their potential gains from redistribution, etc)?
• What about the insurance motive?
• Tax rate and expenditure defined unequivocally
  in reality both taxes and expenditures are more
  complex (also in their incidence!)
• Voters do not take moral standing about
  recipients (what if they do about the deserving
  and the undeserving poor)?
• The political system translates preferences into
  public spending in a straightforward way this
  is not (always) the case
• The redistribution affects the final shape of
  inequalities a great deal (also reverse
  causality..)

6
Theoretical framework
Redistribution
Translation mechanisms (from policies to modified
inequalities) Tax-transfer shemes Regulation,
etc
Translation mechanisms (2) from demand for
redistribution to policies Macro (political
system) Actors (parties, bureaucracies,
etc) Electoral rules (majoritarian,
proportional etc)
Translation mechanisms (1) socio-economics to
redistributive attitudes Micro
(motivations) Perceptions Interests Attitudes
Inequality
7
Research questions

• Q1 What individual socio-economic
  characteristics drive (the formation of
  redistributive preferences?
• Q2 How do various contextual factors (most
  importantly aggregate income inequalities) shape
  redistributive preferences?
• Q3 What effect the structure of inequality has
  on the attitudes of the middle income classes?

8
Data and Definitions
The empirical model used in the analysis

• We want to predict redistributive preference
  (RPI) by individual attributes (X) AND by
  contextual variables (Z)
• RPI a bXij cZj U0j Eij
• i The number of individuals in the analysis
  (Level 1)
• j The number of countries (Level 2)
• a Intercept
• b and c Coefficients at individual and
  country level, respectively
• Eij Level 1 residual
• U0j Level 2 residual
• The effects of individual attributes on RPI were
  predicted with simple OLS regression (with
  clustered standard error)
• RPI a bXij Eij

9
Data and Definitions
Measuring redistribution preference
All the individual level data come from
Eurobarometer (EB 72.1)
Vertical redistribution
10
Data and Definitions
Measuring redistribution preference
Jobs
Education
Social expenditures
Everyone is provided for
11
Data and Definitions
Measuring redistribution preference
RPI is an index coming from principal component
analysis
Corr. with RPI
Qa14_3 (vertical redistribution) 0.59
Qa25_a (providing jobs for the citizens) 0.65
Qa25_b (education finance) 0.53
Qa25_c (social expenditures) 0.12
Qa25_d (everyone is provided for) 0.74
Eigenvalue 1.62
Cumulative Sums of Squared Loadings 32.47
12
Data and Definitions
The mean value of RPI by countries
13
Data and Definitions
Measuring material status
No objective income data was available!!!
1
2
3
4
5
6
missing

• much higher income (qa43) than 2000 Euro/months
  (qa42) 6
• much lower income (qa43) than 500 Euro/month
  (qa42) 1
• make ends meet (qa35) very easy 6
• make ends meet (qa35) with great difficulty
  1

14
Data and Definitions
Independent variables (X) in the regression models
RPI a bXij Eij
I. Basic model Country dummies reference Germany
II. Demography (controls only) Gender male1 female, Variable d10.
II. Demography (controls only) Age 18-30, 31-40, 41-50, 51-60, 61-70 and 70 Variable vd11.
II. Demography (controls only) School less than primary, primary, secondary, higher, no education Variable d8
II. Demography (controls only) Settlement village, small town, large town Variable d25
II. Demography (controls only) Household size. The sum of the variables vd40avd40bvd40c
III. Material self interest Material status index continuous, see the construction above
III. Material self interest Labour market position self employed, employed, not working Variable c14.
15
Data and Definitions
Independent variables in the regression models
IV. Expectations Question used What are your expectations for the next twelve months will the next twelve months be ... when it comes to the financial situation of your household? (qa38) Future expectations better, same, worse Three binary coded variable
V. Failure attribution Question used Why in your opinion are there people who live in poverty? Here are four opinions which is closest to yours? (qa8) Poverty attribution unluck, lazy, injust, part of progress Four binary coded variables.
IV the variable on living standard improvement ?
social mobility. V meaning of the question is
poverty private failure or social failure?
16
Data and Definitions
Independent variables in the regression models
VI. Social context/values Poverty perception Binary coded variable 1, if someone perceive that poverty is very widespread in the country (qa4), the value is zero otherwise
VI. Social context/values Perception of (lot of) conflicts between poor-rich, young-old, managers-workers and between ethnic groups Binary coded variables Questions from qa15_1 to qa15_4,
VII. Inequality sensitivity Binary coded variable 1, if someone totally agreed the question that income differences between people are far too large (qa14_2), and the value is zero otherwise
17
Data and Definitions
Contextual variables (Z) in the regression models
All contextual data come from Luxembourg Income
Study (LIS) We used use distance-based rather
than variance based inequality measures
Contextual variable Definition Number of countries
P95/P5 The income of the person at the 95th percentile of the income distribution divided with the income of the person at the 5th percentile 17
P95/P50 The income of the person at the 95th percentile of the income distribution divided with the income of the median income person 17
P50/P5 The income of the median income person in the income distribution divided with the income of the person at the 5th percentile 17
Gini Gini coefficient  17
RPI a bXij cZj U0j Eij
Countries from LIS wave VI AT, DE, DK, ES, FI,
GR, HU, IT LV, PL, SE, UK Countries from LIS wave
V BE, EE, IE, NL, SI
18
Macro level analysis
Inequalities and redistributive attitudes across
countries
Positive relationship between inequality and
RPI RPI is more influenced by the lower part
(below median) of the income distribution, than
by the upper part (above median).
19
Macro level analysis
Inequalities and redistributive attitudes across
countries
Positive relationship between inequality and
RPI Gini performs weaker than the distance based
measures
20
Q1 individual covariates - multivariate analysis
OLS results at individual level
Country dummies in the model
Gendermale -0.05
Age 18-30 0.05
Age 31-40 0.06
Age 51-60 0.04
Age 61-70 -0.05
Age 71 -0.03
Educ max primary 0.08
Educ tertiary -0.12
Locality village -0.04
Locality lrg town -0.02
Hsize 0.01
Lab. mark selfemp -0.16
Lab. mark notwork 0.1
Lab. mark retired -0.01
Lab. mark student -0.01
Mat. status -0.05
Expects gets better 0
Expects gets worse 0.12
Gets better mat.status -0.02
Gets worser mat.status -0.03
Why poor person lazy -0.24
Why poor soc. unjust 0.23
Why poor byproduct of econ progress -0.07
Around large povety 0.17
Tension rich-poor 0.11
Tension aged 0.01
Tension man/work 0.06
Tension ethnic 0.01
Ineq too large 0.38
Expectations
Demography
Failure
Values
Mat. int.
Reference categories Female, Age 41-50,
Secondary school, Small town, Employed, Future
expectation the same, Failure attribution
unluck.
plt1 plt5, plt10
21
Q1 individual covariates - multivariate analysis
Findings (OLS results)

• People with low material resources have a
  significantly larger appetite for redistribution
• Those expecting a worsening position have a
  significant positive evaluation of redistribution
• People believing that the poor get into poverty
  because of laziness have a much smaller
  redistributive taste
• Those who think poverty is a consequence
  injustice show larger RPI
• People evaluating poverty a problem and/or think
  large tensions between social groups are more
  pro-redistributive

22
Q1 individual covariates - multivariate analysis
Adj. R square change attributed to different
explanatory mechanisms
Robust explanatory variables
3.0
1.9
4.4
23
Q2. The role of contextual factors
Random intercept models, different inequality
measures
In countries with large inequalities, respondent
are more pro-redistribution. Between-country
differences in RPI can partly be attributed to
inequality.
A. B. C.
Inequality measure Inequality measure's estimated fixed effect Proportion of variance attributed to the random between-country effect Proportion of between country variance transmitted through the inequality measure
P95/P5 0.17 5.68 26.95
P95/P50 0.69 6.74 13.32
P50/P5 0.72 4.60 40.89
Gini 5.09 6.74 13.32
Model VI. 7.78
plt1 plt5, plt10
24
Q2. The role of contextual factors
Is the impact of material status different in
various kinds of inequality regimes?
Opinion differences in unequal countries
-0.02
-0.1
-0.05
Opinion differences in equal countries
plt1 plt5, plt10
Low inequalities DK, NL, SE, FI / Middle
inequalities SI, AT, BE, LU, DE, HU, IE / Large
inequalities PL, UK, ES, GR, IT, EE
25
Q2. The role of contextual factors
Standardized regression coefficients of material
status and inequality
The difference between rich and poor respondents
RPI is the largest in countries where
inequalities are in the middle range.
Standardized regression coefficients are
calculated from country level OLS regressions,
using Model VI. The level of significance used in
the grouping (plt0.1)
26
Summary/Conclusion

1. Demand for redistribution, in addition to
   rational self interest, is also driven by general
   attitudes about the role of personal
   responsibility in ones own fate, of general
   beliefs about causes of poverty and the like.
2. The overall levels of income inequalities do
   explain (part of) cross country variance in
   demand for redistribution.
3. Larger aggregate inequalities do correspond to
   larger redistributive demands (on country level).
   
4. In countries having larger level of aggregate
   inequalities the general redistributive
   preference (of the rich, of the middle and of the
   poor) is higher.
5. The slope of this socio-economic gradient seems,
   however, steeper in countries with middle
   inequality levels.

27
Thank you for your attention!
www.tarki.hu
28
Multivariate analysis
One possible explanation on the difference
between rich and poor in various kinds of
inequality regimes
The richer the society, the less do income
explains individuals preferences.
Standardized regression coefficients are
calculated from country level OLS regressions,
using Model VI.
Economic Development and Happiness Evidence
from 32 Nations