Redistribution and Preference Formation

Redistribution and Preference Formation
Pablo Beramendi (Duke University, [email protected])
&
Philipp Rehm (Ohio State University, [email protected])
Abstract
Why is income a good predictor of attitudes toward redistribution in some countries but
not in others? In this paper we argue that the tax-benefit structure of countries plays a role
in shaping social policy attitudes, an aspect largely overlooked by the literature so far.
Attitudes are proportional to expected net benefits, which is determined by what
individuals receive (probability of receiving a transfer; a transfer’s structure) and what
individuals contribute (taxes) to the system. The level of concentration (progressivity)
determines the distribution of both, thereby accounting for cross-national variations in the
distribution of preferences about the welfare state. The paper develops an argument on
the mechanisms driving the relationship between progressivity and preferences. We then
evaluate the argument on the basis of a cross-national design. Our findings indicate, quite
robustly, that the progressivity of the tax-benefit system is a major determinant of the
predictive power of income on preferences for redistribution.
Acknowledgments
Previous versions of this paper were presented at the 2011 MPSA meetings, the political
economy workshop at the University of Oxford, the comparative politics seminar at
Columbia University, the University of Zürich, the Juan March Institute, Madrid, Yale
University, the 2012 CES Meetings at Boston, and the University of Minnesota. We are
grateful for the feedback received in all these occasions. We are also grateful to Ben
Ansell, Lucy Barnes, Larry Bartels, German Feierherd, Herbert Kitschelt, Isabela Mares,
Yotam Margalit, Irfan Nooruddin and David Rueda for their comments on earlier
versions. The usual disclaimer applies.
Redistribution and Preference Formation
Abstract
Why is income a good predictor of attitudes toward redistribution in some countries but
not in others? In this paper we argue that the tax-benefit structure of countries plays a role
in shaping social policy attitudes, an aspect largely overlooked by the literature so far.
Attitudes are proportional to expected net benefits, which is determined by what
individuals receive (probability of receiving a transfer; a transfer’s structure) and what
individuals contribute (taxes) to the system. The level of concentration (progressivity)
determines the distribution of both, thereby accounting for cross-national variations in the
distribution of preferences about the welfare state. The paper develops an argument on
the mechanisms driving the relationship between progressivity and preferences. We then
evaluate the argument on the basis of a cross-national design. Our findings indicate, quite
robustly, that the progressivity of the tax-benefit system is a major determinant of the
predictive power of income on preferences for redistribution.
Acknowledgments
[Redacted to preserve anonymity]
What drives people’s attitudes towards redistribution? As the economic crisis forces countries to
revisit fundamental aspects of their social contract, a better grasp of how citizens form their
views on the scope for the political reallocation of resources seems ever more pressing. In recent
years, comparative political economists have come a long way towards unraveling the
institutional foundations of different public insurance regimes and their implications on
distributive outcomes. From political regimes themselves, to electoral systems, economic
coordination, or federalism, the institutional foundations of redistribution and inequality have
dominated the agenda in recent years. This institutionalist focus, however, has come with a price
tag, namely the relative lack of attention towards the origins of citizens’ preferences for
redistribution and the processes through which these preferences are formed. This seems
particularly unfortunate in light of the suggestion and evidence that public opinion on social
policy matters for social policy outputs (Brooks and Manza 2006a; Brooks and Manza 2006b;
Brooks and Manza 2007).
Fortunately, a recent wave of scholarship has begun to explore the determinants of social policy
attitudes, relying in a variety of theoretical frameworks and methodological bets. Existing
approaches can be classified into three types. First, some scholars rely on self-interest as the
primary source of social policy attitudes – income and its variability (“risk”) are the key
variables in these approaches. See, for example the literature on social upward mobility (Alesina
and La Ferrara 2005; Bénabou and Ok 2001; Piketty 1995), insurance (Cusack, Iversen, and
Rehm 2006; Iversen and Soskice 2001; Moene and Wallerstein 2001; Rehm 2009; Sinn 1995;
Varian 1980), or class-based explanations (Svallfors 2004). Second, other scholars rely on values
and beliefs as sources of social policy attitudes. Norms of “deservingness”, standards of fairness,
beliefs about the causes of inequality, partisan ideology, altruism, national identity, religion, and
many other factors have been explored (Alesina and Angeletos 2005; Bénabou and Tirole 2006;
Fong 2001; Kangas 1997; Kangas 2003; Kangas et al. 1995; Scheve and Stasavage 2006; Shayo
2009). Third, some accounts rely on interpersonal preferences to explain social policy attitudes.
Examples include references to group loyalty (Luttmer 2001), the importance of relative status
(Corneo and Gruner 2000; Lupu and Pontusson 2011; Wilensky 1975) and race or ethnicity
(Alesina, Glaeser, and Sacerdote 2001).
In this paper, we take a closer look at the argument that self-interest shapes attitudes towards
public insurance and redistribution. A core proposition of that argument is that expected net
benefits determine attitudes. However, expected net benefits not only depend on individual-level
characteristics such as income and risk but also on the design and organization of public
insurance systems shapes citizens’ preferences. For example, self-interested rich citizens may
well support social policies if they are regressive – and the degree of progressivity varies hugely
across countries and social policy domains. Existing scholarship has largely neglected this
important point.1
1
To be sure, there is a growing literature that explores the correlation between different welfare regimes (EspingAndersen 1990) and aggregate support for redistribution (Arts and Gelissen 2001; Bean and Papadakis 1998;
Gelissen 2000; Gelissen 2002; Jaeger 2006; Jaeger 2009; Jakobsen 2010; Mehrtens 2004; Svallfors 1997), but this
literature tends not to focus on causal mechanisms.
1
In what follows, we explore theoretically and empirically why income varies in its ability to
predict preferences for redistribution. To give a preview of our core result, we find that income is
a better predictor of preferences towards redistribution in those societies with more progressive
fiscal systems. Throughout the paper we define progressivity as the concentration of net benefits
in the lower strata on the pre-tax and transfers income distribution. This finding has a number of
implications for several literatures. First, it highlights an important institutional component in the
process of preference formation, largely overlooked by the literature so far. Second, it provides a
framework to approach the political consequences of fiscal reforms, an important issue in the
context of the current financial crisis. Third, it identifies the process linking economic patterns
(inequality), macro-responses (progressivity), and political attitudes, thereby shedding light on
the forces behind the self-sustained nature of politico-economic configurations across countries.
The paper proceeds as follows. Section I motivates the question by presenting the key theoretical
puzzle. Section II develops the theoretical argument. Section III outlines the empirical strategy.
Section IV presents the results. Finally, section V summarizes the findings, discusses some of the
limitations in the paper, and points to future lines of inquiry.
I. The Puzzle: Income and Preferences for Redistribution
Insofar as democracies manage to achieve effectively a minimum degree of political
representation, redistributive policies and outcomes will reflect, at least in part, the degree of
support for redistribution existing in a society (Brooks and Manza 2006b; Brooks and Manza
2006a; Brooks and Manza 2007). The latter link has often been treated as a matter of course.
Following layman’s intuition, one should expect income to be closely correlated with
redistributive references. It would appear almost self-evident that poor people like redistribution
for the same reasons that rich people resist it. However, available evidence suggests that the
issue requires some additional thought. Figure 1 displays the size of the income-coefficient when
predicting attitudes on the following ISSP question:
“On the whole, do you think it should be or should not be the government’s responsibility
to: Reduce income differences between the rich and poor” [1. Definitely should not be; 2.
Probably should not be; 3. Probably should be; 4. Definitely should be]
- Figure 1 about here The bars indicated the size of the income coefficient from regressing redistribution attitudes on
income and a small set of controls (gender, education, age); these coefficients refer to 2006, and
are recovered from a multi-level model.2 The layman’s expectation appears correct in that
income almost always has a negative effect on support for redistribution across all nations, but
2
To produce these slopes, we estimate hierarchical linear models predicting social policy attitudes with income and
a set of controls (education, gender, and age), with random intercepts and random slopes. We then recover countryspecific income-slopes (income-gradients) and their standard errors from best linear unbiased predictions (BLUPs).
While there is some variation over time within countries, the main differences remain fairly stable over time. In the
remainder of the paper, our analysis refers to roughly 2006 (ISSP Research Group 2006). We limit our sample to
respondents aged 18-65 since our macro-level measure of progressivity refers to the same working age population.
2
the diversity in the magnitude of the effect is striking (Dion and Birchfield 2010; Dion 2010).
Why is income such a strong predictor of redistributional attitudes in New Zealand, but not in
Portugal? What explains the fact that income seems to have the same effect on preferences for
redistribution in Sweden and the United States? Why does income seem to be a worse predictor
of support for redistribution in France or Finland? Why are richer respondents in Portugal more
likely than poorer respondents to express support for redistribution?
One obvious candidate to account for cross-national differences in the importance of income as a
predictor of preferences is the level of pre-tax inequality. Such expectation follows quite directly
from median voter accounts of preferences, most notably Meltzer and Richards’ (Meltzer and
Richard 1981): As the gap between the rich and the poor increases, so does the gap between the
median voter income and the average income in society, thereby fostering support for higher
taxes and redistribution. Under those circumstances the degree of polarization around the welfare
state is likely to increase, rendering income a stronger predictor of preferences for redistribution.
Interestingly, a partisan logic would also yield similar predictions (Hibbs 1977). Given an
increase of inequality, left parties will find it in their interest to target their mobilization efforts to
low income voters, whereas right parties will work to protect the interests of high income voters.
The expected outcome becomes, again, a polarization in terms of redistributive preferences along
the income scale.
Yet, however compelling in their simplicity, these logics do not seem to work. Figure 2 plots the
Gini coefficient of pre-tax income inequality (top panel) and post-tax income inequality (bottom
panel) against the size of the income effect on the citizens’ preferences for the government’s role
in bridging the differences between the rich and the poor, as reported earlier in Figure 1. Clearly,
there appears to be no systematic relation between the incidence of pre-tax inequality and the
importance of income as a predictor of redistribution.
- Figure 2 about here Likewise, the observable patterns of association between the degree of contention over
redistribution and the size of the welfare state do not make much intuitive sense. Figure 3 plots
the income slopes reported in Figure 1 against the effort devoted to social spending expressed as
a percentage of GDP (top panel) and the importance of social policy, measured by the percentage
share of public cash transfers in household disposable income (bottom panel). One may reason
that the larger the amount of transfers at stake in fiscal policy, the higher the importance of
income as a determinant of preferences. But this does not seem to be the case: there is no
correlation between the income-gradient and the size of the welfare state (measured as total
public social expenditure as % of GDP) and there is actually a positive correlation between
income gradients and the importance of transfers for household disposable income, suggesting
that income is a worse predictor of redistributional attitudes in larger welfare states.
- Figure 3 about here Figures 2 and 3 reveal the limitations of theoretical models based exclusively on redistributive
motives. They simply do not go very far in accounting for the variation in income slopes on
attitudes toward redistribution. A possible alternative builds on the notion that welfare states are
3
primarily insurance systems tailored to solve inter-temporal trade-offs in economies organized
around different types of skills (Estevez-Abé, Iversen, and Soskice 2001; Iversen and Soskice
2001; Mares 2003). Those economies organized around the production of goods intensive in
specific skills by firms competitive in international markets develop welfare states to ameliorate
the risks incurred by workers with little transferability to other sectors during economic
downturns. In contrast, workers with general skills face less risk, and their skill’s transferability
reduces their demand for comprehensive insurance programs. According to this logic, in those
countries with labor markets where general skills are prevalent, we would expect income to be a
stronger predictor of redistributive preferences. Figure 4 sets out to explore this expectation by
displaying the relationship between the income slopes (as defined above) and the incidence of
vocational training (Iversen and Soskice 2001, 888), i.e. the share of people of a certain agecohort that goes through vocational training programs. Once again, Figure 4 contains a nonfinding: the variation in income slopes does not reflect differences in the composition of skills
within the labor force across countries.
- Figure 4 about here Taken together, Figures 1 to 4 present a systematic pattern of variation that does not sit well with
the dominant theoretical approaches in the field. By drawing from and combining these and other
theoretical approaches, we put forward a different logic that points to the degree of progressivity
in the fiscal system as a key mechanism largely overlooked by the literature so far.
II. The Argument: Progressivity, Redistribution, and Preferences
Different welfare states rely on different mixes of taxes/contributions and benefits/entitlements,
and different societies are characterized by different income and risk distributions. In
conjunction, these factors influence who benefits and who loses from social policy. The
distribution of expected net benefits, in turn, should have stark consequences for the politics of
the welfare state. Some systems pitch winners against losers, while others are much less zero
sum.
The size of the welfare state – arguably the most prominent dependent variable in the social
policy literature – has no obvious connection to the distribution of expected net benefits, and
therefore no clear link to interests and politics. In fact, any given distribution of expected net
benefits can be the result of any number of combinations of transfers, taxes, inequality and risk.
For example, some social policy programs are largely paid for by the rich and largely used by the
poor. These (progressive) programs are characterized by a very uneven distribution of expected
net benefits, and we would expect them to cleave citizens along income lines. In other programs,
premiums are scaled to risk – these programs follow actuarial principles with a much more even
distribution of expected net benefits; income should be much less important in shaping attitudes.
And some public policies are financed by everybody but are disproportionally consumed by the
rich. These regressive policies should cleave citizens’ attitudes along income – but the rich
should actually favor these policies, while the poor should oppose them.
4
Of course, that the politics of welfare states isn’t just about the Benjamins has been recognized
before. The welfare regime literature uncovered different types of welfare states (EspingAndersen 1990), which finance and distribute benefits in very different ways. Social policy
preferences shape (Brooks and Manza 2007) or are being shaped by these regimes (Arts and
Gelissen 2001). The social insurance literature shows how incorporating risk into the analysis
changes our understanding of welfare state politics (Baldwin 1990; Mares 2003; Iversen and
Soskice 2001; Rehm, Hacker, and Schlesinger 2012; Moene and Wallerstein 2001; Moene and
Wallerstein 2003). Risk exposure offers a motive for affluent citizens to support social policy
programs. The small literature on taxation and social policy uncovered what became to be known
as “the paradox of redistribution” (Korpi and Palme 1998).3
However, our rich understanding of welfare state politics at the macro-level has not been
leveraged for explaining its micro-level underpinnings. We argue that the complex macro-level
interplay of financing (taxes/contributions), generosity (benefits/entitlements), inequality and
risk leads to different distributions of expected net benefits (across programs; across countries;
over time), which in turn shape individual level social policy attitudes. We further argue that the
complex macro-level mix can be approximately summarized by the concentration of benefits and
taxes across the income distribution – something we call progressivity. In particular, in more
progressive systems,4 we expect income to be more closely connected to social policy attitudes
(income slopes are more negative).
To illustrate the logic, we next present a model that builds on the canonical Meltzer-Richard
formalization (Meltzer and Richard 1981) [MR]. Citizens have an exogenously given level of
income (wi), which is taxed by function ti. Taxes are collected and handed out as a flat-rate
benefit c. Taxation leads to disincentive effects (labor supply decreases as taxation increases),
which we capture (indirectly) by function L. Individuals’ utility is then:
𝑈! = 1 − 𝑡! 𝑤! + 𝑐 − 𝐿
[1]
This formulization departs from the MR model in two ways. First, it simplifies the original
model by taking incomes as exogenous. Second and more interestingly, ti varies by individuals.
In particular, we follow de Donder and Hindrik (De Donder and Hindriks 2003) and set:
𝑡! = 𝛼 + 𝛽𝑤!
[2]
With this quadratic income tax function, individual i’s tax payment is:
𝑤! 𝑡! = 𝛼𝑤! + 𝛽𝑤! !
[2’]
Parameter alpha captures a proportional tax rate (everybody pays, say, 20% of their income as
taxes). Parameter beta is the progressivity tax parameter, with β > 0 indicating a (marginally)
3
“The more we target benefits at the poor only and the more concerned we are with creating equality via equal
public transfers to all, the less likely we are to reduce poverty and inequality” (Korpi and Palme 1998, 681–682)
4
Benefits are concentrated at the bottom of the income scale, costs are concentrated at the top of the income scale;
inequality is high and income-risk correlations are negative and high.
5
progressive income tax and β < 0 representing a (marginally) regressive one.5 In the MR model,
β=0, i.e. taxes are proportional.
Balanced budgets require that the sum of benefits equals the tax take, which implies:
𝑐 = 𝛼𝑤 + 𝛽𝑤 !
[3]
where w is the average wage.
Note that tax policies are bidimensional: benefit c depends on α and β. Without strong
assumptions it is not possible to find a median voter type solution to such a model (it is always
possible to find a majority coalition to defend the status quo, leading to cycling). However, this
simple set up allows us to explore the impact of progressivity (β) on preferred spending levels
(c). Rewriting the utility function6 and taking the derivative with respect to zero leads to the
following expression:
!!!
!"
!!
!
= − !! + 1 − !!! (𝛽𝑤𝑤! − 𝛽𝑤 ! + 𝑐)
[4]
Setting this expression to zero lets us solve for optimal c, which we call c*:
!
𝑐 ∗ = 𝑤(𝑤 − 𝑤! )(𝛽 + ! ! )
[6]
!
Ignoring w (since it is constant) leads to a simplified expression of c*:
!
𝑐 ∗ ∝ (𝑤 − 𝑤! )(𝛽 + ! ! )
[7]
!
Comparative statics on this nice expression reveal the following:
!! ∗
!!!
Or
!! ∗
!!!
with
!
!!
!
!!
= −𝛽 + ! ! (1 −
= −𝛽 + 𝜖
!
!!
!
!!
𝜖 = ! ! (1 −
)
[8]
[8’]
)
[9]
For positive values of β (i.e. in progressive systems), citizens with above average incomes
(wi>w) want zero benefits (no taxes, no transfers), while those below it want a positive benefit;
the poorer a person, the higher the preferred benefit. In somewhat regressive systems (for fairly
5
6
De Donder and Hindrik set c≥0, 0≤α≤1, and –(α/2)≤β≤(1-α)/2.
Modeling disincentive effect as L=(witi)2/2, solving [3] for α and substituting the expression into [1].
6
negative values of β, i.e. β+ε<0), the relationship between preferred benefits and income flips to
positive.7 The pogressivity parameter mediates the preference gap between different income
levels: the gap between the most and least preferred benefit level increases as β raises.8 In other
words, the income differences in preferred levels of redistribution increase in β. In more
progressive systems, income is a better (more negative) predictor of redistribution preferences.
More formally, the cross-partial with respect to β is negative:
!! ∗
!"!!!
= −1
[9]
The results in [8] and [9] summarize our argument. For illustration, we use [6] to compute c* for
different income groups, for different values of progressivity.9 Figure 5a plots optimal levels of c
against income, for different levels of progressivity (β). The fitted lines are income slopes for
these (hypothetical) scenarios. In turn, Figure 5b displaysfour of these income slopes (resulting
from different βs), in one panel.
[Figure 5a and 5b here]
Our model and the simulations based on it show how progressivity and income slopes are
connected, as suggested by our theoretical framework. Of course, our model is very abstract and
simplifies as much as possible. Moreover, we model progressivity via the tax function. In reality,
the progressivity of a social policy program is the result of many factors, as mentioned above
(financing, benefit, inequality, income-risk correlations, etc.). In our empirical analysis below we
will employ measures of progressivity that take these various sources into consideration.
We also assume the status quo level of progressivity to be exogenous and given to the
individual’s process of preference formation. We believe this to be a reasonable theoretical
assumption. This paper analyzes the effect of progressivity on income slopes, not on the origins
of progressivity. And the latter is clearly a given for any individual entering politics: what
individuals experience as the status quo are the results of previous compromises on the politics
of risk management and fiscal burden allocation.
Why does progressivity vary across countries and over time? This is an interesting question, but
not one we can comprehensively address in this manuscript. In all brevity, let us sketch our
understanding of the variation in progressivity, our key exogenous variable. Large welfare states
are funded primarily on labor and consumption taxes.10 As the fiscal burden increases, too much
progressivity in the distribution of net benefits becomes an issue for the internal stability of proredistributive coalitions.11 Workers in the upper half of the earnings distribution, a central
7
The flipping point is not β=0, but somewhat to the left of that (loosely speaking, that’s because of the disincentive
effects of taxation, which are summarized by ε)
8
Note that for positive β the most preferred level of c is zero; the gap still increases because poor citizens prefer a
positive level of c. Progressivity parameter β myltiplies that gap.
9
For simplicity, we allow for negative c.
10
For evidence backing this point see Bengtsson, Holmlund, and Waldenstrom (2012), Beramendi and Rueda
(2007), and Cusack and Beramendi (2006).
11
A related way of thinking about the implications of too high, too progressive taxes on the upper end of the
earnings distribution is in terms of a violation of Director’s Law (Stigler 1970). To the extent that an increasing
7
element of Social Democratic constituencies, grow increasingly resistant to bear the bulk of the
cost of the welfare state alone and threatend with exit towards more accommodating options.
Incumbents cannot afford to ignore the concerns of groups dominated by swing voters with high
political clout (Dixit and Londregan 1995; Dixit and Londregan 1998). Hence, a straightforward
response to ensure electoral survival and prevent/revert this potential exit of core members of the
coalition is to tax labor income less progressively and lower the levels of concentration of
transfers. This new compromise, however, only makes sense in societies with overlapping
distributions of income and risk:12 low income citizens accept access of middle and high income
citizens to benefits and agree to pay a share of the cost, whereas middle and high income citizens
accept a moderately progressive system and an increase in the size of the benefits. By contrast,
when risks and scarcity are concentrated at the bottom end of the income distribution, this
compromise becomes politically unfeasible. As a result, citizens will come into politics under
fiscal systems with high levels of concentration and relatively smaller benefits.13
Our argument suggests that these institutional differences translate directly into the process of
preference formation. What matters for understanding the role of income in preference formation
is not the level of inequality per se, nor the level of generosity, but the set of rules that govern the
allocation of costs and benefits across income groups. Pre-existing levels of progressivity shape
individual’s preferences formation. Speaking directly to the questions above, we derive the
following testable empirical implication:
The impact of income on preferences over redistribution is higher in systems with higher
levels of progressivity
The remainder of the paper offers empirical assessments of this expectation.
share of the middle classes bears a large share of the fiscal burden of the welfare state, the purported advantage of
middle income groups, regardless of the institutional environment, dwarfs relative to the rich (de facto sheltered
from taxation) and the poor (net beneficiaries of progressive tax schemes).
12
For a recent analysis of the variation in the distributions of income and risk and their implications for the welfare
state, see Rehm, Hacker, and Schlesinger (2012).
13
For systematic evidence on this patterns, see Korpi and Palme (1998), OECD (2008), Prasad and Deng (2009) and
Wilensky (2002).
8
III. Empirical Strategy
Our argument is that the tax-benefit structure in combination with the distribution of income and
risk influences the predictive power of income when it comes to attitudes towards social
insurance programs. We test this claim in four different ways:
(1) Across countries: do income-slopes (from predicting redistributional / social insurance
attitudes) vary systematically with the degree of progressivity?
(2) Across social policy domains, across countries: do income-slopes vary with the degree of
progressivity in different social policy domains in different countries?
(3) A dynamic analysis of the relationship between tax progressivity and income slopes in
Sweden during the period 1968-2010
(4) A natural experiment, based on the German case before and after re-unification, where we
can exploit an exogenous change in the degree of progressivity and assess its implications on
income slopes
Our dependent variable is the predictive power of income for social insurance attitudes. To
compute a reliable measure, we estimate hierarchical linear models predicting social insurance
attitudes with income and a set of controls (education, gender, and age), with random intercepts
and random slopes. We then recover country-specific income-slopes (and their standard errors)
from best linear unbiased predictions (BLUPs). These income-slopes are our dependent variable
(and we use the inverse of the standard errors as weights in all further country-level estimates).
Figure 1 above displayed these slopes.
Larger (more negative) coefficients indicate that social policy issues are a more salient cleavage
in a society. One advantage of this approach is that it allows us to take into account other
relevant factors: the partial correlation coefficient of social policy attitudes and income is net of
control variables. Perhaps the largest challenge for the income-slope approach is that the income
data in our public opinion survey are of problematic quality and not necessarily comparable
across countries. It is therefore reassuring that the estimated income-gradients from alternative
data-sources with better and more comparable income variables are comparable16
16
We convert the country-specific family income variables in the ISSP surveys into income noviles. But not all
countries report detailed income data. For Portugal, for example, we only have six income categories, and we
therefore drop that country from our analysis involving survey data. To get a sense of the robustness of our estimates
of the dependent variables, we also relied on the European Social Survey (ESS) (ESS 2008). In particular, the ESS
2002, 2004, 2006, and 2008 contain the following survey item: “Using this card, please say to what extent you agree
or disagree with each of the following statements: The government should take measures to reduce differences in
income levels. [The answer categories are 1 “Disagree strongly”, 2 “Disagree”, 3 “Neither agree nor disagree”, 4
“Agree”, 5 “Agree strongly”]. While the income data in the ESS 2008 survey are better suited for our purposes
(since they are reported in national income deciles), the ESS sample is restricted to European countries. We
therefore prefer to use the ISSP surveys. However, a comparison of the estimates of contestation from the two
different data sources shows a fair degree of overlap. Results are available upon request.
9
Our main source for measuring our dependent variable is the International Social Survey
Program’s (ISSP) “Role of Government” (RoG) module IV (ISSP Research Group 2006). While
income-slopes do show some variation over time (a point we will come back to below), they tend
to be fairly stable. We therefore restrict our analysis to 2006. To construct our dependent
variable, we recover income slopes on the following attitudinal items:
•
On the whole, do you think it should or should not be the government's responsibility to ...
o Reduce income differences between rich and poor
o Provide decent standard of living for the old
o Provide decent standard of living for the unemployed
o Provide health care for the sick
o Provide decent housing for those who can’t afford it
The answer categories are 1 “Definitely should not be”, 2 “Probably should not be”, 3
“Probably should be”, and 4 “Definitely should be.”
Our main focus is on the redistribution item (government responsibility to reduce income
difference between rich and poor). In turn, for the comparative analysis of contestation across
social policy domains, we rely on the items asking about government responsibility for the old,
unemployed, sick, and those in need of housing.
Our key explanatory variable is “progressivity” – parameter β in the model presented above.
How should it be measured? One prominent measure in the literature is the redistributive effect
taxes and transfers have, which is typically measured as the (proportional) reduction in the gini
coefficient comparing market and disposable income distributions (Bradley et al. 2003;
Kenworthy and Pontusson 2005).17
Standard measures capturing the overall reduction in inequality due to taxes and transfers, such
as the difference in Gini coefficients before and after taxes and transfers, provide a summary of
the scope of redistribution but do not speak to the directionality of the policy effects behind
observable redistribution nor about the subspace of the income distribution in which the
reallocation of resources actually takes place. For instance a 10 percent reduction in pre-tax
inequality may reflect transfers from the top to the middle, from the middle to the bottom, or
from the top to the bottom. These are three very different scenarios in terms of progressivity and
the politics of social policy – and yet the overall reduction in the Gini coefficient for pre-tax
inequality offers no leverage to distinguish between them. This is not to say though that
redistribution and progressivity are unrelated.
Following Kakwani and Lambert the overall redistributive impact of the fiscal system –
measured by the difference between market and disposable income gini coefficients – can be
broken into two components (Kakwani 1977; Kim and Lambert 2009): the scope of the effort
and its progressivity. The relationship can be formally stated as follows:19
17
In the literature on measurement of inequality, this is known as the Reynolds-Smolensky index.
The first equality has been established by Kakwani (Kakwani 1977, equation 3.2), the second by Lambert (Kim
and Lambert 2009, equation 3).
19
10
𝐺𝑖𝑛𝑖 !"#$%& − 𝐺𝑖𝑛𝑖 !"#$%#&'() =
!
!!!
𝛽=
!! ! !! ! !
!!!!!
[10]
where t denotes the tax level, b denotes the benefit level and βT and βB indicate the progressivity
of taxes and benefits. Assuming balanced budgets (t=b=α), we can rewrite this equality as:
𝐺𝑖𝑛𝑖 !"#$%& − 𝐺𝑖𝑛𝑖 !"#$%#&'() = 𝛼(𝛽 ! + 𝛽 ! )
[11]
In words: the overall reduction in inequality due to taxes and transfers can be decomposed into
the product of the size of the welfare state (α) and the progressivity of its taxes (βT) and benefits
(βB). It is our contention that progressivity is a central ingredient of welfare state politics.
The literature has followed Kakwani in measuring βT and βB. Kakwani defines progressivity as a
tool to “measure deviations of the tax [or transfer] system from proportionality” (Kakwani 1977,
74), which are commonly captured by concentration curves. The concentration of taxes is
derived by plotting the share of taxes paid against rank-ordered income groups. In a progressive
system, those at the bottom of the income scale payer a lower share of taxes and the the
concentration curve is below the 45-degree line, while it is above the 45-degree line for richer
income groups. The concentration coefficient of taxes (βT) sums the area between the
concentration curve and the 45-degree line in a way that more positive values indicate more
progressive systems.
Progressivity of benefits is measured analogously. The concentration of benefits is derived by
plotting the share of benefits received against rank-ordered income groups. In a progressive
system, those at the bottom of the income scale receive a higher share of benefits and the
concentration curve is above the 45-degree line, while it is below the 45-degree line for richer
income groups. The concentration coefficient of benefits (βB) sums the area between the
concentration curve and the 45-degree line, where the area above the 45-degree line has a
negative sign, while the area below the 45-degree line has a positive sign. More negative values
indicate more progressive systems.
To arrive at a measure of overall progressivity, we take the concentration coefficient of taxes
(βT) and subtract the absolute value of the concentration coefficient of benefits (βB). Higher
values now indicate more progressive systems.
In the empirical analysis below, we will capture progressivity by the concentration of benefits
(βB) as well as the combined progressivity measure (βT-|βB|). In both cases, a value of zero
indicate proportionality. With respect to the concentration of benefits, negative values indicate
more progressive benefits systems, and we expect a positive correlation between income slopes
(welfare state contestation) and that measure. With respect to the combined measure of
progressivity – taking into account taxes and benefits – positive values indicate more progressive
systems, and we expect a negative relationship between income slopes and overall progressivity.
We take these measures from the OECD (Förster and Whiteford 2009; OECD 2008), and they
refer to the mid-2000s. The OECD also provides concentration measures of cash benefits for
various social policy domains. Together with an indicator of the importance of benefits
11
(percentage share of public cash transfers in household disposable income), these measures are
displayed in Table 1.
-
Table 1 about here Our theoretical framework predicts that the importance of social policy as a political cleavage
will be larger, the higher the levels of progressivity, that is the higher the concentration of net
benefits in the lower end of the pre-tax and transfer distribution. It also indicates that we need to
control for the size of benefits and taxes (alpha) to isolate the impact of progressivity on income
slopes. 20 But, clearly, the tax-benefit structure is not the only plausible factor influencing how
contested welfare states are. We already noted that three plausible explanatory variables are
income inequality (higher inequality should lead to more contestation, per Meltzer-Richard), the
magnitude of the welfare state (the more is at stake, the more contested should welfare states be),
and the incidence of vocational training (the large the incidence of vocational training, the less
contested social policy). Neither of these variables performed well in bivariate analysis (see
Figures 2-4), but it nevertheless may make sense to include them as control variables.
While American Politics scholars focus a lot on contestation of preferences (Evans 2003; Evans
2003), parties (the literature is extremely voluminous; for a review, see Hetherington 2009) or
both, there are only few comparative contributions that explore polarization of social policy
attitudes (Rehm, Hacker, and Schlesinger 2012). However, there is a growing literature exploring
the link between income and voting (De La O and Rodden 2008; Huber and Stanig 2007), which
offers some clues regarding potential determinants of income. The main argument in this
subfield focus around the claim that a second dimension “distracts” the poor from their
“objective” interests and/or limits the formation of comprehensive pro-redistributive coalitions
by introducing group differentials what otherwise would be homogenous class/income groups
(Shayo 2009).
The usual suspect explanatory variables for second dimension are religion and ethnic
fractionalization. We therefore include a control variable that measures religious
fractionalization (Alesina et al. 2003). We also explored measures of religiosity that contain the
share of confessional groups (such as Catholics or Protestants), with data from the Religion and
State Project (Fox 2004; Fox 2008), with very similar (non-)results. We also control for ethnic
fractionalization (Alesina et al. 2003) to account for the possibility that ethnic issues alter the
link between income and redistributional preferences. Additional robustness checks with control
variables containing information on immigration stocks and flows (such as inflow of asylum
seekers, inflow of foreigners, stock of foreign-born population, etc.) do not change the results we
report below.
Since it is plausible that attitudes towards government intervention are shaped by dominant (left)
parties and unions, we include control variables for left party dominance (cumulative share of
cabinet posts for left parties) and trade union density.
20
Equation [11] above suggests that we should control for the overall size of the welfare state, which we do when
possible.
12
Finally, the overall level of risk in a society may shape attitudes towards social policy and have
an impact on the link between income and redistribution preferences. To take this possibility into
account, we control for the unemployment rate. Table 2 lists all variables and their sources.
- Table 2 about here IV. Findings
We now turn to the results. We are primarily concerned with the correlation of welfare state
contestation and the structure of tax-benefit systems (the who-gets-what-at-which-prize
question). Because the probability of receiving benefits and/or the size of benefits vary more
across countries and social-policy domains than the financing of these benefits, and in order to
keep the number of figures at a manageable number, we will concentrate on two explanatory
variables only: the concentration of cash benefits and the concentration of net benefits, namely
joint impact of benefit and financing concentration (“overall progressivity”).
Figure 6 displays bivariate correlations between these two core explanatory variables and the
income-slopes. The figures show that the more redistributive benefits, the more contested they
are. This is what we expected from our theoretical framework. There are also good reasons why
the results are generally stronger when we solely look at the benefit structure (top panel) as
opposed to both the benefit and financing structure (bottom panel). As mentioned above, benefit
structures vary more across countries than tax structures. This certainly does not imply that
financing structures can or should be neglected. But it implies that in our empirical
investigations, their effects are smaller than those generated by the structure of benefits.
- Figure 6 about here The close fit between income-slopes and the concentration of (taxes and) benefits, as displayed
in Figure 6 is remarkable. We now explore whether these correlations withstand the inclusion of
(more or less plausible) control variables. As elaborated above, we include the following
controls: income inequality; magnitude of transfers; incidence of vocational training; religious
fractionalization; ethnic fractionalization; left party dominance; trade union density; and the
unemployment rate. We include these control variables one at a time. Table 3 displays the results
when we predict income-slopes with the concentration of cash benefits, while Table 4 has the
results when we predict income-slopes with overall progressivity (concentration of taxes –
concentration of benefits).
- Table 3 and Table 4 around here The results are easy to report: the key explanatory variables in Table 3 and Table 4 – the
concentration of benefits and overall progressivity, respectively – turn out to be statistically
significant in all models, while none of the control variables is found so. The substantive impact
of cash benefit concentration is also significant. The estimated slope of around 0.2 suggests that
a one standard deviation (0.19) change in the concentration of benefits changes the income slope
13
by about 0.038. Since the income-slope ranges from about to -0.149 to about -0.016, this is a
23% change.21
Based on our theoretical framework, we expected cash benefit concentration and overall
progressivity to be statistically and substantively significant predictors of the saliency of social
policy as a political issue. In contrast, the finding that none of the more or less standard variables
in the comparative political economy literature took us by surprise.
We turn now to the analysis disaggregated by policy domain (unemployment, pensions, health
care, and housing). To this end, we need to match survey items on social policy domains to
concentration measures of social policy domains. While there is often more than one possible
match, many mappings are straightforward. In particular, the ISSP RoG surveys include the
following social policy attitudinal items22 that can be easily matched with concentration
measures of benefits:
• Provide decent standard of living for the unemployed → concentration of unemployment
benefits
• Provide decent standard of living for the old → concentration of old age pensions
• Provide health care for the sick → concentration of disability benefits
• Provide decent housing for those who can’t afford it → concentration of housing benefits
When we explore the correlation between income-slopes (from multi-level models predicting a
social policy attitude) and concentration measures within each of these domains, we find the
expected positive correlations (the more concentrated benefits towards the poor, the more
contested the social policy area). Figure 7 shows the results, pooling the four social policy
domains together.
- Figure 7 about here Each symbol in the scatter-plot represents one of four social policy domains, for a given country:
unemployment (U), pensions (P), disability benefits (D), or housing benefits (H). As can be seen
from the Figure, there is a positive correlation between the concentration of benefits and the
income-slopes, within each domain (dotted lines) and overall (solid line). The overall correlation
is quite strong, and statistically significant (whether we include fixed effects for social policy
domains or not), as was the case with the overall progressivity measures.
To recapitulate, we have tested our hypothesis that social policy polarization (measured in terms
of the strength of income as a predictor of attitudes) can be explained by the progressivity of the
tax-benefit system. Our findings suggest, in line with our theoretical expectations, a strong
impact of the level of progressivity of the tax-transfer system. This finding is robust across
measures of concentration (cash benefits vs. benefits and taxes), income measures, and policy
domains, as well as to the inclusion of relevant control variables.
21
Slopes: Mean=-.072; SD=.042; Min=-.149; Max=.016; N=21. Concentration of cash benefits: Mean=-.11;
SD=.19; Min=-.431; Max=.315; N=21
22
These are based on ISSP survey items, with the following wording stem: “On the whole, do you think it should or
should not be the government's responsibility to ...”. The answer categories are 1 “Definitely should not be”, 2
“Probably should not be”, 3 “Probably should be”, and 4 “Definitely should be.”
14
These are solid empirical grounds. Yet, however robust, the joint endogeneity between
progressivity and support for redistribution continues to loom large over these findings. None of
the results identify in full the causal effect of progressivity on preferences for redistribution.
Sadly, the field lacks enough data points to undertake the type of analysis that would bring us
closer to that point in a systematic and generalizable way. Two potential concerns stand out:
first, a necessary yet insufficient condition for a causal effect of progressivity on income slopes
is that these two variables co-vary over time; second, for causality to be identified, an exogenous
increase (decrease) of progressivity should lead to a subsequent increase (decrease) in the
magnitude of income slopes. We address each of these two points in turn.
Unfortunately, data limitations prevent us from performing a viable time-series crosssectional analysis. However, in some cases, there are enough data on both progressivity and
income slopes to evaluate the relationship over time. We make use of a time series of
progressivity indicators in Sweden from 1968-2009. Bengtsson et al. (2012) compute these
indicators on the basis of LINDA, a dataset that captures a 3.35 percent random sample of the
Swedish population and incorporates demographic information as well as entries for all tax
payments and deductions registered by tax authorities and population censuses. This source
allow us to trace the evolution of income tax progressivity in Sweden for over forty years.
During this time, two major reforms took place. The first one, in 1971, increased significantly
the levels of income tax progressivity. Before 1971 the Swedish tax system had two distinctive
features: first. couples and singles were taxed under two different schedules; second,
proportional local income taxes were deductible from national income taxes. The 1971 reform
made all income from employment taxable on the basis of a single scale, regardless of marital
status, and eliminated the possibility to deduct local taxes from national taxes, thereby
significantly increasing the levels of progressivity. The second major reform, enacted by the first
conservative government to gain office in decades, took place in 1991 and worked in the
opposite direction. The 1991 tax reform enacted a massive reduction of the top marginal tax rate
from 80 to 50 percent and a simplification of the tax code such that about 85 percent of the
population were no longer required to file an income tax (Steinmo 2002: 850; Bengtsson et al.
2012). Instead, all tax payers pay a proportional tax of 20% (which eventually increased to 30%
in most districts) and the top 20% of income earners pay an additional national tax rate or either
20 or 25 % depending on their pre-tax income level. In addition, capital and corporate marginal
tax rates dropped significantly. Overall, the reform reduced massively the pre-existing levels of
progressivity in the system.
Accordingly, we expect the income slopes to increase after the 1971 reforms and to decrease
after 1991. Figures 8 and 9 map the co-evolution between different measures of progressivity
and the evolution of income slopes on preferences for redistribution in Sweden. Figures 8 and 9
plot the evolution of various measures of progressivity and redistribution between 1968 and 2009
along with the income slopes on preferences for redistribution in twelve Swedish National
Election Studies between 1964 and 2006 on the level of progressivity.23 In turn, Figure 10
23
The dependent variable comes from a question that asks respondents to react to the following statement: “Social
reforms in this country have gone so far that the state ought to reduce rather than increase social benefits and support
for people”. Respondents have four options: 1 "agree completely"; 2 "agree on the whole"; 3 "disagree on the
15
regresses the income slopes on the progressivity indicators to test if the expected relationship
holds.
[Figures 8,9, and 10 here]
The longue durée analysis of the Swedish case reveals a pattern consistent with our theoretical
expectations. As the progressivity of the fiscal system increased between 1972 and 1990, so did
the income slopes. By contrast, the 1991 reform caused a change in trends that reduced and
ultimately stabilized the importance of income as a predictor of preferences for redistribution.
The change in trends is less sharp than one would ideally like to see for two reasons. First, the
change in progressivity is contemporaneous to a major economic crisis that triggers automatic
staibilizers and reduces the tax contributions of low income citizens (via unemployment effects).
Hence, the effects of the reform did take some time to become apparent. Second, the return of the
Social Democrats to power in 1994 implied a marginal adjustment of the 1991 Tax Reform.
While the fundamental spirit of the 1991 reform, that is the need to preserve high wage earners
for paying an exceedingly high share of the tax burden, remained in place, the new government
introduced a series of changes that scaled back the scope of the reform. These adjustments
included a 5% increase in the top marginal tax rate and a drop in the VAT for food of 50%
(Steinmo 2002:852). These changes slowed down the reduction of progressivity and its impact
on income slopes. Overall, however, there is clear, strong and negative relationship between the
degree of progressivity24 and the impact of income slopes in Sweden over time, as captured by
figure 10. Higher levels of progressivity imply a stronger negative impact of income, as
predicted by our argument.
This result, however, does not address the joint endogeneity between progressivity and
preferences for redistribution over time as neither the 1971 nor the 1991 reforms can be said to
be exogenous with respect to the pre-existing distribution of preferences. To address this
problem we exploit German Reunification (1989-1990) as a natural experiment where
by virtue of sharp increases in the pool of recipients from the East and ad hoc modifications of
the tax and transfers system the levels of progressivity increased sharply. The West incorporated
five new poorer Länder and twenty million new and largely poorer citizens. The fundamentals of
the fiscal system did not change, despite a massive alteration in the geography of income and
labor markets (Beramendi 2012). Rather, it assimilated the new members in a short period of
time, triggering an unprecedented redistributive effort from the West to the East. The levels of
progressivity increased through multiple mechanisms, including changes in the tax code (e.g. the
introduction of a Solidarity Surcharge of 7.5% on personal and corporate income in 1996, met by
Western tax payers) and the allocation of massive additional benefits (b in the model above) to a
massive amount of individuals incapable of contributing to the revenue pool. Accordingly, our
theory predicts increase in the income slope on preferences for redistribution in re-unified
Germany.
whole"; 4 "disagree completely" A negative slope for income indicates that wealthier respondents are more likely to
agree with the statementIn turn, income is measured as a five scale variable: 1 "low income"; 2 "fairly low income";
3 "neither low nor high income"; 4 "fairly high income"; 5 "high income". The analysis also controls for age,
gender, marital status, education, and whether the respondent is affiliated with a union.
24
As measured by the Kakwani Index for taxes (source: Bengtsson et al. 2012)
16
To assess the validity of this prediction, we make use of the German General Social Survey.
In particular, we explore the predictive power of income on two variables for which we have
consistent questions before and after Re-unification. The first one asks whether social benefits
should be cut or extended (1-3 scale, with 3 being pro-expansion); the second one asks whether
the state should secure income levels in times of crisis (1-4 scale, with 4 reflecting strong
agreement with the statement). Figures 11 and 12 present the income slopes on both questions.
The analysis breaks the sample into East and West Germany.
[Figures 11 and 12 here]
The evolution of income slopes in Germany after Re-unification validates our theoretical
expectation. The income slope becomes a much stronger predictor in both East and West
Germany in response to the shift in the levels of progressivity exogenously triggered by the
incorporation of the East. Noticeably, there is a consistent difference in the size of the income
slope between the East and the West which suggests that richer citizens oppose redistribution in
both regions, but they do so to a much lesser extent in East Germany. This gap may reflect the
cumulative effect of several factors: citizens in East Germany being the net beneficiaries of
territorial transfers, or simply more exposed to the negative externalities of inequality and hence
more tolerant towards palliative measures (in large part funded by the West). Interestingly,
though, it also suggests that citizens understand well the implications of the fiscal system given
their geographical context, and adjust their preferences accordingly.
V. Conclusion
We began this paper by noting the absence of relation between the level of pre-tax inequality and
the importance of income as a predictor of redistribution and a slight negative relationship
between the size of redistributive transfers and the predictive capacity of income. Both patterns
are puzzling from the perspective of standard models of redistribution: if anything, one would
expect income to be a better predictor of redistribution the higher the level of inequality and the
higher the stakes in terms of redistribution. What explains these puzzling findings?
Addressing this issue requires establishing the conditions under which income is a good
predictor of redistributive preferences, that is to say, the conditions under which preferences
about income redistribution are more polarized. Our approach in this paper is straightforward:
the key factor that explains contentions over redistribution is the design of the tax and transfers
system. Progressivity is inversely related to the size of redistributive transfers and positively
related to the level of conflict about the welfare state. A higher concentration of taxes and
transfers produces higher levels of polarization over redistribution, boosting up the importance of
income as a predictor of redistributive preferences. At the same time, higher levels of
concentration come at the price of a reduction in the pool of resources available for
redistribution.
These findings are quite robust and suggest significant payoffs to modeling preferences, in
contrast to earlier contributions, as a joint function of risk, benefits (transfers) and cost (taxes).
17
Nonetheless, neither the logic nor the findings in this paper are devoid of limitations. In closing
the paper we highlight three important issues as pointers for future research efforts.
First, the shadow of endogeneity continues to loom large over this paper. The idea that
progressivity may itself reflect pre-existing levels of support for redistribution motivates this
concern. We have presented evidence on the US and Germany that reveals an impact of
progressivity on preferences in circumstances where the case for exogeneity is plausible. This
evidence would suggest then that the concerns about endogeneity apply only in the long run. A
sizeable stock of research indicates that fiscal institutions have remained pretty sticky since their
consolidation between the inter-war and the post-war period. They do change, but very slowly
and within their own logic, in ways heavily constrained by their own feedback effects on
political coalitions. Thus, we believe that our assumption that citizens evaluate the expected net
benefits given a rather stable pre-existing design remains plausible. Having that said, however,
progressivity is itself the outcome of political contentions in the medium- to long-run. A natural
next step in this agenda is to explain the political origins of different equilibria in terms of
concentration and generosity.
Second, we have argued that the expected benefits of social policy are captured by the product of
p (risk, or share of time on benefits) and b (the size of the benefit), while its costs can be
understood as the product of t (taxes and social contributions) and the probability (or share of
time) of not experiencing a bad event (1-p). We measure the former with the concentration of
benefits, while the latter is measured by the concentration of taxes and contributions. These
measures strike us as good proxies, but future research should disentangle and measure the three
parameters (p, b, t) directly. At a minimum, this would give a richer picture of who gets what at
which prize.
Third, while our paper makes use of individual-level data to measure the dependent variables and
our core explanatory variables, it does not directly test the micro-level mechanisms that our logic
assumes in terms of preference formation. In particular, we did not shed light on the question
whether our three parameters (p, b, t) have the postulated effects on citizens’ social policy
attitudes. To be sure, previous research has established some of these individual-level links, but –
as far as we know – not all of them simultaneously. We doubt that there are cross-national
surveys that provide agent-specific estimates for p, b, and t. But we think these parameters can
be estimated at cohort-levels (such as income groups, or occupations), and then merged into
existing public opinion surveys. This seems one fruitful way to pursue the mechanisms
underpinning the link between material interests and preference formation.
18
References
Alesina, Alberto, and George-Marios Angeletos. 2005. “Fairness and Redistribution.” American
Economic Review 95 (4): 960–980.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain
Wacziarg. 2003. “Fractionalization.” National Bureau of Economic Research Working
Paper Series No. 9411. http://www.nber.org/papers/w9411.
Alesina, Alberto, and Eliana La Ferrara. 2005. “Preferences for Redistribution in the Land of
Opportunities.” Journal of Public Economics 89 (5-6): 897–931.
Alesina, Alberto, Edward Glaeser, and Bruce Sacerdote. 2001. “Why Doesn’t the United States
Have a European-Style Welfare State?” Brookings Papers on Economic Acticity 2001
(2): 187–254.
Armingeon, Klaus, Panajotis Potolidis, Marlène Gerber, and Philipp Leimgruber. 2009.
Comparative Political Data Set 1960-2007. Berne, Switzerland: Institute of Political
Science, University of Berne.
Arts, Wil, and John Gelissen. 2001. “Welfare States, Solidarity and Justice Principles: Does the
Type Really Matter?” Acta Sociologica 44: 283–299.
Atkinson, Anthony. 1980. “Horizontal Equity and the Distribution of the Tax Burden.” In The
Economics of Taxation, ed. Henry Jacob Aaron and Michael J. Boskin, 3–18. Brookings
Institution Press.
Baldwin, Peter. 1990. The Politics of Social Solidarity: Class Bases of the European Welfare
State, 1875-1975. Cambridge, MA: Cambridge University Press.
Bean, Clive, and Elim Papadakis. 1998. “A Comparison of Mass Attitudes Towards the Welfare
State in Different Institutional Regimes, 1985-1990.” International Journal of Public
Opinion Research 10 (3) (September 1): 211–236. doi:10.1093/ijpor/10.3.211.
Bénabou, Roland, and Efe A. Ok. 2001. “Social Mobility and the Demand for Redistribution:
The POUM Hypothesis.” Quarterly Journal of Economics 2001 (May): 447–487.
Bénabou, Roland, and Jean Tirole. 2006. “Belief in a Just World and Redistributive Politics.”
The Quarterly Journal of Economics 121 (2) (May 1): 699 –746.
doi:10.1162/qjec.2006.121.2.699.
Bengtsson, Niklas, Bertil Holmlund, and Daniel Waldenstrom. 2012. “Lifetime Versus Annual
Tax Progressivity: Sweden, 1968-2009.”
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2098702.
Beramendi, Pablo. 2012. The Political Geography of Inequality. New York: Cambridge
University Press.
Beramendi, Pablo, and David Rueda. 2007. “Social Democracy Constrained: Indirect Taxation in
Industrialized Democracies.” British Journal of Political Science 37 (04): 619–641.
Bradley, David, Evelyn Huber, Stephanie Moller, François Nielsen, and John D. Stephens. 2003.
“Distribution and Redistribution in Postindustrial Democracies.” World Politics 55 (2):
193–228.
Brooks, Clem, and Jeff Manza. 2006a. “Why Do Welfare States Persist?” Journal of Politics 68
(4): 816–827.
———. 2006b. “Social Policy Responsiveness in Developed Democracies.” American
Sociological Review 71 (June): 474–494.
———. 2007. Why Welfare States Persist: The Importance of Public Opinion in Democracies.
University Of Chicago Press.
19
Corneo, Giacomo, and Hans Peter Gruner. 2000. “Social Limits to Redistribution.” American
Economic Review 90 (5): 1491–1507.
Cusack, Thomas, and Pablo Beramendi. 2006. “Taxing Work.” European Journal of Political
Research 45 (1): 43–73.
Cusack, Thomas, Torben Iversen, and Philipp Rehm. 2006. “Risks at Work: The Demand and
Supply Sides of Government Redistribution.” Oxford Review Economic Policy 22 (3):
365–389. doi:10.1093/oxrep/grj022.
Dion, Michelle L. 2010. “When Is It Rational to Redistribute? A Cross-national Examination of
Attitudes Toward Redistribution.”
http://polmeth.wustl.edu/mediaDetail.php?docId=1232.
Dion, Michelle L., and Vicki Birchfield. 2010. “Economic Development, Income Inequality, and
Preferences for Redistribution1.” International Studies Quarterly 54 (2): 315–334.
doi:10.1111/j.1468-2478.2010.00589.x.
Dixit, Avinash, and John Londregan. 1995. “Redistributive Politics and Economic Efficiency.”
American Political Science Review 89 (4): 856–866. doi:10.2307/2082513.
———. 1998. “Ideology, Tactics, and Efficiency in Redistributive Politics.” Quarterly Journal
of Economics 113 (2): 497–529.
De Donder, Philippe, and Jean Hindriks. 2003. “The Politics of Progressive Income Taxation
with Incentive Effects.” Journal of Public Economics 87 (11) (October): 2491–2505.
doi:10.1016/S0047-2727(02)00051-8.
Esping-Andersen, Gosta. 1990. The Three Worlds of Welfare Capitalism. Princeton, NJ:
Princeton University Press.
ESS. 2008. European Social Survey Round 4 Data. Data File Edition 4.0. Norway: Norwegian
Social Science Data Services. http://ess.nsd.uib.no/ess/conditions.html.
Estevez-Abé, Margarita, Torben Iversen, and David Soskice. 2001. “Social Protection and the
Formation of Skills: A Reinterpretation of the Welfare State.” In Varieties of Capitalism  :
The Institutional Foundations of Comparative Advantage, ed. Peter A. Hall and David
Soskice, 145–183. Oxford University Press.
Evans, John H. 2003. “Have Americans’ Attitudes Become More Polarized? An Update.” Social
Science Quarterly 84 (1): 71–90.
Fong, Christina. 2001. “Social Preferences, Self-interest, and the Demand for Redistribution.”
Journal of Public Economics 82: 225–246.
Förster, Michael, and Peter Whiteford. 2009. “How Much Redistribution Do Welfare States
Achieve? The Role of Cash Transfers and Household Taxes.” CESifo DICE Report 7 (3).
CESifo DICE Report: 34–41.
Fox, Jonathan. 2004. The Religion and State Project. Ramat Gan, Israel: Bar Ilan University.
http://www.thearda.com/ras/.
———. 2008. A World Survey of Religion and the State. 1st ed. Cambridge, MA: Cambridge
University Press.
Gelissen, John. 2000. “Popular Support for Institutionalised Solidarity: a Comparison Between
European Welfare States.” International Journal of Social Welfare 9 (4): 285–300.
doi:10.1111/1468-2397.00140.
———. 2002. Worlds of Welfare, Worlds of Consent? Public Opinion on the Welfare State.
Brill.
Hetherington, Marc J. 2009. “Review Article: Putting Polarization in Perspective.” British
Journal of Political Science 39 (02): 413–448. doi:10.1017/S0007123408000501.
20
Hibbs, Douglas A. 1977. “Political Parties and Macroeconomic Policy.” American Political
Science Review 71 (December): 1467–1487.
Huber, John, and Piero Stanig. 2007. “Why Do the Poor Support Right-wing Parties? A Crossnational Analysis.”
ISSP Research Group. 2006. International Social Survey Programme (ISSP): Role of
Government IV. Distributor: GESIS Cologne Germany ZA4700, Data Version 1.0 (200808-18).
Iversen, Torben, and David Soskice. 2001. “An Asset Theory of Social Policy Preferences.”
American Political Science Review 95 (December): 875–895.
Jaeger, Mads Meier. 2006. “Welfare Regimes and Attitudes Toward Redistribution: The Regime
Hypothesis Revisited.” European Sociological Review 22 (2): 157–170.
———. 2009. “United But Divided: Welfare Regimes and the Level and Variance in Public
Support for Redistribution.” European Sociological Review 25 (6) (December 1): 723–
737. doi:10.1093/esr/jcn079.
Jakobsen, Tor. 2010. “Welfare Attitudes and Social Expenditure: Do Regimes Shape Public
Opinion?” Social Indicators Research. doi:10.1007/s11205-010-9666-8.
http://dx.doi.org/10.1007/s11205-010-9666-8.
Kakwani, Nanak C. 1977. “Measurement of Tax Progressivity: An International Comparison.”
The Economic Journal 87 (345) (March 1): 71–80. doi:10.2307/2231833.
Kangas, Olli E. 1997. “Self-Interest and the Common Good: The Impact of Norms, Selfishness
and Context on Social Policy Opinions.” Journal of Socio-Economics 26 (5): 475–494.
———. 2003. “The Grasshopper and the Ants: Popular Opinions of Just Distribution in
Australia and Finland.” Journal of Socio-Economics 31: 721–743.
Kangas, Olli E., David Miller, Per Arnt Petterssen, Stefan Svallfors, and Peter Taylor-Gooby.
1995. In the Eye of the Beholder. Opinions on Welfare and Justice in Comparative
Perspective. Umea: Scandbook.
Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of
Redistribution in Affluent Countries.” Perspectives on Politics 3 (3): 449–471.
Kim, Kinam, and Peter J Lambert. 2009. “Redistributive Effect of U.S. Taxes and Public
Transfers, 1994-2004.” Public Finance Review 37 (1) (January 1): 3–26.
doi:10.1177/1091142108324423.
Korpi, Walter, and Joakim Palme. 1998. “The Paradox of Redistribution and Strategies of
Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries.”
American Sociological Review 63 (5): 661–687.
De La O, Ana L., and Jonathan A. Rodden. 2008. “Does Religion Distract the Poor?”
Comparative Political Studies 41 (4-5) (April 1): 437 –476.
doi:10.1177/0010414007313114.
Lambert, Peter. 2001. The Distribution and Redistribution of Income: Third Edition. Manchester
University Press.
Lupu, Noam, and Jonas Pontusson. 2011. “The Structure of Inequality and the Politics of
Redistribution.” American Political Science Review 105 (02): 316–336.
doi:10.1017/S0003055411000128.
Luttmer, Erzo F. P. 2001. “Group Loyalty and the Taste for Redistribution.” Journal of Political
Economy 109 (3): 500–528.
Mares, Isabela. 2003. The Politics of Social Risk. Business and Welfare State Development.
Cambridge, MA: Cambridge University Press.
21
Mehrtens, F. John. 2004. “Three Worlds of Public Opinion? Values, Variation, and the Effect on
Social Policy.” International Journal of Public Opinion Research 16 (2): 115–143.
doi:10.1093/ijpor/16.2.115.
Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.”
Journal of Political Economy 89 (October): 914–927.
Moene, Karl O., and Michael Wallerstein. 2001. “Inequality, Social Insurance, and
Redistribution.” American Political Science Review 95 (4): 859–874.
———. 2003. “Earnings Inequality and Welfare Spending. A Disaggregated Analysis.” World
Politics 55 (July): 485–516.
OECD. 2008. “How Much Redistribution Do Governments Achieve? The Role of Cash
Transfers and Household Taxes.” In Growing Unequal? Income Distribution and Poverty
in OECD Countries, ed. OECD. Paris: OECD.
Piketty, Thomas. 1995. “Social Mobility and Redistributive Politics.” Quarterly Journal of
Economics 110 (3): 551–584.
Prasad, Monica, and Yingying Deng. 2009. “Taxation and the Worlds of Welfare.” SocioEconomic Review 7 (3): 431 –457. doi:10.1093/ser/mwp005.
Rehm, Philipp. 2009. “Risks and Redistribution: An Individual-Level Analysis.” Comparative
Political Studies 42 (7): 855–881. doi:10.1177/0010414008330595.
Rehm, Philipp, Jacob S. Hacker, and Mark Schlesinger. 2012. “Insecure Alliances: Risk,
Inequality, and Support for the Welfare State.” American Political Science Review 106
(2): 386–406.
Scheve, Kenneth, and David Stasavage. 2006. “Religion and Preferences for Social Insurance.”
Quarterly Journal of Political Science 1 (3): 255–286.
Shayo, Moses. 2009. “A Model of Social Identity with an Application to Political Economy:
Nation, Class, and Redistribution.” American Political Science Review 103 (02): 147–
174. doi:10.1017/S0003055409090194.
Sinn, Hans-Werner. 1995. “A Theory of the Welfare State.” Scandinavian Journal of Economics
97 (4): 495–526.
Stigler, George J. 1970. “Director’s Law of Public Income Redistribution.” Journal of Law and
Economics 13 (1): 1–10.
Svallfors, Stefan. 1997. “Worlds of Welfare and Attitudes to Redistribution: A Comparison of
Eight Western Nations.” European Sociological Review 13 (3): 283–304.
———. 2004. “Class, Attitudes and the Welfare State: Sweden in Comparative Perspective.”
Social Policy & Administration 38 (2): 119–138.
Varian, Hal R. 1980. “Redistributive Taxation as Social Insurance.” Journal of Public
Economics 14: 49–68.
Wilensky, Harold L. 1975. The Welfare State and Equality: Structural and Ideological Roots of
Public Expenditures. University of California Press.
———. 2002. Rich Democracies: Political Economy, Public Policy, and Performance.
University of California Press.
22
Figures and Tables
Figure 1: Income as predictor of redistributional attitudes
NZL
DNK
AUS
CAN
NLD
USA
NOR
CZE
GBR
SWE
DEU
FIN
FRA
CHE
JPN
POL
IRL
HUN
ESP
KOR
PRT
-.15
-.1
-.05
0
Slope of income from regression:
Redistribution = Income + Controls
.05
Note:
Shown are BLUPs (Best Linear Unbiased Predictors) from multi-level models.
Based on ISSP 2006.
Preferences for redistribution: “On the whole, do you think it should be or should not be the government’s
responsibility to reduce income differences between the rich and poor” [1. Definitely should not be; 2. Probably
should not be; 3. Probably should be; 4. Definitely should be].
Controls are education, gender, and age. Sample is restricted to employed respondents aged 18-65.
Standard errors are as follows: AUS (0.014), CAN (0.016), CHE (0.016), CZE (0.016), DEU (0.014), DNK (0.014),
ESP (0.011), FIN (0.015), FRA (0.014), GBR (0.018), HUN (0.017), IRL (0.017), JPN (0.015), KOR (0.013), NLD
(0.015), NOR (0.013), NZL (0.016), POL (0.014), PRT (0.014), SWE (0.014), USA (0.012).
23
Figure 2: Income inequality and income slopes
.05
PRT
Income slopes
0
KOR
IRL
-.05
POL
JPN
CHE
FIN
FRA
DEU
SWE
-.1
CZEGBR USA
NOR
NLD CAN
AUS
DNK
NZL
-.15
.3
.35
.4
.45
.5
Gini coefficient (before taxes and transfers) age 1865
.55
Coef=.07, SE=.19, t=.4, adj. R2=-.05, N=19
Income: Family income in noviles
.05
PRT
Income slopes
0
KOR
ESP
HUN
IRL
-.05
POL
JPN
CHEFRA
FIN
DEU
SWE
CZENOR
NLD
-.1
DNK
GBR
CAN
USA
AUS
NZL
-.15
.2
.25
.3
.35
Gini coefficient (after taxes and transfers) age 1865
.4
Coef=.31, SE=.22, t=1.4, adj. R2=.05, N=21
Income: Family income in noviles
24
Figure 3: Welfare state size and income slopes
.05
Income slopes
0
KOR
ESP
HUN
IRL
-.05
POL
JPN
CHE
-.1
USA
CAN
FIN
DEU
SWE
CZE GBR
NOR
NLD
AUS
FRA
DNK
NZL
-.15
5
10
15
20
25
Total Public Social Expenditure (%GDP)
30
Coef=0, SE=0, t=-1, adj. R2=0, N=20
Income: Family income in noviles
.05
PRT
Income slopes
0
ESP
HUN
IRL
-.05
POL
JPN
FIN
CHE
FRA
DEU
-.1
USA
GBR
CAN
SWE
NLD
AUS
-.15
NOR
CZE
DNK
NZL
10
15
20
25
30
35
Percentage share of public cash transfers in household disposable income
Coef=0, SE=0, t=2, adj. R2=.13, N=20
Income: Family income in noviles
25
Figure 4: Incidence of vocational training and income slopes
.05
PRT
Income slopes
0
ESP
IRL
-.05
JPN
FRA
FIN
DEU
SWE
-.1
USA
CAN
AUS
CHE
GBR
NLD
NOR
DNK
NZL
-.15
0
10
20
30
40
Incidence of vocational training indicator (Iversen & Soskice)
Coef=0, SE=0, t=-.1, adj. R2=-.07, N=17
Income: Family income in noviles
26
Figures 5a and 5b: Optimal Benefits, Progressivity, and Income Slopes.
Optimal transfer c* as function of income | beta
-.5
-.4
-.3
-.2
-.1
0
.1
.2
.3
.4
.5
1 2 3 4 5 6 7 8 9 10
1 2 3 4 5 6 7 8 9 10
10
5
0
Optimal transfer c*
-5
10
5
0
-5
1 2 3 4 5 6 7 8 9 10
10
5
0
-5
1 2 3 4 5 6 7 8 9 10
Income deciles
Graph by different values of beta
Optimal transfer c* as function of income | beta
Optimal transfer c*
4
2
0
-2
-4
1
2
3
4
5
6
Income deciles
beta=-0.5
beta=+0.1
7
8
9
10
beta=-0.1
beta=+0.5
27
Figure 6: Concentration of benefits / benefits and taxes and income slopes
.05
PRT
Income slopes
0
KOR
ESP
HUN
IRL
POL
-.05
JPN
FIN
CHE
FRA
DEU
SWE
CZEUSA
NOR
NLD CAN
GBR
-.1
AUS
DNK
NZL
-.15
-.4
-.2
0
.2
Progressivity public cash benefits, working age
.4
Coef=.18, SE=.03, t=6.2, adj. R2=.65, N=21
Income: Family income in noviles
.05
Income slopes
0
-.05
KOR
IRL
POL
JPN
FRA
-.1
CHE
FIN
DEU
SWE
NORCZE
USA
NLD
CAN
GBR
AUS
DNK
NZL
-.15
.2
.4
.6
.8
1
Overall progressivity (concentration of taxes – concentration of benefits)
Coef=-.12, SE=.03, t=-3.8, adj. R2=.44, N=18
Income: Family income in noviles
28
Figure 7: Correlation between concentration of benefits in various social policy domains
and income-slopes in matched social policy domains
P
Income-slope predicting attitude (ISSP)
.02
D
P
H
-.02
-.06
P
H
H
H
H H
H
H
-.6
P
D
U
U
H
-.4
DH
D
P
U
U
D
D
P
P
P
P
P
D
0
-.04
D
DU D
UD
P
P U P
P P
U U
U
U
UH
U
U
P
P
P
D
D
U
U
H
P
D
P
H
U
U
U
-.2
0
Concentration of benefits
.2
.4
D=Disability, H=Housing, P=Pensions, U=Unemployment
Coef=.029, SE=.009, t=3.09, adj. R2=.116
Note:
The unit of analysis is a social-policy-domain in different countries. The solid line is the pooled
regression line (for equation, see note of figure). The four dotted lines are separate regression lines for
each of the four social policy domains. The four social policy domains are disability (D), housing (H),
pensions (P), and unemployment (U). The mapping of social policy attitudes25 and concentration of
benefit domains is as follows:
• Provide decent standard of living for the unemployed → concentration of unemployment benefits
• Provide decent standard of living for the old → concentration of old age pensions
• Provide health care for the sick → concentration of disability benefits
• Provide decent housing for those who can’t afford it → concentration of housing benefits
25
These are based on ISSP survey items, with the following wording stem: “On the whole, do you think it should or
should not be the government's responsibility to ...”. The answer categories are 1 “Definitely should not be”, 2
“Probably should not be”, 3 “Probably should be”, and 4 “Definitely should be.”
29
Figure 8: Redistribution and Progressivity in Sweden (1965-2010)
30
Figure 9: The Evolution of Income Slopes in Sweden (1965-2010).
31
Figure 10: Progressivity and Income Slopes in Sweden (1965-2010)
32
Figures 11 and 12: Income and Social Policy Preferences in Germany before and after
Reunification
-.07
Income-coefficient
-.06
-.05
-.04
-.03
G_taxspend1 = SOCIAL BENEFITS: CUT OR EXTEND?
1980
1990
2000
2010
year
West
East
-.08
Income-coefficient
-.06
-.04
-.02
0
G_unempl = STATE:SECURE INCOME IN TIMES OF HARDSHIP
1980
1990
2000
2010
year
West
East
33
Table 1: Concentration measures
Transfers
Concentration of …
HH taxes
cash
% DI
HH taxes
–cash
benefits
benefits
AUS (Australia)
14.3
-0.43
0.49
0.92
CAN (Canada)
13.6
-0.17
0.47
0.64
CHE (Switzerland)
16
-0.18
0.21
0.39
CZE (Czech Republic)
24.3
-0.15
0.42
0.57
DEU (Germany)
28.2
-0.07
0.44
0.50
DNK (Denmark)
25.6
-0.30
0.33
0.63
ESP (Spain)
21.3
0.10
.
.
FIN (Finland)
14.4
-0.26
0.42
0.68
FRA (France)
32.9
0.10
0.35
0.26
GBR (United Kingdom)
14.5
-0.35
0.49
0.83
HUN (Hungary)
35.1
-0.03
.
.
IRL (Ireland)
17.7
-0.20
0.53
0.74
JPN (Japan)
19.7
0.02
0.36
0.34
KOR (South Korea)
.
0.04
0.36
0.32
NLD (Netherlands)
17.1
-0.22
0.44
0.66
NOR (Norway)
21.7
-0.18
0.35
0.53
NZL (New Zealand)
13
-0.33
0.49
0.82
POL (Poland)
35.8
0.17
0.38
0.21
PRT (Portugal)
25.5
0.31
.
.
SWE (Sweden)
32.7
-0.15
0.33
0.48
USA (United States)
9.4
-0.12
0.55
0.66
Source: OECD 2008, Table 4.3, 4.4; Figure 4.2.
Concentration of cash benefits for …
Unemplo
Old age
Housing
Disability
yment
pensions
Benefits
Benefits
Benefits
-0.47
-0.44
.
-0.35
-0.11
-0.06
.
.
-0.19
-0.15
.
.
-0.11
-0.28
-0.66
-0.06
0.10
-0.28
0.00
.
-0.49
-0.22
-0.58
-0.18
0.04
0.02
0.48
0.11
-0.44
-0.24
-0.61
0.07
0.25
0.08
-0.55
0.14
-0.21
.
.
-0.20
0.01
-0.25
.
.
-0.32
-0.07
-0.46
-0.27
0.02
-0.11
.
.
.
.
.
.
-0.16
0.03
-0.65
-0.11
-0.27
-0.12
-0.65
-0.06
-0.32
-0.38
-0.37
-0.35
0.26
0.13
-0.26
0.04
0.33
0.20
0.13
0.03
-0.19
-0.10
-0.66
0.25
-0.04
0.07
.
.
34
Table 2: Sources of variables
Variable
Concentration of public cash benefits
(working age)
Concentration of HH tax (working age)
Overall progressivity
HH market income inequality (gini)
Total Public Social Expenditure as %
of GDP
Percentage share of public cash
transfers in household disposable
income
Incidence of Vocational Training
Religious fractionalization
Ethnic fractionalization
Cumulative left parties in percentage of
total cabinet posts, weighted by days
Trade union density (OECD)
Rate of Unemployment as % of
Civilian Labour Force
Source / Comment
OECD 2008, Tables 4.3, 4.4
Calculated as [concentration of HH tax – concentration of
public cash benefits]
OECD‘s ―Gini coefficient based on equivalised household
market income, before taxes and transfers (18–65 years only)
(http://stats.oecd.org/wbos/Index.aspx?DataSetCode=INEQUA
LITY)
Social Expenditure Database (SOCX) - OECD
www.oecd.org/els/social/expenditure
OECD 2008, Figure 4.2
Data from Torben Iversen (Iversen and Soskice 2001, 888)
Data available at:
http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/p
apersum.html (Alesina et al. 2003)
Calculated from the Comparative Political Dataset (Armingeon
et al. 2009)
Based on variable gov_left1. Variable is the cumulative cabinet
share of left parties since 1990, for 2005.
OECD
(http://stats.oecd.org/Index.aspx?DataSetCode=UN_DEN),
Mildly interpolated.
OECD (http://stats.oecd.org)
35
Table 3: Predicting income-slopes with the concentration of cash benefits
Concentration of cash benefits
(1)
(2)
(3)
0.18**
(0.03)
0.20**
(0.03)
-0.23#
(0.12)
0.18**
(0.03)
HH market income inequality (gini)
(4)
(5)
(6)
(7)
Income-slope of regression:
preferences for redistribution=income + controls (ISSP) a
0.18**
(0.03)
0.17**
(0.03)
0.15**
(0.03)
0.17**
(0.03)
(8)
(9)
0.18**
(0.05)
0.19**
(0.03)
0.00
(0.00)
Percentage share of public cash transfers
in household disposable income
Incidence of vocational training
0.00
(0.00)
0.00
(0.00)
cumulative Left parties in percentage of
total cabinet posts, weighted by days
Religious fractionalization
-0.03
(0.02)
-0.02
(0.03)
Ethnic fractionalization
0.00
(0.00)
Trade union density
Unemployment rate
Constant
-0.05**
(0.01)
21
0.652
0.04
(0.05)
19
0.678
-0.05*
(0.02)
20
0.644
-0.05**
(0.01)
17
0.647
-0.06**
(0.01)
20
0.642
-0.04**
(0.01)
20
0.683
-0.05**
(0.01)
20
0.655
-0.06**
(0.01)
16
0.411
-0.00
(0.00)
-0.04#
(0.02)
21
0.638
No of cases
R2
Note: Standard errors are in parentheses. # p<0.1, * p<0.05, ** p<0.01.
a
ISSP 2006. Preferences for redistribution: “On the whole, do you think it should be or should not be the government’s responsibility to: Reduce
income differences between the rich and poor” [1. Definitely should not be; 2. Probably should not be; 3. Probably should be; 4. Definitely should
be]. Controls are education, gender, and age. Sample is restricted to employed respondents aged18-65.
Table 4: Predicting income-slopes with the overall progressivity
Overall progressivity (concentration of
taxes - concentration of cash benefits)
HH market income inequality (gini)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Income-slope of regression:
preferences for redistribution=income + controls (ISSP) a
-0.12**
(0.03)
-0.12**
(0.03)
-0.13
(0.13)
-0.11**
(0.04)
-0.09*
(0.04)
-0.10**
(0.03)
-0.08*
(0.03)
-0.09**
(0.03)
(8)
(9)
-0.12*
(0.04)
-0.12**
(0.03)
-0.00
(0.00)
Percentage share of public cash transfers
in household disposable income
Incidence of vocational training
-0.00
(0.00)
-0.00
(0.00)
cumulative Left parties in percentage of
total cabinet posts, weighted by days
Religious fractionalization
-0.03
(0.02)
-0.02
(0.03)
Ethnic fractionalization
-0.00
(0.00)
Trade union density
Unemployment rate
Constant
-0.02
(0.02)
18
0.445
0.04
(0.06)
18
0.446
-0.01
(0.04)
17
0.373
-0.03
(0.03)
15
0.246
-0.02
(0.02)
17
0.407
-0.03
(0.02)
17
0.432
-0.03
(0.02)
17
0.388
-0.01
(0.03)
15
0.336
-0.00
(0.00)
-0.01
(0.03)
18
0.412
No of cases
R2
Note: Standard errors are in parentheses. # p<0.1, * p<0.05, ** p<0.01.
a
ISSP 2006. Preferences for redistribution: “On the whole, do you think it should be or should not be the government’s responsibility to: Reduce
income differences between the rich and poor” [1. Definitely should not be; 2. Probably should not be; 3. Probably should be; 4. Definitely should
be]. Controls are education, gender, and age. Sample is restricted to employed respondents aged18-65.
37