Abstract
Increased income inequality, and policies that can limit its further growth, are an important issue for citizens, politicians and the media. Numerous empirical studies have measured political support for redistributive policies by asking whether the government should equalize gaps between rich and poor. They try to discover whether ‘redistributive’ policies are supported by public opinion and are therefore politically feasible. This research note argues that the standard instrument measures diffuse support for more equality, but gives rather vague hints if this support is transformed into a political mandate for redistributive programmes. With regard to the way in which the political demand for state redistribution to reduce income inequality has been raised so far, methodological critique and innovation is largely lacking. This article therefore tests the validity of the standard item. It argues that the conventional measurement only captures a general ‘inclination’ towards the idea of equality. However, since the item phrasing is unspecific, other orientations confound the answers, so that ultimately the predictive power in terms of political behaviour is low. The standard item measures egalitarian preferences with inconsistence. This limits its reliability, so that hardly any conclusions can be drawn regarding voting for left parties or support for redistributive programmes.
Introduction
Increased income inequality, and policies that can limit its further growth, are an important issue for citizens, politicians and the media. Numerous empirical studies have measured political support for redistributive policies by asking whether the government should equalize gaps in income. 1 Their aim is to discover whether voters want corrections and whether ‘redistributive’ policies are supported by public opinion – and are therefore politically feasible. 2 But is it always clear what is being measured when respondents give a rating on whether the state should reduce the gaps between rich and poor, or between high and low incomes? This research note casts doubt on this and asks: how much egalitarianism is there in the answers given by the widely used survey item? How validly does the standard item measure any existing preferences for a political correction of income inequality in the population?
For some time now there has been a certain unease about measuring (social) policy attitudes with generic items on whether the state should spend more money on old-age security, health or the unemployed. These and other common questions from international surveys on the responsibility of the state for the material security of its citizens in the areas mentioned are considered too unspecific to provide reliable information. Instead, trade-off questions are being tested to capture policy demand more realistically. By requiring choices to be made between alternative programmes, conjoint analyses tries to induce reflexion and weighting up among respondents (Busemeyer and Garritzmann, 2017; De Vries and Hofmann, 2015; Häusermann et al., 2019). However, methodological innovation with regard to the measurement of political demand for redistribution is largely lacking, although dissatisfaction has been expressed. 3 This article neither promotes conjoint analysis nor a special revised item. It just scrutinizes the validity of the standard item to encourage innovation above and beside the conventional item, which we need for time trends and comparative purposes. However, the conventional measurement only captures a general ‘inclination’ towards the idea of equality. Since the item phrasing is unspecific, other orientations confound the answers. The general wording of the questions allows a lot of leeway that ultimately makes the measurement inconsistent and limits its reliability. So, conclusions regarding support for concrete public policies can hardly be drawn.
As attitudinal research has shown, the high level of approval achieved with generic questions collapses as soon as respondents are presented with concrete programmes, whose beneficiaries and financiers are more clearly outlined, and when financial burdens of redistribution for private households, government spending or the economy are mentioned. Thus, a certain double standard is found in attitudes to public policies: respondents are in favour of higher social benefits but at the same time lower taxes (Converse, 1964: 5). A large majority calls for the reduction of inequality, but most refuse to pay higher taxes themselves in favour of improved poverty policies (Eurobarometer 56.1, own analysis). Recognized goals find verbal support, which shrinks at the prospect of having to contribute oneself (Bechtel and Liesch, 2020). ; 91% of respondents think more recognition is needed for cleaning staff, but 57% would not pay a cleaner more than €12 an hour (Die Welt 2019, p. 10). of those who want more equality through redistribution, 46% would cut the benefits of unemployed people who refuse job offers (ESS, 2016; own analysis). In light of these results, the question arises: does the standard item only measure cheap talk? How much support would higher basic security benefits or greater wealth and top income taxes receive by those advocating redistribution? How ‘sustainable’ is the political demand for redistribution measured by the standard item? Can it be used to identify potential support for concrete political programmes with redistributive effects?
Testing the validity of the instrument means asking: does it measure what it is supposed to measure? This formulation presupposes that the object to be measured can be clearly delimited. Like other social science concepts, inequality and the judgement in favour of a need for political action is a rather fuzzy object of social reality. In such cases, the object becomes tangible – operationalized – through the measurement itself. After all, the item clearly formulates that it is about political measures to reduce income gaps. Therefore, I ask: does the instrument measure not only diffuse support for more equality, but also a political mandate for action?
State of research
How citizens’ wishes for redistribution develop, whether they react at all to more inequality, and which socio-cultural factors influence them beyond that, have been the subject of many social-scientific attitude analyses from research on political economy, comparative welfare state and public opinion. Hardly any other variable has been used as often as the item ‘the government should take measures to reduce differences in income levels’. The vague wording of the question offers room for interpretation in various directions (Kim et al., 2016). The questionnaire may be intended to measure the population’s discomfort with inequality and the associated support for political intervention. In addition, a generic formulation bypasses the different organization and designation of specific programmes depending on the country. But does this support go beyond a diffuse ideal of equality? From the citizens’ point of view, redistribution is not necessarily a transfer from the rich to the less well-off households, but also means the contribution-financed safety-net that cushions unemployment or health risks in the life cycle (Weakliem and Biggert, 2013).
In terms of policy instruments, the item remains imprecise. What do respondents actually mean when they support reduced income disparities as a task that should be carried out by the state? A probing question ‘What were you thinking of when you answered the question?’ posed directly after the redistribution item showed that pensions, followed by education and benefits for families spontaneously came to mind (Dallinger, 2021). Answers to the question posed in Germany in autumn 2019 are obviously influenced by the introduction of a basic pension for low-income earners, about which the media reported intensively. Measured redistributive preferences are presumably always influenced by existing policies and institutions as well as by current reform debates in the respective country that are present in the media. This is not problematic in itself if this context is controlled. However, this is done patchily. Easily accessible characteristics such as unemployment or welfare state design show up many times, but media debates and political communication hardly show up at all (see also Cavaillé and Trump, 2014).
The high approval for redistributive policies – in Germany it is 73% according to data from the European Social Survey (ESS) 2018, in many European countries even higher (for example, Italy 86%), in liberal countries as the United Kingdom it is lower (66%) – is often ‘taken literally’ as approval for concrete programmes, without reflecting on whether a ‘something-for-nothing phenomenon’ drives approval upwards: as long as no costs are visible, redistribution demand is expressed with a ‘light hand’ (Naumann, 2018). Funding is an important dimension of preferences and divides those financing and those who benefit (Cavaillé and Trump, 2014).
Questions about the validity of the construct also arise because variants of the item lead to different results. For example, the European Values Study uses a 10-point scale for measurement, the extreme values of which are labelled as two poles: ‘Incomes should be made more equal’ and ‘There should be greater incentives for individual effort’. With the explicit mention of personal responsibility, the item wording sets a different anchor for respondents than the standard item, which generates much less agreement on income equalization than the ESS item: 46.6% in Germany. Nevertheless, the results are not comparable because of the different question wording. Pontusson et al. (2020) also criticize the item used in the ESS and International Social Survey Program (ISSP) as vague and biased in favour of more support for redistribution than is actually there. Their Inequality and Politics Online Survey uses a wider 0–10 response scale that allows respondents to differentiate. A comparison of only the affirmative answers – that is, only the scale values above the middle value – shows that the percentage of affirmative answers is lower compared to the 5-point scale, in Germany by about 10 percentage points. The broader scale apparently makes it easier for respondents to also evaluate redistribution negatively. With about 62% of the respondents in Germany, the answers are mostly in agreement (Pontusson et al., 2020: 24), but less clear. With the variation of the response scale, however, the scope to elicit and replace distorting effects of the standard item has not yet been exhausted. 4
Our considerations on the validity of a generic item on redistribution preferences can draw on classical studies on the acceptance of public policies (Cook and Barrett, 1992; Kangas, 1997), according to which unspecific questions achieve more agreement than ratings of specific programmes or policies since the former offer little room for attack. Questions about policy-specific programmes (indirectly) provide more clues about their target group or funding, which feed into preference formation. This does not necessarily make preferences more rational, as prejudices are also activated. Items more specific to political programmes avoid disadvantages of the generic item. However, it is doubtful whether citizens have the information level to adequately assess specific policies. The observation that respondents often mix up redistribution through higher taxes or through progressive taxes (Barnes, 2015) confirms the assumption that citizens have diffuse ideas about political instruments and their effects. 5 The fact that support for progressive taxes and for redistribution correlate only slightly (Berens and Gelepithis, 2019) also points in this direction. Respondents do not necessarily associate progressive taxes with attitudes to the redistribution item. Especially in countries with a social insurance-welfare state, redistribution is likely to be associated with social insurance schemes, which rather balance income risks over the course of an individual lifetime than redistribute between social groups. Since support decreases when the welfare state is geared towards the poor, whereas it is broad in welfare states that are also geared towards the middle class (Sumino, 2018), it is doubtful whether respondents really mean redistribution to the poor when they speak out in favour of equalization of income gaps by governments.
The points raised are by no means problems only of the standard item, but characteristics of political attitudes as a whole. Attitude research is confronted with these and must seek empirical solutions. Policy-specific items do not generally solve problems of diffuseness and are answered on the basis of little knowledge. But they are beneficial in eliciting the political will of respondents.
Consistence of redistributive preferences was questioned because respondents simultaneously support more redistribution and the principle of competition in free markets (Sudo, 2020). However, this ambivalence is not just due to the arbitrariness of citizens’ preferences. It also expresses a norm compromise and ‘muddling through’ in societies that combine free market economy and a social system that partly equalizes income and living conditions. A social market economy combines divergent ideological traditions and its residents will not find these contradicting each other. This means that ‘people generally hold multiple, often contradictory, ideologies that become more or less accessible as a function of chronic and situational factors’ (Brown-Iannuzzi et al., 2015: 15).
Political attitudes seem unstable because they react to political communication or contextual conditions such as changing levels of unemployment or national debt. Public opinion research has long known that attitudes towards policy issues are not logically structured. General, ‘honourable’ principles find high approval, but are not consistently adapted to specific circumstances (Chong and Druckman, 2007: 103). Judgements about policy issues are formed under the influence of agents from politics and interest groups who have their say in the media. Views are influenced by aspects, information and considerations made available in the media, which activate specific evaluation criteria (Converse, 1964; Zaller, 1992). These peculiarities of public opinion serve to confuse when criteria such as response stability and coherence are applied. However, inconsistency or lack of information and enlightenment is no reason to declare (social) political preferences obsolete. 6
Data and methods
The concept of validity aims to determine whether the range of characteristics to be measured is covered to a sufficiently precise degree by the survey instrument (Rammstedt, 2010). Yet support for redistribution is not a clearly limited characteristic, but rather a highly controversial field. In the case of diffuse concepts, measurement often provides a definition of the ‘thing itself’ in the first place. There is no agreement among researchers on a universe of items that would equally well define the range of characteristics. People have been working pragmatically for years on the measurement of political demand for redistribution using a single variable, which limits possible testing strategies. The validation of the item applied here makes use of the consistency and inconsistency of individual criterion variables.
According to criterion validity, the quality of the measurement results from the degree of agreement between the results obtained with the item and the results of other instruments. The idea of correspondence of results assumes, for example, that environmental awareness goes hand in hand with a willingness to engage in environmentally friendly consumption or behaviour (waste separation). If this is not the case, one would conclude that the measurement of environmental awareness is insufficient. The question of how validly the standard item measures the demand in favour of more income equality is examined on the basis of criterion variables that express a consistent or an inconsistent value attitude. 7 Consistency is determined by criterion variables that clearly express egalitarian values. Inconsistency occurs when respondents support redistribution and meritocratic distribution principles at the same time.
Suitable data are scarce, so we rely on some items of the ESS 2016 that are considered as criterion variables for the validity of attitudes towards redistribution. 8 On the one hand, items measuring the equality dimension are used to test the extent to which these are incorporated into redistribution preferences. On the other hand, meritocratic orientations aim to shed light on how much an inconsistent, contrasting dimension flows into the answers. One variable testing egalitarianism is: ‘For a society to be fair, differences in people’s standard of living should be small.’ (Answers on a scale from 1: agree strongly, to 5: disagree strongly.) The orientation towards the meritocracy principle is measured by the item ‘Large differences in people’s incomes are acceptable to properly reward differences in talents and efforts’. A third criterion offers an item on the approval of unconditional basic income. 9 It is the only item that can map the concordance between the responses to a specific policy instrument and the generic redistribution item. How consistent is support for redistribution and for a basic income? 10
Results
Redistribution and equality (column percentages).
Percentages of all respondents.
Data ESS 2016; weighted, own calculations.
Should we now call the measurement valid? The standard item undoubtedly captures a political demand for redistribution underpinned by ideals of equality. Whoever says A also says B – whoever finds equality just also wants policies that reduce inequality. But besides that, the inconsistencies are considerable. (Therefore, Cronbach's alpha for both items is only 0.564.) If many of those who demand redistribution at the same time support a meritocratic distribution according to individual performance, the demand for redistribution turns out to be even less normatively ‘flawless’ and equality-driven. This could be irrelevant, but after the last step of analysis in this article we see that inconsistency affects voting behaviour. How much are preferences for reduced income gaps confounded with valuing individual achievement?
Redistribution and meritocratic orientation (column percent, weighted).
Percentages of all respondents.
Data: ESS 2016, weighted, DE only.
Combinations of redistribution and meritocracy (% of respondents).
Notes: weighted. Data ESS8 2016.
After this assignment to five levels between the extreme values ‘redistribution without ifs and buts’ and ‘meritocracy without restriction’, 9.9% of the respondents are still unassigned. This group consists of ambiguous respondents who answered ‘neither’ or ‘reject’ in both items. I also include among the ambiguous respondents those who combine rejection of meritocracy with undecidedness about redistribution. These groups do not seem to be able to use the categories provided to classify their own attitudes and cannot locate their preferences on scales for meritocratic or egalitarian ideas. Even the group of ambivalent respondents (see above) is not so ‘homeless’, as they simply find equality and meritocracy equally important.
According to Table 5 on the size of the single assigned groups, the ambivalent group in favour of both meritocracy and equalization of income gaps by the state (31.8%) is the largest among the respondents. Almost a third of Germans reconcile meritocracy and redistribution. In terms of the validity discussed in this article, this means that one-third of the answers to the redistribution item reflect non-egalitarian attitudes. What makes this contradictory group plausible? First, one might explain ambivalence with erratic response behaviour or political ignorance. However, the size of this group speaks against this. Second, ambivalence might arise due to the generic formulation of the item covers a broad, dazzling field of policy options among respondents. More equality can be associated with more equal opportunities in the education system – that is, with meritocracy – or with recognition for ‘equal’ work. Third, this ambivalent response combination is simply an expression of muddling through in a social market economy, as argued above. But the ambivalence of responses remains unsatisfactory.
Which consequences does the diffuse measurement have in terms of political behaviour and especially on voting behaviour? Is support for redistribution sufficiently connected to the equality dimension that the answers explain the election of parties with a redistributive line up? The support for redistribution measured with the standard item does ‘explain’ highly significantly the probability of voting for a left-wing party (average marginal effect 5%). But how well can the election of left-wing parties be explained if one takes into account the ambivalence in the standard item? Voters of the Social Democratic Party and Die LINKE are grouped with left-wing parties because their election programmes are more oriented towards equality and the expansion of the welfare state than other parties. 12
Since the item achieves little precision because a considerable proportion of supporters of redistribution also support meritocracy, it is not very clear what respondents associate with the equalization of income gaps and to which political programmes they attribute ‘redistributive’ effects (see Kim et al., 2016). The consequences of the fuzziness are illustrated below by using the classification variable constructed in Table 5 as a predictor in a logistic regression on the probability of voting for a left-wing party. It is to be expected that significant effects occur only for groups with clear positions (that is, 1 and 5 in Table 5). The remaining groups (2, 3 and 4) contain people with ambivalent attitudes, which does not imply a clear orientation towards left parties.
Average marginal effects, logistic regression, voting for left-wing parties.
Notes: weighted. Data ESS8 2016. Reference group ambivalent and ambiguous. N = 2852; logistic regression results see Supplemental appendix Table A1 in the Supplemental Appendix.
There is nothing new about the finding that voting for left-wing parties becomes more probable with a clear preference for smaller gaps in income, but decreases with a meritocratic orientation. What is interesting is the result in connection with the question: what does the standard item on redistribution preferences measure? Only respondents whose value horizon does not contain meritocratic elements clearly vote more often for left-wing parties. If the meritocratic principle is given even some importance (‘partly agree’), voting for left-wing parties becomes less likely. Even those in groups 2 and 3 who vote in favour of redistribution (45% of the population, see Table 5) do not necessarily vote for left-wing parties, even though their programmes promise more equality. So, it is not at all certain that those who support lower income gaps in the standard item are really behind more ‘equality’ in the light of concrete programmes.
Figure 1 illustrates the effects from Table 6 graphically. Only unambiguous positions in the distribution opinions determine the probability of voting for one of the left-wing parties. Here the coefficients also become highly significant (marked in the graph). The three groups with ambivalent attitudes do not show any specific left-wing party preference (neither in terms of identification nor in terms of voting behaviour). Vote for left wing parties – average marginal effects.
If a positive vote on reduced income gaps influences voting behaviour only to a limited extent on a narrow group, it can hardly be concluded from the unspecific answers on the redistribution item that egalitarian programmes are supported. This can be illustrated by the unconditional basic income. In public discourse, it is treated as a redistributive instrument that creates more equality, since it is not linked to contributions. In the context of this article, the item serves as a further criterion variable that can test the extent to which supporters of redistribution are also fans of an unconditional basic income.
Redistribution and unconditional basic income.
Weighted. Percentage of all respondents.
Unfortunately, there are no more items on concrete policies to show how redistribution preferences behave in the presence of specific clues on financing, beneficiaries, and so on. According to Kangas (1997), general questions that offer few substantial specifics produce socially desirable but arbitrary answers, while concrete policies always perform worse. But the results on basic income already indicate that general, desirable answers do not stand up to the level of concrete policy programmes.
Conclusion
The starting point was the unease in socio-political attitude research that generic items such as those on government responsibility for social security or on additional expenditure for the unemployed, the sick or pensioners offer little valid measurement. Citizen support for policies to reduce income gaps is another item that measures too arbitrarily, despite the fact that it has been used very often given the high interest in whether rising income inequality is ‘processed’ into political demand. Therefore, the article sheds light on what the standard item on redistribution actually measures. How validly does the standard item measure a true political will? Can one derive support for specific government action?
We saw that the standard item measures a hybrid mix of motives. Egalitarian ideals undoubtedly also play an important role, but by no means solely or dominantly. The political demand for redistribution is partly supported by an ideal of equality. However, since the general item gives respondents a lot of leeway for their own interpretation, their answers include opposite orientations, such as meritocracy. Consequently, respondents are in favour of reduced inequality by the state, but not necessarily in favour of a basic income. The fact that basic income is highly controversial, contributes to this break. However, it is also a result of a measurement of support for redistribution allowing a high degree of inconsistency.
As a result, the answers from the standard item are ambivalent: one cannot conclude from affirmative answers that people support policies or parties that implement more equality. In the case of inconsistent redistribution preferences, respondents do not tend towards parties with a left-wing programme, since meritocratic orientations match party programmes with a quite different alignment. The standard item cannot capture a specific political demand because of its open formulation. This fits to effects of inconsistent policy preferences (Stötzer, 2017: 52): The more inconsistency, the less precise voting behaviour can be predicted, and the less connected are people’s own left–right position with the platform of the preferred party. The item leaves much room for respondents to define redistribution exactly as they see fit. Researchers should control this mental process of the interviewees more strongly. It is important to improve the predictive power of the generic standard item. Diverse approaches can be starting points for innovation (for example, the think-aloud method, separating opinion direction and strength, such as Krosnick and Berent, 1993; policy-specific survey questions) without abolishing the standard item, which is required for longitudinal trend analysis and comparative analyses.
Supplemental Material
sj-pdf-1-esp-10.1177_09589287211066469 – Supplemental Material for On the ambivalence of preferences for income redistribution: A research note
Supplemental Material, sj-pdf-1-esp-10.1177_09589287211066469 for On the ambivalence of preferences for income redistribution: A research note by Ursula Dallinger in Journal of European Social Policy
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
