Selective Solidarity?
Explaining the variation in unionization rates
between migrant and native workers in Europe
By: Maria Kranendonk
Research Master Social Sciences
Supervisor: Prof. Dr. Paul de Beer
Second reader: Prof. Dr. Herman van de Werfhorst
Date of submission: Amsterdam, 04-08-2014
Abstract. This article explores the difference between migrant and native workers in their unionization rates in 23 European countries. It explores whether individual characteristics or contextual factors explain the variation across countries in the degree of trade unions’ inclusion of migrant workers. This study finds that individual characteristics cannot fully explain the variation of the difference between migrant and native workers in unionization rates. Characteristics of the country of destination affect trade unions’ inclusion of migrant workers and the institutional embeddedness of trade unions is found to be of particular importance.
Introduction
‘The Union Makes Us Strong’
The statement above, belonging to the English Trade Union Congress, is just one of the many that trade unions utilize to express the solidarity between workers and to express how trade unions can improve the position of workers.1 Trade unions should attempt to promote the welfare of all workers according to their social democratic role (Harcourt et al., 2008: 99). However, how inclusive are trade unions actually when it comes to ‘newcomers’ to the labour market? The feminization, flexibilisation and diversification of the workforce have resulted in declining trade union membership rates in most European countries. The declining rates illustrate how trade unions face difficulties in recruiting women and migrants, among other social groups (Ebbinghaus, 2002: 465). By exploring the relationship between trade unions and migrant workers this study questions whether trade unions are inclusive of all workers or whether trade unions express solidarity for a select group of workers, belonging to the ‘traditional’ workforce.
During the last decades, scholars of migration and integration as well as scholars of industrial relations have started to acknowledge the relationship between trade unions and migrant workers as an important topic. Studies indicate that the inclusion of migrants in trade unions is beneficial for both parties regarding several aspects, however, trade union membership rates are assumed to be lower among migrant workers compared to native workers. Penninx and Roosblad argue that the membership rate of migrants is an important indicator for the formal openness of trade unions towards migrants (2000: 193). According to several scholars, the relationship between trade unions and migrants cannot be explained by individual characteristics, such as the over representation of migrants in occupations that are less organised. They claim that contextual characteristics affect how inclusive trade unions are towards migrant workers (Some examples: Gorodzeisky & Richards, 2013; Eldring et al., 2012; Marino, 2012; Krings, 2009; Lucio & Perrett, 2009; Wrench, 2004; Cachón & Valles, 2003; Penninx & Roosblad, 2000). This study connects to the latter by exploring to what degree individual characteristics, and contextual factors explain the difference between migrants and natives in their unionization rates in European countries.
Scholars have applied mostly qualitative methods when researching contextual characteristics and their presumed association with the inclusion of migrants in trade unions, with the exception of the recent study of Gorodzeisky and Richards (2013) (Some examples: Eldring et al., 2012; Marino, 2012; Krings, 2009; Lucio & Perrett, 2009; Wrench, 2004; Cachón & Valles, 2003; Penninx & Roosblad, 2000). This study contributes to the debate by applying a quantitative research design. It considers more European countries in a longer time-frame in multi-level logistic regression analyses using the six rounds of the European Social Survey (2002-2012), which adds to the certitude about the existence and strength of contextual impact. Moreover, it considers several - quantitatively not yet tested – contextual characteristics that impact the likelihood of unionization, which, according to existing scholarly research is different for migrant workers compared to native workers. Lastly, this study considers differences between migrant groups, such as variation in the experience with trade unions in the country of origin, and their likelihood to unionize.
Besides academic relevance, studying the relationship between trade unions and migrant workers has also important societal relevance. It is essential that every individual has equal access to important institutions from the perspective of democracy, and it is crucial to study the barriers towards
1
http://www.unionhistory.info/
2
equal access. Furthermore, the legitimacy of trade unions as labour institutions also depends on their ability to represent all workers.
This study explores whether contextual characteristics affect trade unions’ formal openness towards migrants. Firstly, characteristics of the migrants themselves affect their likelihood of being a member of a trade union. Secondly, the characteristics of the country of origin affect the relationship between trade unions and migrant workers through providing migrant workers with specific institutional experiences. Lastly, institutional, social and political circumstances in the migrant’s country of destination affect this relationship (Penninx & Roosblad, 2000: 8-11). This article explores whether the cross-national variation of the difference between migrant and native workers in the likelihood to unionize is explained by compositional factors, thus by individual characteristics or migrants’ experience with trade unions in the country of origin, or caused by contextual characteristics of the country of destination which affect migrants differently compared to natives.
To what degree do compositional factors explain the cross-national variation in the difference between migrant and native workers in unionization rates, and to what degree do contextual characteristics of the country of destination cause this variation?
Theoretical framework
Migrants and Trade Unions
According to scholars of migration and integration and scholars of industrial relations, the inclusion of migrant workers in trade unions leads to benefits for both parties. Scholars of migration and integration, such as Penninx and Roosblad, claim that trade unions, as stakeholders in the socio-economic integration process of migrants, can promote the social, political and socio-economic societal inclusion of migrant workers (2000: 3). In addition, amongst other factors, the composition of union members, according to their employed sector or nationality, affects trade unions’ policy decisions (EC, 2010: 18). Therefore, inclusion of migrants in trade unions is needed in order for migrants’ needs to be represented in collective bargaining and trade unions’ policy decisions. Scholars of industrial relations claim that migrant workers are a substantial ‘pool’ of potentially new members that could be mobilized in order to revitalize the trade unions’ power (Marino, 2012: 5). In addition, employers can resort to migrant workers who are not unionized for competitive strategies (Hardy et al., 2012: 356, 359). Thus, organizing employees is important for trade unions as it permits them to control the labour supply and increase their leverage in negotiations (Avci & McDonald, 2000: 191, 192, 198).
Though scholars indicate benefits for both trade unions and migrant workers, the relationship between trade unions and migrant workers has regularly been troubled. Trade unions considered the impact of migrant workers on labour relations and questioned whether labour migration would be detrimental to the local workforce. Migrant workers felt that their demands were hardly represented. In addition, trade unions have had difficulties with creating unity among employees, especially among those divided by language and culture (Cachón & Valles, 2003: 473: Krings, 2009: 50). Therefore, the relationship between trade unions and migrant workers is less straightforward than it might seem at first glance.
Scholars whose research surrounds this topic attribute different degrees of autonomy to the ability of trade unions to make their own decisions with regard to migrant workers. Some scholars highlight the autonomy of trade unions, while others underline the importance of the institutional embeddedness of trade unions (Penninx & Roosblad, 2000: 202, 203; Marino, 2012: 6). The former group of scholars highlight the importance of different responses to the problems faced by trade 3
unions. Important aspects by which actors can affect the functioning of the trade union are in the internal restructuring, in coalition building with other social movements, in starting partnerships with employers, in organizing political action and lastly by supporting international links. These efforts can boost membership rates amongst other factors. In addition, these scholars contribute more importance to internal union characteristics and dynamics, such as union structure and communication processes, thus they bestow more agency upon involved actors (Frege & Kelly, 2003: 9; Marino, 2012: 6, 15, 16). Though the individual characteristics of migrants contribute to the degree of organization among migrants, Penninx and Roosblad claim that structural and institutional factors concerning the functioning of trade unions are more important in explaining cross-national differences regarding the union mobilization of migrant workers (2000: 193, 200). Trade unions face three major dilemmas with regards to migrant workers. Firstly, trade unions decide whether to cooperate or resist migration policies. Secondly, they determine whether to include or exclude migrant workers and lastly, they decide whether to represent migrant specific interests or to pursue general policies for all workers (De Jongh, et al. 1984: 218; Roosblad, 2013: 35).
This study connects to this group of scholars by exploring the contextual impact on the unionization rates among migrants and focuses mainly on the second dilemma, namely whether trade unions include migrant workers. However, it touches upon the third dilemma by exploring how national contexts affect trade unions’ ability to represent migrant specific needs. However, first the individual characteristics and contextual factors, which are expected to affect unionization rates in general, are discussed before turning to the contextual factors that are expected to affect the unionization rates among migrants differently compared to natives.
Determinants for Trade Unionization
There are several theoretical approaches which can be used to explain individual decisions to join a trade union. The rational choice approach assumes that individuals are utility-maximizers, thus it assumes that individuals join trade unions when expected advantages exceed expected disadvantages. From this perspective, collective bargaining provides trade unions with the free-rider problem, as non-union members also benefit from the activities of trade non-unions. Negative incentives, such as obliging trade union membership, and positive incentives, such as providing services to union members specifically, can counteract this free-rider problem. The sociological approach attributes explanatory power to Weber’s four categories of social action, claiming that next to rational considerations, values, norms, emotions and traditions also affect the individuals’ decision to become a member of a trade union (Ebbinghaus, 2011: 108, 109). Lastly, the ‘social custom’ theory explains that the choice of becoming a member of a trade union is dependent upon whether others join trade unions. Since joining a trade union would deliver the individual with a positive reputation in regard of the opinions of their peers and family (Schnabel & Wagner, 2005: 4, 5).
An example of how individual characteristics affect unionization rates among social groups is how lower and higher income groups are less likely to join trade unions. Lower income groups are likely to be disappointed by egalitarian policies, while higher income groups are expected to be deterred by these redistribution aspects of trade unions’ policies. In addition, lower income groups are generally less inclined to participate in organizations that involve costs (Checchi et al., 2010: 85; Schnabel, 2003: 31). Another example is how attitudes towards inequality are likely to affect an individual’s choice to unionize, since trade unions ‘expound a philosophy of equality and advertise their actions as contributing to more fairness in the treatment of workers and more equality in opportunities and rewards’ (Checchi et al., 2010: 85). It is therefore expected that employees who are more supportive of egalitarian policies are more likely to become a member of a trade union.
The examples of income and of attitudes towards equality demonstrate how some social groups are more likely to unionize compared to others. The following section argues that there are several migrant specific characteristics that cause variation with regard to several indicators of integration, amongst other factors the likelihood to unionize.
Pathways of Integration
In the research of integration there are two important strands of literature. Some scholars explain the variation between migrants and natives on several integration indicators as resulting from differences on the individual level, while others acknowledge the importance of the context of reception in explaining positions of migrant groups. According to the former, the difference between migrants’ and natives’ position in society can be explained by the migrants’ social status and labour market position, amongst other factors. They assume that the societal position of migrants changes once migrants adapt to the countries of destinations’ norms and values and if they participate actively in society. The integration of migrants is hence expected to be an individual level process by which migrants adapt to the country of destination (Soysal, 1994: 30).
One example, and related to the topic of trade unions, is the research to the labour market position of migrant workers. Migrant workers often have smaller incomes and lower occupational statuses compared to their native counterparts (Kogan, 2006: 698). Several scholars of integration attribute the divergence between native and migrant workers in incomes and occupational statuses to individual characteristics, such as human capital. For example, skills obtained in the country of origin are frequently not easily transferred to the country of destination. Nevertheless, the labour market position of migrant workers is expected to improve by the increasing length of stay, by acquiring proficiency in the country of destination’s language, by acquiring the country of destination’s credentials and by gaining labour market experience (Chiswick et al., 2011: 3; Chiswick & Miller, 2009: 162). The theorized improvement of the labour market position of migrants could also affect the difference between migrant and native workers in unionization rates.
Some scholars argue that individual characteristics are not sufficient to explain migrants’ positions and participation in society, and more importantly, they are not sufficient in explaining its cross-national variation. The following section explains how context affects different indicators of integration among migrant groups and across countries.
Context of Reception
Scholars on migrants’ integration, such as Soysal, and Portes and Böröcz among others, argue that it is not only the individual characteristics that cause different societal positions among migrants and natives. They argue that the macro-level process of migrant incorporation, meaning the process by which migrants become part of the country of destinations’ polity, goes beyond the individual level adaptation of migrants. Nation states invoke existing schemes of membership in the incorporation of migrants, thus, discourses, practices, institutions and organizations that persist over time cause cross-national differences in migrant incorporation regimes (Soysal, 1994: 30, 31, 36; Portes & Böröcz, 1989: 614, 615, 618).
Returning to the example of migrants’ labour market position, scholars, such as Kogan, Lin and Model, claim that the country of destination’s labour market characteristics affect the labour market incorporation of migrants. Unemployment, occupational and income inequality, and labour market regulations affect migrant workers differently in comparison to native workers. The theorized main causal mechanism explains how employers are more inclined to favour native workers when jobs are scarce or when it is costly to fire employees (Lin & Model, 2002: 1065; Kogan, 2006: 698-700).
The theory concerning skills and economic incorporation are of interest to this study, as it provides information about what characteristics of migrant workers are expected to affect the divergence between migrants and natives in the participation in organizations. The studies to the context of reception show how contextual characteristics affect migrants’ position and participation in society differently compared to natives. Trade unions themselves are affected by the national context as well, as is discussed in the next section.
Context and Trade Unions
Scholars such as Ebbinghaus et al. show that national context matters for the individual’s likelihood to unionize, and they demonstrate that it explains cross-national differences in union membership rates (2011: 122). The Scandinavian countries have relatively high trade union membership rates, while the Southern and Eastern European countries have comparatively low membership rates (Jensen, 2006: 67). The composition of the labour market can partly explain cross-national variation in trade union membership rates, nevertheless, context shows to affect this cross-national variation as well, as it discussed below.
Factors that have transformed unionization rates are the feminization of the labour market, because not in all countries are women as likely to unionize as men. Adding to this, the number of part-time jobs has increased and individuals with these jobs receive less benefit from being a member of a trade union compared to full-time workers (Visser, 2006: 47; EC, 2008: 33). In addition, the traditionally well-organised sectors, such as manufacturing, have shrunk, while the traditionally less organised sectors, such as the private sector, have expanded significantly (Gorodzeisky & Richards, 2013: 24). This expansion of the public sector led to an increase in smaller workplaces as well, leading to decreasing membership rates as union costs of organizing are relatively lower in larger establishments and employees may feel an increased need for representation due to the impersonal treatment in larger organizations (Schnabel, 2012: 11).
The institutional context affects levels of trade union membership in addition to the composition of the workforce. The more trade unions are recognized and the less they are opposed by employers, the more trade union membership will be favoured (Ebbinghaus et al., 2011: 101). Nevertheless, the institutionalized position of trade unions, such as the commitment to the socio-economic decision making, can affect trade unions’ recruitment strategies. Unions that are less institutionalized perceive membership as an important means by which they can exert influence on employers and the government. Union members are of less importance if trade unions are institutionalized and thus, for example, have the opportunity to influence policy decisions. In this case, the costs of creating new recruiting programmes to reach unrepresented workers can outweigh the benefits of gaining new members (Marino & Roosblad, 2008: 628, 629; Booth, 1986: 42).
The relevant institutional features cannot be summarized as the trade unions’ involvement in policymaking. For example, bargaining coverage is an important indicator for the different types of industrial relations. The effect of bargaining coverage on trade union membership is debated within the research field of industrial relations. Some scholars claim that bargaining coverage increases the incentive to unionize, while others argue the opposite. Moreover, the position of employers, several bargaining styles as well as the characteristics of the relationship between the state, trade unions and employers, affect the institutionalized position of trade unions (EC, 2008: 47-49; Scheuer, 2011: 68, 69). Lastly, the manner in which trade unions are involved with other institutions also affects unionization rates. For example, trade unions in Scandinavian countries and Belgium provide access to unemployment benefits, which has resulted in relatively high trade union membership rates (Jensen, 2006: 69).
Nonetheless, context affects the unionization rates of the ‘traditional’ workforce often differently compared to ‘newcomers’ to the labour market, such as migrants. The following section will explore this hypothesis further.
Hypotheses
As this study is solely able to assess indirect effects on the likelihood of being a member of a trade union, it cannot be concluded whether the effects are due to the attitudes of migrant workers or due to the attitudes of trade unions. Moreover, it cannot prove that the migrant workers’ decision to join a trade union is solely informed by rational considerations, such as taking the expected advantage versus the unexpected advantage into consideration, or by social custom (Checchi et al., 2010: 86, 87; Ebbinghaus et al., 2011: 109). Most of the following hypotheses are characterized by cross-national interaction effects, because the contextual characteristics are expected to impact on the likelihood to unionize differently for migrant workers compared to native workers.
Firstly, characteristics of migrants themselves affect 1) how trade unions perceive them and 2) how migrants themselves perceive trade unions (membership). The incentive for becoming a member of a trade union depends on whether the migrant is planning a temporary stay, whether he or she has any knowledge about the institutional features of the country of destination, and on their level of education (Roosblad, 2002: 8-11). Some cross-national differences can be explained by the unequal distribution of migrant workers with certain individual characteristics. Gorodzeisky and Richards argue that part of the lower membership rates of migrant workers can be explained by labour market segregation (2013: 242). This study takes the labour market position and relevant characteristics of workers into account in order to explore compositional effects.
H1: Compositional effects cannot fully explain cross-national variation in the difference between migrant and native workers in unionization rates.
In addition, the institutional features of the country of origin are expected to influence the difference between migrants and natives. As mentioned before, the country of origin provides migrants with certain institutional experiences. Thus, migrants from countries where trade unions are well established are more likely to join a trade union in the country of destination.
H2: More experience with trade unions in the country of origin decreases the difference between migrant and native workers in unionization rates.
Penninx and Roosblad draw some cautious conclusions about contextual characteristics that influence the attitudes of trade unions towards migrant workers on the basis of comparative case-studies in Western European countries. They claim that trade unions that are not strongly committed to the central socio-economic decision making tend to participate more as partners of migrants in social movements, and to oppose employers and the government. Lucio explains how migrant workers’ experiences of racism, social exclusion, economic disadvantage and political alienation produce specific needs (2009: 327,328). The less committed trade unions are to the socio-economic decision making, the more freedom trade unions have in adopting alternative attitudes towards social issues, such as issues concerning migrant workers. On the other hand, unions that are strongly committed to the central socio-economic decision making act as if their hands are tied and tend to keep to the core activities of the trade union, as they are obliged to adopt attitudes on all kind of social issues. (Penninx & Roosblad, 2000: 196, 202, 203).
However, trade unions could also make use of their institutionalized position in order to
defend and represent migrant workers’ rights, and consequently migrants could be more inclined to join the trade union. Therefore, it is expected that trade unions that actively seek to promote migrants’ rights are better able to recruit migrants. From the migrants’ perspective, the advantage of joining a trade union increases if a trade union is able to represent their more specific demands, which lies beyond the core tasks of the trade union.
H3: The difference between migrant and native workers in unionization rates is larger in countries where trade unions are more committed to the socio-economic decision making.
H4: The difference between migrant and native workers in unionization rates is smaller in countries where trade unions are more committed to the socio-economic decision making.
However, the commitment to socio-economic decision making does not fully capture the effect of industrial relations on the position of trade unions, as the section above demonstrates. Bargaining coverage, arrangements for consultation between the state, the employers and trade unions, trade unions’ involvement in providing unemployment benefits, and more factors decide the institutional embeddednness of trade unions. Trade unions that are less institutionally embedded perceive trade union membership as an important means by which they can exert influence on employers and the government. Union members are of less importance if trade unions already are institutionally secured (Marino & Roosblad, 2008: 628, 629; Booth, 1986: 42). Gorodzeisky and Richards indicate, for example, that the difference between migrant and native workers in unionization rates is bigger in Western European countries where trade unions experience a high degree of organizational security2 (2013: 251). Also Marino finds partial support for the hypothesis that trade unions that are more institutionally embedded perceive migrant workers’ unionization as less relevant in her comparison between Italy and the Netherlands (2012: 18).
Nevertheless, trade unions that are strongly institutionally embedded do have better means and more possibilities to recruit migrants and to represent migrants’ needs due to the presence of corporative bodies, which in turn can decrease the difference between migrants and natives in their membership rates (Marino & Roosblad, 2008: 636). This study considers both the negative and positive effect of institutional embeddedness on migrants’ union membership rates.
H5: The difference between migrant and native workers in unionization rates is larger in industrial relations regimes where trade unions are more institutionally embedded.
H6: The difference between migrant and native workers in unionization rates is smaller in industrial relations regimes where trade unions are more institutionally embedded.
Also, it is argued that economic and labour market conditions affect the formal inclusion of migrant workers in trade unions. Unemployment underlines competitiveness between native and migrant workers, thus affecting the trade unions’ ability to represent both (Penninx & Roosblad, 2000: 14). Observed from the perspective of migrant workers, it is expected that they are less inclined to join a trade union when they perceive competitiveness with native workers, since native workers comprise the majority of union members.
H7: The difference between migrant and native workers in unionization rates is larger in countries with higher unemployment rates.
2
Illustrated by state financing and centralized union confederation.
8
Lastly, societal attitudes towards migrants are expected to affect trade unions’ formal openness towards migrants. The general discourse concerning migrants, and national and institutional attitudes towards migrants, affect whether trade unions perceive migrant workers as valuable members in order to revitalize the unions’ organizational strength (Roosblad, 2002: 10). It also affects how migrant workers perceive the openness of institutions in the country of destination. Constant et al. claim that negative attitudes towards migrants provide barriers to the socio-economic integration of migrants (Constant et al., 2008: 3). Wrench concludes that the public discourse and attitudes are of particular importance in understanding trade unions’ attitudes towards migrants on the basis of his comparison between trade unions in the United Kingdom and Denmark (2004: 27). The assumption will be tested whether negative societal attitudes towards migrants affect the relationship between trade unions and migrant workers.
H8: The difference between migrant and native workers in unionization rates is larger in countries with negative societal attitudes towards migrants.
This eighth hypothesis loosely connects to the seventh hypothesis. Native workers may perceive migrant workers as more of a competitive threat when economic conditions deteriorate. These perceptions of threats can result in increased negative attitudes towards migrant workers (Gorodzeisky et al., 2006: 427). However, this study considers personal attitudes as well as unemployment rates since negative attitudes are not solely explained by a perceived economic competitiveness. Although contextual characteristics are important, the general contextual explanations only have a limited explanatory power as there is often an exception to the rule. Though membership rates are an important indicator, they do not guarantee equality, as migrants can - for example - be treated as second-rank members of trade unions (Penninx & Roosblad, 2000: 205). This study will have to answer for the assumed indirect causal link between the independent variables and the dependent variable. An example of one of the underlying assumptions is that all workers are aware of the tasks and position of trade unions in society.
Data, Operationalization and Methods
Data
The data is obtained from six rounds – 2002, 2004, 2006, 2008, 2010, 2012 – of the European Social Survey. This survey was established in order to monitor changing social and political attitudes and their interactions with institutions, and to improve cross-national survey measurement3. It is necessary to include multiple waves since the samples do not contain many respondents with a migrant background. The number of migrants is of special concern since this study is only interested in employees, by which it dismisses the self-employed and the unemployed. The data were analysed in order to clear outliers and impossible combinations, such as the age being smaller than the total number of years of education completed. Also, the following countries were excluded from the
3 www.europeansocialsurvey.org Effect of being a migrant worker, compared to being a native worker Likelihood to unionize Characteristics of the country of destination:
Trade unions’ institutional embeddedness, trade unions’ commitment to the socio-economic decision making, unemployment,
attitudes towards migrants
Mediating individual level effects: Migrant specific: experience with trade unions in country of origin, citizenship, years since arrival, knowledge about the
country of destination
Work related: age, education, gender, occupation, hours worked per week, type of
contract, income, establishment size, attitude towards equality.
10
analyses because of their specific institutional experiences, or because they were only present in less than three rounds: Croatia, Israel, the Russian confederation, Turkey, Iceland, Kosovo, Ukraine, Italy and Luxembourg. The dataset is not balanced, meaning that not all countries were observed in every round. The analyses include 6 rounds, 156,107 employees (N1) and 23 countries (N2)4.
Information about several contextual variables, such as unemployment and trade unions’ commitment to socio-economic decision making, is obtained from the OECD5 database and the database on Institutional Characteristics of Trade unions, Wage Setting, State Intervention and Social Pacts (ICTWSS) by Jelle Visser. The ICTWSS database covers 34 countries between 1960 and 2012 on issues concerning industrial relations and other aspects of the political economy.6
Operationalization
Dependent variableThe variable of interest is whether or not the respondent is a member of a trade union. The answer categories of no, yes previously and yes were recoded into a binary variable with 0 being no (no and yes previously) and 1 being yes7.
Independent variables Individual level
Individual characteristics that affect the likelihood to unionize are: gender, age and education. Work-specific characteristics included are: type of contract, establishment size, number of hours worked per week overtime included, occupation, income, perception of income and attitude towards equality. In order to avoid the unnecessary deleting of respondents who did not answer one item, or who recorded an unlikely answer such as ‘having completed more than 25 years of education’, these respondents will be modelled as ‘missing’.
Women and younger employees are a minority among trade union members, which is accounted for by controlling for age and gender (Visser, 2006: 47; EC, 2008: 33). Since the sample contains only employees, respondents older than 75 were excluded from the analyses.8 Education is measured as the total of years of education completed. Respondents who indicated to have completed more than 25 years are modelled separately as a dummy. A quadratic term is included for age and education to account for the non-linear effects on the likelihood of being a union member. The type of contract (unlimited, limited or no contract) and the size of the organization employed (number of respondents who are employed at the workplace) will be regarded. The ESS includes a measure for the total number of hours worked per week which includes overtime. Respondents who indicated to work more than 160 hours per week on average are deleted from the sample. Also, respondents who worked less than 8 hours or more than 60 hours a week are modelled as a dummy. Again, a quadratic term is included to account for a non-linear effect on union membership. Due to lack of information about the employed sectors in the six included waves, this study chooses to analyse the major groups according to the 1st digits of the ISCO88 for practical
4 See table 7 in appendix for the distribution of respondents over countries 5 The organization for Economic Co-operation and Development
6 http://www.uva-aias.net/207 7
See table 8 in appendix for membership rates among participating countries.
8 The average effective age of retirement, compared to the official retirement age, is higher in all European
countries
(http://www.oecd.org/els/emp/ageingandemploymentpolicies-statisticsonaverageeffectiveageofretirement.htm), these analyses take into account that some people might retire during a later stage in their life.
11
reasons, which categorize skill levels and skill specialization9.
Income is measured as household income based on deciles of the actual household income in the concerning country.10 Since household income generates many missing cases due to shortcomings of the questionnaire, an alternative indicator is chosen to measure the impact of income on the likelihood of being a member of a trade union in addition to household income (Hoffmeyer-Zlotnik & Warner, 2006: 298, 305). Namely the respondent is asked: ‘How you feel about your household's income nowadays?’ (Living comfortably, coping, difficult and very difficult). Lastly, the attitude towards equality is operationalized with the question of responding to the following statement: ‘Important that people are treated equally and have equal opportunities’ (scale of five answering categories ranging from ‘very much like me’ to ‘not like me at all’). Social desirability is expected to affect the responses and therefore ‘very much like me’ and ‘like me’ are joined together and the other answering categories are joined together. 11
Migrant specific variables
Migrant-specific characteristics important for membership of trade unions included in the analyses are: citizenship, country of birth, years that passed since migration, trade union density in the country of origin and whether the country of origin was previously a colony of the country of destination. Having citizenship may be an indicator for whether a migrant is planning a temporary or permanent stay. Nevertheless, in some countries it is easier to gain citizenship than in other countries and exclusionary measures with regard to migrants without citizenship vary between countries, which likely affects unionization rates of migrants. Unfortunately, it is not possible to compare migrant groups from all the specific countries of origin due to the small sample size of the migrant population. The migrants are grouped according to the regions of non-western countries, western countries and post-communist eastern European countries.
Whether the migrant respondents speak the mother tongue of the country of destination is also an important characteristic. However, this criterium is biased and underestimated since the European Social Survey is only provided in the language of the participating countries. Therefore, this variable is not included in the analyses. It can even be questioned whether the self-selection of migrants, of those who have sufficient language-proficiency to answer the survey, results in an underestimation of the difference between migrants and natives in unionization rates.
The amount of years since migration is categorized as less than 5 years ago, between 6 and 10 years ago and more than 10 years ago. It is expected that the language proficiency of migrants will improve over time, as well as their knowledge about the institutional features of the country of destination (Chiswick et al., 2011: 3). Experience with trade unions in the country of origin is operationalized as the average union density rate, net union membership as a proportion of wage and salary earners in employment, in the country of origin around the time of migration (Codebook ICTWSS, 2013: 23)12. Concerning the trade union density in the country of origin this study distinguishes between the following: ‘within the same year’, ‘1 to 5 years ago’, ‘6 to 10 years ago’, ‘11 to 15 years ago’, ‘16 to 20 years ago’ and ‘more than 20 years ago since migration’. For example, the average trade union density in the country of origin is calculated between 2011 and 2007 for a migrant who migrated 1 to 5 years ago. This indicator is not without its shortcomings, as trade union density rates are only available for a select amount of countries of origin, namely 49 countries which
9
www.ilo.org
10 www.europeansocialsurvey.org
11 See table 9 and table 10 in appendix for descriptive statistics.
12 If not available, the union density date, net union membership as a proportion wage and salary earners
in employment, as in national household or labour force surveys is used (Codebook ICTWSS 2013: 23).
12
include all OECD countries. The other countries will be recoded as ‘not applicable’ in order to avoid decreasing the sample size.13 Lastly, this study takes the migrants’ previous experience with the country of destination into account by including a variable that indicates whether the country of origin was a colony of the country of destination in the past. This study expects that migrants from former colonies have more knowledge about the institutional features, culture, and language compared to migrants from other countries (Roosblad, 2002: 8-11).
Contextual level Industrial relations
Countries are categorized according to characteristics such as union and employer organizations, levels and styles of bargaining, state intervention in employee-employer relations, the power balance and the social partners’ commitment to public policy making into the following types of regimes: organised corporatism, social partnership, polarised, liberal pluralism and fragmented14 (EC, 2008: 48, 49). The following paragraphs describe the key characteristics of the industrial relations regimes that are relevant for this study.
The regimes differ qualitatively along different dimensions and are not to be measured along one dimension, or with one statistic. Therefore, this study categorizes countries according to these regimes in order to explore the effect of trade unions’ institutional embeddedness on their attitude towards migrants. The role of social partners in public policy is institutionalized in organised corporatism and social partnership regimes. Social partners are irregular involved, and dependent on the ruling political parties for their involvement in polarised and fragmented industrial relations regimes, while they are rarely, and mostly event driven involved in countries that are characterized as liberal pluralism. Labour dominates the power balance between the state, trade unions and employers in social corporatism regimes, while it is dominated by employers in the liberal pluralism and fragmented regime. The power balance is balanced in the social partnership regime and alternating in the polarised regime. Collective bargaining is integrated in the social corporatism and the social partnership regimes, while it is conflict orientated in the polarised and liberal pluralism regimes. It is characterized by compliance in the fragmented regime, whereas workers’ rights appear to be inferior to liberal market policies. In general, the Eastern European countries (fragmented industrial relations regime) have weak tripartite institutions for the cooperation between the state, employees and employers (EC, 2008: 50; Martin & Cristescu-Martin, 2004: 635, 642). Lastly, trade unions are involved with providing unemployment benefits in Scandinavian countries (social corporatism regime) and in Belgium, thus trade union membership is favoured in these countries.
To summarize, trade unions’ position appears most favourable in countries that are characterized by a social corporatism regime, followed by the social partnership regime. Trade unions in the polarised and liberal pluralism regime enjoy a less favourable position compared to the before mentioned regimes, while trade unions in the fragmented regime enjoy the least institutionally embedded position.15
The commitment to socio-economic policymaking is operationalized as ‘involvement of unions and employers in government decisions on social and economic policy’, according to the ICTWSS database. The categories are: regular and frequent involvement, irregular and infrequent involvement, and lastly, involvement is absent and rare (Visser, 2013: 18).
13
See appendix table 11 for descriptive statistics of migrant specific variables and see table 12 for the individual level corresponding questions as stated in the European Social Survey.
14 See appendix table 13 for the list of corresponding countries.
15 Other characteristics of the different types of industrial relations and the corresponding categorization of
countries can be observed in table 13 in the appendix.
13
In addition, the trade union density rates, also obtained from the ICTWSS database, are included in the analyses. It is expected that trade union density rates influence individuals’ likelihood of being a union member, since individuals are more likely to join trade unions if it is common to do so. The industrial relations regimes coincide to some extent with trade union density rates. Nevertheless, informed by theory, it is expected that the industrial relations regimes have an effect, next to the effect of trade union density rates, on the difference between migrants and natives in their unionization rates. In order for the effects of the industrial relations regimes to emerge it is necessary to control for this particular characteristic which is likely to dominate variation between countries. In addition, it is expected that it is more likely that there is a larger difference between migrants and natives in countries where the trade union density rates are relatively high (compared to low). Therefore, a quadratic term is included to account for the non-linear moderating effect of trade union density rates for migrants.
Labour market
Since it is not possible to determine how long it takes unemployment rate changes to affect the differences in unionization rates, this study chooses to assess the impact of structural unemployment rates (NAIRU)16 obtained from the OECD database as an indicator for labour market circumstances. Structural unemployment rates illustrate the fundamental mismatch between demand and supply of labour. This type of unemployment is of a more structural nature and less easy to solve. The structural unemployment rates allow considering and accommodating the more robust cross-sectional differences in unemployment rates and its effects.17
Attitudes towards migrants
This dataset solely allows for consideration of the attitudes of the individual respondents in the sample. It is not within the scope of this study to include the political and trade unions' discourses concerning migrants. The attitude towards migrants, as derived from the sample of respondents who were born in the concerning country, will function as a proxy for the general attitude towards migrants and discourse involving migrants. Whether migrants are considered to make the country a better (indicated by the score of 10) or worse (indicated by the score of 0) place to live is aggregated to the national level.18
Methods
Analysing individuals within different countries requires specific methods that are designed to process different levels in data. The assumption that there is no correlation between the error terms is violated, since individuals are nested within countries. Additionally, establishing claims about cross-level interaction effects demands a multi-cross-level design. This problem is resolved by including an estimate of the random variance of the second level intercept (country level), thus accounting for the possibility of clustered observations for individuals within a similar country. This estimate accounts for unobserved heterogeneity between countries by allowing individuals within the same country to be more similar regarding their likelihood of being a union member (Rabe‐Hesketh & Skrondal, 2005: 116).
16 The NAIRU rates were not available for Cyprus and Bulgaria. The average unemployment rates from 2002 to
2012 were used.
17 See appendix table 14 for the NAIRU unemployment rates per country year combination. 18
See appendix table 15 for the aggregate attitudes towards migrants per country.
14
Secondly, the binary dependent variable violates the assumption of normality and the assumption of homoscedastic errors. The residual distribution is binomial because the dependent variable is dichotomous. The generalized linear model resolves this by including a necessary transformation and including the choice of error distribution in the statistical model (Hox, 2010: 112, 113). The following analyses will be conducted with the logit link, which produces a linear model for the log of the odds and thus implies a multiplicative model for the odds themselves. With every unit increase in a x variable, the log odds are multiplied by the exponential function (approx. 2,72) of the coefficient β1 (2,72β1) (Rabe‐Hesketh & Skrondal, 2005: 102).
Firstly, models including the estimates of the effect of being a migrant worker and the impacts of the ESS rounds are analysed. The uj models the random variation of the second level’s intercept.
𝐿𝑜𝑔𝑖𝑡�𝑀𝑒𝑚𝑏𝑒𝑟𝑖𝑗� = 𝛼 + 𝛽1𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖𝑗+ 𝛽2𝐸𝑆𝑆𝑟𝑜𝑢𝑛𝑑𝑠 + 𝑢𝑗
Secondly, individual level variables are included in the models.
𝐿𝑜𝑔𝑖𝑡�𝑀𝑒𝑚𝑏𝑒𝑟𝑖𝑗� = 𝛼 + 𝛽1𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖𝑗+ 𝛽2𝐸𝑆𝑆𝑟𝑜𝑢𝑛𝑑𝑠 + 𝛽3𝐼𝑛𝑑. 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑗+ 𝑢𝑗
Thirdly, the contextual effects are included in the models.
𝐿𝑜𝑔𝑖𝑡�𝑀𝑒𝑚𝑏𝑒𝑟𝑖𝑗� = 𝛼 + 𝛽1𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖𝑗+ 𝛽2𝐸𝑆𝑆𝑟𝑜𝑢𝑛𝑑𝑠 + 𝛽3𝐼𝑛𝑑. 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑗+ 𝛽4𝐶𝑜𝑛𝑡𝑒𝑥𝑡𝑗+ 𝑢𝑗
Fourthly, the contextual effects and the cross-level interaction effects with the variable of being a migrant worker, compared to being a native worker, are included.
𝐿𝑜𝑔𝑖𝑡�𝑀𝑒𝑚𝑏𝑒𝑟𝑖𝑗� = 𝛼 + 𝛽1𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖𝑗+ 𝛽2𝐸𝑆𝑆𝑟𝑜𝑢𝑛𝑑𝑠 + 𝛽3𝐼𝑛𝑑. 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖𝑗+ 𝛽4𝐶𝑜𝑛𝑡𝑒𝑥𝑡𝑗+
𝛽5𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖𝑗∗ 𝐶𝑜𝑛𝑡𝑒𝑥𝑡𝑗+ 𝑢𝑗
The degrees of freedom on the contextual level are limited, since this study only includes 23 countries. Testing the contextual variables separately would produce biased results, as the contextual effects are theoretically expected to be interconnected. The contextual effects that are not significant will therefore be omitted from the analyses. The cross-level interaction terms are only included for the contextual effects that prove to significantly affect the likelihood of being a trade union member. The results are presented in log odds that illustrate the log odds of for example, of yes vs. no, of success vs. failure. Log odds smaller than 0 illustrate negative effects and log odds larger than 0 illustrate positive effects.
Results
This section describes the results from the multi-level logistic regression analyses. To start with, descriptive statistics are demonstrated, after which the multi-level logistic models are presented. Firstly, only the effect of being a migrant, compared to being a native worker, is estimated. Consequently, individual characteristics are included. Following this, the contextual effects and the cross-level interaction effects are included. Secondly, the migrant specific characteristics are included to explore whether the effect of being a migrant in the previous model is due to the fact of being a migrant ‘an sich’ or whether there is variation between migrant groups in their difference with regard to natives in their likelihood unionize. Again, a model is shown with solely the migrant specific characteristics, followed by a model that includes other individual characteristics and concluded with a
model that adds the contextual and cross-level interaction effects. Thirdly, the impact of the contextual effects is estimated separately for migrants.
The table below demonstrates the unionization rates of migrants, natives and the difference between them according to country. Table 1 demonstrates that the differences are greatest in countries that are characterized as having an organised corporatism regime, followed by countries that are considered as having social partnership, liberal and polarised regimes. Belgium, Slovenia and Portugal stand out, because the difference is smaller or even in favour of migrants in contrast to their regime counterparts. The countries that are labelled as having fragmented industrial relations regimes have the smallest differences between migrants and natives in unionization rates in comparison to the other countries.
Table 1: Trade union membership rates among native workers and migrant workers in percentages Industrial relations
regime Country Natives (%) Migrants (%)
Difference (Natives-Migrants) (%) Organised corporatism Denmark 71.0 61.9 -9.1 * Finland 63.2 56.7 -6.5 Norway 53.4 43.1 -10.4 * Sweden 64.9 58.6 -6.2 * Social partnership Austria 26.4 16.3 -10.1 * Belgium 42.8 42.0 -0.8 Germany 14.9 10.1 -4.8 * Netherlands 22.5 18.1 -4.5 * Slovenia 24.0 27.7 3.7 Switzerland 14.9 9.7 -5.3 * Liberal pluralism Cyprus 27.8 18.2 -9.6 * Ireland 22.2 15.0 -7.2 * United Kingdom 19.6 14.6 -5.0 * Polarised France 7.8 5.4 -2.4 Greece 11.0 5.9 -5.1 * Portugal 8.8 10.4 1.6 Spain 12.6 5.6 -7.0 * Fragmented Bulgaria 6.4 4.7 -1.8 Czech Republic 9.1 6.9 -2.2 Estonia 6.9 6.9 0.0 Hungary 11.0 6.8 -4.2 Poland 9.2 7.6 -1.6 Slovak Republic 10.0 10.1 0.1 Total 26.1 20.4 -5.6 *
*Significant difference in unionization rate between migrants and natives with a confidence interval of 95% is indicated with *
Source: European Social Survey
Multi-level analyses
Table 2: The difference between migrant and native workers in the likelihood to be a member of a trade union
Model 1 Model 2 Model 3 Model 4
Log odds St. Error Log odds St. Error Log odds St. Error Log odds St. Error
First level
Origin
Migrants (ref. Natives) -0.30 (0.04)*** -0.36 (0.03)*** -0.36 (0.03)*** -0.23 (0.17)
Contextual effects
Industrial relations regimes (ref. social
partnership, liberal & polarised)
Organised Corporatism 1.39 (0.22)*** 1.41 (0.22)*** Fragmented -0.27 (0.15)* -0.29 (0.15)*** Attitudes towards migrants 0.21 (0.03)*** 0.22 (0.03)*** Trade union density 0.07 (0.01)*** 0.07 (0.01)*** Trade union density2 -5.75e-04 (0.75e-04)*** -5.84e-04 (0.76e-04)***
Little involvement socio-economic decision
making (ref. No and regular involvement) -0.27 (0.03)*** -0.28 (0.03)***
Cross-level interaction effects Organised Corporatism*migrants -0.46 (0.13)*** Fragmented*migrants 0.31 (0.13)**
Attitudes*migrants -0.10 (0.06)*
Trade union density*migrants -0.01 (0.01) Trade union density2*migrants 2.03e-04 (1.05e-04 )*
Little involvement*migrants 0.08 (0.07)
Individual characteristics No Yes Yes Yes
Migrant specific individual characteristics No No No No
ESS rounds Yes Yes Yes Yes
Random effects second level
Variance intercept (estimate & std. error) 1.11 (0.33)** 1.18 (0.35)** 0.08 (0.03)** 0.08 (0.03)**
Prob> Chi2 0.000 0.000 0.000 0.000 Log likelihood -72,052 -64,973 -64,876 -64,858 N first level 156,107 156,107 156,107 156,107 N second level 23 23 23 23 Df 8 50 56 62 *p<0.10 ** p< 0.05 ***p<0.01 Source: European Social Survey
Table 3: Variation between migrant groups and their likelihood to unionize
Model 5 Model 6
Log Odds Std. Error Log Odds Std. Error
First level
Origin
Migrants (ref. Natives) 0.07 (0.10) -0.03 (0.11)
Origin groups (ref. Western migrants )
Non-Western migrants -0.09 (0.07) 0.03 (0.08) Eastern European migrants -0.08 (0.07) 0.05 (0.08)
Colonies (ref. migrants from former colonies)
Not from former colony -0.12 (0.08) -0.13 (0.09)
Experience trade unions (ref. more than 50 percent)
No info available 0.08 (0.07) 0.04 (0.08) less than 10 percent -0.34 (0.18)* -0.44 (0.19)** 10 to 15 percent -0.31 (0.16)** -0.34 (0.16)** 15 to 20 percent -0.33 (0.12)*** -0.43 (0.13)*** 20 to 25 percent 0.17 (0.12) 0.06 (0.13) 25 to 30 percent -0.01 (0.13) -0.17 (0.14) 30 to 40 percent -0.04 (0.09) -0.07 (0.09) 40 to 50 percent -0.08 (0.10) -0.12 (0.11)
Years ago since migration (ref. more than 10 years ago)
To 5 years ago -0.89 (0.09)*** -0.80 (0.09)*** 5 to 10 years ago -0.36 (0.08)*** -0.45 (0.08)*** No citizenship (ref. having citizenship) -0.13 (0.05)** -0.12 (0.05)**
Individual characteristics No Yes
Migrant specific individual characteristics Yes Yes
ESS rounds Yes Yes
Random effects second level
Variance intercept (estimate & std. error) 1.11 (0.33)** 1.18 (0.35)** Prob> Chi2 0.000 0.000 Log likelihood -71,935 -64,873 N first level 156,107 156,107 N second level 23 23 Df 22 64 *p<0.10 ** p< 0.05 ***p<0.01 Source: European Social Survey
Table 4: The difference between migrant and native workers in the likelihood to be a member of a trade union, including migrant specific characteristics
Model 7 Model 8
Log odds St. Error Log odds St. Error
First level
Origin
Migrants (ref. Natives) -0.04 (0.11) -0.16 (0.20)
Contextual effects
Industrial relations regimes (ref. social partnership, liberal & polarised)
Organised Corporatism 1.38 (0.22)*** 1.41 (0.22)*** Fragmented -0.27 (0.15)* -0.29 (0.15)* Attitudes towards migrants 0.21 (0.03)*** 0.21 (0.03)*** Trade union density 0.07 (0.01)*** 0.07 (0.01)*** Trade union density2 -5.80e-04 (0.75e-04)*** -5.77e-04 (0.76e-04)***
Little involvement socio-economic decision
making (ref. No and regular involvement) -0.28 (0.03)*** -0.28 (0.03)***
Cross-level interaction effects
Organised Corporatism*migrants -0.40 (0.14)*** Fragmented*migrants 0.33 (0.14)** Attitudes*migrants -0.04 (0.06) Trade union density*migrants 0.01 (0.01) Trade union density2*migrants -0.01e-04 (1.10e-04)*** Little involvement*migrants -0.02 (0.07)
Individual characteristics Yes Yes
Migrant specific individual characteristics Yes Yes
ESS rounds Yes Yes
Random effects second level
Variance intercept (estimate & std. error) 0.08 (0.03)** 0.08 (0.03)** Prob> Chi2 0.000 0.000 Log likelihood -64,777 -64,766 N first level 156,107 156,107 N second level 23 23 Df 70 76 *p<0.10 ** p< 0.05 ***p<0.01 Source: European Social Survey
Model 1 demonstrates that compared to being a native worker, being a migrant worker decreases the relative odds of being a member of a trade union. Migrants are also less likely to be a member of a trade union when the individual characteristics are included that were described in the operationalization section. The difference between migrants and natives in unionization rates can thus not be explained by individual characteristics, for example by the over-representation of migrants in occupations or jobs that are less well organised.
The third model includes the significant contextual effects19. This model also shows that the difference between migrants and natives cannot be explained by the composition of the workforce in the participating countries. The main effect of being a migrant worker still significantly and negatively affects the likelihood to unionize. Therefore, the difference between migrants and natives in the likelihood to unionize cannot be considered the result of the settlement patterns of migrants, such as more migrants living in countries where unionization rates are low.
Model 4 includes the cross-level interaction effects in order to study whether context affects the likelihood of being a member of a trade union differently for migrants in comparison to natives. The main effect of being a migrant loses significance. It shows that the difference between migrants and natives in unionization rates larger is in the social corporatism regime, compared to the liberal, polarised and social partnership regime. This difference is smaller in countries characterized by a fragmented industrial relations regime, compared to a liberal, polarised and social partnership regime. Surprisingly, a more positive attitude towards migrants negatively moderates the likelihood of being a union member for migrants. Lastly, the interaction effect of the squared term of trade union density proves to be significant and positive.
To summarize, H1 is supported by the analyses. The difference in the likelihood to unionize cannot solely be explained by compositional factors, either by the individual characteristics of migrants or by their pattern of settlement. Whether migrant specific characteristics help to explain the difference between natives and migrants, H2, is tested in the following paragraph. This study found no support for H3 and H4, since the indicator for commitment to socio-economic policy making does not moderate the relative odds of being union members for migrants. The analyses provided partial support for H5, since the industrial relations regimes that significantly affect the difference between migrants and natives in the likelihood to unionize are the ones that are the most contrasting when it comes to the degree to which trade unions are institutionalized. Nevertheless, the analyses show no variation of the effect of the industrial relations regimes in between these two extremes. Still, the difference between migrants and natives in unionization rates is larger in the countries where trade unions enjoy the most favourable position (social corporatism regime) and is smaller in countries where trade unions are least institutionally embedded (fragmented regime), which provides partial support for H5 and provides no support for H6. H7 is refuted, as unemployment demonstrates not to affect individuals’ likelihood to unionize. H8 is refuted, as it was expected that more positive attitudes towards migrants would decrease the difference between migrants and natives in unionization rates. However, the analysis shows that this works the other way around, on which will be elaborated further in the discussion section.
19 NAIRU unemployment rates did not significantly affect the relative odds of being a union member and it was
therefore omitted from the analyses. Polarised and liberal regimes did not differ significantly from social partnership regimes in their effect on the individuals’ likelihood of being a union member, thus they were joined together as the reference category. Regular involvement in socio-economic policy making did not significantly differ from no involvement in policymaking with regard to its effect on the likelihood of being a member of a trade union, thus these were joined together in the reference category.
20
Migrant specific variables
Table 3 shows the effects of the migrant specific individual characteristics20. Model 5 demonstrates that the main effect of being a migrant worker, in comparison to being a native worker, is no longer significant. The table presents the variation between migrant groups and how it affects the likelihood to unionize. The shorter the time since migration, the lower the chance is of being a member of a trade union. Also, workers who migrated from countries with low trade union density rates (lower than 20 percent) are less likely to join trade unions compared to their counterparts who migrated from countries with high trade union density rates (more than 50 percent). Lastly, individuals without citizenship, compared to individuals with citizenship, are less likely to be a member of a trade union. Model 6 demonstrates that the inclusion of the other individual characteristics does not alter the effects of the migrant specific characteristics significantly in sign, size or significance. Thus, H2 is supported, as more experience with trade unions in the country of origin increases the likelihood of being a union member in the country of destination.
Model 7 shows that the size, sign and significance of the contextual effects are similar to model 3 without the migrant specific characteristics. Model 8 demonstrates, in correspondence with model 4, that the difference between migrants and natives in unionization rates larger is in the social corporatism regime and smaller in the fragmented regime, in comparison to the other industrial relations regimes. The migrants’ likelihood to unionize is not moderated by the attitudes towards migrants in this model. Lastly, the interaction effect with the squared term of trade union density is significant, though now negative. These analyses sustain the previous findings and provide support for the same hypotheses. To summarize, the contextual characteristics that moderate the migrants’ likelihood to unionize demonstrate that the difference between migrants and natives in unionization rates cannot be explained by the settlement patterns of migrants with certain migration backgrounds. For example, the cross-national variation in this difference cannot be explained by the settlement patterns of migrants who migrated less than 5 years ago, and their smaller likelihood to unionize compared to workers who migrated more years ago.
The table below demonstrates the difference between migrant and native workers, who share similar individual characteristics, in their predicted probability of being a member of a trade union per country.21 The migrant specific variables were set to the reference categories. The table shows a modest estimation of the difference between migrants and natives, since the predicted probabilities are calculated for migrants who migrated more than 10 years ago, and from countries with more than 50 percent trade union density rates, while fewer years since migration and less experience with trade unions decreases the likelihood to unionize significantly.
The table demonstrates that countries that are characterized as having a social corporatism regime demonstrate the largest differences between migrants and natives in their predicted probabilities to unionize. Norway shows the largest difference, whereas migrants are 4.6 percent less likely to unionize, given they share similar characteristics as natives such as occupation and gender, amongst other factors. Countries characterized as a social partnership, liberal pluralism or polarized regime show small differences between migrants and natives in their likelihood to unionize, with the exception of Belgium and Cyprus. These countries illustrate that migrants are around 1.7 percent more likely to be a member of a trade union compared to natives. According to these predictions, migrants
20 Native workers are among the reference category for all migrant specific categories, thus multicollinearity
could have been problematic. In order to test whether multicollinearity affected the estimates, the reference categories were interchanged with one of the other categories. This provided no problems and resulted in similar estimates.
21 The individual characteristics were set on the reference categories, or set to the mean in the case of continuous
variables for both migrant and native workers.
21
are somewhat more likely to be a member of a trade union, compared to natives, in countries that are categorized as having a fragmented industrial relations regime.
Table 5: Predicted probabilities of being a member of a trade union compared, in percentages. Industrial relations regime Country Difference predicted probabilities (migrants-natives) (%) Organised corporatism Denmark -1.95 Finland -1.32 Norway -4.62 Sweden -1.59 Social partnership Austria -0.13 Belgium 1.74 Germany -0.50 Netherlands -0.69 Slovenia -0.43 Switzerland -0.72 Liberal pluralism Cyprus 1.65 Ireland 0.19 United Kingdom -0.24 Polarised France -0.54 Greece 0.06 Portugal -0.22 Spain -0.67 Fragmented Bulgaria 0.68 Czech Republic 0.53 Estonia 0.17 Hungary 0.59 Poland 0.35 Slovak Republic 0.50
Source: European Social Survey
Analyses for migrants separately
Table 6 shows the analyses separately for migrants, as the previous models were foremost dominated by the large number of native workers. The migrant specific characteristics show the same sign, significance and size as in the previous models, with the exception of having migrated from a previous colony of the country of destination. In the models below, migrants from no former colonies are less likely to unionize compared to migrants from former colonies. Model 9 shows that solely trade union density rates in the country of destination, and its non-linear effect, affect the likelihood of being a union member concerning the contextual effects. The higher the trade union density rates, the more
likely migrants will unionize. When all the non-significant effects are omitted from the analyses in model 10, solely the main linear effect of trade union density rate remains significant.
Model 9 illustrates that none of the industrial relations regimes significantly affect the relative odds of being a union member for migrant workers. This is remarkable as the main effects of living in a country characterized as a fragmented or social corporatism regime were significant in the previous analyses, which were dominated by native workers. Thus, the positive effect of living in a social corporatism regime, and the negative effect of living in a fragmented industrial relations regime (compared to living in a social partnership, liberal or polarized regime) do not apply to migrants. Thus, migrants’ likelihood to unionize is not affected by the institutionalized position of trade unions in the concerning countries. Trade unions that enjoy more favourable conditions (social corporatism regime) are not better able to recruit migrants compared to trade unions with less favourable conditions, even though they showed to be better able to recruit workers in the analyses that were dominated by natives. The results below support the previous findings, namely those industrial relations regimes affect the likelihood of being a union member differently for migrant workers compared to native workers.
Table 6: Analyses for migrants separately
Model 9 Model 10
Log Odds Std. Error Log Odds Std. Error
Contextual effects
Industrial relations regimes (ref. social partnership regime)
Organised Corporatism 0.06 (0.23) Polarized 0.01 (0.22) Liberalism -0.18 (0.15) Fragmented 0.31 (0.22) NAIRU Unemployment -0.02 (0.02) Attitudes towards migrants 0.13 (0.09)
Involvement socio-economic policymaking (ref. None)
Little involvement -0.08 (0.17) Regular involvement 0.09 (0.16)
Trade union density 0.07 (0.01)*** 0.05 (0.00)*** Trade union density2 -2.85e-04 (1.56e-04)*
Random effects second level
Variance intercept (estimate & std. error) 0.01 (0.01)** 0.03 (0.01)**
Individual characteristics Yes Yes
Migrant specific individual characteristics Yes Yes
ESS rounds Yes Yes
Prob> Chi2 0.000 0.000 Log likelihood -4,905 -4,912 N first level 13,045 13,045 N second level 23 23 Df 73 64 *p<0.10 ** p< 0.05 ***p<0.01 Source: European Social Survey
