Social inequality and voter turnout - evidence from EU elections

Download Report

Transcript Social inequality and voter turnout - evidence from EU elections

Social inequality and voter
turnout - evidence from EU
elections
Daniel Horn
TÁRKI
GINI Year-One Conference
February 4-5. 2011. Milan
Introduction
• Why people vote?
– Individual explanations
• rational choice (Downs 1957, Tulock 1967, Riker and Ordershook
1968, Muller 2003)
• social capital (Putnam 2000), social status (Lijphart 1997)
• civic resources (Verba, Schlotzman, Brady 1995)
– Macro explanations (see e.g. Geys 2006)
• socio economic (population size, - stability, - concentration, homogenity)
• political variables (closeness, campaign spendings, fragmentation)
• institutional variables (electoral system, compulsory voting,
concurrent elections, registration requirements)
But how does inequality associate with voter turnout? Does higher
inequality drive more people to cast a vote? Or does it hinder
this form of political participation?
Inequality and turnout
Daniel Horn
2
Additional motivation
• The link between inequality and redistribution is
thought to be well established (Meltzer and
Richard 1981).
– lower income people prefer more redistribution
– higher turnout is likely to be driven by higher turnout
of lower income people, who demand more
redistribution (the median voter is moving to the „left”)
– thus the association between inequality and
redistibution is mediated by voter turnout (see
Lacrinese 2007)
So the association between inequality and voter turnout
is interesting as a part of the governmental
redistribution topic as well.
Inequality and turnout
Daniel Horn
3
Possible explanations
• Three possible ways of asociation bw/
inequality and voter turnout
– negative
– positive
– none (less interesting)
Inequality and turnout
Daniel Horn
4
Possible explanations - negative
• Changing political agenda (Solt 2010, Schattschneider
1960)
– rich are better able to set the political agenda. When social
inequality is high, this „advantage” is higher, hence voter turnout
of the poor is lower
• Social norms (Lister 2007)
– institutions (such as the welfare state) influence social norms,
and hence individual behavior. In more equal societies (c.f.
universalist welfare states) we should observe higher levels of
political participation
• Reverse causation – class bias (Muller and Stratmann
2003, but see the logic of Meltzer and Richard 1981)
– Upper classes (rich) participate more (Lijphart 1997). Lower
voter turnout benefits upper classes, which leads to socially
unjust policies and thus social inequality.
Inequality and turnout
Daniel Horn
5
Possible explanations - positive
• Heterogeneity increases voter turnout (Zimmer 1976)
– If the government performs redistributive actions, homogeneous
groups with high political power can gain much (many
homogeneous groups = heterogeneous society)
• Or alternatively:
– If social tensions are high, relatively homogeneous groups of
people (poor, elderly, students… etc.) are easier to mobilize
(citation needed!)
– „societies with deep divisions between social groups that are
linked to specific parties are more likely to see high turnout than
societies in which the decline of cleavage politics has reduced
the importance of these links” (Franklin et al., 1992, citation from
van der Eijk et al 1996).
Inequality and turnout
Daniel Horn
6
Empirical evidence so far:
All negative! Increasing inequality associates with
decreasing voter turnout.
Lister (2007) – Comparative Welfare States dataset: Macro data for 15
countries, 1963-1993.
Mueller and Stratmann (2003) – Macro level data for 76 countries, 19601990.
Solt (2010): US Gubernatorial
Elections
– Individual dataset (~240
thousand, in 6 elections
1978-2000)
– post-tax Gini as income
inequality
Inequality and turnout
Daniel Horn
7
Data
•
•
5 European elections (European Election Survey 1989-2009 – PIREDEU
database)
Disadvantage: „second-order election” (Reif and Schmitt 1980)
– turnout is lower than in national parliamentary elections, affected by
• the closeness of the next parliamentary election
• complusory voting
• and other „endogeneous” factors, like pro-European feeelings
– and continuously falling (1979: 63%, 1989: 58,5%, 1999: 49,4%, 2009: 43%)
but
– it seems that „voters go to the polls because of a desire for political influence.
Moreover there is no evidence of different processes at work in different
countries.” (Franklin, van der Ejik and Oppenhuis 1996, p. 330)
– and „The second-order election model implies that it is more likely that national
elections will effect European elections than vice-versa, but in reality we find
influences running in both directions.” (van der Eijk et al. 1996)
•
Advantage:
– similar everywhere (smaller institutional effects)
• must be PR system (~UK)
• can only be subdivided to regions if this does not affect the PR nature
• maximum 5% threshold
– taken at the same time
– no unexpected effects (e.g. early elections)
Inequality and turnout
Daniel Horn
8
Data (individual level)
• European Election Study
– phone interviews
– after elections
– approx. 150.000 individuals in 5 years (~3000
people/country until 1994; ~1000 in 1999,
2004 – 2009 is in preparation)
– questions about participation and individual
background (no income) – and much more
(controls: age, sex, education, class,
settlement, church, voted last parl. elect.)
Inequality and turnout
Daniel Horn
9
Data (country level)
• National data
– Next election in days/365 (wikipedia)
– Compulsory voting (IDEA, www.idea.int)
– GDP/capita (Eurostat 1995-2004) and Nationmaster (WDI
database - 1989)
– Nationality – consider herself belong to other nationality (EES
2009)
– Inequality
• (post-tax) Gini for each year (Eurostat - SILC)
• pre-tax Gini (Social Situation Observatory Annual Montioring Report
2009)
– Gini indices of the distribution of equivalent household market income
among the working age 2004
• MDMI (Lancee - v. d. Werfhorts 2011)
• poverty - Population at risk of poverty or social exclusion 2005
(Eurostat) (Bulgaria 2006, Romania 2007)
Inequality and turnout
Daniel Horn
10
Voter turnout - EES
With weights
Std. Dev.
0,49
0,50
0,50
0,50
0,50
EES study
1989
1994
1999
2004
Total
Mean
0,58
0,56
0,54
0,43
0,50
EES study
1989
1994
1999
2004
2009
Total
No weights
Mean
Std. Dev.
0,74
0,44
0,72
0,45
0,70
0,46
0,60
0,49
0,71
0,45
0,68
0,47
Inequality and turnout
Freq.
13278
15794
16823
37418
83313
Obs.
10503
12357
12980
26790
62630
Freq.
10503
12357
12980
26790
26908
89538
Daniel Horn
11
Own calculations
1
Descriptive
LU
.8
IT
IE
.6
CY
NL
.4
DK FI
PT
ES
FR
AT
CZ
UK
HU
LV
SE
EE
SI
.2
turnout
GR
PL
SK
20
Inequality andyear=2004
turnout
25
30
Gini
Daniel Horn
35
40
12
Own calculations
.4
.5
.6
.7
.8
Lowess smoothing
20
Inequality and turnout
25
30
Gini
35
real voting
probability of voting
fixed effects Daniel Horn
40
13
Methods
• Pooled logit with clustered standard errors
• two 2-step probits
– fixed effect:
• 1st step: individual controls and country fixed marginal
effects
• 2nd step: country controls and inequality indicators
– separate country regressions
• 1st step: individual controls for each country and probabilities
are predicted (for 42 year old middle class women, who live in mid-size
settlement, with 12 years of education, and voted in parl. elections.)
• 2nd step: country controls and inequality indicators
• Multilevel mixed effect logit (no weighting)
Inequality and turnout
Daniel Horn
14
ORs
(se)
1.021** (0.00173)
0.897** (0.0261)
0.927
(0.0981)
0.783** (0.0558)
0.886+
(0.0642)
0.873*
(0.0588)
0.910
(0.0714)
1.070
(0.0790)
1.089
(0.0796)
1.402**
(0.121)
1.353**
(0.141)
1.552**
(0.101)
1.116** (0.0462)
1.077
(0.0521)
0.866** (0.0419)
1.063
(0.0647)
2.138**
(0.576)
1.869**
(0.428)
1.054
(0.238)
4.182**
(0.456)
2.611**
(0.577)
age
female
educ==up to 14 years
educ==15 years
educ==16 years
educ==17 years
educ==19 years
educ==20 years
educ==21 years
educ==22 years and more
educ==still studying
go to church
settlement, rural area/village
settlement, large town
Working class
Upper class
1989
1994
2004
Voted in parl. elections
Compulsory voting
distance from the next national
elections, years
0.771** (0.0569)
% of other nationalitis
1.009
(0.00805)
Constant
0.0860** (0.0264)
Observations
55,423
Inequality and turnout
Daniel Horn
Robust clustered se in parentheses
** p<0.01, * p<0.05, + p<0.1
Base model
All variables here
are included in
each regression
below.
Country level variables
15
Pooled logit
gdp
gini
Weighted
vote
1.007*
(0.00306)
1.088**
(0.0261)
gini^2
Constant
0.00431**
(0.00394)
Pooled logit
Non-weighted (plus2009)
vote
vote
vote
1.006*
1.008**
1.008**
(0.00267)
(0.00230)
(0.00216)
1.796**
1.082**
1.369
(0.390)
(0.0192)
(0.281)
0.992*
0.996
(0.00359)
(0.00345)
3.80e-06** 0.00713**
0.000257**
(1.23e-05) (0.00480)
(0.000747)
Observations 55,423
55,423
Robust clustered se in parentheses
** p<0.01, * p<0.05, + p<0.1
Inequality and turnout
Daniel Horn
55,148
55,148
16
Pooled logit – other indicators
ODDS RATIOS
Gini
Gini^2
(1)
(2)
1.088**
(0.0261)
1.796**
(0.390)
0.992*
(0.00359)
pre-tax Gini
(3)
(4)
1.089**
(0.0361)
0.852
(0.594)
1.003
(0.00817)
pre-tax Gini^2
MDMI
(9)
(10)
1.039**
(0.0103)
1.258**
(0.0888)
0.998**
(0.000742)
MDMI^2
Poverty
(8)
1.059*
(0.0239)
Poverty^2
Constant
(7)
0.00431** 3.80e-06** 0.000886** 0.143
(0.00394) (1.23e-05) (0.00151) (2.129)
Observations
55,423
55,423
Robust
clustered
in parentheses
Inequality
andse
turnout
** p<0.01, * p<0.05, + p<0.1
41,862
41,862
Daniel Horn
1.246**
(0.0748)
0.997**
(0.000927)
0.00801** 0.000116** 0.0116** 0.00124**
(0.00568) (0.000191) (0.0112) (0.00147)
55,423
55,423
55,423
55,423
17
2step probit
1
LU
BE
GR GR GR
IT
.8
LU
IE
DE
ES
IE
ES ES
.6
CY
PT PT PT
FR FR
DK
DK
UK
UK
CZ
AT HU
.4
FI
LV
FI
NL
NL
SE SE
SI
EE
.2
SK
20
Inequality and turnout
25
30
Gini
Daniel Horn
35
40
19
.8
IE
.6
IE
ES
UK
ES
PT
LT
LV
FR
AT
PT
FI
.4
AT HU
CZ
SE
DK
SI
FR
NL NL
SK
PL
SE
.2
UK
FI
20
25
separate country regressions
Inequality and turnout
30
Gini
Daniel Horn
35
40
20
Quick tests of „negative”
theories
Social norms
.6
DK
DK
.4
FI
SE SE
DK
FI
FI
NL NL
NL
NL
SE
.2
fixed-effects
.8
1
• universal welfare state -> more equal and higher
turnout
20
Inequality and turnout
25
30
Gini
Daniel Horn
35
40
22
Changing political agenda /Class bias
.2
.4
.6
.8
1
If inequality increases rich are more likely to vote than the poor
20
25
30
Gini
35
40
Working class
Middle class
Upper class
fractional-polynomial prediction
Inequality and turnout
Daniel Horn
23
Conclusion
• „conventional” results for the base model (compulsory
voting, voted last election… etc.)
– so results for european „second-order” elections work as firstorder elections
• but unconvetional results for the inequality dimesion:
– Inequality and voter turnout associates positively
or
in a quadratic relation
– Negative association is less likely to hold
• Social norms theory does not explain low levels of turnout for
universal welfare states (e.g. Sweden, Finland, Denmark,
Netherlands)
• Political agenda/ Class bias theories does not explain coparatively
higher rising levels of turnout for the working class
– Theories of positive association are yet to be confirmed
Inequality and turnout
Daniel Horn
24
Questions
• Whether it is due to
– the quality of the data
– the „second-order” nature of the EU elections
– the fact that aggregate data works differently
than individual (ecological fallacy)
– the fact that US inequality is higher (cf. Solt
2010) and very low inequality and very high
inequality both predicts low turnout compared
to the middle
are all open to further research.
Inequality and turnout
Daniel Horn
25
Thank you for your attention!
Daniel Horn
[email protected]
250
150
200
LUXEMBOURG
IRELAND
100
NETHERLANDS
AUSTRIA
SWEDEN
DENMARK
BELGIUM
FINLAND
FRANCE
UK
SPAIN
CYPRUS
SLOVENIA
ITALY
GREECE
PORTUGAL
CZECH REPUBLIC
50
HUNGARY
SLOVAKIA
20
year=2004
Inequality and turnout
25
30
Gini
Daniel Horn
ESTONIA
POLAND
LITHUANIA
LATVIA
35
40
27
60
50
POLAND
HUNGARY
LITHUANIA
IRELANDUK
ESTONIA
40
BELGIUM
FRANCE
SLOVENIA FINLAND
CZECH
REPUBLIC
LUXEMBOURG
SLOVAKIA
AUSTRIA
NETHERLANDS
SWEDEN
DENMARK
20
30
CYPRUS
20
25
30
(post-tax) Gini
35
40
year=2004
Inequality and turnout
Note: ----------- X=Y line
Daniel Horn
28
50
LATVIA
POLAND
40
LITHUANIA
30
HUNGARY
SLOVAKIA
GREECE
CYPRUS
20
BELGIUM
IRELAND
ITALY
UK
SPAIN
PORTUGAL
ESTONIA
CZECH REPUBLIC
FRANCE
SLOVENIA
LUXEMBOURG
DENMARK
FINLAND
AUSTRIA
NETHERLANDS
10
SWEDEN
20
year=2004
Inequality and turnout
25
30
(post-tax) Gini
Daniel Horn
35
40
29
80
70
LITHUANIA
60
PORTUGAL
IRELAND
GREECE
50
UK
HUNGARY
LUXEMBOURG
CYPRUS
SPAIN
LATVIA
ESTONIA
POLAND
ITALY
40
BELGIUM
SLOVAKIA
FINLAND
FRANCE
CZECH REPUBLIC
AUSTRIA
NETHERLANDS
SLOVENIA
30
SWEDEN
DENMARK
20
year=2004
Inequality and turnout
25
30
(post-tax) Gini
Daniel Horn
35
40
30
Country fixed effects
ODDS RATIOS
Gini
Gini^2
(1)
Dependent variable: Country fixed effects (Weighted GLS)
(2)
(3)
(4)
(5)
(6)
(7)
1.022*
1.095
(0.00815) (0.0842)
0.999
(0.00130)
pre-tax Gini
1.025+
(0.0137)
pre-tax Gini^2
0.888
(0.183)
1.002
(0.00249)
MDMI
1.010**
1.051+
(0.00327) (0.0277)
1.000
(0.000268)
MDMI^2
Poverty
Poverty^2
Constant
(8)
0.806
(0.213)
year fixed effects
y
Observations
55
R-squared
0.654
Robust clustered se in parentheses
** p<0.01,
* p<0.05,
+ p<0.1
Inequality
and
turnout
0.307
(0.338)
y
55
0.663
0.516
(0.308)
y
39
0.674
Daniel Horn
9.712
(41.46)
y
39
0.684
0.986
(0.164)
y
55
0.703
0.401
(0.252)
y
55
0.729
1.020*
1.077**
(0.00773) (0.0207)
0.999**
(0.000331)
0.949
0.456*
(0.232)
(0.141)
y
y
55
55
0.656
0.738
31
Separate country regressions
ODDS RATIOS
Gini
Gini^2
(1)
(2)
(3)
(4)
1.019
(0.0135)
1.441*
(0.199)
0.996*
(0.00150)
(5)
(6)
pre-tax Gini^2
MDMI
1.009**
1.020
(0.00208) (0.0149)
1.000
(0.000139)
MDMI^2
Poverty
Poverty^2
year fixed effects
Observations
R-squared
Robust seeform in parentheses
** p<0.01, * p<0.05, + p<0.1
Inequality and turnout
(8)
1.021*
1.076
(0.00718) (0.0849)
0.999
(0.00129)
pre-tax Gini
Constant
(7)
0.723
(0.193)
y
26
0.470
0.350
(0.421)
y
26
0.480
0.515 0.000276* 0.815
(0.339) (0.000863) (0.130)
y
y
y
21
21
26
0.402
0.588
0.582
Daniel Horn
0.637
(0.250)
y
26
0.590
1.009
1.058*
(0.00890) (0.0252)
0.999+
(0.000370)
0.995
0.519
(0.433)
(0.236)
y
y
26
26
0.274
0.446
32
Multilevel logit
ODDS RATIOS
Gini
1
2
1.107**
(0.0124)
1.094
(0.0871)
1.000
(0.00132)
Gini^2
pre-tax Gini
3
4
0.999
(0.0351)
1.593
(0.772)
0.995
(0.00551)
pre-tax Gini^2
MDMI
5
6
1.002
(0.0130)
1.096
(0.111)
0.999
(0.000984)
MDMI^2
Poverty
8
0.979+
(0.0117)
Poverty^2
Constant
7
0.00688** 0.00819** 0.0614+ 2.84e-06 0.120**
(0.00289) (0.00994) (0.0941) (2.97e-05) (0.0771)
year fixed effects
Observations
55,148
R-squared
27
se in parentheses
** p<0.01,
* p<0.05,
p<0.1
Inequality
and+turnout
55,148
27
39,349
19
39,349
19
Daniel Horn
53,245
24
0.0148+
(0.0360)
53,245
24
0.962
(0.0593)
1.000
(0.000870)
0.276**
0.353
(0.117)
(0.340)
55,148
27
55,148
27
33
Changing political agenda /Class bias
• If inequality increases rich are more likely to vote than the poor
ODDS RATIOS
Compulsory vote
Time until next election
% of other nationality
GDP/capita in % of EU27
Gini
Gini^2
Constant
Dependent variable: Country fixed effects (Weighted GLS)
(1)
(2)
(3)
(4)
(5)
(6)
Lower class
Middle class
Upper class
1.261**
1.235*
1.241*
1.223*
1.188*
1.184*
(0.0914)
(0.0923)
(0.0977)
(0.102)
(0.0862)
(0.0877)
0.973
0.969+
0.974
0.971
0.971
0.971
(0.0171)
(0.0165)
(0.0178)
(0.0172)
(0.0169)
(0.0169)
0.997
0.995
0.997
0.996
0.999
0.999
(0.00373) (0.00336) (0.00423) (0.00410) (0.00416) (0.00414)
1.002*
1.002*
1.002*
1.002*
1.001
1.001
(0.000616) (0.000574) (0.000682) (0.000662) (0.000610) (0.000624)
1.024**
1.121
1.022*
1.091
1.008
1.025
(0.00817) (0.0795) (0.00882) (0.0925) (0.00714) (0.0750)
0.998
0.999
1.000
(0.00120)
(0.00143)
(0.00125)
0.733
0.207
0.768
0.308
1.362
1.079
(0.194)
(0.210)
(0.217)
(0.373)
(0.346)
(1.110)
year fixed effects
Observations
55
R-squared
0.697
Robust
seeform in parentheses
Inequality and turnout
** p<0.01, * p<0.05, + p<0.1
55
0.712
55
0.634
Daniel Horn
55
0.642
55
0.476
55
0.477
34