Navigation – Plan du site

AccueilNuméros38VariaBrahmin left versus merchant right?

Varia

Brahmin left versus merchant right?

An analysis of political and policy preferences in Western democracies
Gauche brahmane versus droite marchante ? Une analyse des préférences politiques et des politiques publiques dans les démocraties occidentales
¿Izquierda brahmánica versus derecha mercantil? Un análisis de las preferencias políticas y de las políticas públicas en las democracias occidentales
Bruno Amable et Thibault Darcillon

Résumés

Résumé : L’objectif principal de l’article est d’examiner si les changements dans les stratégies politiques observées en France peuvent être observés dans d’autres pays développés. À cet effet, nous relions les différences de soutien individuel aux partis politiques (gauche/droite et partis spécifiques) avec les différences de soutien à diverses politiques publiques, telles que la redistribution, les allocations de chômage, la perception du niveau d’inégalité, l’immigration/mondialisation, les dépenses en éducation et la réglementation des entreprises. Nous utilisons un vaste ensemble de données d’enquête provenant de différentes vagues de l’International Social Survey Programme (ISSP) de 1985 à 2019 dans 11 pays de l’OCDE en mettant l’accent sur les caractéristiques spécifiques des pays du soutien à la politique et aux politiques publiques. Nos résultats montrent que la transformation de la division gauche-droite en une compétition entre ce que Piketty a appelé une gauche brahmane et une droite marchande ne s’est pas produite partout, reflétant une forte hétérogénéité transnationale dans les préférences en matière de soutien aux partis politiques. Les préférences en matière de politiques publiques présentent certaines caractéristiques communes entre les pays, en particulier lorsqu’on examine les préférences en matière de redistribution. Par conséquent, on constate que les préférences politiques sont faiblement associées aux préférences en matière de politiques publiques.

Haut de page

Texte intégral

We would like to thank all of the participants of various conferences/workshops at SASE 2022 in Amsterdam, AFEP 2022 in Amiens, and at the University of Paris VIII during the “Inequalities and vote” workshop in December 2022, the two anonymous referees and the Editor for valuable comments on previous versions.

1. Introduction

  • 1 See Amable et al. (2012) and Amable (2021a).
  • 2 See also Amable (2021b).

1A series of contributions1 studied the transformations of socio-economic models in relation to changes in the social bases that supported the political strategies associated with them. Analyses of political support and strategies in terms of a dominant social bloc were proposed by Amable and Palombarini (2005; 2018; 2023; 2024), who highlighted a strategy of left-wing parties, in particular the French Socialist Party, which consisted in proposing a political platform including neoliberal reforms by seeking the support of a social bloc that abandoned the working classes in favour of bourgeois classes that belonged to the traditional electorate of both the left and the right.2 This bourgeois bloc strategy has been identified in France (Amable & Palombarini, 2018) and Italy (Palombarini, 2003; Amable et al., 2012).

  • 3 Such a shift has led to the emergence of a multidimensional political space, with the addition of a (...)

2The generalisation of neoliberal transformations in models of capitalism in developed countries raises questions about the political strategies underlying these developments. Are we seeing in OECD countries the kind of political evolution that has been highlighted in France and Italy? Some recent contributions, such as those by Piketty (2018) or Gethin et al. (2021; 2022), give indications that could point in this direction. Gethin et al. (2021; 2022)’s main findings may be summarised as follows. Over the post-World War II period, Western democracies have undergone a slow change affecting the electoral base of the left. The traditional constituency of left parties was mostly composed of voters with relatively lower education and income levels; educated and well-off individuals voted for conservative/right-wing parties. This gave a one-dimensional representation of the political space opposing the left to the right, the working class to the bourgeoisie, the economic policies in favour of redistribution and public investment to those promoting less taxes and regulation of business activities. Gethin et al. (2021; 2022) show with the help of electoral survey data for 21 countries between 1948 and 2020 that this divide has gradually subsided, and left vote has increasingly been associated with a high level of education.3 Since high income individuals kept on voting for right parties, the new political landscape tends to oppose what Piketty (2018) calls a “merchant right” to a “brahmin left.”

3However, Amable and Darcillon (2021) produced some contrasted findings regarding the importance of the educational cleavage and the socioeconomic divides that would oppose different groups such as the so-called brahmin left, the merchant right or the bloc bourgeois to popular classes. Above a certain education level, when controlling for income, support for the left increases with the education level, suggesting that the left may become the party of the skilled though not necessarily affluent individuals. However, left leaning is declining with education for the less educated individuals, implying that left support can also be found among the low educated. The left could thus become mostly the party of the educated because one part of its constituency, that with a low education level, is dwindling as the average education level of the population increases, a trend that, in reverse, reinforces the other part of the constituency. One sees also that left support decreases with income more steeply at high education levels, suggesting that the left could have become the party of the middle-income skilled. Moreover, rather than a political competition opposing a brahmin left to a merchant right, one could consider the possibility of a socio-political alliance between the better-off part of the brahmin left and the high-educated fraction of the merchant right, united in a bloc bourgeois.

4Turning to the political supply, the explanation put forward by Gethin et al. (2022), as well as a sizeable fraction of political scientists and some economists (Noury & Roland, 2020), is that the left, at least part of it, has turned towards so-called “progressive” policies and away from protection and redistribution issues that used to be at the core of the left wing political supply. This is supposed to have put off a large part of the left’s working class constituency, which would have found in “populist” right-wing parties a support for their conservative, anti-immigration, views on sociocultural issues (Houtman et al., 2008). The issue of policy preferences as well as the discrepancy between the expectations of the traditional social base of the left, and the policy choices made by the different parties, are therefore at the centre of the phenomenon affecting the transformation undergone by the traditional left-right divide. Amable and Palombarini (2018) have stressed the role of the European integration issue for the emergence of a divide between the educated middle classes and the working classes and the reorganisation of the political supply.

5The aim of this article is twofold. First, is the divide between a brahmin left and a merchant right relevant for the analysis of the contemporary political divides in all Western democracies or only for a subset of them? The consideration of two dimensions (income and education) and two modalities (low, high) give four possible cases for individuals or social groups. In this perspective, the multi-elite configuration described by Gethin et al. (2022) is not the only possible party system. Second, the transformation of the political divides is linked to the policy preferences of individuals and the way political parties take them into account. It is therefore necessary to see whether the two main dimensions of political differentiation mentioned so far, education and income, influence policy preferences and can explain part of the transformation of the left-right divide. To analyse both issues, we use a large survey dataset taken from different International Social Survey Programme (ISSP) waves from 1985 to 2020 in 11 OECD countries. In addition, in order to capture changes in political supply, we document the change in overall political parties’ ideological position over time (to connect with changes in political preferences), and this for specific issues (to connect with changes in policy preferences) using the Manifesto Project Database (Lehmann et al., 2022).

6Following Amable and Darcillon (2021) who used a pooled sample of 20 OECD countries, we focus on the country-specific characteristics of the support for the various issues. Our results show that the transformation of the left-right divide into a competition between a “brahmin left” and a “merchant right” has not taken place everywhere. In many countries, the left has kept a popular support, even if there are some brahmin elements found in the broad left, mostly expressed through the support for Green parties. Then, the policy preferences do not everywhere conform to the expectations of a prominent divide between a brahmin left somewhat indifferent to redistribution issues and more sensitive to the openness of the economy, and a merchant right.

7The paper is organised as follows. The next section briefly surveys the literature. The following section presents the data used in the estimations and the empirical strategy. The empirical results are then discussed followed by some robustness checks. Finally, a last section interpreting the results concludes.

2. Theory

8Piketty (2018, 2019) and Gethin et al. (2021; 2022) have proposed an analysis of the transformation of the left-right political divide that revisit themes that have been discussed in the political science literature for a few decades (Ford & Jennings, 2020). The argument can be summed up as follows. The left-right divide used to reflect a traditional class- or income-based cleavage in the immediate post-war decades. It has gradually turned into a multi-dimensional divide based on income and education. This change has several origins, including the rise in the average education level of the population, which has made it possible for those who succeed in the so-called meritocratic competition to form a new “elite” possibly different from that based on income or wealth. The new elite is Piketty (2018)’s “brahmin left”, the old, income- or wealth-based one is the “merchant right.”

9But while the transformation of the left-right divide appears in the writings of Piketty and his co-authors as a consequence of the rise in the general level of education, Amable and Palombarini (2018) have insisted on the deliberate strategies of certain political parties, particularly on the left, to modify their social base by changing their political supply in favour of a more neo-liberal orientation.

2.1. A “new” cleavage

10Following Piketty (2018), individuals with a high education level are likely to demand less redistributive policies and more public investment in education. Also, having less to fear from direct or indirect foreign competition (through foreign trade, relocation, etc.) on the labour market, they are prone to be more “open” to globalisation or immigration. The increasing importance of the so-called brahmin groups in the left parties’ constituency would explain their growing neglect of redistribution issues and their turn towards “progressive” socio-cultural values and international openness. This fosters the growing disaffection of the popular classes for left parties, reinforcing the predominance of the interests of the brahmin groups in left party politics.

11Numerous political science contributions have dealt with these issues and the two-dimensional representation of the political space is now more widespread than the simple one-dimensional left-right divide. The traditional economic divide centred on the redistribution or public expenditure issue was supplemented with a “new dimension” which could represent the opposition between materialist and postmaterialist values (Inglehart, 1977; 1990; 1997), or libertarian/cosmopolitan versus authoritarian/nationalist/nativist views (Kistchelt, 1994; Hooghe & Marks, 2009). Education plays a role in this new dimension because the rising access to higher education among younger generations (Duch & Taylor, 1993) would foster the reject of traditional hierarchies, the promotion of individual freedom and openness towards ethnic diversity (Kitschelt, 1994). One may consider these issues as “cultural” (Oesch, 2008a), with Bornschier (2010) pointing out the conservative cultural “backlash” against the rise of multiculturalism in the last decades (Norris & Inglehart, 2019), but the new divide is also related to the position of individuals in the economic structure, explained by the difference between winners and losers of globalisation for instance (Kriesi et al., 2008; 2012) or the growing precariousness among the workforce (Marx, 2014).

12Economic and cultural aspects are linked (Noury & Roland, 2020) because the losers of globalisation, those in direct competition with immigrants in the labour market and whose jobs are threatened by globalisation, are supposed to be more receptive to authoritarian, or nationalist messages (Kriesi, 2010). For Oesch (2008b), right-wing populist parties are disproportionately supported by blue collars and small business owners, i.e. likely losers of globalization. These elements are taken to have affected the political party structure (Hooghe & Marks, 2018), with new “populist” parties benefitting from the emergence of the new cleavage as the traditional parties on the left became more turned towards openness both economically and culturally (Kriesi et al., 2008). Changes in the political party structure also affects the left. Rodrik (2018) finds that “populism” is mostly a right-wing phenomenon on Europe but mostly a left-wing one in Latin America. The element usually taken to be common to all aspects of the new cleavage is education. For Kriesi et al. (2012) the open and cosmopolitan stance is a characteristic of highly educated people, who are also on the winning side of globalisation whereas those unskilled or with lower education are on the losing side.

2.2. Changes in the party system

  • 4 The so-called Duverger’s law holds that in majoritarian political systems, the political competitio (...)

13The left-right divide was reflected in an opposition between left/social-democratic and right/conservative/Christian-democratic parties. The emergence of a new cleavage opens the possibility for new parties to emerge if the political institutions allow for it, i.e. if Duverger’s law4 does not apply. Considering the diversity of electoral systems in Western democracies (Lijphart, 1999), one may expect that the consequences of the transformations in the political cleavages will have different consequences across countries.

14Piketty (2018, 2019) sums up the political transformations as follows. The left/right divide which reflected a traditional class-/income-based cleavage has turned into an “identity-based conflict” because the increased foreign competition and migration has fostered a divide on the desirability of globalisation. The traditional redistributive policy of left parties has become increasingly difficult to implement and the economic policy debate has shifted to a conflict over the limits to international opening, particularly regarding immigration. Also, the increasing level of education of the population has made it possible for those who succeeded in the so-called meritocratic competition to join the “elite” groups, which opened a new dimension of inequality supplementing the old wealth- or income-based dimension. The capacity to face the globalisation challenge is supposed to depend on meritocratic success. One obtains therefore a two-dimensional representation of inequality: financial and educational. Each dimension has its hierarchy and there are two elite groups sharing a common interest in the pursuit of globalisation but diverge on redistribution and public expenditure. Piketty calls these elite groups the “brahmin left” and the “merchant right” and predicts the emergence of a multiple-elite party system. This would explain the growing disinterest of left parties for the demands of the population with low levels of both income and education which in return would feed into the growing disaffection of the popular classes for left parties, reinforcing the predominance of the interests of the brahmin groups in left party politics.

2.3. New socio-political alliances

15Crossing the two economic (income) and education dimensions, one may envisage the existence of four social groups (Table 1). The brahmin group mentioned by Piketty (2018) can be thought of as having a high education level and a relatively low income level. The merchant would symmetrically be characterised by a high income level and a relatively low education level. The popular group would have low levels of both education and income. Finally, the bourgeois group would be endowed of high levels of education and income. Depending on the political traditions and institutions of each country, the political parties may be organised to capture the support of one or the other group, or a coalition of these groups.

Table 1. Four social groups

 

Low level income

High level income

Low level education

Popular

Merchant

High level educatio

Brahmin

Bourgeois

16Majoritarian political institutions do not favour the co-existence of many parties addressing specific social groups’ political demands (Duverger’s law). One may then envisage two possibilities. The traditional parties extend their political supply in the direction of brahmin groups, running the risk of losing part of their traditional constituency, or they keep their traditional position, running the risk of leaving the brahmin groups to other parties (or the other party in a two-party system). Proportional systems are more flexible in this respect and lead to a more differentiated party system. More specifically, for the left, it would be possible for the traditional social-democratic party to reorient its political supply towards the brahmin group, possibly leaving to other parties the more popular fraction of its traditional base. Alternatively, a social-democratic party keeping its traditional working class base would open the possibility for other parties to represent the brahmin groups.

2.4. Testable propositions

17In addition to the mentioned political preferences, the different social groups may be expected to be more or less in favour of specific public policies. As mentioned before, education is considered by most scholars to predispose favourably individuals towards different aspects of economic or cultural openness. A high income level can be expected to make individuals less in favour of redistribution or public investment financed by taxation and more in favour of pro-business policies.

18On the basis of these considerations, one would expect the popular groups to be pro-redistribution/social protection and possibly anti-immigration/globalisation. The brahmin groups would be in favour of investment in education, less in favour of redistribution and not necessarily prone to support the pro-business policies that would please the bourgeois groups. The merchant groups would be hostile to redistribution and possibly to immigration too. Finally, the bourgeois groups are expected to be anti-redistribution and pro-openness. These propositions can be hypothesised as follows.

  • Political preferences:
    • P1: Support for the left should increase with the education level; this is the key hypothesis behind the existence of a brahmin left.

    • P2: Support for the left should decrease with income, more steeply at low education levels; this would split apart the brahmin left from the merchant right and the less educated part of this latter group from popular classes.

    • P2b: Support for the right should increase with the income level more steeply at low education levels.

    • P3: Support for the right should decrease with the education level at high income levels; this would favour the unification of bourgeois groups (brahmin left and educated right)

  • Redistribution:
    • R1: Support for redistribution should decrease with income more at low than at high levels of education; this would explain the emergence of a brahmin left less concerned about inequalities.

    • R2: Support for redistribution should decrease with education at low levels of income and should be constant at high level of income; this would support the separation of the brahmin left from the popular classes.

  • Public investment in education:
    • PI1: Support for investment in education should increase with the education level more strongly for low income than high income; this would support the existence of the brahmin left.

    • PI2: Support for investment in education should decrease more strongly with income at low education levels than at high education levels. This would point at the possibility to unite the brahmin left and the merchant right in a bloc bourgeois.

  • Globalisation:
    • G1: Support for globalisation of business activities should increase with the education level at low income levels and stay constant at high income levels; this would point to the importance of the globalisation divide for the split between the popular classes and the brahmin left.

    • G2: Support for globalisation of business activities should increase with income, more strongly at low education levels; this would separate the merchant right from other social groups.

  • Immigration:
    • I1: Support for immigration should increase with the education level at all income levels; this would unite the bloc bourgeois.

    • I2: Support for immigration should be constant with income at all education levels; this would support the existence of an education-based globalisation divide.

  • Business regulation:
    • B1: Support for less business regulation should increase with income. This would characterise a merchant right.
    • B2: Support for less business regulation should increase with education. This would point to the existence of a bourgeois right.

19Whether these hypotheses are validated and whether the left-right divide and more generally the political party positions correspond to policy preferences is empirically analysed in what follows.

3. Data on political and policy preferences

  • 5 Compared to the initial sample of 20 OECD countries from Amable and Darcillon (2021), some countrie (...)

20To investigate individuals’ preferences on political and policy preferences, we use a unified survey database—by combining different ISSP waves—for 11 OECD countries (Australia, France, Germany, Ireland, Japan, New Zealand, Norway, Spain, Sweden, United Kingdom and United States) from 1985 to 2020.5 Table 2 shows country-level descriptive statistics for all the variables used.

Table 2. Descriptive statistics

  Australia France Germany Ireland Japan New Zealand Norway Spain Sweden United Kingdom United States Total
Support for politics
Left 41.34 33.46 34.59 10.69 4.57 20.91 33.43 26.81 35.59 23.20 27.83 28.42
Right 41.04 26.34 24.97 14.96 29.73 28.06 28.86 16.65 22.65 21.52 25.59 25.97
Support for redistribution
Strongly disagree (SD) 12.32 6.51 5.34 5.65 11.12 14.90 5.26 3.84 8.09 7.15 20.58 9.20
Disagree (D) 27.81 11.29 17.81 11.96 13.72 23.10 18.01 8.01 17.60 15.01 28.20 18.42
Indifferent (Ind) 7.20 7.44 5.40 0.00 15.96 9.26 8.22 4.06 10.89 7.75 9.42 7.69
Agree (A) 33.55 28.19 40.93 33.22 31.47 30.62 35.86 38.25 32.54 37.34 26.07 34.22
Strongly agree (SA) 19.12 46.57 30.52 49.17 27.73 22.11 32.65 45.84 30.88 32.74 15.73 30.48
Support for more spending on unemployment benefits
Spend much less 14.62 12.63 2.36 2.15 5.93 17.81 4.21 1.66 3.51 5.28 4.67 6.86
Spend less 32.20 24.68 12.10 9.68 9.20 34.80 17.61 6.88 15.53 17.46 14.73 18.41
Spend the same as now 41.46 45.95 49.45 37.65 52.94 39.21 58.88 35.00 46.75 43.63 49.94 45.63
Spend more 9.52 12.29 25.92 32.33 20.30 6.21 15.65 43.87 25.45 25.33 21.95 21.49
Spend much more 2.20 4.46 10.17 18.19 11.62 1.97 3.64 12.58 8.77 8.30 8.70 7.61
Support for the opinion that income differences are too large
Strongly disagree (SD) 1.49 0.82 0.73 0.73 3.84 1.71 2.26 0.21 2.97 1.17 2.96 1.70
Disagree (D) 13.90 3.36 5.80 5.80 6.00 11.97 13.38 3.47 9.35 7.60 12.20 9.07
Indifferent (Ind) 17.85 6.53 8.26 8.26 16.63 15.47 16.44 6.06 18.69 12.62 18.00 13.58
Agree (A) 47.39 23.97 47.47 47.47 32.64 43.51 49.04 56.25 39.84 47.95 42.34 43.31
Strongly agree (SA) 19.37 65.32 37.75 37.75 40.89 27.34 18.87 34.02 29.15 30.67 24.49 32.34
Support for the opinion that large income differences necessary for prosperity
Strongly disagree (SD) 8.17 26.53 17.47   27.13 16.12 13.64 14.15 13.39 10.39 8.91 13.99
Disagree (D) 34.56 37.91 37.78   13.56 45.73 46.73 37.72 31.78 43.54 34.28 37.39
Indifferent (Ind) 32.63 18.68 22.73   31.68 19.96 22.95 20.20 30.36 23.58 27.25 25.55
Agree (A) 21.05 13.78 17.97   17.94 15.36 14.91 23.13 20.60 18.85 24.09 19.06
Strongly agree (SA) 3.60 3.10 4.05   9.69 2.81 1.77 4.80 3.86 3.64 5.47 4.00
Support for taxing more richer people
Strongly disagree (SD) 4.39 12.03 5.39 - 23.78 10.59 4.23 12.74 9.62 3.90 6.63 7.62
Disagree (D) 10.47 24.34 11.21 0.31 24.00 23.75 13.56 24.74 18.26 10.78 13.34 15.18
Indifferent (Ind) 31.83 21.26 13.64 17.13 21.03 23.83 23.74 18.23 22.17 24.04 24.75 21.92
Agree (A) 40.84 28.64 38.48 50.15 18.50 27.32 45.39 29.80 36.41 43.25 34.67 36.55
Strongly agree (SA) 12.46 13.73 31.28 32.42 12.69 14.50 13.08 14.49 13.54 18.03 20.61 18.74
Support for the opinion that benefits make individuals lazy
Strongly disagree (SD) 8.47 15.61 6.07     7.38     15.30 6.18 7.98 11.33
Disagree (D) 20.10 13.42 20.23     19.11     21.46 15.11 19.18 18.57
Indifferent (Ind) 20.49 20.00 27.24     22.55     30.60 25.88 23.17 26.01
Agree (A) 30.74 23.33 31.29     32.25     23.82 35.23 35.42 28.13
Strongly agree (SA) 20.20 27.63 15.17     18.71     8.83 17.61 14.25 15.97
Support for the opinion that it is unjust that rich people can buy education
Very just, definitely right 7.79 3.30 2.00   9.90 13.73 3.11 3.52 3.76 19.94 13.30 7.99
Somewhat just, right 22.70 3.93 7.38   19.26 19.78 12.82 11.86 8.23 21.02 20.10 14.32
Neither just nor unjust, mixed feelings 18.78 16.44 13.92 25.31 22.44 19.27 13.16 15.93 22.65 20.47 18.92
Somewhat unjust, wrong 30.88 30.82 40.96   17.39 24.81 28.37 35.12 25.50 17.62 22.64 27.61
Very unjust, definitely wrong 19.85 45.51 35.74   28.14 19.24 36.43 36.34 46.58 18.77 23.48 31.15
Support for spending education
Strongly disagree (SD) 1.74 4.22 2.03 0.25 4.84 2.61 0.56 3.97 0.52 1.38 4.17 2.50
Disagree (D) 1.89 5.09 3.61 1.11 3.04 2.28 2.47 1.37 2.44 1.70 4.96 2.81
Indifferent (Ind) 24.14 30.74 32.03 23.03 32.78 24.88 44.82 16.02 36.87 24.26 20.90 27.74
Agree (A) 43.01 29.50 34.35 37.60 29.57 36.42 41.11 42.34 33.74 41.53 38.58 37.62
Strongly agree (SA) 29.23 30.45 27.99 38.00 29.76 33.81 11.04 36.31 26.43 31.14 31.38 29.34
Support for the opinion that immigration is good for the economy
Strongly disagree (SD) 3.23 14.90 5.50 3.73 11.61 2.43 9.71 5.28 10.08 6.39 3.70 7.43
Disagree (D) 14.03 19.82 20.02 24.90 9.94 13.69 23.92 20.90 16.53 25.68 22.50 19.54
Indifferent (Ind) 22.40 32.44 33.96 23.79 41.58 27.98 33.93 32.57 35.51 37.16 30.18 32.20
Agree (A) 52.83 26.49 37.14 42.98 29.03 50.68 29.88 34.86 32.45 28.40 36.93 35.48
Strongly agree (SA) 7.51 6.34 3.37 4.60 7.85 5.22 2.55 6.39 5.43 2.37 6.69 5.35
Support for the opinion that no immigration-related job substitution
Strongly disagree (SD) 13.42 12.34 8.94 8.12 7.05 11.93 4.29 11.19 5.40 17.99 11.93 9.38
Disagree (D) 29.87 15.32 19.34 31.88 12.76 27.32 12.34 33.41 8.06 34.20 31.42 21.86
Indifferent (Ind) 20.72 24.50 28.80 16.52 34.68 24.38 26.55 18.27 29.91 25.33 22.80 24.89
Agree (A) 29.67 24.80 31.73 38.11 21.25 31.84 46.38 28.03 38.07 18.34 28.19 31.74
Strongly agree (SA) 6.32 23.03 11.20 5.37 24.26 4.52 10.45 9.11 18.56 4.14 5.66 12.13
Support for free trade
Strongly disagree (SD) 5.12 3.36 0.42 3.44 1.65 2.06 1.27 1.43 0.36 2.53 2.47
Disagree (D) 15.13 15.77 6.41 5.93 11.75 10.18 7.83 6.24 4.97 14.02 10.46
Indifferent (Ind) 31.31 31.63 15.27 27.19 21.55 28.22 21.25 34.44 29.82 27.58 27.07
Agree (A) 38.54 41.04 67.87 39.04 54.43 53.01 53.26 48.34 54.67 45.90 48.54
Strongly agree (SA) 9.90 8.20 10.03 24.40 10.62 6.52 16.40 9.55 10.18 9.97 11.46
Support for large companies
Strongly disagree (SD) 39.74 13.16 9.40 17.34 17.64 9.61 20.72 7.75 17.04 20.66 18.63
Disagree (D) 38.32 37.72 31.95 27.52 35.59 33.32 45.27 31.18 41.30 42.42 36.55
Indifferent (Ind) 14.00 26.63 20.84 32.46 30.27 34.94 21.19 39.33 29.82 23.54 26.02
Agree (A) 6.10 18.19 32.42 11.88 15.03 19.85 10.68 17.25 9.94 11.94 15.20
Strongly agree (SA) 1.84 4.31 5.40 10.80 1.46 2.28 2.14 4.49 1.89 1.44 3.59
Support for less regulation
Strongly disagree (SD) 2.55 6.85 7.00 2.26 7.44 4.97 4.36 8.01 6.91 4.81 3.72 5.52
Disagree (D) 18.49 19.87 20.05 17.78 28.50 19.42 19.66 28.96 24.16 20.16 17.82 21.30
Indifferent (Ind) 27.74 22.59 29.19 25.23 31.54 31.73 32.86 28.25 35.01 37.77 32.09 30.46
Agree (A) 37.31 25.78 27.38 39.00 18.83 28.97 33.74 24.95 23.90 27.97 30.86 28.80
Strongly agree (SA) 13.92 24.92 16.38 15.72 13.68 14.91 9.39 9.83 10.01 9.29 15.51 13.92
Income
Relative income to median (mean) 108.39 101.82 105.83 115.92 108.68 108.05 109.98 112.3 103.18 112.51 174.17 115.92

21Survey data are used to assess the expectations and demands of individuals. The first point to note is that the data we use do not come from “opinion polls” but from surveys conducted for scientific purposes. Unlike opinion polls, their results are not reported or commented on in the media, so their influence on the formation of “public opinion” is virtually non-existent. Bourdieu (1973) has analysed the social conditions framing this formation—individuals are surveyed on specific questions on which they do not necessarily have a precise opinion; the “public opinion” measured by survey data can be seen as a social construct. In the same line, Champagne (1990) analyses surveys as tools used by policymakers, political parties or media to guide their actions, influence perceptions, and legitimize certain positions. Besides, some methodological biases—such as how the sample is chosen or how questions are formulated—can contribute to shaping the public opinion. More recently, Frinault et al. (2023) argue that the public opinion is shaped by different heterogeneous actors—such as political leaders, specific social groups—to analyse and criticize the existing power. But the fact that opinions are socially constructed is not a problem for our investigation because we are not concerned with the “true”, “objective”, preferences, assuming that such things could exist, but with the preferences upon which individuals decide of their policy and political inclinations. One could go as far as saying that the data we need is precisely the social constructions resulting from the ideological work carried out by political actors and other “opinion makers” (Amable & Palombarini, 2024).

3.1. Party preferences

  • 6 We exclude from our analysis different categories available in the initial PARTY_LR variable: “did (...)

22Political preferences are derived from the variable PARTY_LR which indicates the self-placement of individuals on a left-right scale (Far Left/Left/Center/Right/Far Right/Other or no preference). For simplicity, we create a first variable capturing the support for left (respectively right): 1 if the respondents declares to vote for a left (respectively right) or far-left party (respectively far-right), 0 otherwise.6 Table 2 shows that the distribution in in the support for left and the support for right largely differs across countries. Then, we investigate the support for specific parties beyond the simple dichotomy between left and right. We use the “nat_PRTY” variable which provides information on country specific party voted in last general election. Only main political parties have been selected (Figure 1). As for the general variable on the support for left or right, we create a binary variable of support for each many party which equals 1 if the respondent voted for the specific party in the last general election and 0 otherwise.

Figure 1. Distribution of party affiliation by country

Figure 1. Distribution of party affiliation by country

3.2. Preferences for redistribution

23We use different variables to measure individuals’ attitudes towards redistribution, and more specifically towards fairness and effort. Each variable has five different categories of answer: strongly disagree (SD), disagree (D), indifferent (Ind), agree (A) and strongly agree (SA).

24The first variable, labelled “redistribution”, measures the support for the opinion that the government has the responsibility to reduce income inequality. As shown in Table 2, support for redistribution is relatively higher in France, Germany, Ireland, Norway, Spain and the United Kingdom than in other countries.

25In addition to this general variable, we investigate the determinants of the support for more spending on unemployment benefits which allows us to capture the “effort” issue of redistribution, i.e. whether redistribution does reduce effort to work or not. Support for more spending on unemployment benefits is relatively higher in Germany, Ireland, Japan, Spain, Sweden, the United Kingdom and in the United States (Table 2).

26Then, we use additional variables (when available) on redistribution capturing the fairness issue (Is redistribution fair or not?), with all of them five different categories (from SD to SA):

  • Support for the opinion that income differences are too large. Support for such an opinion is higher in France, Germany, Ireland and Spain.
  • Support for the opinion that large differences in income are necessary for prosperity (in the respondents’ country). Australia, Japan, Spain, Sweden and the United States show higher values on support for this statement.
  • Support for taxing more rich people, which is particularly strong in Ireland and in Germany.
  • Support for the opinion that social benefits make people lazy. Finally, such support is particularly strong in Australia, France, Germany, New Zealand, the United Kingdom, and the United States.

3.3. Preferences on public investment in education

27Then, we use two different variables for the preferences on public investment in education. The first variable measures the support for education spending by government. A second variable evaluates whether respondents consider just (or unjust) that people with higher incomes can buy a better education for their children than people with lower incomes.

28Support for the opinion that it is unjust is higher in France, Germany, Norway, Spain and Sweden whereas Australia, Ireland, New Zealand, United Kingdom and United States show particular strong support for spending education (Table 2).

3.4. Preferences on globalisation and immigration

29We use two variables on immigration, again with five different categories: the first variable measures the support with the opinion that immigrants are generally good for the economy of the respondents’ country whereas the second variable refers to the support for the opinion that immigrants take jobs away from people who were born in the respondents’ country. When looking at the positive opinion on the effect of immigration in the economy (Table 2), we find a relatively higher support in Australia, Ireland, New Zealand, Spain, and in the United States. Then, when focusing on the opinions on the effect of immigration on the specific issue of employment, France, Japan, Norway, and Sweden show higher levels of negative opinions.

30Then, we use two variables on trade globalisation: a first variable is referring to the support for free trade, and more particularly the support to the opinion that free trade leads to better products becoming available in their own countries. A second variable measures the support for the opinion that large international companies are not causing more and more damage to local business. Support for free trade is higher in Ireland, Japan, New Zealand, Spain and United Kingdom whereas Germany, Ireland, Japan, Norway, and Sweden show higher values of support for large international companies (Table 2).

3.5. Preferences on business regulation

31We use a variable measuring the support for less business regulation, taken an indication of the preferences for pro-business policies. As displayed in Table 2, support for less business regulation is higher in Australia, France, Germany, Ireland, New Zealand, Norway and the United States.

4. Data on political supply

32We document the changes in political supply in terms of ideological position (general and by specific theme) over time using the Manifesto Project Database (Lehmann et al., 2022). Based on the coding of election programmes in a large number of countries, this database provides different variables for political parties reflecting the relative frequency of propositions in political party platforms on economic and non-economic issues.

  • 7 Our index slightly differs from the “rile”-index developed by Laver and Budge (1992). In their inde (...)
  • 8 Please note that Appendix M in Online Appendix provides more detailed figures on the evolution of e (...)

33First, to capture the position of each political party, we use the Manifesto Project Database, which provides different variables reflecting the relative frequency of statements in party manifestos on economic and noneconomic topics. Based on the broad Left/Right classification proposed by Amable et al. (2006), left-wing issues refer to positive statements on market regulation [per403], economic planning [per404], Keynesian demand management [per409], controlled economy [per412], Marxist analysis [per415], social justice [per503], Welfare State expansion [per504], social harmony [per606], labour groups [per701], and underprivileged minorities [per705] and to negative statements on the military [per105] and traditional morality [per604]. Right-wing issues refer to positive statements on the military [per104], governmental and administrative efficiency [per303], political authority [per305], free enterprise [per401], incentives [per402], economic orthodoxy [per414], Welfare State limitation [per505], traditional morality [per603], and law and order [per605] and negative statements on labour groups [per702].7 Then, we compute for each political party a score measuring their relative position on a left-right axis. To do this, we first calculate the “right” (respectively “left”) score by adding the frequency of the variables considered right-wing (respectively “left”) issue and those considered left-wing issue. Then, the party’s score is simply obtained by the (unweighted) difference between the right score and the left one. Left panel in Figure 2 shows the evolution between the 1980s and the 2010s the relative position on the political spectrum for each main political party in each country.8

  • 9 In addition, we produce two separate indicators on the specific issue on education spending: a firs (...)

34Second, we capture the ideological position of each political party for each specific policy issue for which we analyse the preferences. To measure the ideological position on redistribution, we consider the positive statements on equality [per503] and on welfare state [per504], and then we subtract with the negative statements on welfare [per505]. Same as for the ideological position on the Left/Right axis, we plot for each main political party in each country the relative position on the specific issue of redistribution in the right panel in Figure 2. Then, we calculate an indicator on the globalisation issue by subtracting the positive statements on protectionism [per406] from the negative statements on protectionism [per407]. The ideological position of each main political party is shown on the left panel in Figure 3. Finally, to reflect political parties’ position on business regulation, we compute the positive statements on the free market economy and subtract the positive statements on market regulation [per403] and economic planning [per404]. The right panel in Figure 3 shows the score of each party on this specific issue.9

Figure 2. Left-Right ideological position and on Redistribution (source: Manifesto Project Database)

Figure 2. Left-Right ideological position and on Redistribution (source: Manifesto Project Database)

Figure 3. Ideological position on Globalisation and on Business (source: Manifesto Project Database)

Figure 3. Ideological position on Globalisation and on Business (source: Manifesto Project Database)

5. Empirical strategy and results

5.1. Empirical specification

  • 10 We preferred to use a continuous variable compared to quantiles to produce more precise estimates, (...)
  • 11 As shown in Figure 4, country-specific variables have been used for Australia, France and Norway. F (...)

35The main objective of this paper is to investigate the role of income and education levels in shaping political and policy preferences. For that purpose, we use the information provided by the ISSP survey on post-tax family income. To compare the effect of the income level on different preferences, we compute the relative income as a proportion of median income in each country for each year.10 Then, different thresholds, 40%, 60%, 100%, 150%, 200% and 250%, have been used. Table 2 provides the average value of the relative (to median) income variable for each country. Second, in order to capture the role of the education level, we use the variable on the (country-specific) degree—which has been harmonised over time in each country.11 Figure 4 shows the distribution of the education level for each country over the entire period of time. Large cross-country differences can be observed.

Figure 4. Distribution of degree by country

Figure 4. Distribution of degree by country
  • 12 Our estimation strategy differs from Amable & Darcillon (2021) where pooled panel data were used.

36When focusing on the first dimension of politics, due to the dichotomous nature of our variables expressing the support for left/right or specific political parties, we first run binary logit regressions for each country.12

(1)
  • 13 Correlation matrices in Table N.1 in Appendix N in Online Appendix show that partisan and policy pr (...)

37Where denotes the support for a specific party, and are respectively the degree and relative income of individual i. Finally, X is a vector of several individual socio-demographic characteristics, i.e. gender, age, age squared (to allow for concavity), marital status and the presence of children in the household.13 Finally, Year is a vector of year dummies, and  is the error term. We use robust standard errors to deal with the heteroskedastic error terms. In this respect, we follow the specification used in Amable & Darcillon (2021). As our empirical model includes an interactive term between income and education, we compute marginal effects, indicating how the support for specific items varies with income (respectively education) level for each education (respectively income) level.

38Then, because our variables other than those on political leaning all increase with the individual respective support, we perform ordered logit regressions, allowing us to check whether the support for a specific opinion is symmetrical with the opposition to that opinion. This method also allows us to explore the specific determinants of support (agree and strongly agree) and opposition (disagree and strongly disagree). In particular, we have:

if and and (2)

39with a variable which is not directly observable but related to another variable that takes values between 1 and 5.

40Finally, the issue of endogeneity between our different dependent variables and our main independent variables (education/income) can potentially arise. However, finding valid instruments to directly explain income and education levels—without having direct effects on the various explained variables—remains challenging. For instance, the level of income/education of one individual can be considered to be related to the level (of education or income) of the father/mother/spouse of the respondent, which are supposed to be also correlated with individual’s ideological positioning. This assumption violates the exclusion criterion necessary to validate the instrument. Given these challenges, this issue of endogeneity between ideological support and public policy variables and income/education variables can be considered as relatively low, as suggested by the correlation matrices (in Online Annex N.1) which do not suggest a strong direct correlation between these variables.

5.2. Results on preferences on politics

  • 14 Additional figures representing the estimations of the marginal effects with the error margins for (...)

41Table 3 summarizes our main results on the first four testable propositions.14 For each proposition, we start with the general label of left and right, and then we focus on individual parties (which are gathered under traditional party families).

Table 3. Preferences on political leaning (main results)

 

Proposition

Validated

Unvalidated

P1

Support for the left should increase with the education level

Australia (for high incomes); France; Germany; Japan; New Zealand

Ireland (under 150% of median income); Sweden

  • Radical Left

Germany; Japan

France (under median income)

  • Social-democratic Left

France; Japan

Australia and New Zealand (under 150% of median income); Germany; Ireland (Sinn Fein) and United Kingdom (under 100% of median income); Norway; Spain; Sweden

  • Greens

Australia; France (except for high incomes); Germany; Ireland; New Zealand; Sweden

 

P2

Support for the left should decrease with income, more steeply at low education levels

Australia (except for university); Ireland (for low/median education levels); New Zealand; Sweden; United Kingdom; United States

Spain

  • Radical Left

France; Germany

 

  • Social-democratic Left

Australia; Ireland (Sinn Fein); New Zealand; Spain (for low and middle education levels); Sweden; United Kingdom (except for the highest education level); United States

 

  • Greens

Sweden (for high education levels)

Ireland (except for high education levels)

P2b

Support for the right should increase with the income level more steeply at low education levels

Australia (except for university degree) ; France; Germany and Spain (for low to middle education levels) ; Ireland; New Zealand; Sweden; United Kingdom; United States

 

 

  • Center Right

Australia; France; Germany (for low to median education levels); Ireland (except for high education); New Zealand; Spain; Sweden; United Kingdom

 

 

  • Far Right

 

Australia (for high education levels); Germany

P3

Support for the right should decrease with the education level at high income levels

Ireland; New Zealand (under 100% of median income) ; Spain; Sweden; United States

Australia and Germany (above 150% of median income)

  • Center Right

Ireland (Fiana Fail)

Australia, France, Germany, New Zealand, Spain, United Kingdom (under median income); Ireland (Fine Gail); Norway; Sweden; United States

  • Far Right

Australia; France; Germany; Norway

 

  • 15 This contrasts with Gethin et al. (2021)’s result that left vote is positively related with educati (...)

42P1 and P2 are the two hypotheses that test most clearly the relevance of the brahmin left assumption. Looking at the left in general, a few countries validate P1, Australia (partly), France, Germany, Japan and New Zealand (Figure 5) and a few others validate P2 exclusively: Ireland, the United Kingdom, the United States and Sweden (Figure 6). Spain is an exception, with the support for the left increasing with income but not with the education level, making it almost a merchant left.15 Therefore, without being an overwhelming phenomenon, because the left or some fractions of it keep a popular support in some countries, the existence of a brahmin left is confirmed. The support for the existence of a merchant right is less clear. Without surprise, right support increases with income in most countries of the sample, but a strictly merchant right (in the sense of P3) cannot be found in most countries, at least for the centre right.

Figure 5. Support for Left (P1)

Figure 5. Support for Left (P1)

Figure 6. Support for Left (P2)

Figure 6. Support for Left (P2)

43Crossing the income and education dimensions, one finds a popular left in Ireland and Sweden, a bourgeois right in Spain, Ireland, Sweden and the US, a merchant right in Australia and Germany, a brahmin left in Australia and New Zealand. The picture is more accurate when one looks at individual parties.

5.2.1. A popular left

  • 16 By contrast, Gethin et al. (2021) found that education influenced positively and income negatively (...)

44As shown in Figure 7, a popular left is represented by PCF-LFI in France, by Sinn Fein in Ireland (whose support decreases with the education level under 150% of the median income and with income except for highly educated individuals) and by the Spanish Socialist Workers’ Party (PSOE) which appears to benefit from a traditional left-wing base with a support decreasing with education and with income (at low and middle education levels). A popular social-democratic left also found in Australia (Australia Labour Party, ALP), New Zealand, Sweden, and the United Kingdom (Labour).16

Figure 7. Evidence of a popular left (P1)

Figure 7. Evidence of a popular left (P1)

45Some of these findings are compatible with changes in political supply captured by the manifesto data (Figure 2). The popular support of Spain’s PSOE is compatible with the leftward move indicated by the comparative manifesto data. The rightward move of the social-democratic parties in Australia and the United Kingdom have been reverted in the recent period considered here, i.e. before Labour’s comeback to Downing Street in 2024. This has most certainly helped Labour to keep a popular social base. The social democratic left in New Zealand appears distinctly left of the centre-right parties. The situation is less clear in Sweden, where the Swedish social democrats appear to be relatively close to the Christian democrats on the left-right dimension. By contrast, in Ireland, the popular support of Sinn Fein is at odds with the rightward move of this party (whereas Fianna Fáil has shown a leftward move).

5.2.2. A brahmin left

  • 17 An additional brahmin dimension characterising the support for the Green party and, to a limited ex (...)

46Then, as displayed in Figure 8, the brahmin left is embodied in the radical left in Germany (die Linke), and the Green in Sweden as well as in Australia (in the cases of validation of P1 and P2).17 If the relative position on the left-right axis of Australia’s Green party with respect to other Australian parties is not too dissimilar from the relative position of die Linke in Germany, the Swedish Green, by contrast, appear relatively close to the social democrats. The brahmin left party in Japan appears to be the Social Democratic party, whose position on the left-right axis appears close to that of the Communist party.

Figure 8. Evidence of a Brahmin left (P1)

Figure 8. Evidence of a Brahmin left (P1)
  • 18 We find that the support for the centre-left (SPD) decreases with the education level; the SPD bene (...)
  • 19 Rey (2004), Rennwald (2015).

47The brahmin left is also represented by the Green in France (except for high incomes), Germany18, Ireland, and in New Zealand where the support for left-centre and the Green incorporates some brahmin elements. The proximity of the German Greens and the SPD on the left-right axis indicates that there is a de facto brahmin/popular split in social support between the two parties, most likely structured around environmental themes. This indicates that there is a “natural” split of the potential left electorate in Germany between the Green and SPD. The same proximity is found in France but the same conclusion as for Germany cannot be drawn because the French Socialist Party (PS) has a weak popular support. This weak popular electoral support has been widely documented19 and the strategy of the socialists over the past decade was to try to change its social and electoral base in the direction of more affluent and educated social groups (Amable & Palombarini, 2018). The data from the comparative manifesto project show that PS did not change its position on the left right axis (Figure 2), which represents in fact a move to the right in comparative terms when one compares the leftward move made by the centre-right and far right parties. Different from the German situation, the French Green and PS are in competition for the same electorate.

48The situation in New Zealand and Ireland looks similar to the German one. The popular left electorate supports the social democratic party and the brahmin left electorate supports the Green.

5.2.3. A bourgeois right

49There is a bourgeois right or centre-right in Australia, France, Germany, Ireland, New Zealand, Spain, Sweden, and the United States (Figure 9). The Tory party appears more merchant than bourgeois in the United Kingdom. In Japan, it is interesting that right-wing parties, including the Liberal Democratic party (LDP), appear as catch-all political formations, with no clear links between their respective social bases and either the income or education level. In Ireland, the support for Fine Gael increases with the education level and with income, although this latter relationship wanes as the education level increases. One may note the existence of a popular far right in Australia (National Party) and Germany (AfD, NPD, Republikaner).

Figure 9. Evidence of a bourgeois right (P3)

Figure 9. Evidence of a bourgeois right (P3)

5.3. Results on policy preferences

50Then, Table 4 summarizes our main results on the additional propositions on policy preferences.

Table 4. Policy preferences (main results)

Proposition

Validated

Unvalidated

R1

Support for redistribution should decrease with income more at low than at high levels of education

Australia; France; Germany; Ireland; Japan; New Zealand; Spain; Sweden; United Kingdom and United States

 

 

  • More spending on unemployment benefits

Australia; Germany; Ireland; Japan; New Zealand; Spain; Sweden; United Kingdom and United States

 

 

  • Income differences are too large

Australia; France; Japan; Sweden; United Kingdom

 

 

  • Inequality for prosperity

 

Japan; Norway; Sweden

 

  • Study to earn a lot

United Kingdom (except for university graduates)

 

 

  • Taxing the rich

Germany; Norway

 

R2

Support for redistribution should decrease with education at low levels of income and should be constant at high level of income

France; Germany and Japan (under 150 % of median income); Ireland; New Zealand; Norway; Spain; Sweden; United Kingdom (except for university graduates); United States

 

  • More spending on unemployment benefits

Ireland and Norway and United States (under 150 % of median income); Spain; Sweden; United Kingdom (except for university graduates under median income)

 

  • Income differences are too large

Japan; Norway; Sweden

 

  • Inequality for prosperity

 

 

  • Study for years

Norway

 

  • Study to earn a lot

Norway; United Kingdom

 

  • Hard work

 

Spain (above median income)

  • Benefits lazy

France

 

PI1

Support for investment in education should increase with the education level more strongly for low income than high income.

 

 

 

  • Spending education

Germany; Norway (under median income)

 

 

  • Unjust education

 

Norway

PI2

Support for investment in education should decrease more strongly with income at low education levels than at high education levels.

 

 

 

  • Spending education

France; Germany (at low education levels); United Kingdom

Norway

 

  • Unjust education

France (above secondary education level); Japan; Sweden

 

G1

Support for globalisation of business activities should increase with the education level at low income levels and stay constant at high income levels

 

 

 

  • Large companies

New Zealand (under median income)

 

G2

Support for globalisation of business activities should increase with income, more strongly at low education levels.

 

 

 

  • Large companies

New Zealand

 

 

  • Free trade

Japan

Germany (for median and high levels)

I1

Support for immigration should increase with the education level at all income levels

 

 

 

  • Immigration good for the economy

Australia, Japan, Sweden and United Kingdom (above median income); France (except for high incomes); Germany (under 200 % of median income); Ireland (above median income); Spain; United States (U-shaped)

 

 

  • No job substitution with immigration

Australia and Ireland (above median income); France (except for high income); Spain (except above 200 % of median income); Sweden (under median income)

 

I2

Support for immigration should be constant with income at all education levels

 

 

 

  • Immigration good for the economy

 

Ireland (for the lowest and highest education levels); Spain (except at the lowest education level)

 

  • No job substitution with immigration

 

Australia (except at the lowest education level); New Zealand; Spain (at low education levels)

B1

Support for less business regulation should increase with income

New Zealand (for middle and high education levels); Norway

Germany

B2

Support for less business regulation should increase with education

Japan

Norway

5.3.1. Support for redistribution

51Turning to policy preferences, one finds, without surprise, a popular support for redistribution, decreasing with both income and education levels, in almost all countries (Figure 10). The R1 and R2 hypotheses account for the existence of a brahmin left less concerned with redistribution and different from the popular left. Both hypotheses are broadly validated with the exception of a few countries.

Figure 10. Support for redistribution (R1 and R2)

Figure 10. Support for redistribution (R1 and R2)

52This broad pattern of preferences should lead to a popular support for those left parties that emphasise the redistribution issue. The clear differentiation in Australia, New Zealand, Spain, Sweden, the United Kingdom and the United States, as well as to some extent in Sweden, should allow left-wing parties to maintain a popular base. This is indeed the case (Figure 2). France is, however, a special case. The early periods show a clear left-right difference on the redistribution issue, which has gradually diminished over time so as to make the far right, right, green and left parties very similar to one another on this topic. The presence of the redistribution issue in all French parties has made it more difficult for the left to keep a popular support on the basis of this issue alone. The weakness of the popular support for the PS, which did not affect the PCF-LFI alliance, is partly associated with its lack of specificity on the redistributive issue. In addition, one may notice that Norway is the only country of our sample where there is no clear evidence of a support for redistribution related with the level of individual income. In this respect, if one puts the far right Progress party apart, Norwegian parties appear close to one another on the redistribution issue, especially in the most recent period. A reason could be that the social-democratic character of economic policy is so entrenched that redistribution is part of the DNA of most political parties. For this reason, our result may suggest that left-wing parties are focusing on themes other than redistribution to differentiate themselves from other parties.

5.3.2. Support for public investment in education

53Turning to public investment in education (Figure 11), PI1 is validated for Germany only, which points to the importance of this issue for the brahmin left there. In this country, the liberal party FDP has expressed over time a strong support as well as a strong opposition to public investment in education according to the party manifesto data. The conservative CDU/CSU has had a similar but more muted attitude.

Figure 11. Support for investment in public education (Pl1 and Pl2)

Figure 11. Support for investment in public education (Pl1 and Pl2)

54PI2 has a wider support (France, Germany, Japan, Sweden and the United Kingdom), which suggests the possibility of having a bourgeois bloc united on the education issue. The support for public education spending appears to be of the bourgeois type in Norway, which may explain why right parties have expressed a degree of support not to dissimilar to that of the other parties. However, the far-right progress party distinguished itself by an opposition to public investment in education, which is coherent with the fact that the support for this party declines with the level of education.

5.3.3. Support for immigration / globalisation

55The support patterns for other policy issues is far less clear. Openness, whether trade or immigration, is an area where the brahmin left is expected to differ from a more popular left. One finds the support for immigration associated with higher education levels (I1) in almost all countries of the sample, New Zealand being an exception (Figure 12). This points to a unification of the bourgeois categories on these issues rather than to the existence of a brahmin left. I2, on the other hand, is never validated, which casts some doubts on the existence of a strictly education-based openness divide.

Figure 12. Support for immigration with a good effect on the economy (I1)

Figure 12. Support for immigration with a good effect on the economy (I1)

56In most countries, the Green party, when it exists, receives the most a brahmin-type support: in Australia, France, Germany, New Zealand, and Sweden. One would then expect that the Green in these countries would choose the immigration issue to stand apart from other parties. The limited information we have only provides a partial support to this hypothesis (Figure in Online Appendix M). The Green party is not always the party the most favourable to immigration, and sometimes other parties appear close to the Green on this issue. Only Germany conforms clearly to the initial hypothesis. One notices that immigration does not appear to be a theme of particular importance for the Green in Norway and New Zealand.

57The lack of support for G1 and G2 (except for New Zealand or Japan) also question the importance of education as a unique determinant of preferences on globalisation. The latter receives a merchant-type support in Germany but is more bourgeois in Japan and New Zealand. Unsurprisingly, globalisation is most strongly supported by the liberal FDP whereas other German parties do not seem to differ from one another on this issue (Figure 3). With the exception of the 1990s period, Japanese parties do not signal a strong differentiation either. The position of parties in New Zealand follows a broad left-right divide, with the left and the Green expressing more opposition whereas support comes from the right parties. This is coherent with the bourgeois/merchant support that these latter parties receive.

5.3.4. Support for less regulation

58Finally, liberalisation (less regulation) policies receive a brahmin support in Japan, a bourgeois support in New Zealand and a merchant support in Norway (Figure 13). In Japan, LDP and Kameito express some support for liberalisation whereas other parties oppose it (Figure 3). The brahmin support observed for the Japanese social democratic party does not seem to be related to the party’s position on liberalisation. In New Zealand, Act and the National party are pro-deregulation while the other parties are against it. The pattern of party support, a bourgeois/merchant right and a popular/brahmin left, is coherent with this. In Norway, the far-right Progress party and, to a lesser extent, the Conservatives support less regulation for business. This seems relatively coherent with the support for the Progress party, which does not decrease with income but with the education level.

Figure 13. Support for less regulation (B1 and B2)

Figure 13. Support for less regulation (B1 and B2)

5.4. Summing up

59The transformation of the left-right divide into a competition between a brahmin left and a merchant right has not taken place everywhere. In some countries (Australia, Germany, New Zealand, Spain, Sweden), the left has kept a popular support, even if there are some brahmin elements found in the broad left, mostly represented by Green parties. In other countries, such as France, brahmin left parties are supplemented with others that keep a popular support. In fact, only Japan appears to have a strictly brahmin left, and this is a country where the left is particularly weak. Elsewhere, both brahmin and popular support for the left coexist. The support for the right is also mixed, with either merchant (Australia, Germany, New Zealand, Spain, United Kingdom) or bourgeois elements (Ireland, Norway, Sweden, United States) dominating. Education often exerts a negative influence on redistribution (France, Ireland, New Zealand, Norway, Spain, Sweden, United Kingdom, United States), sometimes for individuals with relatively modest income levels (Germany, Japan), which may point to the existence of an aspiration to upward social mobility.

60The preferences concerning more precise redistribution-related policies are more diverse according to the country considered. The fact the higher the income the less individuals support redistribution is validated when one looks at the answers to the question on the tolerance towards inequality or the opinion on unemployment benefits. Other public expenditures (education) may be seen as a social investment and are sometimes favoured by individuals with high income and education levels (Norway).

61Immigration-related issues are usually taken as the most obvious indicator of the possible existence of a brahmin left. In most, but not all, cases examined, education is indeed related with a broadly positive opinion on immigration. Some country cases present original characteristics though. New Zealand does not show an expression of immigration-related policy preferences connected with education. Only income influences positively the opinion that immigration does not cause job substitution. In the United States, the influence of education on the opinion regarding the positive consequences of immigration for the economy exhibits a U-shaped pattern (decreasing from low to average levels and increasing thereafter). In Australia, Japan, Ireland, and the United Kingdom, education influences positively the favourable opinion on immigration for individuals above the median income level only. This may point to the existence of a perceived competition from immigrants for individuals with modest income or education. In general, the results regarding the links between policy preferences for immigration and education or income levels broadly confirm that the competition with immigrants is perceived negatively by a low income/education population, which points out that the positive appreciation of immigration by educated or better-off individuals is the result of the absence of such perception.

5.5. Robustness checks

  • 20 We decided not to plot these results for a lack of space (including in the Online Appendix). Result (...)

62First, our database allows us to distinguishing the support for a specific item from the opposition to it, and this with a characterisation of the intensity of the support or the opposition. We find in general very symmetrical effects of the support and the opposition for/to each issue.20

63Second, as the distribution of individuals’ political leaning is strongly dissymmetrical in some countries, logit estimations can produce biased estimates (King & Zeng, 2001) especially when the share of the support for one political side is small (for instance the support for the left in Japan). To deal with this issue, we rerun our baseline regressions for left support (as displayed in Figures 5 and 6) using the “relogit” Stata command (Appendix O in Online Appendix). Regarding the effect of education (highest degree obtained) conditional on income (Appendix O.1), we find that left support increases with education in all countries (as in Figure 5), except for Australia and New Zealand. We find more robust results when investigating the effect of income for different levels of education (Appendix O.2).

64Third, we investigate how the effect of the interaction between income and education on political leaning and for the preferences concerning different policies has evolved over time (Appendix L in Online Appendix) in line with Piketty (2018) and Gethin et al. (2022) who have extensively exploited the temporal dimension to analyse changes in the political cleavages. Although our database does not allow us to analyse long run trends, it is possible to focus on a more limited time span. To do this, we replicate a baseline logit model with an interactive term between education, income levels and a time variable considering three periods (1995-1999; 2003-2010; 2011-2018).

  • 21 We focus on France and the United States in reference to the initial work of Piketty (2018). Then, (...)

65Because a complete analysis of the time dimension would go beyond the limits of a single article, we limit ourselves to three countries (France, Germany, and the United States) and two specific issues (support for left/right and support for redistribution).21 When comparing the three countries, the support for the left is found very stable over time except for high incomes (and more particularly for individuals with low education levels) in France whereas the support for left has slightly declined over time in Germany, especially among low-educated individuals (except among the highest incomes). The support for the left is more unstable over time in the United States, again among high incomes. We find symmetrical effects when looking at the evolution in the support for right over time. Finally, support for redistribution in France has shown a stable evolution among low incomes (and a large increase among lowest and highest educated and high-income individuals). In Germany as in the United States, the support for redistribution has shown a relatively decline in the 2000s and a rise in the 2010s, and this more evidently among middle and high incomes.

6. Conclusion

66The phenomenon analysed by Gethin et al. (2022), the emergence of a “multiconflictual” or “multi-elite” party systems in most Western democracies, in which higher-educated voters vote for the “left,” whereas high-income voters still vote for the “right” is partially confirmed by the empirical analysis performed in this contribution but must be qualified. The traditional class cleavage opposing a popular left to a bourgeois right has not disappeared everywhere. In fact, only a few countries could be regarded as epitomising the opposition between a brahmin left and a merchant right. Depending on the political history and institutions of the countries considered, the different coalitions between the four broad social groups of Table 1 may give more diverse structures of political competition, with consequences for the policy choices that will result from the compromises between social groups. The existence of a brahmin left appears to be a necessary condition for the creation of a bloc bourgeois. In contrast, a strong popular left makes a bloc bourgeois strategy much more difficult to implement.

67The education cleavage has not the same importance in every country and does not determine unilaterally political preferences for the left or the right, nor the policy preferences. This questions the generally assumed collinearity between the broad cultural and policy issues associated with the education divide: cultural and economic openness, economic and cultural liberalism, etc. The consideration of the determinants of policy preferences reveals that the multidimensionality of the political conflict may be higher than is commonly taken into consideration, i.e. a political space spanned by an economic and an educational divide. This suggests that the political strategy based on the breakup of the left and right social blocs in order to create a bloc bourgeois may hinge on the priority given to various divides, depending on the socio-economic context and the national history of the political parties. In addition, in line with Amable and Palombarini (2005), the macroeconomic dynamics which is influenced by the political mediation affects social expectations and the social structure. In this respect, the evolution of the political and policy preferences has been certainly influenced by economic crises (e.g. the Global Financial Crisis in 2008 or the sovereign debt crisis in Europe in 2010-2012). This will be the topic of further research.

Haut de page

Bibliographie

Amable B. (2021a), La résistible ascension du néolibéralisme : Modernisation capitaliste et crise politique en France (1980-2020), Paris, La Découverte.

Amable B. (2021b), “In search of the bloc bourgeois”, Revue de la Régulation, no 31, URL: https://journals.openedition.org/regulation/20350

Amable B. & T. Darcillon (2021), “The Brahmin Left, the Merchant Right, and the Bloc Bourgeois”, Review of International Political Economy, vol. 29, no 4, p. 1342‑1367.

Amable B., Gatti D. & J. Schumacher (2006), “Welfare-State Retrenchment: The Partisan Effect Revisited”, Oxford Review of Economic Policy, vol. 22, no 3, p. 426‑444.

Amable B. & S. Palombarini (2005), L’économie politique n’est pas une science morale, Paris, Raisons d’Agir.

Amable B. & S. Palombarini (2018), The Last Neoliberal. Macron and the Origins of France’s Political Crisis, London & New York, Verso Books.

Amable B. & S. Palombarini (2023) « Économie et politique : le bloc social dominant » in Boyer R., J.-P. Chanteau, Labrousse A. & T. Lamarche (dir.), Théorie de la régulation, un nouvel état des savoirs, Malakoff, Dunod, p. 51‑56.

Amable B. & S. Palombarini (2024), Blocs sociaux et domination. Pour une économie politique néoréaliste, Paris, Raisons d’Agir.

Bourdieu P. (1973), « L’opinion publique n’existe pas », Les Temps Modernes, no 318, p. 1292‑1309.

Champagne P. (1990), Faire l’opinion. Le nouveau jeu politique, Paris, Les Éditions de Minuit.

Duch R. M. & M. A. Taylor (1993), “Postmaterialism and the economic condition”, American Journal of Political Science, vol. 37, no 3, p. 747‑779.

Ford R. & W. Jennings (2020), “The Changing Cleavage Politics of Western Europe”, Annual Review of Political Science, vol. 23, p. 295‑314.

Frinault T., Karila-Cohen P. & E. Neveu (2023), Qu’est-ce que l’opinion publique ? Dynamiques, matérialités, conflits, Paris, Gallimard.

Gethin A., Martınez-Toledano C. & T. Piketty (2021) (eds), Political Cleavages and Social Inequalities. A Study of Fifty Democracies, 1948–2020, Cambridge, MA, Harvard University Press.

Gethin A., Martınez-Toledano C. & T. Piketty (2022), “Brahmin Left Versus Merchant Right: Changing Political Cleavages in 21 Western Democracies, 1948-2020”, Quarterly Journal of Economics, vol. 137, no 1, p. 1‑48.

Hooghe L. & G. Marks (2009), “A postfunctionalist theory of European integration: From permissive consensus to constraining dissensus”, British Journal of Political Science, vol. 39, no 1, p. 1‑23.

Hooghe L. & G. Marks (2018), “Cleavage theory meets Europe’s crises: Lipset, Rokkan, and the transnational cleavage”, Journal of European Public Policy, vol. 25, no 1, p. 109‑135.

Houtman D., Achterberg P. & A. Derks (eds) (2008), Farewell to the Leftist Working Class, New Brunswick, Transaction.

Inglehart R. F. (1977), The Silent Revolution: Changing Values and Political Styles Among Western Publics, Princeton, NJ, Princeton University Press.

Inglehart R. F. (1990), Culture Shift in Advanced Industrial Society, Princeton, NJ, Princeton University Press.

Inglehart R. F. (1997), Modernization and Postmodernization: Cultural, Economic, and Political Change in 43 Countries, Princeton, NJ, Princeton University Press.

King G. & L. Zeng (2001), “Logistic Regression in Rare Events Data”, Political Analysis, vol. 9, no 2, p. 137‑163

Kitschelt H. (1994), The Transformation of European Social Democracy, Cambridge, Cambridge University Press.

Kriesi H. (2010), “Restructuration of partisan politics and the emergence of a new cleavage based on values”, West European Politics, vol. 33, p. 673‑685.

Kriesi H., Grande E., Dolezal M., Helbling M., Höglinger D., Hutter S. & B. Wüest (2012), Political Conflict in Western Europe, Cambridge, Cambridge University Press.

Kriesi H., Grande E., Lachat R., Dolezal M., Bornschier S. & T. Fray (2008), West European Politics in the Age of Globalization, New York, Cambridge University Press.

Lehmann P., Burst T., Matthieß T., Regel S., Volkens A., Weßels B. & L. Zehnter (2022), “The Manifesto Data Collection”, Manifesto Project (MRG / CMP / MARPOR), Version 2022a, Berlin, Wissenschaftszentrum Berlin für Sozialforschung (WZB), URL: https://manifesto-project.wzb.eu/datasets/MPDS2022a

Lijphart A. (1999), Patterns of Democracy. Government Forms and Performance in Thirty-six Countries, New Haven, Yale University Press.

Noury A. & G. Roland (2020), “Identity Politics and Populism in Europe”, Annual Review of Political Science, vol. 23, no 1, p. 421‑439.

Oesch D. (2008a), “Explaining Workers’ Support for Right-Wing Populist Parties in Western Europe: Evidence from Austria, Belgium, France, Norway, and Switzerland”, International Political Science Review, vol. 29, no 3, p. 349‑373.

Oesch D. (2008b), “The changing shape of class voting: An individual-level analysis of party support in Britain, Germany and Switzerland”, European Societies, vol. 10, no 3, p. 329‑355.

Palombarini S. (2003), Dalla crisi politica alla crisi sistemica: interessi sociali e mediazione pubblica nell’Italia contemporanea, Milan, Franco Angeli.

Piketty T. (2018), “Brahmin Left vs Merchant Right: Rising Inequality and the Changing Structure of Political Conflict”, WID.world Working Paper 2018/7, URL: http://piketty.pse.ens.fr/files/Piketty2018.pdf

Piketty T. (2019), Capital et idéologie, Paris, Seuil.

Rennwald L. (2015), Partis socialistes et classe ouvrière. Ruptures et continuités du lien électoral en Suisse, en Autriche, en Allemagne, en Grande-Bretagne et en France (1970-2008), Neuchâtel, Éditions Alphil-Presses universitaires suisses.

Rey H. (2004), La gauche et les classes populaires. Histoire et actualité d’une mésentente, Paris, La Découverte.

Haut de page

Notes

1 See Amable et al. (2012) and Amable (2021a).

2 See also Amable (2021b).

3 Such a shift has led to the emergence of a multidimensional political space, with the addition of at least a “second” dimension supplementing the traditional left-right economic/redistributive divide. Different interpretations for this new divide have been proposed in the political science literature: “materialists” versus “postmaterialists” (Inglehart, 1977); “winners” versus “losers” of globalisation (Kriesi et al., 2008); “libertarian/cosmopolitan” versus “authoritarian/nationalists/nativists” (Kistchelt, 1994; Hooghe & Marks, 2009). A common feature of these characterisations is their strong link with the education divide, so that whatever the interpretation given to the second dimension, the political space remains spanned by income and education.

4 The so-called Duverger’s law holds that in majoritarian political systems, the political competition tends to take place between two main parties (e.g. the US). In political systems with proportional representation, the political competition takes place between a larger number of parties.

5 Compared to the initial sample of 20 OECD countries from Amable and Darcillon (2021), some countries for which we failed to observe a clear relationship between our different explanatory variables and the variables of income/education have been excluded from our analysis.

6 We exclude from our analysis different categories available in the initial PARTY_LR variable: “did not vote”, “center, liberal”, “other, no specific”, “no party, no preference”.

7 Our index slightly differs from the “rile”-index developed by Laver and Budge (1992). In their index, some categories are defined as right-wing and others as left-wing. The classification of items of the Manifesto Project as left or right has been used in other publications. Social harmony is a typical theme of classical (economic) liberalism: market exchange with free competition is supposed to bring social harmony (as opposed to redistribution resulting from class conflict). Efficiency of the public sector is a typical right/neoliberal theme, political authority is more typically associated with right governments, as is militarism.

8 Please note that Appendix M in Online Appendix provides more detailed figures on the evolution of each main party for each sub-period (1985-1989; 1990-1999; 2000-2009; 2010-2021).

9 In addition, we produce two separate indicators on the specific issue on education spending: a first indicator referring to the positive statements on education [per506] and a second indicator measuring the negative statements on education [per507]. Ideological differences in immigration are captured by the difference between the positive statements on immigration [per602_2] and the negative statements on multiculturalism [per608_1]. Data on the specific issue on immigration/multiculturalism are only available for 2017. Figures for the three indicators are shown in Section M of the Online Appendix.

10 We preferred to use a continuous variable compared to quantiles to produce more precise estimates, given that we run specific-country regressions.

11 As shown in Figure 4, country-specific variables have been used for Australia, France and Norway. For the other countries, we use instead the general degree variable due to large changes in country-specific variable over time.

12 Our estimation strategy differs from Amable & Darcillon (2021) where pooled panel data were used.

13 Correlation matrices in Table N.1 in Appendix N in Online Appendix show that partisan and policy preferences are strongly related to each other in most countries, and the two main explanatory variables, i.e. degree and median income, are not very strongly correlated.

14 Additional figures representing the estimations of the marginal effects with the error margins for each country are in Sections A to K of the Online Appendix.

15 This contrasts with Gethin et al. (2021)’s result that left vote is positively related with education but negatively with income.

16 By contrast, Gethin et al. (2021) found that education influenced positively and income negatively left (Labour) vote.

17 An additional brahmin dimension characterising the support for the Green party and, to a limited extent, the centre-left, was also present in Gethin et al. (2021).

18 We find that the support for the centre-left (SPD) decreases with the education level; the SPD benefits thus from a popular support and is therefore far from being a brahmin left party. This label would better characterise the Green party, whose support is positively related with education, and this in line with Gethin et al. (2021).

19 Rey (2004), Rennwald (2015).

20 We decided not to plot these results for a lack of space (including in the Online Appendix). Results are available upon requests.

21 We focus on France and the United States in reference to the initial work of Piketty (2018). Then, we compare these results with Germany which can share some institutional/political characteristics with France.

Haut de page

Table des illustrations

Titre Figure 1. Distribution of party affiliation by country
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-1.png
Fichier image/png, 143k
Titre Figure 2. Left-Right ideological position and on Redistribution (source: Manifesto Project Database)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-2.png
Fichier image/png, 118k
Titre Figure 3. Ideological position on Globalisation and on Business (source: Manifesto Project Database)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-3.png
Fichier image/png, 121k
Titre Figure 4. Distribution of degree by country
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-4.png
Fichier image/png, 185k
Titre Figure 5. Support for Left (P1)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-5.png
Fichier image/png, 270k
Titre Figure 6. Support for Left (P2)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-6.png
Fichier image/png, 229k
Titre Figure 7. Evidence of a popular left (P1)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-7.png
Fichier image/png, 316k
Titre Figure 8. Evidence of a Brahmin left (P1)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-8.png
Fichier image/png, 305k
Titre Figure 9. Evidence of a bourgeois right (P3)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-9.png
Fichier image/png, 303k
Titre Figure 10. Support for redistribution (R1 and R2)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-10.png
Fichier image/png, 528k
Titre Figure 11. Support for investment in public education (Pl1 and Pl2)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-11.png
Fichier image/png, 179k
Titre Figure 12. Support for immigration with a good effect on the economy (I1)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-12.png
Fichier image/png, 283k
Titre Figure 13. Support for less regulation (B1 and B2)
URL http://journals.openedition.org/regulation/docannexe/image/25121/img-13.png
Fichier image/png, 383k
Haut de page

Pour citer cet article

Référence électronique

Bruno Amable et Thibault Darcillon, « Brahmin left versus merchant right? »Revue de la régulation [En ligne], 38 | 1er semestre |Spring 2025, mis en ligne le 27 octobre 2025, consulté le 13 novembre 2025. URL : http://journals.openedition.org/regulation/25121 ; DOI : https://doi.org/10.4000/15150

Haut de page

Auteurs

Bruno Amable

Professor of political economy, department of History, Economy and Society, Yniversity of Geneva, Unimail, boulevard du Pont d’Arve 40, 1205 Geneva; bruno.amable@unige.ch

Articles du même auteur

Thibault Darcillon

Associate Professor, Department of Economics and Management, LED? University of Paris 8, Saint-Denis, France, 2 rue de la Liberté, 93200 Saint-Denis; thibault.darcillon@univ-paris8.fr

Articles du même auteur

Haut de page

Droits d’auteur

CC-BY-NC-ND-4.0

Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Search OpenEdition Search

You will be redirected to OpenEdition Search