I stván György Tóth – Tamás Keller:

István György Tóth – Tamás Keller:

Income distributions, inequality perceptions and redistributive claims in European societies

WP5: Political and cultural impactsDraft of Discussion paper 5.2.3.

Prepared to the Y1 meeting in Milan 3-5 February 2011

I. Introduction II. Research questionsIII. Data and definitionsIV. Inequalities, their perceptions and redistributive

attitudes across countries (macro perspectives) V. Micro- and socio-economic correlates (multivariate

analysis, individual and contextual effects) VI. Summary and conclusions

Outline of the paperOutline of the paper

Country U is more unequal than country E. Therefore, it redistributes more (tU > tE)

BUT: Empirically, this is not really the case. The evidence is rather mixed!

againstfor mean

Median (E)Median (U

tEtU

The proposition by The proposition by Meltzer & RichardMeltzer & Richard ( (19811981): ):

No

of p

erso

ns

incomes

Broad frame of understanding: Broad frame of understanding:

Inequality voting redistributionInequality voting redistribution

Translation mechanisms (1): socio-economics to redistributive attitudes

Micro (motivations):PerceptionsInterestsAttitudes

Inequality

Redistribution

Translation mechanisms (from policies to modified inequalities)

Tax-transfer shemesRegulation, etc…

Translation mechanisms (2): from demand for redistribution to policies

Macro (political system):Actors (parties, bureaucracies, etc)Electoral rules (majoritarian,

proportional etc)

An even broader frame of understandingAn even broader frame of understanding

• 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..)

A list of factors why empirics might deviate from MR A list of factors why empirics might deviate from MR predictionspredictions

Theoretical frameworkTheoretical framework

Translation mechanisms (1): socio-economics to redistributive attitudes

Micro (motivations):PerceptionsInterestsAttitudes

Inequality

Redistribution

Translation mechanisms (from policies to modified inequalities)

Tax-transfer shemesRegulation, etc…

Translation mechanisms (2): from demand for redistribution to policies

Macro (political system):Actors (parties, bureaucracies, etc)Electoral rules (majoritarian,

proportional etc)

Theoretical frameworkTheoretical framework

• 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?

Research questionsResearch questions

Data and DefinitionsData and DefinitionsThe empirical model used in the analysisThe empirical model used in the analysis

We want to predict redistributive preference (RPI) by individual attributes (X) AND by contextual variables (Z)

RPI = a + bXRPI = a + bXijij + cZ + cZjj +U +U0j0j + E + Eijij

i = The number of individuals in the analysis (Level 1)j = The number of countries (Level 2)a = Interceptb and c = Coefficients at individual and country level, respectivelyEij = Level 1 residualU0j = Level 2 residual

The effects of individual attributes on RPI were predicted with simple OLS regression (with clustered standard error)

RPI = a + bXRPI = a + bXijij + E + Eijij

Data and DefinitionsData and DefinitionsMeasuring redistribution preferenceMeasuring redistribution preference

Vertical Vertical redistributionredistribution

All the individual level data come from Eurobarometer (EB: 72.1)


JobJobss

EducationEducation

Social Social expendituresexpenditures

Everyone is Everyone is provided forprovided for


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%

RPI is an index coming from principal component analysis

CorrCorr.. with RPI with RPI

-0,8

-0,6

-0,4

-0,2

0,0

0,2

0,4

0,6

0,8

1,0

NL BE LT CZ DK SI SE UK FR LU AT DE EE FI EU SK PL PT RO IT MT IE LV ES BGHU CY GR

Data and DefinitionsData and DefinitionsThe mean value of RPI by countriesThe mean value of RPI by countries

Data and DefinitionsData and DefinitionsMeasuring material statusMeasuring material status

No objective income data was availableNo 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

Data and DefinitionsData and DefinitionsIndependent Independent variablesvariables (X) (X) in the regression in the regression

modelsmodelsI. Basic model Country dummies [ reference: Germany]

II. Demography

(controls only)

Gender: male=1 [female], Variable d10.

Age: 18-30, 31-40, [41-50], 51-60, 61-70 and 70+; Variable vd11.

School: less than primary, primary, [secondary], higher, no education; Variable d8

Settlement: village, [small town], large town; Variable d25

Household size. The sum of the variables vd40a+vd40b+vd40c

III. Material self interest

Material status index : continuous, see the construction above

Labour market position: self employed, [employed], not working; Variable c14.

RPI = a + bXRPI = a + bXijij + E + Eijij

Data and DefinitionsData and Definitions

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.

IndependentIndependent variables variables in the regression modelsin the regression models

IV: the variable on living standard improvement IV: the variable on living standard improvement ≠≠ social mobility.social mobility.V: meaning of the question: isV: meaning of the question: is poverty private failure or poverty private failure or social social failurefailure??

Data and DefinitionsData and Definitions

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

Perception of („lot of”) conflicts between poor-rich, young-old, managers-workers and between ethnic groups

Binary coded variablesQuestions 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

IndependentIndependent variables variables in the regression modelsin the regression models

Data and DefinitionsData and DefinitionsContextual variablesContextual variables (Z) (Z) in the regression modelsin the regression models

Contextual variable Definition

Number of

countries

P95/P5The 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/P50The income of the person at the 95th percentile of the

income distribution divided with the income of the median income person

17

P50/P5The 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 17Countries from LIS wave VI: AT, DE, DK, ES, FI, GR, HU, IT LV, PL, SE, UKCountries from LIS wave VI: AT, DE, DK, ES, FI, GR, HU, IT LV, PL, SE, UKCountries from LIS wave V: BE, EE, IE, NL, SICountries from LIS wave V: BE, EE, IE, NL, SI

All contextual data All contextual data ccome fromome from Luxembourg Luxembourg Income StudyIncome Study (LIS) (LIS)We used use distance-based rather than variance based inequality measuresWe used use distance-based rather than variance based inequality measures

RPI = a + bXRPI = a + bXij ij + cZ+ cZjj +U +U0j 0j + E+ Eijij

Macro level analysisMacro level analysisInequalitiesInequalities and redistributive attitudes across countriesand redistributive attitudes across countries

UKSISE

PL

NL

LU

IT

IE

HU

GR

FI

ES

EE

DK

DE

BE

AT

R2 = 0.4806

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1.9 2.1 2.3 2.5 2.7 2.9 3.1

P50/P5, (LIS wave V. and wave VI.)

Red

istr

ibut

ive

Pref

eren

ce In

dex

(RPI

, fac

tor

load

ings

from

PC

A),

EB.7

2.1,

200

9.

UKSISE

PL

NL

LU

IT

IE

HU

GR

FI

ES

EE

DK

DE

BE

AT

R2 = 0.2331-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1.5 2 2.5 3


Red

istr

ibut

ive

Pref

eren

ce In

dex

(RPI

, fac

tor

load

ings

from

PC

A),

EB.7

2.1,

200

9.

Positive relationship between inequality and RPIPositive relationship between inequality and RPIRPIRPI is more influenced by the lower part (below median) of the income is more influenced by the lower part (below median) of the income distribution, than by the upper part (above medistribution, than by the upper part (above mediandian).).

Macro level analysisMacro level analysisInequalitiesInequalities and redistributive attitudes across countriesand redistributive attitudes across countries

UKSISE

PL

NL

LU

IT

IE

HU

GR

FI

ES

EE

DK

DE

BE

AT

R2 = 0.3049

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0.2 0.25 0.3 0.35 0.4

Gini, (LIS wave V. and wave VI.)

Red

istr

ibut

ive

Pref

eren

ce In

dex

(RPI

, fac

tor

load

ings

from

PC

A),

EB.7

2.1,

200

9.

UKSISE

PL

NL

LU

IT

IE

HU

GR

FI

ES

EE

DK

DE

BE

AT

R2 = 0.3742

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

3 4 5 6 7 8


Red

istr

ibut

ive

Pref

eren

ce In

dex

(RPI

, fac

tor

load

ings

from

PCA

), EB

.72.

1, 2

009.

Positive relationship between inequality and RPIPositive relationship between inequality and RPIGini performs weaker than the distance based measuresGini performs weaker than the distance based measures

Q1: Q1: individual covariatesindividual covariates - m - multivariate analysisultivariate analysis

Gender:male -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***

Dem

ogra

phy

Dem

ogra

phy

Mat

. int

. M

at. i

nt.

Expe

ctat

ions

Expe

ctat

ions

Failu

reFa

ilure

Val

ues

Val

ues

OLS results at individual levelOLS results at individual level

*** p<1%; ** p<5%,; * p<10

Reference categories: Female, Age 41-50, Secondary school, Small town, Employed, Future expectation: the same, Failure attribution: unluck.

Country dummies in the modelCountry dummies in the model


Findings (Findings (OLS resultsOLS 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

6.9%8.6%

11.1% 11.3%

15.7%17.6%

20.6%

0%

5%

10%

15%

20%

25%

Basic model Demography Material selfinterest

Expectations Failureattribution

Social context/ value

Inequalitysensit ivity

Adj. Adj. R square change attributed to different explanatory mechanismsR square change attributed to different explanatory mechanisms Robust explanatory variables

4.4%4.4%1.9%1.9%

3.0%3.0%


Q2. The Q2. The role of corole of conntextual factorstextual factorsRandom intercept models, Random intercept models, different inequality measuresdifferent inequality measures

A. B. C.

Inequality measureInequality 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%

*** p<1%; ** p<5%,; * p<10

In countries In countries with largewith large inequalities, respondent are more pro-redistribution inequalities, respondent are more pro-redistribution. . BBetween-country differences in RPI can etween-country differences in RPI can partly partly be attributed to inequalitybe attributed to inequality..

Model VI. 7.78%Model VI. 7.78%

-.50

.5Pr

edic

ted

RPI

1 2 3 4 5 6Material status index

Low inequalities Middle inequalitiesHigh inequalities

Low inequalitiesLow inequalities: DK, NL, SE, FI: DK, NL, SE, FI / / Middle inequalitiesMiddle inequalities: SI, AT, BE, LU, DE, HU, IE: SI, AT, BE, LU, DE, HU, IE / / Large inequalitiesLarge inequalities: PL, UK, ES, GR, IT, EE: PL, UK, ES, GR, IT, EE

Opinion differences in equal countries

Opinion differences in unequal countries

Is Is the impact of material status differthe impact of material status differentent in various kinds of in various kinds of inequainequallity regimesity regimes??

-0.05-0.05

-0.1***-0.1***

-0.02**-0.02**

*** p<1%; ** p<5%,; * p<10*** p<1%; ** p<5%,; * p<10

Q2. The Q2. The role of corole of conntextual factorstextual factors

Standardized regression coefficients of material status and inequalityStandardized regression coefficients of material status and inequality

HU

UK

SE

AT

NL

LU

IT

ES

BE

SI

PLEE

IEFI

GR

DE

DK

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

1.9 2.1 2.3 2.5 2.7 2.9 3.1

P50/P5

Stan

dard

ized

regr

essi

on c

oeff

icie

nt o

f m

ater

ial s

tatu

s in

mod

el V

I.

Not significant Significant

HU

UK

SE

AT

NL

LU

IT

ES

BE

SI

PL EE

IEFI

GR

DE

DK

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

1.5 2 2.5 3

P95/P50

Stan

dard

ized

regr

essi

on c

oeff

icie

nt o

f m

ater

ial s

tatu

s in

mod

el V

I.

Not significant Significant

Standardized regression coefficients are calculated from country level OLS regressions, using Model VI.Standardized regression coefficients are calculated from country level OLS regressions, using Model VI.The level of significance used in the grouping (p<0.1)The level of significance used in the grouping (p<0.1)

Q2. The Q2. The role of corole of conntextual factorstextual factors

TThe difference between he difference between rich and poorrich and poor respondents’ respondents’ RPIRPI is is the the largelargestst in countries where inequalities are in the middle range. in countries where inequalities are in the middle range.

1. Demand for redistribution, in addition to rational self interest, is also driven by general attitudes about the role of personal responsibility in one’s 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.

Summary/ConclusionSummary/Conclusion

Thank you for your attention!

www.tarki.hu

Multivariate analysisMultivariate analysisOne possible explanationOne possible explanation on the difference between rich and on the difference between rich and

poor in various kinds of inequality regimespoor in various kinds of inequality regimes

UK

SI

SE

PL

NL

LU

IT

IE

HU

GR

FI

ES

EE

DK

DE

BE

AT

R2 = 0.3545

0.22

0.24

0.26

0.28

0.3

0.32

0.34

0.36

0.38

0 20 40 60 80 100

Self reported high income status (%)

Stan

dard

ized

regr

essi

on c

oeff

icie

nt o

f mat

eria

l st

atus

in m

odel

VI.

(abs

olut

e va

lue)

TThe richer the society, the less dohe richer the society, the less do income explainincome explainss individuals’ preferencesindividuals’ preferences..

*Economic Development and Happiness: Evidence from 32 Nations

UK

SI

SE

PL

NL

IT

IE

HU

GR

FI

ES

EE

DK

DEBE

AT

R2 = 0.2441

0.22

0.24

0.26

0.28

0.3

0.32

0.34

0.36

0.38

10000 15000 20000 25000 30000 35000

GDP in Purchasing Power Standard per inhabitant (2009)

Stan

dard

ized

regr

essi

on c

oeff

icie

nt o

f mat

eria

l st

atus

in m

odel

VI.

(abs

olut

e va

lue)

Standardized regression coefficients are calculated from country level OLS regressions, using Model VI.

Documents

I stván György Tóth – Tamás Keller: