Do European Trade Unions Foster Social Solidarity? Evidence from Multilevel Data in 18 Countries

Trade unions have been analysed quantitatively primarily in their role as vested interest organisations, attempting to quantify the excludable benefits they provide to members rather than examine their wider impact in an institutional context. Power resource theory acknowledges unions as social agents but assumes the willingness to oppose neoliberalism is constant, limited only by scarce power resources. Whilst true in general terms, this fails to explain trends of increasing labour market dualism in resource&#8208;rich industrial relations regimes. This article examines social solidarity as a union power resource, measuring the impact of trade union membership on social attitudes of solidarity. Data were collected from the 2016 European Social Survey for 18 countries, grouped into five distinct industrial relations regimes. The findings suggest that, at European level, union membership still has a significant effect on all dimensions of social solidarity, but these relationships vary significantly across industrial relations regimes.


INTRODUCTION
Trade unions are socio-economic actors with a dual purpose: to serve the interests of their members and to promote a more egalitarian social model in society as a whole. The latter is the foundation of their representative legitimacy and the basis on which their role became institutionalised in industrial relations (IR) as we know them (Flanders, 1970). In recent decades, a generalised decline in union membership severely limited their ability to fulfil their roles.
It is common for unions and union performance to be analysed as organisations of vested interest, primarily protecting their members, even at the expense of outsiders. 1 Whilst this analysis explains some trends of union decline, particularly when situated in the increasingly limiting legal framework in which unions must operate, it still fails to explain the initial rise and institutionalisation of unions. Power resource theory (PRT) links the decline in employment protections to the decline in power resources which unions need in order to oppose the interests of capital (Korpi, 1983). Whilst PRT acknowledges the collective social role played by unions, its analysis takes this role for granted. The focus is on union ability to resist capital, the assumption being that the willingness to resist is absolute. Whilst there is evidence that overall, labour is more successful at reducing inequality where it has more power resources (Pontusson, 2013), PRT fails to account for trends of dualism, where seemingly resource-rich union movements can become complicit in the polarisation of the labour market, leaving precarious workers increasingly unrepresented (Rueda, 2007).
This article examines social solidarity as a power resource. Whilst most of the IR literature accepts that unions require their role of social agents in order to survive (Bourdieu, 2003;Gumbrell-McCormick and Hyman, 2013;Hyman, 2005;Munck, 2004;Urban, 2012), union performance continues to be primarily measured through individual outcomes and rarely focuses on unions as social agents. Social solidarity in union ranks affects a union's ability to fight back against neoliberal pressures (Pontusson, 2013), and it is affected by individual characteristics such as earnings (Checchi et al., 2010). Using data from 18 countries in the 2016 European Social Survey (Centre for Comparative Social Surveys, 2016), we test the relationship between union membership and four dimensions of social solidarity (inequality, welfare state, diversity and immigration). We assess this impact at European level by grouping countries into five distinct IR regimes and investigating how different institutional frameworks impact the nexus between union membership and social solidarity.
The article begins by discussing the decline of union power resources and union density. The following section explores unions as agents of social change, reviewing the literature on union performance and noting its limitations in explaining unionisation trends. We distinguish between five European IR regimes characterised by shared institutions. Our empirical analysis will assess the impact of individual union membership on attitudes towards inequality, the welfare state, diversity and immigration. In Section 5, we outline our methodology, variables, model specification and limitations. We review our findings in Section 6 and discuss their implications. Finally, we conclude by summarising our findings and explore how they might relate to and contribute to our understanding of trade union analysis.

DECLINE IN POWER RESOURCES
Union objectives vary significantly across countries; however, the literature highlights four main forms of union power resources with which they can be achieved, namely, structural, associational, organisational and institutional (Gumbrell-McCormick and Hyman, 2013). Structural power resources originate from individual workers' specialised skillset or position in the production process (Silver, 2003). These have been considerably eroded outside the public sector. Fixed-term contracts, temporary agency workers and competitive pressures and structural changes following globalisation led to a standardisation of skillsets and work processes (Hyman, 1999). Associational and organisational resources are two sides of the same coin, both necessary to harness the power resulting from membership. Having a large member base is important, but membership alone is not sufficient without involvement and unity among union members. Organisational power is the ability to mobilise members to action, which in turn requires an internal 'social capital' (Nissen and Jarley, 2005).
Institutional power resources refer to union support from the state and legislative system. The literature is more divided on the merits of institutional power. On the one hand, state support is vital to defend social solidarity against neoliberal influence (Thelen, 2012), and institutional power may even substitute organisational power resources (Dörre et al., 2009;Frege and Kelly, 2003). At the same time, institutionalised security can create complacency and diminish union effectiveness (Hassel, 2007). Unions may face conflicts of interest in their strategy when choosing between defending institutional power or representing the labour force (Dufour and Hege, 2011). Strong labour movements supported by strong social democratic parties have actually increased this dualist inequality in France and Germany (Rueda, 2007), and outside of Europe, established unions in Japan have sometimes fostered anti-solidarity sentiments (Watanabe, 2018). In the long run, poor inclusivity and the resulting loss of representative legitimacy can also undermine institutional support for organised labour (Bispinck et al., 2010). Finally, even when institutions protecting trade unions remain strong, they may lose the ability or willingness to pursue their original goals (Streeck, 2009;Thelen, 2012), which means they cease to act as defenders of labour interests, contrary to what PRT would predict.
Union membership has decreased in nearly all affluent countries during the last few decades (Kjellberg, 2011;Schnabel, 2013). Sectorial shifts in employment moved workers from high to low union density sectors (Polachek, 2003). However, empirical evidence across Europe countries suggests that purely sectorial explanations are too narrow (Blanchflower and Bryson, 2009;Checchi and Corneo, 2000;Fitzenberger et al., 2011;Ours, 1992;Schnabel, 2013), and instead the key structural factor is workplace size (Ebbinghaus et al., 2011;Kirmanoğlu and Başlevent, 2011;Scheuer, 2011;Schnabel and Wagner, 2005). Intergenerational issues play a role as well, as unionisation rates in the under-25 group are universally lower in affluent countries, partly due to the difference in young people's employment patterns (Gomez et al., 2004) and to changes in ideologies and preferences (Bain and Elsheikh, 1976;Blanden and Machin, 2003;Ebbinghaus, 2002). Whilst it remains true that unionising the social media generation poses a unique set of challenges, the evolution of young people's employment conditions is driven by the same underlying trends eroding worker protections universally. Neoliberal policies and pushback from the state and employers has seriously compromised union ability to recruit and retain members, particularly in Anglophone countries (Smith and Morton, 2001;Standing, 1999). However, we should not overstate or generalise the impact of neoliberalism at European level. Evidence suggests there is no convergence towards union decline following the pattern of liberal market economies (Checchi and Lucifora, 2002;Schmitt and Mitukiewicz, 2012;Schnabel, 2013).

SOCIAL SOLIDARITY: A NECESSARY POWER RESOURCE
There has been considerable analysis of unions as vested interest organisations. Insider-outsider theory states they defend workers at the expense of the unemployed (Lindbeck and Snower, 1988), leading many to try and calculate a wage premium associated with union membership (Blakemore et al., 1986;Budd and Na, 2000;Hildreth, 2000) or measure the free-rider problem of individual benefits from collective bargaining (Booth and Bryan, 2004;Sobel, 1995). Evidence of a union wage premium for members vis-à-vis non-members is mixed at best, even non-wage benefits such as job security and influence on workplace decisions show no significant effects through individual membership and are instead mediated by union influence at workplace level (Furåker and Bengtsson, 2013). Furthermore, the channels through which these individual benefits are supposedly secured remain unclear. Affluent countries tend to legislate away closed shop agreements 2 and wage discrimination based on union membership. Unions cannot deliver higher wages to their members exclusively, nor can they coerce 'free riders' into paying membership fees. Analysing unions as agents of vested interest leaves significant questions unanswered and fails to explain why popular support for trade unions remains solid across Europe and is higher now than in the early 1980s (D'Art and Turner, 2008).
Some scholars do analyse unions as agents of social solidarity. Notably, Pontusson (2013) measures the macroeconomic impact of union density on earnings inequality and income redistribution. He highlights the central role that union member solidarity still plays in determining objectives and outcomes at national level. A key finding was that the impact of union density on redistribution and inequality has weakened, which is attributed to union members moving higher up the income distribution in recent decades. The implied causal link is that wealthier individuals have lower levels of social solidarity, a relationship further explored by Checchi et al. (2010) where a correlation is found between normative concerns about inequality and union membership and it is affected by earnings. However, whilst acknowledging the role of trade unions as agents of social solidarity, and even how this might appeal to prospective members, neither study allows unions an active role in fostering social solidarity internally. More importantly, neither examines whether the relationship between union membership and social solidarity is affected by institutional factors and individual characteristics.
One of the limitations of PRT is the underlying assumption that organised labour will always use its power resources to push back against capital. The ability to resist is the focus of the analysis, whilst the willingness to resist is assumed. This leaves PRT unable to fully explain trends of dualism and erosion (Rueda, 2007;Thelen, 2012), where resource-rich labour movements have sometimes legitimised, rather than fought against increasingly unequal outcomes and polarised labour markets, sometimes with the help of social democratic governments. Because we know that the willingness of union members to perform a social function can change (Pontusson, 2013), we argue that social solidarity is in itself a power resource. By relaxing the PRT framework to include solidarity as a determinant of the willingness of labour to resist capital, we can explain trends of dualism and erosion, and why seemingly 'strong' unions can revert to a role of agents of vested interest and cease to perform their social function. The relationship between union membership and social solidarity is not exogenously given. Instead, unions play a role in the creation of an internal narrative of solidarity, and power resources determine its effectiveness.

INDUSTRIAL RELATIONS REGIMES
Creating categories of IR regimes and 'styles of unionism' requires a degree of simplification and should complement, rather than substitute, individual country analysis. Nonetheless, it allows us to identify a middle ground between generalised European-level analysis, which ignores an extensive underlying institutional heterogeneity, and individual country analysis, which overlooks consistent exogenous patterns extending beyond the national framework. We draw from the classifications used by Ebbinghaus (1998), Furåker and Bengtsson (2013), Gumbrell-McCormick and Hyman (2013), van Ruysseveldt and Visser (1996) and Visser (2009) to define our five European IR regimes, namely, Nordic, Central, Southern, Anglophone and Eastern.
In Nordic countries, unions and employers are both strong and have relatively cooperative relations, resulting in low labour market conflict. Regulation is agreed upon through negotiations, so the state rarely needs to step in, allowing both sides a great deal of independence. Union density is high, further reinforced by the Ghent system of unemployment insurance (Gumbrell-McCormick and Hyman, 2013;Larsson et al., 2012). Unions draw significant institutional power resources from their close association with the welfare state (Korpi, 2006) and their historic connection with social democracy. Union membership is individually advantageous due to the Ghent system and does not necessarily imply active participation as it does in other groups (Muiiler-Jentsch, 1985). As a result, we expect Nordic countries to have high levels of social solidarity overall, but low impact of union membership on individual attitudes.
Central European countries are a particularly heterogeneous group, whilst still tied together by strong common traits. Relations between unions, the state and employer organisations are corporatist, but as in the Nordic model they enjoy significant independence and the state rarely needs to intervene directly. During the 19th and 20th century, competing political and religious identities were formed, and labour was organised in this context. This resulted in unions being divided along political and religious lines (Penninx and Roosblad, 2002). Unions enjoy solid and institutionalised relations with employers, and their role in shaping public policy is accepted. The welfare states in Central countries tend to be well developed but inegalitarian, whilst unionisation is generally low in density (except for Belgium) and high in bargaining coverage (Gumbrell-McCormick and Hyman, 2013). Within this group, countries have experienced dualisation through 'drift': a growing divide between highly protected groups of unionised workers and an increasingly vulnerable and unregulated group of precarious workers, despite resilient labour institutions (Thelen, 2012). We expect union members to be defensive of their institutional power resources, but be as strong on inequality, diversity and immigration.
The Southern regime shows the effect of polarised politics and a state-heavy approach. Relations between employers and unions are more hostile than in other groups, which led to increased state intervention. Consequently, employment conditions are subject to extensive state regulation rather than collective agreements. There is a wide framework of institutions to represent workers, but its impact is usually intangible and mediated through the government. IR are thus heavily politicised, whilst producing weaker outcomes. Union density is low, and the union movement is highly fragmented and divided by politics and religion. Countries in the southern group have higher national debt-to-GDP ratios than other groups, all above 100 per cent at the time of the survey (OECD, 2018). Due to the antagonistic, polarised nature of IR in this group, we expect a strong relationship between union membership and social solidarity, particularly on politically contentious issues.
Anglophone countries are liberal market economies, a small group characterised by British IR. The key features are a conflictual relationship between employers and labour, a non-interventionist state leading to a voluntarist style of IR and a general focus on market-based policy open to foreign trade and investment (Ebbinghaus, 1998). The role of the market in particular is perceived as vital, meaning that employers have access to far more institutional power resources than unions (Gumbrell-McCormick and Hyman, 2013). The doctrine of laissez faire allowed mobilisation without interference early on, until the anti-union offensive in the 1980s, where legislation became actively antagonistic towards organised labour (Howell and Givan, 2011). Union organisation is fragmented, characterised by low density and low collective coverage as a result of legal restrictions on worker organisation and bargaining, which happens predominantly at company level. This is the IR regime where neoliberal pressures have 'won' most of the major battles. We expect social solidarity, particularly on economic issues, to be very low. It follows that internal union identity should be in stark contrast to the neoliberal consensus and have a significant impact on solidarity, provided their power resources are not too depleted to create a collective sense of identity.
Finally, economies in the Eastern European model are influenced from a shared post-communist transition process. The strong hand of the state remains a central feature, with the government driving top-down reform in the transition towards a market economy, EU access and compliance (Kohl and Platzer, 2007). Small medium enterprises are widespread and exist as a union-free zone. Unionisation is low; collective bargaining is fragmented and mainly at company level; and coordination is very weak, yielding a low coverage. As in the Anglophone group, government policy is predominantly business friendly and free market oriented. Worker rights are often championed by populist right wing parties (Meardi, 2007) who tend to push against diversity and immigration. The exit-voice decision (Freeman and Medoff, 1984) is exercised here at national and individual level, through abstention and migration in response to poor worker representation. Many employer concessions have been obtained by labour shortages following 'exit' rather than worker pressure from collective bargaining (Meardi, 2007). As a result, we expect social solidarity to be low overall, with weak influence of union membership on individual attitudes.

DATA
We use survey data from 15,606 respondents in the European Social Survey, 2016 (Centre for Comparative Social Surveys, 2016) collected in 2016 and 2017. Our analysis focuses on members of the labour force, so our sample is limited to respondents aged 16-65 who are currently employed. Incomplete data and responses such as 'do not know' or 'do not wish to disclose' were removed. Not every country in the European Social Survey falls under our specified union regimes, so we limit the sample to 18 European countries. The Nordic group consists of Finland, Norway and Sweden and the Central group Austria, Belgium, Germany and the Netherlands. The Southern group includes France, Italy, Portugal and Spain. Anglophone countries are the UK and Ireland, whilst the Eastern group includes Czech Republic, Estonia, Hungary, Poland and Slovenia. Design and population size weights provided within the survey were applied (European Social Survey, 2016).
Our eight dependent variables are divided into four categories: attitudes towards inequality, the welfare state, diversity and immigration. These were chosen to reflect willingness to resist neoliberal pressure (inequality and welfare state) and willingness to build an inclusive labour movement with representative legitimacy (diversity and immigration). Diversity includes attitudes towards gay rights and ethnicity. Our dependent variables represent qualitative responses. Respondents are presented with a statement, for example, 'large differences in income are acceptable to reward talents and efforts', and they are asked to respond on a scale of 1 to 5 ('strongly agree', 'agree', 'neither agree nor disagree', 'disagree' and 'strongly disagree') or 0 to 10 (0 for a very negative view and 10 a very positive view). To analyse responses that signal social solidarity, we dichotomise these discrete variables into binary dummy variables, assigning a value of 1 to responses that signal social solidarity and 0 otherwise. Table 1 summarises our dependent variables, the questions they are based on and the answers which signal a social solidarity response.
Our analysis focuses on the independent variable of union membership, because we wish to test whether being trade union members makes respondents more likely to express social solidarity. Respondents can be union members, former union members or non-members. We control for former members and use non-membership as our reference group. Our models include both macro and individual level controls. We control for national unemployment at macro level, whilst at individual level, as well as past union membership, we control for sex, age, income decile, industry, workplace size, native status, working hours and contract type. Our analysis is based on three regression models. The base model is a logistic regression of all 18 countries, controlling for IR regimes through dummy variables in addition to our previously listed control variables. Our second model is a multilevel mixed-effects logistic regression with random slope coefficients (Gelman and Hill, 2006;Hox et al., 2017). Individual observations are nested within countries, and the relationship between our dependent variables and union membership is allowed to vary across countries. 3 The first two models measure overall effects at European level. The multilevel model is more accurate 4 ; however, the logistic model shows whether our results are sensitive to changes in specification. If our results are of similar magnitude and significance, the correlation between union membership and social solidarity is robust to changes in specification, suggesting a consistently strong underlying relationship. Our third model replicates the initial logistic regression for each IR regime, allowing us to observe whether the effect of union membership on social solidarity varies across groups, and in what way.
Some limitations in our dataset and methodology must be taken into account before we can draw conclusions from our results. Whilst our overall sample for our European level analysis is of considerable size, with data from 15,608 individual respondents, when we replicate the logistic regressions in each IR regime our sample sizes are greatly reduced, particularly in the case of the Anglophone model with only 1,890 observations. When interpreting lack of correlation between union membership and social solidarity, we bear in mind that the individual IR regime samples are much smaller than the generalised European one, which means we might miss some correlations that do exist nonetheless. Despite this, our results will confirm that some aspect of social solidarity has a significant relationship within each IR regime. Therefore, even as we acknowledge the limitations incurred by reduced sample sizes, IR regime regressions still highlight the relative strength of these correlations and how 3 A likelihood ratio (LR) test suggests this specification provides a better fit at 1 per cent significance level compared to keeping coefficients constant. 4 A LR test suggests the multilevel model provides a better fit at 1 per cent significance level compared to the corresponding non-hierarchical logistic model. Personal view on immigration they vary in different institutional frameworks. No survey questions address attitudes towards ethnic minorities directly. Instead, we used responses to statements about immigration, which makes the two difficult to separate. Our 'support entry of ethnic minorities' variable signals solidarity to both immigration and ethnic minority background. Because we already have variables to measure attitudes towards immigration, we chose to only count the strongest support responses ('allow many') as social solidarity towards ethnic minorities. This reduces false positives by excluding responses that primarily signal general support for immigration. Albeit slightly arbitrary in design, we argue that this variable is effective at signalling unambiguous solidarity towards ethnic minority groups, even if it may contain some false negatives by failing to capture the full extent of social solidarity on this issue.

Summary of responses in our key variables
We summarise the average responses for each group in Table 2. At the time of the survey, approximately 25 per cent of respondents were unionised. The Nordic group had the highest unionisation rate of 62 per cent and the Eastern group had the lowest at 10 per cent, consistent with our expectations. An overall 17 per cent of respondents claimed to have been union members in the past, with relatively low variation across groups. All dependent variables are designed to capture a positive response to social solidarity, so higher values represent higher solidarity. Nordic respondents showed strong solidarity in all dimensions, having the highest level of solidarity in nearly all fields. Even so, only 40 per cent of respondents disagreed that the welfare state 'places too much strain on the economy', and only 37 per cent disagreed that it 'makes people lazy'. The Central group is average for most responses, but scores highly in support for the welfare state. Southern countries also responded close to the average on most issues but had a stronger opposition to inequality. Nearly half respondents disagreed that meritocracy justifies inequality, and 71 per cent of respondents believed that 'for a fair society, differences in the standard of living must be small'. Anglophone countries scored highly on immigration. However, as we would expect, there is very low solidarity towards the welfare state and little opposition to inequality, with only a quarter of respondents disagreeing that meritocracy justifies inequality, that the welfare state is unaffordable and that it makes people 'lazy'. The Eastern group scores significantly lower than average in nearly all aspects of solidarity, particularly on diversity issues, whilst responding within the European average towards the welfare state.

Logistic regression analysis: Union members
We summarise the effect of union membership in Table 3, displayed as odds ratios. From a sensitivity analysis perspective, the coefficients appear highly robust to changes in specification. The standard logistic regressions (Logit) and the multilevel mixed-effect logistic models (Multilevel) yield consistent results in sign, magnitude and statistical significance for nearly all variables. Overall, there is a highly significant correlation between union membership and attitudes towards inequality and immigration, and a significant link to attitudes towards the welfare state and diversity. The nexus between social solidarity and union membership is strong overall, but these  seemingly consistent relationships conceal a high degree of heterogeneity across IR regimes.
In the Nordic group, union membership is linked with opposition to inequality and lower solidarity towards immigrants. Whilst the overall level of support for immigration is higher than other groups, union members are only 82 per cent as likely as non-members to have a positive view of immigration. No other aspects of social solidarity appear to be influenced by union membership. The lower propensity for solidarity towards immigrants, albeit only significant at the 10 per cent level, may signal a fear of dilution of worker rights. This issue is particularly relevant in the Nordic group, where the relative strength of social democracy is highest. Evidence from the 1994 Norwegian referendum suggests that concerns over worker rights created hostility towards increased EU integration at the grassroots level, despite strong support from union leadership (Geyer, 1997). The impact on attitudes towards inequality is worth mentioning, as it contradicts the conventional view that high union density and reliance on institutional power resources weakened the collective identity within the union movement in this regime.
Union members in our Central sample value the welfare state. They are significantly more likely to disagree that social benefits and services 'make people lazy' and slightly more likely to disagree that they 'place too great a strain on the economy'. There is also a correlation between union membership and opposition to inequality; however, it is weaker than in the Nordic model, and it does not extend to more equal societal outcomes. There is a weak effect on attitudes towards immigration, and no effect on views towards diversity. The overall support for these issues in the Central group is within average, so there would be opportunities for unions to create an alternative narrative, but there is no evidence in our sample that this has occurred. Our Central group responses suggest a weakened relationship between union membership and social solidarity consistent with a trend of dualism. Union members are significantly more likely to defend institutional power resources such as the welfare state, but they are not more inclusive, nor do they feel more strongly about more equal societal outcomes than non-members.
In our Southern sample, our results suggest a strong relationship between union membership and solidarity in all dimensions. Opposition to inequality is high overall, and union members are significantly more likely to disagree that rewarding 'effort and talent' justifies 'large differences in income'. Immigration and diversity attitudes both show a highly significant correlation with union membership. In this heavily politicised arena of IR, union membership comes with significant ideological connotations compared to other IR regimes. Unionised respondents in the Southern regime are more willing to be inclusive and oppose the interests of capital, but the labour movements here have not delivered the outcomes we see in the Nordic regime. It follows that internal union narrative is necessary, but not sufficient. It needs to be complemented by a growing popular consensus outside the labour movement, as well as a favourable institutional framework. Union members in our sample are more likely than non-members to reject the idea that the welfare state 'makes people lazy', however not that it 'places too great a strain on the economy'. Only 32 per cent of respondents from the Southern group reject the idea that the welfare state is unaffordable, and union membership has no impact on this view. Whilst it may appear that respondents from this group have more objective reasons to question the affordability of social benefits than those in other groups, we argue that this represents an ideological victory of neoliberal narrative rather than a factual review of national finances.
Economic issues faced by Southern countries are not direct or inevitable outcomes of welfare provision, because they do not affect countries in the Nordic and Central groups with more generous welfare states. Economic competitiveness need not dictate low social standards, as shown in Sweden and the Netherlands (Hyman, 2005). Instead, economic downturns have been politicised to strengthen neoliberal pressure on the welfare state (Streeck, 2004). Much like appeals to competition and globalisation have been used to push for deregulation in the past (Hirst et al., 2015); low growth and high deficits have been used to normalise the idea that social benefits 'place a strain on the economy'. It is therefore noteworthy that even in this deeply polarised IR regime, where union membership carries significant ideological weight, unions were unable to create an opposing narrative on this issue, even internally.
Unionised respondents from the Anglophone group are not more likely to oppose inequality nor defend the welfare state, and the level of solidarity on these issues is very low, both overall and compared to the European average. Union membership also has no effect on attitudes towards diversity, but there is high overall solidarity on these issues to begin with. The only significant effect observed in our sample is towards immigration. It is evident from this group in particular that the relationship between union membership and social solidarity is not constant across IR regimes. We can see further reasons why there has not been a European convergence towards liberal market economy (LME) style union decline. Whilst inclusive on issues of immigration, unions in this IR regime have been unable to foster solidarity to fight the neoliberal consensus. As a result, it is the only regime where union membership has no correlation with any economic aspect of social solidarity, accompanied by a low popular support for these issues at national level.
Respondents in the Eastern group show low levels of social solidarity across nearly all dimensions we examined. Union members are more likely to favour more equal outcomes, but no other significant effects from union membership are observed. Social solidarity is low at national level, and trade union membership seems to have no effect on it. These findings are consistent with the combined effects of a reactionary socio-political transition following institutional collapse and high levels of worker migration. The exit-voice decision (Freeman and Medoff, 1984) happens at both individual and national level here (Meardi, 2007), with many workers choosing to emigrate rather than engage in IR. It therefore makes sense for union members who remain to feel strongly about more equal outcomes, as our sample suggests, however union impact is limited, and the influence of populist nationalist movements reduces the willingness of organised labour to promote solidarity towards diversity and immigration.

CONCLUSION
This article has analysed the role unions play in fostering social solidarity among their members, comparing the impact of union membership across five European IR regimes. We have found that the nexus between union membership and social solidarity is highly significant at European level but heterogeneous across IR regimes. Not a single aspect of social solidarity (inequality, welfare state, diversity and immigration) has a consistent relationship with union membership in every European IR regime.
Nordic countries display high levels of social solidarity in nearly all dimensions, and union members are still more opposed to inequality than non-members, showing a persistent internal ideology despite high union density and reliance on institutional power resources. In the Central group, defence of institutional power resources stands out. Union members are more likely to defend the welfare state but are not more likely to show solidarity on issues of diversity and immigration, and the effect on inequality is weaker than in Nordic countries. This is consistent with ongoing trends of labour market dualism. In Southern countries, union membership has the strongest overall effect on solidarity. As a result of highly politicised IR, union members are significantly more likely to have solidarity responses towards inequality, the welfare state, diversity and immigration. Interestingly, however, union membership does not influence individual opinions about the affordability of the welfare state. In line with qualitative literature, we find that even in highly politicised and strongly ideological IR regimes, support for social benefits is vulnerable to downward neoliberal pressures. Our findings for the Anglophone regime reflect a deeper impact of neoliberal ideology on economic solidarity than originally expected. The ability of unions to create an opposing narrative has been severely undermined, to the point that union membership has no significant effect on views towards inequality or the welfare state. Whilst it is unsurprising that overall popular support for these issues is low, the inability of unions to create a significant opposition, even internally, suggests unions in the Anglophone system face even greater, challenges than much of the literature suggests. Finally, organised labour in the Eastern group bears the scars of the transition to market economies and the influence of nationalist populism on organised labour. Although union members are more likely to prefer more equal outcomes, there is no effect on any other aspect of solidarity, particularly on diversity and immigration, where solidarity is not actively pursued to begin with.
Literature on trade union performance, both in labour economics and IR, offers insight on patterns of union membership decline as a result of poor union performance in their role as organisations of vested interests. However, they fail to account for the second face of unions, as a 'sword of justice' (Flanders, 1970), which played a central role in early union growth and development, and continue to be the basis for their representative legitimacy and institutional support. Ignoring the role of unions as social agents also fails to address why popular support for unions remains high and has increased in recent decades (D'Art and Turner, 2008). PRT, in turn, recognises unions as agents of social justice, but it does not offer a satisfactory explanation for trends of labour market dualism. Evidence still suggests that overall, unions with more power resources and more allies tend to be more effective at fighting inequality than those with fewer. However, their willingness to fight for these outcomes depends on the stock of social solidarity among their members (Pontusson, 2013). Previous empirical findings show that normative views towards inequality can be influenced by personal earnings (Checchi et al., 2010). Building on this previous body of research, we have analysed normative views of social solidarity on a broader range of socio-economic issues. We have provided evidence that the stock of solidarity resources varies substantially across regimes and that the relationship between union membership and social solidarity is heterogeneous and affected by institutional factors as much as by individual characteristics.
The ability of unions to shape an internal narrative and foster social solidarity in each regime paints a picture of the institutional arrangements unique to that style of unionism which is generally consistent with the qualitative literature. A few stylised findings stand out. First, despite reliance on institutional power resources, Nordic unions retain an internal ideology against inequality. The sense of identity resulting from union membership might be less strong than in the Southern group, but it remains significant. Second, even a strong internal sense of identity rooted in politics and ideology as observed in the Southern regime does not prevent the gradual rationalisation of lowering social protections on economic grounds. This suggests that even ardent opposition to capital does not exempt the welfare state from the pressure exerted by neoliberal ideology, particularly under unfavourable economic circumstances. Our findings are consistent with the view that European countries are not converging towards a trend of union decline as liberal market economies. We acknowledge the nearly ubiquitous trend of decline in union density across Europe, as well as the severe limitations of using cross-sectional data to explain trends. However, the evidence presented here suggests that union membership in Anglophone countries has significantly different solidarity implications from the rest of Europe. Whilst not sufficient to reject a convergence hypothesis on their own, our findings further support this existing view in the literature (Checchi and Lucifora, 2002;Schmitt and Mitukiewicz, 2012;Schnabel, 2013). Finally, further analysis is required on the effect of social solidarity on union outcomes. Whilst our results suggest that a lack of solidarity is a significant impediment to an organised labour movement (e.g. Anglophone and Eastern regimes), there are cases where seemingly strong internal union narratives do not result in equally favourable outcomes for labour (e.g. Southern regime).
Trade unions face a diverse and growing set of challenges in today's increasingly fluid and globalised labour markets. They may be less equipped with traditional power resources than they were in previous decades; however, the narrative of union decline is not all-encompassing. Overall unions still foster a culture of social solidarity within their ranks. In most European countries, being a trade union member still carries some significance.

CONFLICT OF INTEREST STATEMENT
The author has no conflicts of interest to declare.