Political Shades of ‘we’: sociotropic uncertainty and multiple political identification in Europe

ABSTRACT

Identity complexity in the sense of people holding multiple, concurrent affiliations, is taken for granted in modern societies. Yet we know very little about what factors stimulate or instead mitigate identification with multiple political communities. It is usually assumed to go hand-in-hand with globalization and other macro-level trends, but empirical research on this issue is scarce. In this article, we examine how sociotropic uncertainty mitigates multiple political identification. We investigate multiple identification with respect to three political communities: the city, the country and the EU. We show, using Eurobarometer data from 25 EU countries, that political distrust, negative economic expectations and societal pessimism all lead individuals to identify with less political communities concurrently. The implication is that the kinds of multiple, cross-cutting identities often invoked by theorists of globalization are hindered by uncertainty and pessimism about the enterprise.

KEYWORDS:

• Multiple identification
• political identity
• uncertainty
• political trust
• economic expectations
• societal pessimism

Introduction

Although Europeans belong to several political communities, they do not necessarily identify with all to the same extent. Some scholars expect globalization to stimulate identification with multiple groups (i.e. multiple identification), resulting in hybrid combinations of identities (Arnett 2002; Geschiere and Meyer 1998; Hermans and Dimaggio 2007). On the other hand, recent decades have witnessed an increasing politicization of identity in Europe that stresses exclusivity. For example, populist movements in many countries emphasize nationalism at the expense of broader commitments to the EU (Duyvendak 2011; Mudde 2007). Furthermore, movements crafted around regional identities, such as those for Scottish and Catalonian independence, have fueled national debate and unrest (Huici et al. 1997; Paasi 2009). And, of course, doubts spring eternal over the prospects for European identification (Lubbers and Scheepers 2010).

We know that political identities are often mobilized, sometimes in complementary fashion (encouraging identity overlap) and sometimes against each other (encouraging a kind of ‘hunkering down’). That said, little systematic research exists on such interactions. All agree that people can hold multiple identities (Brewer 2001; Citrin et al. 2001; Díez Medrano and Gutiérrez 2001; Kohli 2000; Risse 2004), but beyond this recognition, few ask what conditions stimulate or mitigate what we refer to here as multiple identification.

We aim here to understand one specific aspect of multiple identification: namely, the extent to which it is undermined by sociotropic uncertainty, a concern about the state of the collectivity. Over the past decade, European countries have experienced repeated terrorist attacks, economic recession, ‘Brexit’, and a euro crisis. Based on theories regarding the psychological effect of uncertainty, primarily ‘uncertainty-identity theory’ (Grant and Hogg 2012; Hogg 2000; Hogg et al. 2010a, b; Hohman et al. 2010; Reid and Hogg 2005) which focuses on personal, egotropic, types of uncertainty, and most often a single identity, we show that, in addition, sociotropic uncertainty is associated with a reduction in multiple identification. When in doubt, people ‘retrench’, in the sense that they will identify with a smaller profile of political communities. We are agnostic about which identities individuals will fall back to; instead, we identify a general pattern in which concerned individuals are likely to retreat to a smaller set of political identities and allegiances. Using data from the Eurobarometer survey series that cover 25 European countries, we investigate the effect of three types of sociotropic uncertainty: political distrust, negative economic expectations, and societal pessimism. We show that all three mitigate multiple political identification.

Theoretical framework

Conceptualization of collective identity

There is a massive literature on the psychology of social identity, with streams of research focusing on ethnic, racial, religious, gender and geographical identity, among others. These social identities are ‘we’ identities, identifications of an individual with a group or a category, as opposed to ‘me’ identities, which are identifications of the individual with social types, roles or characteristics (Brewer 2001; Simon and Klandermans 2001). In this article, we are interested in political identities, which are spatially defined collective groups that can be called integrative because they connect people within a politically unified territory who embody a wide variety of the ascriptive and acquired identities studied in the literature. In this article, we study identification with towns, countries, and the EU, all of which are politically integrated units (albeit to varying degrees) that contain within them all kinds of other social diversity.

Political ‘identity’ can be described as ‘a subjective or internalized sense of belonging’ (Huddy and Khatib 2007: 65). Although the example of national identity is the best developed in the literature, we can apply the same lens to one’s town or city identity – a phenomenon some in the literature refer to as ‘place attachment’ (Hidalgo and Hernandez 2001; Twigger-Ross and Uzzell 1996) – and EU identity.

The literature on identification distinguishes content-based (or normative), cognitive, and affective dimensions (e.g. Bellucci et al. 2012; Citrin et al. 2001). The first refers to idiosyncratic characteristics (Abdelal et al. 2009): what does it mean to be Parisian, French or European? Second, the cognitive dimension reflects the extent to which people think of themselves as a member of a group. Social identity theory, for instance, points to the influence of identifying as a member of or belonging to a social group (Tajfel and Turner 2004 [1986]). However, this identification as a member of a group does not reveal the importance of the identity for individuals’ self-image. This is captured in the third, affective, dimension, which is referred to as identification with (e.g. Citrin and Sears 2009). Here, the idea is that people can identify with a group to a greater or lesser extent. The affective dimension thus describes the strength or intensity of the identification. This conceptualization dovetails with self-categorization theory, which introduces the influence of salience and context (Turner et al. 1987). This means that self-categorization acknowledges variability in the strength of identification. The affective dimension also gives room to acknowledge that identity is becoming more a subject of choice, which is also referred to as a change from ascribed identities to acquired identities (Huddy 2001). As a result of individualization, the ascribed identities of for instance religion, class and family can be easier fashioned than in previous times (Giddens 1991), although this does mean to say such social categories lose all significance. It is the affective component that is of greatest interest here because we want to examine the level of identification with different groups concurrently, not identification with a group as an all-or-nothing game.

Multiple identification, which can be defined as the concurrent identification with multiple socially or geographically defined groups, has not yet been studied in much depth. Focus so far has fallen on the extent to which people can identify with multiple groups concurrently (Díez Medrano and Gutiérrez 2001). For instance, ethnic minorities such as Hispanics and blacks choose a combination of American and ethnic identity if that option is offered, meaning (presumably) that they identify with both groups at the same time (Citrin et al. 2001). Research on European identification also shows that it does not necessarily compete with national or regional identities (Brewer 2001; Kohli 2000; Risse 2004), which illustrates the point that multiple identities need not be zero-sum or competitive. Questions remain, however, about what types of forces lead people to adopt multiple identities instead of simpler ones. We argue in this article that one factor to take into account is sociotropic uncertainty.

Uncertainty and identity

It is clear that uncertainty is psychologically difficult for people. In the words of Van den Bos ‘people have a fundamental need to feel certain about their world and their place within it’ (2009: 200). Several theories suggest that people deal with uncertainty by placing greater emphasis on their identity or world-views. What is lacking however in all these literatures is research on situational, sociotropic uncertainty. Below we first discuss the four theories at hand, after which we turn to the types of uncertainty studied so far and gaps in that literature. Out of this discussion we develop an argument for the relationship between sociotropic uncertainty, on the one hand, and reduced multiple identification on the other.

Uncertainty-identity theory argues that people react to uncertainty by identifying with a group. It is an offshoot of social identity theory, which has individuals identifying with groups because they seek positive self-esteem (Tajfel and Turner 1979). This happens through depersonalization, or ‘seeing [people] not as unique individuals but as more or less prototypical group members’ (Hogg 2007: 79). Depersonalization takes place when we categorize people into social groups and assign them characteristics of that group, but when we identify ourselves as part of a group. ‘In this way, group identification very effectively reduces self-related uncertainty. It provides us with a sense of who we are, that prescribes what we should think, feel and do’ (Hogg 2007: 80). Empirical studies confirm the uncertainty-identity theory and show that uncertainty increases identification of various types, from identification of students with their university (Grant and Hogg 2012), and identification with student action groups (Hogg et al. 2010a), to identification with a political party (Hogg et al. 2007; Hohman et al. 2010).

Uncertainty-identity theory is not the only game in town. Those working from the auspices of the uncertainty management model also assume that people deal with personal uncertainty by finding a way to make it cognitive manageable; the difference is that people do so by placing emphasis on their culture and the values that are important to them (e.g. Van den Bos 2009). Another loosely-related body of work, terror management theory, focuses on personal mortality salience and how this increases the need for self-esteem and faith in one’s conception of reality (Greenberg et al. 1997). In both literatures, researchers have tested and confirmed the relationship between uncertainty and political attitudes if not identity per se. For example, research shows that personal uncertainty increases peoples’ hostility towards religious values that contradict their own (Van den Bos et al. 2006). And a recent study finds that mortality salience intensified the moral differences between liberals and conservatives (Bassett et al. 2015).

Despite their differences, all three theories (uncertainty-identity theory, uncertainty management model and terror management theory) predict that uncertainty increases the need for people to seek reassurance from the collectivity, either via group identification per se or the adoption of group-derived values and norms.

Three sociotropic sources of uncertainty

Where does uncertainty come from? In general, the literatures referenced above deal in egocentric (or personal) sources of uncertainty. For example, Hogg locates it in ‘contextual factors that challenge people’s certainty about their cognitions, perceptions, feelings and behaviors, and ultimately, certainty about and confidence in their sense of self’ (2007: 77). Some try to prime uncertainty about respondents’ cognitive capabilities (Reid and Hogg 2005), others sow doubts about peoples’ personal financial situation (Hogg et al. 2010a), and still others simply ask study participants to focus on whatever aspects of their lives make them feel uncertain (Hogg et al. 2007). Finally, sometimes uncertainty is invoked in a broader sense, by (for example) asking respondents to read a speech or a newspaper article and underline the words or sentences that made them feel uncertain about themselves, their place in the world and their future (Grant and Hogg 2012; Hohman et al. 2010). Still, despite a wide variety of sources and explicit recognition of contextual influences, these studies all emphasize uncertainty in/about peoples’ personal lives. There is to our knowledge one exception: Kim and Ng (2008) broaden the logic of the uncertainty-identity theory by studying the effect of uncertainty of social change. In this article, we build on this study and examine the impact of specific types of sociotropic uncertainty.

Before diving into the specifics, let us briefly turn away from ‘cause’ and back to ‘effect’. To our knowledge, only two previous studies have explicitly taken on the question of multiple identification, and none have focused on multiple identification. Hohman et al. (2010) examine the effect of personal uncertainty on both party identification and identification as an American. They find that uncertainty strengthens party identification among both Democrats and Republicans, while at the same time undermining American identification among Democrats. The Kim and Ng study referenced earlier (2008) shows that uncertainty about social change increases the likelihood of choosing a single political identity (Hongkonger or Chinese) instead of a dual one (Hongkonger and Chinese or Chinese and Hongkonger).

We argue that a similar logic can be expanded to both various types of sociotropic sources of uncertainty and a multiple identification. Our main innovations are: (1) expanding possible sources of uncertainty (on which more below), and; (2) looking at different types of sociotropic uncertainty. The literature assumes and shows that identification with a group that provides a positive self-image and a clear common faith that decreases uncertainty. This is because it encourages selectivity with respect to political identity choice. Building on that assumption, we expect that a smaller number of political identities, or even a singular group, offer more clear direction and meaning, which is what people are looking for to deal with uncertainty. We are agnostic about which identities individuals will fall back to; this is likely to differ across hometown’s, nations and social groups. Therefore, the overall hypothesis H1 reads as follows: sociotropic uncertainty mitigates multiple identification (H1).

Which types of sociotropic uncertainty are likely to matter? Three types of attitudes seem worthy contenders: political distrust, negative economic expectations and societal pessimism. These attitudes are all reflections of the state of society, rather than egocentric considerations about cognitive ability, economic position, and mortality. They reflect, instead, uncertainty about expressly societal developments.

Political trust has been explicitly tied to identity formation, especially at the national level (Citrin et al. 2014; Kim and Robertson 2002; McLaren 2012), but also at the European level (Hooghe and Verhaegen 2017). And it has been defined in various ways (e.g. Easton 1975; Hardin 1999; Miller and Listhaug 1990; Norris 2011; Van der Meer 2010). Two central elements in those definitions are the performance of the political system and its responsiveness. With respect to performance, it is often argued that political (dis)trust results from evaluations of prior performance in terms of policy success or failure (Easton 1975; Hardin 1999; Norris 2011). Others point to responsiveness, as in Miller and Listhaug’s (1990) famous definition of political trust as ‘a summary judgement that the system is responsive and will do what is right even in the absence of constant scrutiny’. Trust also stems from perceived accountability, or the belief that citizens can hold politicians responsible for their actions (Van der Meer 2010). These elements of political trust, (a lack of) political performance, responsiveness and accountability, are important with respect to sociotropic uncertainty. We argue that it makes sense to view political distrust as a dimension of sociotropic uncertainty. After all, if you believe politicians are not tackling problems appropriately, and/or you think they do not even care about the societal problems to find most troubling, you are both uncertain and concerned about what is to be expected societally. Therefore, we hypothesize that political distrust mitigates multiple political identification (H1a).

The literature on economic expectations uses different labels for what are very similar attitudes, e.g. consumer sentiments, general economic outlook and inflation expectation (De Boef and Kellstedt 2004; Haller and Norpoth 1994; Huth et al. 1994). In this literature, both egotropic and sociotropic considerations are in play. The latter often involves asking what people expect for the next year or a different period in terms of economic development, prices, and unemployment. We expect that economic uncertainty mitigates multiple political identification (H1b).

In contrast to political (dis)trust and economic expectations, societal pessimism has received very little attention. Societal pessimism can be defined as a sentiment among citizens that their society is in decline (Steenvoorden 2015). It refers to a perceived unmanageable deterioration of society, and collective powerlessness to change things for the better. At the same time, it is both abstract and content free, meaning it is agnostic about which aspects of society citizens underlie societal pessimism.¹ By this definition, a large part of the citizenry in European countries western democracies can be described as pessimistic about society (European Commission 2013; Steenvoorden and van der Meer 2017). In line with the other hypotheses, we expect that societal pessimism mitigates multiple political identification (H1c).

Social attitudes about the state of society, politics and the economy are not strictly exogenous to peoples’ various political identities. Indeed, one might easily imagine a reciprocal relationship wherein pessimism about a political body might lead one to ‘de-identify’ with it. While it is less clear how this kind of relationship might emerge in our particular case – that is, where the emphasis is on the extent to which people identify with multiple groups rather than level of identification with a particular body – we have to bear in mind that we are testing associations, as in virtually every cross-national comparative study of an ‘attitude-to-attitude’ variety.

Data and method

To examine the effect of sociotropic uncertainty on multiple identification, we use pooled data from Eurobarometers 68.1 and 77.3 (fielded in 2007 and 2012, respectively). These surveys include a rare combination of items tapping both identification and the three types of sociotropic uncertainty that we consider: societal pessimism, political trust, and economic expectations. Because EU identity is one of the key integrative identities examined, we included only respondents from the 25 countries that have been EU member states since 2004.² Although panel data could provide the best test of the assumed mechanism between sociotropic uncertainty and multiple identification, we think that examining the mechanism with cross-sectional data provides a solid indication of whether or not such a mechanism exists.

To measure identification with the nation, the EU and the town/city, we use items that ask how attached respondents feel to each of these three polities: ‘not at all’, ‘not very much’, ‘fairly’ and ‘very’ attached. These items are plausible indicators of identity’s affective dimension. Moreover, because respondents are asked about each identity independently, patterns of multiple identification are much easier to observe than with ‘prioritization’-style measures such as rankings. We recoded these variables into dichotomies: low identifiers (collapsing ‘not at all’ and ‘not very much’) and high identifiers (collapsing those who claim to be ‘fairly’ and ‘very’ attached). These dummies are then combined to create 4 categories: low identifiers (low identification with town/city, country and EU), single identifiers (high identification with only one out of three), dual identifiers (high identification with two out of three) and multiple identifiers (high identification with all three). Note that this measure is entirely agnostic with respect to the geographic hierarchy of these identities.

Table 1 shows the four categories and their size. Multiple identifiers and dual identifiers are the two largest groups, comprising 42% and 48% of all respondents (respectively). Eight percent are single identifiers, and only 3% are low identifiers with respect to all three collective groups. The single and dual identifiers can be further divided into three subgroups, as shown in Table 1. For the most part, dual identifiers are EU-averse (43%), identifying instead with their town and nation: ‘anti-nationalists’ are a very rare species, with only approximately 1% of respondents belonging to this group. For the single identifiers, the country is twice as common an object of allegiance as the town, and EU-only identity is vanishingly rare (1%). From Table 1 it follows that the difference between multiple identification and dual identification is mostly based on a lack of EU identification. Therefore, we performed robustness checks to make sure we are not solely measuring a lack of EU identification instead of multiple identification per se.

Table 1. Identification categories (%)^a

Low identifiers	3
Single identifiers	8
Town-only identifiers	3
Nation-only identifiers	5
EU-only identifiers	1
Dual identifiers	48
Town avers	3
Nation avers	1
EU avers	43
Multiple identifiers	42

^aNumbers are rounded.

Political distrust is measured with three items about national politics, and two about European politics: trust in political parties, the national government, the national parliament, the European parliament and the European commission. Each of these questions allowed respondents to opt for ‘tend to trust’ or ‘tend not to trust’. Because it can be expected that national political distrust affects multiple identification in a different way than EU-level political distrust, we created both a national and an EU-level scale of political distrust. Taking the sum score of the three national and two EU-level items on political distrust results in two scales that range from 0 to 3 and 0 to 2, respectively.³

For economic negative expectations, we rely on perceptions of the national and European economy to ensure, as best we can, that respondents are providing sociotropic rather than egocentric judgments (Hooghe and Marks 2005; Verhaegen et al. 2014). For the national level, we summarized two variables: the respondent’s expectations for the next 12 months with respect to the national economic situation and national employment. For the EU level, we use an item on the expectation for the next 12 months with respect to the European economic situation. The answer options for the three economic items are better, worse or the same. To focus the analyses on sociotropic uncertainty only, and to enable comparison between variables, we dichotomized both economic items into better/same versus worse. For the national level scale, we used the sum score of those answering ‘worse’ on those two items, and ranges from 0 to 2.⁴

Societal pessimism is tapped with a question about the direction citizens think their country is heading, worded as follows: ‘At the present time, would you say that, in general, things are going in the right direction or in the wrong direction, in [your country]’. This measure is used in previous research on societal pessimism (Steenvoorden and van der Meer 2017), and matches the concept of societal pessimism because of its broad emphasis on the country as a whole, as well as the focus on the process (decline or improvement). The answer categories are ‘things are going in the right direction’ (optimists), ‘things are going in the wrong direction’ (pessimists), a spontaneous answer of ‘neither one nor the other’ (undecideds), with ‘don’t know’ treated as missing values. As in the case of the economic items, we dichotomized this item, resulting in a dummy for societal pessimists (wrong direction) versus the optimists and undecideds. The three measures of sociotropic uncertainty do not measure uncertainty as such, but rather negative expectations. However, they do capture concern about what awaits us. And it is not uncommon to use similar negative assessments, which have been fund to affect identification (e.g. Grant and Hogg 2012).

The three types of sociotropic uncertainty are weakly to moderately related, ranging from .13 (Cramer’s V) between national political distrust and expectations about European economy, to .42 (Cramer’s V) between national political distrust and societal pessimism (all significant at p < .001). To enable direct comparison between the coefficients of these three sociotropic uncertainty attitudes, we standardized them all.

At the individual level, we control for basic socio-demographic characteristics, namely, sex, age group (15–24, 25–39, 40–54, 55+), educational level (low, medium high, student), employment status (employed versus unemployed and other) and type of community (rural, small town, large city). We added a dummy variable on life satisfaction (not very or not at all satisfied versus very or fairly satisfied) to further filter out personal circumstances with regards to sociotropic uncertainty. It shows only weak associations with the sociotropic uncertainty measures (highest association is −.22 Cramer’s V in the case of societal pessimism, significant at p < .001), which strikes us as plausible evidence that these attitudes are indeed capturing sociotropic (rather than egocentric) considerations. We also control for personal uncertainty to disentangle that from sociotropic uncertainty. To this end, we include items on the expectations for the next 12 months about the financial situation of one’s household and one’s personal job situation. We created dummy variables (the answer ‘worse’ versus better or the same), which we standardized to enable direct comparison with the coefficients of sociotropic uncertainty.

Because the dependent variable consists of four nominal categories, the appropriate type of analysis is multinomial regression analysis, and with multiple identifiers as a reference group. To account for country and temporal variation, we include dummies for both countries and waves (2007 versus 2012). Additionally, we used clustered robust standard errors to deal with the error structure.

Results

Table 2 shows the results of model 1, including only the sociotropic uncertainty attitudes: national and European political distrust, national and European negative economic expectations, and societal pessimism (for reasons of space, the dummies for countries and wave are not included in the presented models). In almost all instances, sociotropic attitudes show a positive, significant effect for low, single and dual identifiers (compared to multiple identifiers, the reference category), with the exception of negative economic expectations about the EU for low and single identifiers compared to multiple identifiers. This means that being negative about developments on either politics, the economy, or society as a whole increases the likelihood of holding fewer identities concurrently. Furthermore, comparing the attitudes, shows European political distrust to have the largest coefficient, followed by national political distrust. Third in line is societal pessimism, still larger than national economic expectations, and EU economic expectations have the smallest coefficient.

Table 2. Model 1: Effects of attitudes of sociotropic uncertainty on multiple identification^a.

	Low identifiers		Single identifiers		Dual identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.473***	(.063)	.341***	(.039)	.203***	(.028)
Political distrust (EU)	.733***	(.040)	.605***	(.042)	.641***	(.030)
Negative economic expectations (nat.)	.142*	(.058)	.103**	(.040)	.088***	(.020)
Negative economic expectations (EU)	.063	(.039)	.057	(.036)	.062**	(.020)
Societal pessimism	.269***	(.049)	.191***	(.035)	.143***	(.025)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

Table 3 presents the analysis including the control variables. The effects of the sociotropic variables still hold in almost all cases: all else equal, each of the three types of sociotropic uncertainty decrease multiple identification and stimulate a more selective identification pattern. This supports our hypothesis that sociotropic uncertainty mitigates multiple identification.

Table 3. Model 2: Effects of sociotropic uncertainty on multiple identification, including control variables^a.

	Low identifiers		Single identifiers		Dual identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.466***	(.063)	.321***	(.038)	.200***	(.028)
Political distrust (EU)	.727***	(.037)	.606***	(.041)	.618***	(.030)
Negative economic expectations (nat.)	.107	(.058)	.104**	(.037)	.074***	(.021)
Negative economic expectations (EU)	.073*	(.036)	.070	(.041)	.066**	(.020)
Societal pessimism	.213***	(.045)	.188***	(.034)	.122***	(.026)
Control variables:
Male	−.172*	(.083)	.010	(.050)	.059	(.041)
Age (40–54):
15–24	.693***	(.163)	.406**	(.147)	.113	(.104)
25–39	.326**	(.118)	.187**	(.067)	.004	(.046)
55+	−.419***	(.098)	−.283***	(.079)	−.009	(.048)
Educational level (medium):
Education low	−.022	(.184)	−.007	(.096)	.117*	(.051)
Education high	−.217*	(.086)	−.005	(.078)	−.243***	(.046)
Students	.171	(.156)	.436*	(.181)	−.133	(.086)
Occupation (employed):
Unemployed	.291	(.163)	.193*	(.092)	.079	(.049)
Other	−.188	(.130)	−.241***	(.052)	−.076*	(.037)
Community (small town):
rural	.122	(.129)	.035	(.064)	.100*	(.045)
large city	.007	(.098)	−.308***	(.087)	−.147**	(.049)
life satisfaction	−.374***	(.041)	−.237***	(.046)	−.110***	(.031)
Attitudes of personal uncertainty:
negative personal financial expectations	−.037	(.032)	.028	(.033)	.032	(.017)
negative job expectations	.120***	(.036)	−.054**	(.020)	−.027	(.016)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

As uncertainty-identity literature mostly focused on personal uncertainty, it is interesting to compare the coefficients of personal and sociotropic uncertainty in Table 3. We find that a negative expectation about one’s job situation has the expected positive effect in one of the three groups in Table 3, decreasing multiple identification significantly in comparison to low identification. Still, the effect is rather small (.120) compared to national (.466) and European (.727) political distrust and societal pessimism (.213). Of the sociotropic uncertainty attitudes, only negative economic expectations about the EU has a smaller coefficient than job expectations. Furthermore, a negative expectation about one’s financial situation does not reach significance, and the expectation about one’s job has a (just significant) negative effect on a single identification, outcomes that contradict the theories on personal uncertainty. Further analyses indicate that both personal level economic indicators effects’ change when the negative (national and European) economic expectations are not included. In that case, personal financial situation does reach significance for dual and single identifiers (compared to multiple identification), and job expectation shows an effect in the assumed positive direction in the case of low identifiers. As the results do not point to any multicollinearity problems in model 2 of Table 3 (variance inflation factors does not reach more than 1.78, Tolerance does not reach lower than 0.56 for personal financial situation), we interpret the unexpected results in the cases of personal financial situation and job expectations as mediation by negative economic expectations.

Although ideally we would compare a larger range of both types of indictors, especially personal uncertainty, to draw conclusions, these results show sociotropic uncertainty is by no means less important, on the contrary, than personal uncertainty when studying multiple political identification.

Additionally, many of the controls are significant as well (the results for the country and wave dummies can be found in Appendix Table A1). Because there is little work on this topic, it is worth discussing them in some detail. Younger respondents have a higher chance of being single or low identifiers. Educational level also influences multiple identification, with the highly educated found less often among the low and dual identifiers, and the less educated found more often among the dual than the multiple identifiers. Unemployment increases the chance on single identification compared to multiple identification. City-dwellers are less likely to be single or dual identifiers than those from small towns, which indicates that they more often combine different political identities than rural inhabitants. Finally, life satisfaction goes hand-in-hand with multiple identification. These results mirror findings in studies on national and/or European identity which also find an impact of age, education, employment status (Fligstein 2008; Hjerm and Schnabel 2010; Masella 2013; Risse 2010).

To further examine which of the sociotropic uncertainty variables are most important in explaining identification patterns, we compare the predicted probabilities of these effects on being a low, single, dual or multiple identifier in Figures 1–4, where the predicted probabilities are presented for a change from −1 to +1 on the standardized variables (only significant variables are included in the figures). For example, the chance of being a dual identifier (Figure 3) increases from 36% to 57% if one moves from −1 to +1 standard deviation on the European political distrust scale. In the case of European negative economic expectations, the impact is the smallest (45% to 47%).

Figure 1. Predicted probabilities low identifiers.

Figure 2. Predicted probabilities single identifiers.

Figure 3. Predicted probabilities dual identifiers.

Figure 4. Predicted probabilities multiple identifiers.

The figures all show the importance of first of all European political distrust on identity retrenchment, second national political distrust, and a much smaller but significant impact of societal pessimism and both national and European negative economic expectations. However, the four figures paint different pictures, and the differences are worth discussing.

Figures 3 and 4 tell us about the dual and multiple identifiers, and portray that for these two groups, European political distrust is most important for their political identification, although in opposite ways: this relationship is positive for dual identifiers, and negative for multiple identifiers, which is what we would expect. Figures 1 and 2 show that in the case of low and single identifiers, the differences between the attitudes are (much) smaller. Taken together, the figures show that European political distrust is particularly important in differentiating between dual and multiple identifiers, which is very plausible given that dual identifiers are mostly EU averse. For low and single identifiers, the different types of sociotropic attitudes are of more equal importance.

We did three sets of robustness checks, each of which confirms the basic pattern shown in Table 3, and underlines that the economic expectations has the weakest effect of the three types of sociotropic uncertainty. First, we estimated model 2 for all 25 countries separately.⁵ Although not all five operationalizations of sociotropic uncertainty are significant all three comparisons (low, single and dual identifiers versus multiple identifiers), the overall pattern is the same: we find significant positive effects in almost all countries in the case of political distrust, and in a quarter to half of the countries in the cases of societal pessimism and negative economic expectations. In no instance do we find negative effects.

Second, because Table 1 shows that the main difference between the multiple identifiers and the dual identifiers is a lack of identification with the EU, we repeated the analysis using dual identifiers as a reference category (see Appendix Table A2). This replicates most of the findings, with some exceptions. Both national and European negative economic expectations are insignificant (instead of one of them) for both low and single identifiers, and European political distrust is insignificant for single identifiers (all in comparison to dual identifiers). This means our findings do not reflect simply a lack of EU identification, but result from a mechanism that can be found irrespective of the specific combination of identities. Nevertheless, the finding that the main difference between the multiple identifiers and the dual identifiers is a lack of identification with the EU, does mean that the EU identity is one that will be jettisoned most often by those experiencing sociotropic uncertainty. This contributes to the literature on the EU identification. Contrary to findings that dual attachment to both country and the EU is increasing (Citrin and Wright 2014), our results indicate that EU identification is likely to be mitigated most of all political identities during distressing periods.

Third, we looked at the results when comparing two identities in pairwise fashion instead of all three at once. In so doing, we can scrutinize whether our effects are driven by anti-EU feelings. Tables A3, A4 and A5 of the Appendix show the results of comparing identification with the country and the city, the country and the EU, and the city and the EU. Again, the hypothesis that sociotropic uncertainty reduces multiple political identification is supported in the large majority of instances, and all directions (with the exceptions that in some cases, negative economic expectations (national or/and European) and/or societal pessimism are insignificant). This underlines that sociotropic uncertainty leads to less concurrent political identification, which for some people can involve aversion to the nation, for others to the EU, and for still others to the city. For instance, the young and the highly educated are more likely than others to retrench to EU only compared to country & EU and city & EU (tables D &E) than others.

To conclude, although the most prominent divide (in terms of size) appears to be that between those who identify with the EU and those who do not, sociotropic uncertainty consistently drives identity retrenchment across a variety of profiles. Dual, single, and low identifiers always tend to express more sociotropic anxiety than multiple identifiers, which amounts to strong evidence for the mechanism at issue here.

Conclusions and discussion

Although research on political identification recognizes both the ubiquity and importance of multiple identification (Brewer 2001; Citrin et al. 2001; Díez Medrano and Gutiérrez 2001; Kohli 2000; Risse 2004), its causes have been generally assumed rather than studied systematically. In this article, we aimed to show the relationship between sociotropic uncertainty and a more selective and simplified identification pattern with respect to integrative, political-geographical collective groups. We proposed to broaden the insight of the uncertainty literatures, especially the uncertainty-identity literature, that personal, egotropic uncertainty affects identification, to both sociotropic uncertainty and multiple identification. We argued that sociotropic uncertainty mitigates multiple political identification, because it makes people seek certainty by embracing simplified identity profiles that more easily mitigate uncertainty.

Political distrust, negative economic expectations and societal pessimism all go hand-in-hand with selective (rather than multiple) political identification. European political distrust has the largest effect, followed by national political distrust, and national and European negative economic expectations and societal pessimism matter to smaller, similar degrees. These findings persist even when controlling for various demographic and socioeconomic factors. And the sociotropic factors are of larger effect than the indicators of personal uncertainty that are the typical focus in research in the various streams of literature on uncertainty, such as uncertainty-identity theory, but also uncertainty management model, and terror management theory, namely uncertainty about one’s job and financial situation. Robustness checks, which look at single countries, use different reference categories, or different sets of political identification, show the same results. Furthermore, they underline that there is no single pattern of retrenchment. While many citizens drop the EU-identity, some others hold on to the EU identity and drop the country or city identification.

A limitation of our study is our focus on specific types of political identities, namely the city or town, the country and the EU. Because of the dominant role of regional identities in several European countries, that type of political identification should have ideally been included here. That said, while regional identity is an important phenomenon and descriptively interesting, it is unclear how its absence here would bias or undermine the findings we show. Another limitation is the cross-sectional nature of our data, while a more robust test would be provided with longitudinal or experimental data. However, the clear pattern of correlations in all analyses and in all robustness checks seems a solid indication that the assumed mechanism is indeed taking place. A third point of attention is our operationalization of uncertainty with negative attitudes. Although other studies into uncertainty-identity theory also operationalize uncertainty in a way that resembles concern or anxiety about developments taking place (e.g. Grant and Hogg, 2012), future research is needed to examine whether indeed uncertainty is driven by concerns about possible deterioration.

Our contribution to the sprawling empirical literature on identity formation is twofold. First, the insight that sociotropic uncertainty has an impact on multiple political identification expands the insights of the uncertainty-identity literature on personal, egotropic uncertainty. It means not only that concerns about the social context matters no less than personal uncertainties for understanding identification, but the results also indicate that besides from the strength of specific identities, the issue of how complex people’s identity profiles appear to be deserves more attention.

Second, empirical researchers have only recently begun to study multiple political identification, and the results shown here raise numerous new and interesting research questions. As just one example, we did not go into cross-national heterogeneity of identification within geographic groups; it is likely, however, that the make-up of the categories of single, dual and multiple identifiers differs across countries. When strong regional identities are present, such as in Spain, Germany or Italy, the multiple identification pattern is likely to differ from countries such as Denmark, the Netherlands or Sweden. The same goes for cross-national differences in the level of euroscepticism. This and many other questions await further research.

While it would be difficult to make the argument directly in the context of this paper, our findings certainly mesh well with proliferating accounts of ‘white backlash’ in the US (e.g. Abrajano and Hajnal 2017; Cramer 2016; Gest 2016) and so-called populism beyond (e.g. Akkerman, Mudde and Zaslove 2014). The argument there tends to be that uncertainty about and hostility to immigration, globalization, and economic inequality, espoused by putative victims of these trends, drives out-group hostility based on primordial ethnic and national ties. Our argument here – that people seem to fall back on simpler, more meaningful identities in the face of uncertainty about the world – is entirely consistent with these portrayals.

More generally, our results also have implications on the functioning of political communities. It is generally assumed that cultural and economic globalization drive people toward increasingly complex, that is, multiple, identities (Arnett 2002; Geschiere and Meyer 1998; Hermans and Dimaggio 2007). While this may indeed be the case for some, the uncertainty of the kinds explored here seems to push people in the opposite direction. This is a concern because multiple, cross-cutting identifications, particularly with respect to the types of ‘integrative’ groups studied here, are seen as central to the functioning of polities (Easton 1975; Kramer and Brewer 1984; Manzo and Perkins 2006; Miller 1995; Transue 2007). The implications of more selective identity patterns – i.e. choosing a smaller number of identities – are therefore troublesome. If people are less willing to identify with the various political entities to which they belong concurrently, this reticence undermines these political communities just as they are confronted by significant political, social, and economic challenges.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Eefje H. Steenvoorden is a postdoctoral researcher in the Department of Political Science at the University of Amsterdam, the Netherlands. Her research focuses on social divisions in political attitudes and political behavior in European countries. Topics include societal pessimism, political trust, voting for Radical Right Parties and political and civic participation.

Matthew Wright is an Assistant Professor in the Department of Government at American University (Washington, D.C.). In his research, he has explored numerous topics in American and European political behavior. These include: the causes and implications of political identity; immigration, assimilation, and citizenship policies; the politics of ethnic diversity; national identity and patriotism; religion and politics; political culture; social capital, civic engagement, and trust, and US voting behavior.

Notes

1 The definition of pessimism used here is similar to the ones in the few other studies on societal unease or pessimism about society. An example is cultural pessimism, which is “the conviction that the culture of a nation, a civilization or of humanity itself is in an irreversible process of decline” (Bennett 2001). Elchardus & Smits use “a lack of well-being [about being part of] society” (2007: 104), as a definition of societal unease. These can be seen as alternatives for the way pessimism is defined here.

2 Austria, Belgium, Czech Republic, Cyprus, Denmark, Estonia, Finland, France, Germany, Great Britain, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain, and Sweden.

3 Cramer’s V of national level political distrust items ranges between .54 and .74. Cramer’s V of the EU-level political distrust items is .83 (both significant at p < .001).

4 Cramer’s V of the two national economic expectation items is .56 (significant at p < .001).

5 The model did not converge for Malta, therefore, we did could not include it in this robustness check.

Appendix

Table A1. Model 2: Effects of sociotropic uncertainty on multiple identification, including control variables and country and year dummies^a.

	Low identifiers		Single identifiers		Dual identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.466***	(.063)	.321***	(.038)	.200***	(.028)
Political distrust (EU)	.727***	(.037)	.606***	(.041)	.618***	(.030)
Negative economic expectations (nat.)	.107	(.058)	.104**	(.037)	.074***	(.021)
Negative economic expectations (EU)	.073*	(.036)	.070	(.041)	.066**	(.020)
Societal pessimism	.213***	(.045)	.188***	(.034)	.122***	(.026)
Control variables:
Male	−.172*	(.083)	.010	(.050)	.059	(.041)
Age (40–54):
15–24	.693***	(.163)	.406**	(.147)	.113	(.104)
25–39	.326**	(.118)	.187**	(.067)	.004	(.046)
55+	−.419***	(.098)	−.283***	(.079)	−.009	(.048)
Educational level (medium):
Education low	−.022	(.184)	−.007	(.096)	.117*	(.051)
Education high	−.217*	(.086)	−.005	(.078)	−.243***	(.046)
Students	.171	(.156)	.436*	(.181)	−.133	(.086)
Occupation (employed):
Unemployed	.291	(.163)	.193*	(.092)	.079	(.049)
Other	−.188	(.130)	−.241***	(.052)	−.076*	(.037)
Community (small town):
Rural	.122	(.129)	.035	(.064)	.100*	(.045)
Large city	.007	(.098)	−.308***	(.087)	−.147**	(.049)
Life satisfaction	−.374***	(.041)	−.237***	(.046)	−.110***	(.031)
Attitudes of personal uncertainty:
Negative personal financial expectations	−.037	(.032)	.028	(.033)	.032	(.017)
Negative job expectations	.120***	(.036)	−.054**	(.020)	−.027	(.016)
Year (2012)	−.461**	(.160)	−.409***	(.098)	−.247**	(.085)
country (France)
Belgium	.739***	(.024)	.017	(.014)	−.069***	(.009)
The Netherlands	1.972***	(.063)	1.483***	(.040)	1.255***	(.028)
Germany West	−.055	(.038)	−.166***	(.022)	.072***	(.018)
Italy	−.609***	(.082)	−1.163***	(.049)	−.430***	(.036)
Luxembourg	.469***	(.041)	−.208***	(.037)	−.152***	(.022)
Denmark	−.282***	(.078)	.274***	(.058)	.987***	(.031)
Ireland	−.297***	(.055)	−1.343***	(.027)	−.007	(.021)
Great Britain	.730***	(.055)	.618***	(.040)	.743***	(.035)
Greece	−2.072***	(.103)	−1.688***	(.076)	.521***	(.039)
Spain	−1.300***	(.091)	−.637***	(.051)	−.429***	(.037)
Portugal	−1.076***	(.136)	−1.589***	(.078)	−.206***	(.045)
Finland	.094	(.049)	.877***	(.047)	1.370***	(.023)
Sweden	.088	(.065)	.444***	(.053)	.903***	(.027)
Austria	−.207***	(.035)	−.865***	(.033)	.373***	(.022)
Cyprus (Republic)	−.562***	(.060)	−.013	(.032)	.849***	(.016)
Czech Republic	.851***	(.043)	−.204***	(.036)	.314***	(.024)
Estonia	.948***	(.047)	.729***	(.032)	.951***	(.025)
Hungary	−.400***	(.089)	−.752***	(.075)	.039	(.045)
Latvia	−.453***	(.050)	−.738***	(.037)	−.141***	(.033)
Lithuania	.377***	(.041)	.160***	(.046)	.760***	(.037)
Malta	−.092*	(.043)	−.084**	(.029)	.196***	(.018)
Poland	−1.734***	(.053)	−1.888***	(.040)	−.480***	(.031)
Slovakia	.023	(.023)	−.453***	(.031)	.170***	(.017)
Slovenia	−.225***	(.038)	−.702***	(.028)	.129***	(.022)
Constant	−2.465***	(.112)	−1.072***	(.084)	.020	(.053)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

Table A2. Effects of sociotropic uncertainty on multiple identification, including control variables (reference group = dual identifiers)^a.

	Low identifiers		Single identifiers		Multiple identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.266***	(.059)	.121***	(.027)	−.200***	(.028)
Political distrust (EU)	.109**	(.035)	−.012	(.032)	−.618***	(.030)
Negative economic expectations (nat.)	.034	(.047)	.030	(.034)	−.074***	(.021)
Negative economic expectations (EU)	.007	(.036)	.004	(.038)	−.066**	(.020)
Societal pessimism	.091*	(.044)	.066*	(.029)	−.122***	(.026)
Control variables:
Male	−.232**	(.083)	−.050	(.055)	−.059	(.041)
Age (40–54):
15–24	.580***	(.138)	.294*	(.116)	−.113	(.104)
25–39	.322**	(.117)	.183***	(.053)	−.004	(.046)
55+	−.410***	(.099)	−.274***	(.072)	.009	(.048)
Educational level (medium):
Education low	−.139	(.162)	−.124	(.090)	−.117*	(.051)
Education high	.027	(.076)	.238***	(.058)	.243***	(.046)
Students	.304*	(.135)	.569***	(.153)	.133	(.086)
Occupation (employed):
Unemployed	.212	(.150)	.114	(.090)	−.079	(.049)
Other	−.112	(.133)	−.164**	(.054)	.076*	(.037)
Community (small town):
Rural	.022	(.106)	−.065	(.044)	−.100*	(.045)
Large city	.153	(.105)	−.162	(.088)	.147**	(.049)
Life satisfaction	−.265***	(.037)	−.127**	(.041)	.110***	(.031)
Attitudes of personal uncertainty:
Negative personal financial expectations	−.069*	(.029)	−.004	(.023)	−.032	(.017)
Negative job expectations	.147***	(.038)	−.027	(.022)	.027	(.016)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

Table A3. Effects of sociotropic attitudes on dual identification: Country versus City (reference group = dual identifiers)^a.

	Low identifiers		Country only identifiers		City only identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.355***	(.056)	.093**	(.031)	.298***	(.054)
Political distrust (EU)	.173***	(.031)	.054**	(.021)	.116*	(.051)
Negative economic expectations (nat.)	.072	(.046)	.012	(.038)	.165***	(.034)
Negative economic expectations (EU)	−.006	(.032)	.029	(.037)	−.092	(.050)
Societal pessimism	.145***	(.042)	.030	(.029)	.212***	(.034)
Control variables:
Male	−.232***	(.067)	−.011	(.041)	−.251***	(.071)
Age (40–54):
15–24	.709***	(.170)	.335*	(.132)	.557***	(.126)
25–39	.299**	(.107)	.190***	(.057)	.272***	(.079)
55+	−.356***	(.083)	−.179*	(.075)	−.411**	(.127)
Educational level (medium):
Education low	−.111	(.159)	−.232**	(.087)	−.072	(.165)
Education high	.097	(.084)	.297***	(.059)	.187*	(.078)
Students	.358*	(.144)	.384*	(.155)	.483**	(.175)
Occupation (employed):
Unemployed	.305*	(.148)	.243***	(.071)	.027	(.133)
Other	−.170	(.093)	−.144*	(.064)	−.038	(.097)
Community (small town):
Rural	.026	(.113)	.072	(.057)	−.043	(.073)
Large city	.001	(.113)	−.191*	(.081)	−.208*	(.095)
Life satisfaction	−.278***	(.041)	−.141**	(.043)	−.169***	(.047)
Attitudes of personal uncertainty:
Negative personal financialexpectations	−.040	(.032)	.032	(.017)	−.033	(.043)
Negative job expectations	.127***	(.035)	−.046	(.026)	.054	(.044)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

Table A4. Effects of sociotropic attitudes on dual identification: Country versus EU (reference group = dual identifiers)^a.

	Low identifiers		Country only identifiers		EU only identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.458***	(.054)	.202***	(.028)	.296***	(.087)
Political distrust (EU)	.697***	(.034)	.682***	(.026)	−.089	(.071)
Negative economic expectations (nat.)	.140***	(.039)	.066***	(.019)	.191***	(.057)
Negative economic expectations (EU)	.051	(.032)	.086***	(.020)	−.156*	(.071)
Societal pessimism	.242***	(.038)	.123***	(.025)	.222***	(.060)
Control variables:
Male	−.133*	(.061)	.098**	(.037)	−.345**	(.125)
Age (40–54):
15–24	.584***	(.145)	−.025	(.095)	.603*	(.259)
25–39	.277**	(.087)	−.024	(.051)	.162	(.120)
55+	−.434***	(.096)	−.013	(.053)	−.163	(.178)
Educational level (medium):
Education low	.034	(.150)	.151**	(.049)	−.045	(.220)
Education high	−.226**	(.078)	−.339***	(.046)	.337*	(.131)
Students	.311*	(.159)	−.140	(.095)	.343	(.262)
Occupation (employed):
Unemployed	.135	(.114)	.035	(.062)	.297	(.209)
Other	−.172	(.099)	−.090*	(.040)	−.065	(.145)
Community (small town):
Rural	.049	(.086)	.071	(.046)	−.040	(.128)
Large city	−.134	(.087)	−.134**	(.045)	−.177	(.143)
Life satisfaction	−.306***	(.038)	−.109***	(.032)	−.158*	(.073)
Attitudes of personal uncertainty:
Negative personal financial expectations	−.044	(.033)	.031	(.019)	.054	(.056)
Negative job expectations	.063*	(.027)	−.039*	(.018)	.123*	(.054)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***p < .001.

Table A5. Effects of sociotropic attitudes on dual identification: City versus EU (reference group = dual identifiers)^a.

	Low identifiers		EU only identifiers		City only identifiers
	b/beta^b	(se)	b/beta	(se)	b/beta	(se)
Attitudes of sociotropic uncertainty:
Political distrust (nat.)	.302***	(.043)	.174***	(.041)	.219***	(.029)
Political distrust (EU)	.697***	(.031)	−.003	(.033)	.685***	(.027)
Negative economic expectations (nat.)	.068	(.042)	.051	(.055)	.072***	(.020)
Negative economic expectations (EU)	.112**	(.038)	−.014	(.042)	.084***	(.020)
Societal pessimism	.157***	(.033)	.066*	(.034)	.130***	(.026)
Control variables:
Male	.017	(.057)	−.087	(.062)	.090*	(.039)
Age (40–54):
15–24	.351*	(.136)	.677***	(.191)	.057	(.107)
25–39	.209**	(.070)	.193**	(.064)	−.015	(.049)
55+	−.289***	(.071)	−.088	(.118)	−.012	(.050)
Educational level (medium):
Education low	−.022	(.108)	−.312*	(.154)	.138**	(.046)
Education high	−.122	(.074)	.351***	(.080)	−.340***	(.045)
Students	.251	(.170)	.243	(.169)	−.141	(.097)
Occupation (employed):
Unemployed	.247*	(.114)	.385*	(.160)	.039	(.059)
Other	−.227***	(.067)	−.133	(.118)	−.085*	(.038)
Community (small town):
Rural	.083	(.088)	.155	(.087)	.085	(.044)
Large city	−.157	(.086)	−.267*	(.115)	−.150**	(.047)
Life satisfaction	−.294***	(.040)	−.124**	(.045)	−.109***	(.032)
Attitudes of personal uncertainty:
Negative personal financial expectations	.022	(.024)	.022	(.043)	.022	(.019)
Negative job expectations	.004	(.026)	−.019	(.031)	−.042*	(.018)

^aMultinomial regression, reference group = multiple identifiers, coefficients are log odds. N = 28532.

Controlled for country and wave dummies.

^bSince all attitudes are standardized, their reported coefficients are beta’s.

*p < .05; **p < .01; ***; p < .001.