In the last few decades, the political systems of advanced democracies have witnessed momentous changes. Nationalism and populism have gained support almost everywhere, often at the expense of mainstream parties. New dimensions of conflict have emerged – over immigration, globalisation, and civil rights – in place of the classical divide over redistribution (see Figure 1).
Some of these phenomena are correlated with economic changes. The areas of the US that are more exposed to import competition have lost manufacturing jobs and have become politically more polarised and conservative (Autor et al. 2017), and trade shocks or technology shocks account for employment losses and for the rise of populism and anti-immigration sentiment in Europe (Colantone and Stanig 2018a, 2018b, Anelli et al. 2018). Cultural conservatism and support for populist parties have been shown to be strongly correlated with economic insecurity (Guiso et al. 2017, Gidron and Hall 2017). However, the mechanisms behind these correlations are not clear. Why do losers from free trade become nationalist, dislike immigrants, and turn socially conservative? Why do they vote for policy platforms that seem to run counter to some of their interests, such as tax cuts or unsustainable budget deficits?
In a recent paper (Gennaioli and Tabellini 2018), we argue that to address these questions we need to engage with the psychology of group identity and its effect on voter beliefs. The idea that group identity and group-tainted beliefs play a key role in political change permeates the writings of social scientists and revolutionaries. According to Marx and Engels, individual workers should identify with the proletariat, viewing themselves as part of a historical class struggle rather than as carriers of specific cultural or regional traits. For nationalists like Mazzini or Herder, individuals should view themselves as part of the culture or history of an imaginary national community, downplaying more parochial differences. Lipset and Rokkan (1960) describe the evolution of Western party systems as reflecting shifting identities across salient groups such as income classes, religious versus secular groups, the centre versus the periphery, and so on. Changes in the groups of identification can explain why large economic changes can drastically change voters’ beliefs and cause them to become distorted in specific directions. It also helps explain why society becomes strongly polarised across groups, in an ‘us versus them’ conflict.
Social identities and distorted beliefs
Social psychology confirms the link between identity and beliefs. According to the ‘social identity perspective’, the leading theory of groups, individuals routinely identify with social groups of similar people (Tajfel and Turner 1979, Turner et al. 1987) as a way to structure and simplify the social world. Acquiring a social identity, however, also entails an element of ‘depersonalisation’. When an individual identifies with a social group, he enhances differences between outgroups and in-groups. As a result, he stereotypes outgroups but, crucially, he also stereotypes himself by viewing himself more as a typical member of his group than as a unique personality. This implies that identification with a certain group causes voters to move their beliefs in the direction of stereotypes, increasing polarisation and conflict.
To give a political example, the stereotypical ‘cosmopolitan’ favours no controls on immigration and full European integration. Even though such an extreme position is infrequent, it sharply distinguishes the group of those who favour a more open society from the group of those who favour a less open society, say the nationalist group. When someone (say, a member of the European educated elites) identifies himself with the cosmopolitan group, he comes to view himself in contrast to a nationalist. As a result, he perceives nationalists as being more closed than they actually are, and also moves his own beliefs further in the direction of the stereotypical cosmopolitan position. Of course, the same happens when someone holding more closed views identifies with ‘nationalists’ – he thinks of himself as being in contrast to the cosmopolitans, moving his beliefs toward the stereotypical position of his group. As a consequence of this process, these voters hold certain positions not just because of the material benefits they entail, but also because of how they identify their social self.
The distortion of beliefs along group lines has been amply documented by political scientists (Flynn et al. 2017). Recent work shows that belief distortions are systematically shaped by partisanship. US voters exaggerate social mobility, but more so if they identify themselves as right-wing (Alesina et al. 2018a). Voters in the US and Europe overestimate the number of immigrants, but right-wing voters much more than those on the left (Alesina et al. 2018b). Beliefs over global warming also display large partisan differences (Kahan 2014). In our approach, distortions in factual and value judgements naturally arise due to group psychology.
To see the political consequences of this phenomenon, consider the 20th century divide of Western politics: class or economic conflict. Here, a lower-middle-class voter identifying with the working class exaggerates the benefit he draws from redistributive policies because, in thinking about the world, he focuses on the distinguishing feature of his group in contrast to the opposite group – namely, poverty or low upward mobility. The reverse happens for an upper-middle-class voter identifying with a group of rich capitalists. Beliefs about social mobility and policy evaluations become polarised, and redistributive conflict is enhanced relative to a world in which identity does not matter. The same mechanism also explains why polarisation is perceived to be greater than it actually is, in line with the evidence in Westfall et al. (2015).
Changing social identities
But psychology also explains why prevailing social identities may change, so that economic conflict may give room to new cleavages. Indeed, we can all potentially identify with many social groups, defined by our nation, our gender, our social class, our occupation, our cultural traits, and so on. Critically, our political beliefs are shaped by whichever identity is salient at a certain point in time, and this depends on external circumstances.
In our paper we explore the implications of this insight. We show that economic change causes new sources of conflict to become salient, in turn causing identification to change. This creates polarisation along new dimensions, while reducing it along other dimensions. For instance, a trend in globalisation clusters individual interests along exposure to foreign competition or immigration relative to, say, the traditional rich-poor divide. As a result, social identities switch from a poor versus rich conflict to a conflict between globalists and nationalists. Poor or uneducated voters exposed to the costs of immigration or globalisation de-identify with their economic class and identify with the nationalist group. This reduces their demand for redistribution and enhances their demand of external protection. These voters may benefit from greater redistribution, but they do not demand it because they now identify with a group that is more heterogeneous across income classes. Likewise, rich or educated voters benefitting from immigration or globalisation also change their identification to the cosmopolitan group. Their demand for redistribution increases and their demand for openness increases. Overall, social alliances change, individual beliefs about redistribution become less polarised, while beliefs over trade protection or immigration become more polarised. Political conflict over redistribution dampens; conflict over globalisation intensifies.
In other words, through stereotypes and belief distortions, endogenous social identities amplify certain shocks, leading to non-conventional policy responses. A shock induced by trade, or technology, or immigration can have much stronger political consequences if it changes the dimension of social identification. Critically, by inducing people to abandon class identity, these shocks can reduce the demand for redistribution despite increasing income inequality.
Amplification is further enhanced if individual traits and policy preferences are correlated across policy dimensions. Specifically, there is evidence that demand for trade protection and aversion to immigrants are positively correlated across individuals, while these traits are not much correlated with attitudes towards redistribution. This may be an effect of education, or it may be due to an underlying personality or cultural trait that varies across voters. In our theory, this correlation pattern has the following implications. When voters identify with their income class, conflict is predominantly about redistribution. Voters’ beliefs over immigration or trade policy do not polarise politics because income classes – the politically relevant identities – encompass diverse views over globalisation. But now suppose that increased import exposure redefines social identities along a new dimension associated with trade exposure (say, region of residence or occupation). Because views over trade policy are correlated with views over immigration and civil rights, this redefinition of social identities leads to correlated belief distortions over a bundle of policies. This can explain why workers exposed to import competition become anti-immigrant and demand socially conservative policies, and more generally why opinions over different policy issues are now more systematically correlated with partisan identities.
The evidence supports these predictions. In a recent paper, Autor et al. (2013) show that US commuter zones that are more exposed to the rise of imports from China have become politically more polarised and conservative, and were more likely to vote for Donald Trump in the last presidential election. This could reflect a variety of mechanisms, but survey data allow us to explore more specific effects of trade shocks. Exploiting data from the Cooperative Congressional Election Study between 2006 and 2016, we find that individuals more exposed to rising imports from China become more willing to accept cuts in domestic public spending, more averse to immigrants, and consider the issue of abortion to be more important. This is consistent with the idea that the trade shock induced relatively poor respondents to abandon class-based identification and switch to nationalism, given that nationalism is positively correlated with social conservatism. Colantone and Stanig (2018) find similar evidence of the effects of imports shocks in Europe.
France is another interesting case study, because there was a clear shift in the dimensions of political conflicts between 2012 and 2017. This is illustrated vividly in Figure 2. The vertical axis measures attitudes towards immigration, globalisation, and European integration (higher values correspond to more open attitudes); the horizontal axis attitudes towards redistribution and the role of government in protecting workers and regulating the economy (higher values correspond to more right-wing attitudes). Each dot corresponds to an individual. The colours indicate how respondents were split between two clusters estimated from the original questions: in 2013 on the left-hand panel, in 2017 on the right-hand panel. The change in the dimension of political conflict is striking. In 2013 respondents were split between left and right, reflecting the traditional dimension of economic conflict over the role of the state in the domestic economy. In 2017, the cleavage concerned attitudes towards globalisation and immigration.
In our paper we show that this change in the relevant dimension of political conflict is reflected in how people voted and, exploiting panel data, in how their policy preferences changed. Voters who, between 2012 and 2017, abandoned a left versus right identification in favour of a nationalist versus globalist identity moderated their views on redistribution, and became more extreme in their views on globalisation and immigration.
Trade and technology shocks are not the only source of political frictions in modern democracies. Socially conservative cultural views, which were once majoritarian, have gradually been eroded by slow-moving social changes, such as the diffusion of college education and changing gender roles. These cumulative changes created fault lines within traditional political groups defined on the left versus right dimension. Trade and technology shocks increased the relevance of these fault lines, and triggered changes in political and social identities. Socially conservative poor voters, who traditionally identified with left-wing groups despite their social conservatism, are now attracted by nationalism because it appeals to both their trade preferences and their cultural views, and vice versa for voters with opposite political features. As this happens, traditional income- or class-based conflict wanes and is replaced by new political cleavages over correlated dimensions. Political beliefs reflect these new social identities and amplify the effect of these social and economic changes.
If this view of the world is correct, the disruptive political changes that we observe in many democracies are not transitory phenomena, but represent profound and long-lasting transformations of our political systems. One important question in this respect concerns the role of the media. If exposure to social media, such as Facebook or Twitter, strengthens stereotypical thinking, it may cause the consequences of these new political and social identifications to become extreme. At the same time, identities are not biologically ingrained. Because individuals belong to several different groups at the same time, less-polarising identities are always available. As argued by Sen (2007), political platforms reminding people of these alternative identities may reduce polarisation and favour the establishment of a less conflictual political life.