Individual-level determinants of current trade union membership and previous trade union membership in European countries

Abstract

This article aimed to explore the individual-level determinants of current and previous trade union memberships and to question if certain patterns are prevalent among European countries. Accordingly, repeated logit models across 15 European countries are applied drawing on the ninth round of the European Social Survey data. Variables are selected primarily considering the supply–demand theory, which explains the union membership based on the idea that employees are seeking to maximise the utility through cost and benefit analysis. In this context, this article offers a comprehensive analysis of the determinants of trade union membership with findings suggesting that personal, occupational and workplace characteristics affect current membership while occupational and workplace characteristics have effects with inverse directions on current membership and previous membership.

Introduction

De-unionisation is a salient phenomenon throughout the world, primarily depending on de-industrialisation, the increase in non-standard work and the growth of the informal economy, coupled with the changes in employment regulations and the limitation of trade union rights. Europe is not an exception to this phenomenon. Namely, in Western Europe, the union decline began in 1980 and continued after the Great Recession in 2008. When it comes to Southern Europe, the decline intensified after 2008, reversing the slightly positive trend initiated before 2008. Furthermore, the number of union members are declining in Eastern and Central Europe as well as in Southeast Europe and the Balkans (Visser, 2019, pp. 7–17). However, there is still hope for revitalisation. Namely, trade unions can still reverse the trend primarily by contriving to transcend their membership profile. Accordingly, this makes the need for a comprehensive understanding of determinants of trade union membership of vital importance. In this context, this article proposes to examine the individual-level determinants of current union membership and previous union membership in 15 European countries. Seeking to answer if personal, occupational and workplace characteristics are influential in the status of trade union membership based on the supply–demand theory; it explores the divergences and convergences among European countries. Furthermore, it questions if the meticulously selected variables within the scope of personal, occupational and workplace characteristics affect previous membership as much as current membership.

This article can be considered as contributing to the literature as it combines two scarce aspects within the scope of this field. First of all, most of the studies focus on single countries and interpret the determinants through country-specific information (such as political context, institutional arrangements, labour market, etc.) in addition to theoretical explanations. However, this article focuses on 15 European countries and seeks to take advantage of focusing on multiple countries by striving to reveal certain patterns among countries. Second, simply ignoring the probability of non-members being previous members at the same time, the majority of the studies consider the membership statuses limited to members or non-members. However, separating the non-members as never-members and previous members provides a more comprehensive insight into the issue. In this respect, this article not only endeavours to reveal the individual-level determinants of current membership and previous membership separately but also seeks to compare their effects on current and previous memberships and interpret the differences. For this aim, a sample comprised of 14702 current employees, acquired from the ninth round of the European Social Survey (ESS), is utilised.

The structure of this article is as follows: The second section starts with reviewing the literature regarding the theoretical explanations of membership along with similar studies focusing on individual-level determinants. Subsequently, the research questions are set forth. Next, the third section focuses on the methodology of the article. Then, the fourth section starts with descriptive statistics before putting forth the determinants of current and previous memberships along with the discussions indicating how the findings relate to the literature. Afterwards, the reasons for the differences in the effects of determinants on current membership and previous membership are explored. Finally, the fifth section sums up the article and identifies empirical contributions to knowledge.

Literature review

There are different points of view regarding the determinants of trade union membership belonging to a broad range of disciplines from economy to social psychology (Schnabel & Wagner, 2007, p. 7). Among all, economists beginning with Berkowitz (1954), Pencavel (1971) as well as Ashenfelter and Johnson (1972) discussed the factors behind joining within the scope of the supply–demand theory. This theory explains union membership based on the idea that the employees are seeking to maximise the utility depending on the cost and benefit analysis. Namely, it considers that the membership of an employee is a result of a process primarily seeking larger benefits and smaller costs (Ehrenberg & Smith, 2000, pp. 465–466; Hirsch & Addison, 1986, pp. 29–30; Schnabel & Wagner, 2007, p. 7).

Within the scope of the supply–demand theory, Ehrenberg and Smith (2000, pp. 465–466) considered the willingness of employees to be a member of a union as a demand curve, a downward function of the price, on the other hand, considered the willingness of trade unions to supply their services as a supply curve, an upward-sloping function of the price. In this respect, the forces determining the positions of the demand and supply curves are of vital importance as they reflect the determinants of membership. Regarding the demand curve, net benefits from being a trade union member and the appetite for trade union membership are given as the factors to shift the demand curve. On the other hand, any factor affecting the union organising costs, such as new legislation, is argued to shift the supply curve.

Similar to Ehrenberg and Smith (2000, pp. 465–466), Hirsch and Addison (1986, pp. 29–32) analysed the demand for trade union membership and supply of trade union services by means of functions of several variables. In this context, the variables of demand function are given as the price, wealth or permanent income, the wage difference between union members and others, non-pecuniary benefits, substitute services and individuals’ taste for unionism. Besides, the variables within the supply functions are given as the price of union services, costs of union organising, costs of servicing existing members and union goals. In this context, the equilibrium level of unionism is determined in terms of a joint (reduced) function of variables forming the demand and supply functions. Besides, it is highlighted that the researches in this field, including this article, need to define proxy variables (like individual or workplace characteristics) even though the interpretation of these variables is not straightforward as the variables in the joint function cannot be measured directly.

Considering individuals’ evaluation process regarding membership based on utility maximisation, Olson (1965), as cited in Ebbinghaus, Göbel, and Koos (2011, pp. 108–109) highlighted the free-rider problem referring to non-unionised workers’ not being excluded from collective goods like wages and working conditions. Seeking to understand why workers still prefer to be union members rather than being free-riders, the importance of social pressure on employees’ membership decisions came to the forefront (Hartley, 1992; Klandermans, 1986; Witte, 1988; as cited in Visser, 2002, p. 406). In this context, Booth (1985) associated the free-riding with the social custom theory which explained membership with reputation being the only incentive good.

Considering that social and psychological factors, in addition to purely economic ones, affect membership according to the social custom theory, it can be inferred that there are different theoretical approaches stemming from social sciences other than labour economics seeking to explain membership. For instance, sociologists and political scientists evaluate this field within the scope of class consciousness, values, modes of production, the composition of the workforce in the political climate, the role of government income policies and the centralisation and cohesiveness of the labour movement (Schnabel, 2002, p. 9; Schnabel & Wagner, 2007, p. 10). When it comes to social psychology, Klandermans (1986, p. 189) distinguished three theoretical approaches to union participation including frustration-aggression, rational choice and interactionist approaches.

Providing different explanations for the membership, economists and social scientists usually focus on different determinants of unionisation, while some of the determinants are common for distinct disciplines (Schnabel, 2002, p. 11). However, proxy variables are usually operationalised as measuring most of the real factors is impossible or very difficult. Also, it should be noted that most of the empirical analyses in the literature have not directly followed the lines of theoretical research. In this context, combining economic and socio-politic hypotheses and explanations, empirical studies in the literature regarding the determinants of membership (such as Schnabel & Wagner, 2007; Visser, 2002) usually reflect an eclectic approach.

According to Ebbinghaus and Visser (1999, pp. 135–137) and Schnabel (2002, pp. 11–16) most of the empirical analyses focusing on union membership and growth make cyclical, institutional (or configurational) and structural explanations. Firstly, revealing the macro-determinants of the union growth and decline through time-series data, cyclical explanations usually establish models including business cycle components like wage, price changes, employment growth and unemployment along with socio-political variables. Secondly, institutional explanations are made to analyse the cross-national variations in an institutional setting in an economy and a society affecting unionisation. Institutional variables include factors such as the Ghent System, works councils, union access to the workplace, legal protection for union organisers and union members, the centralisation of collective bargaining, the presence of left-wing governments and pro-union legislation. Lastly, structural explanations set forth individual characteristics of union members addressing sectoral and occupational aspects. Most of these studies utilise cross-sectional data and analyse the determinants of unionisation by comparing the characteristics of union and non-union employees.

There are various studies with structural explanations for union membership. Most of them focus on single countries. Among all, Antos, Chandler, and Mellow (1980) focused on the differences between men and women in terms of unionisation in the United States. It was concluded that the important part of the gap between men and women in terms of unionisation remained unexplained and this was evaluated as reflecting discrimination or simply unmeasured determinants. Furthermore, Windolf and Haas (1989) observed the determinants of union membership in West Germany. Size of the firm, union membership of works councillors and employment in public service were found to be important variables, which influence an individual’s propensity to join the union. Other variables such as gender, education and political partisanship (Green Party) were also observed to have significant influences on membership. On the other hand, there are studies primarily focusing on the determinants of leaving trade unions. In this context, Jódar, Alós, and Vidal (2011) revealed that the tendency to leave the union is greater among the unemployed, temporary workers, pensioners, unskilled workers with low seniority, workers in small workplaces, those who have switched jobs and the newest union members. Furthermore, it was observed that job status changes account for the majority of non-union-related reasons for leaving. Also, Leschke and Vandaele (2018) argued that labour market attachment and firm-level characteristics are influential while varying among genders.

Visser (2002) considered the membership separately as joining and leaving. Namely, he sought insight into the dynamics of membership and the differential impact of personal backgrounds, social pressure, workplace organisational and union services on joining or leaving unions in the Netherlands based on the social custom theory. The study revealed that the joining rate was influenced positively by pro-union attitude, perceived pro-union climate in the workplace, perceived level of unionisation of co-workers, attending a union meeting and being approached by the union while the perception of high costs relative to benefits, working in the private sector, the higher number of jobs before joining decreased the odds of joining the unions. On the other hand, the strongest determinants of leaving were observed as the ones due to external events in the working career, in particular, those related to unemployment, childbirth and a change of jobs or status while pro-union attitude and perceived benefits of the union decrease the odds of leaving. Furthermore, Kirmanoğlu and Başlevent (2012) included previous members separately in their study. Combining data comprised of 24 European countries, they found that higher self-transcendence and conservation scores were associated with a greater likelihood of being a current member, while higher openness-to-change and self-enhancement scores have the opposite effect.

Among the studies focusing on the determinants of membership in multiple countries, one of the most prominent ones belongs to Blanchflower (2007) who drew on individual-level micro-data of 38 countries. Most remarkably, the probability of being a union member is found to follow an inverted U-shaped pattern in the age variable in the majority of the countries. Moreover, Schnabel and Wagner (2007) focused on 18 EEU countries. It was observed that personal characteristics played a significant role in explaining union membership in most of the countries studied. The relative importance of individual characteristics, however, varied considerably between countries and there was not one country in which all the personal characteristics included in the model proved to be significant. Furthermore, workplace characteristics as well as employees’ attitudes made significant contributions to explaining union membership in the majority of the countries. Besides, Ebbinghaus, Göbel, and Koos (2011) focused on 19 European countries. In this context, a robust positive significant effect of the Ghent System on union membership was observed while the effect of workplace representation on membership was revealed to be smaller in the Ghent countries than in others. Also, increasing firm size was found to decrease the effect of being a union member due to the Ghent incentive.

Briefly, there is a rich literature on the individual-level determinants of membership. However, most of the studies focus on single countries while a fewer number of studies estimate models of membership for multiple countries. Besides, the majority of the studies focus on the determinants of membership as a whole, reflecting the dichotomy of “member” or “not member”. Considering that an employee may also resign from a union, previous membership is separately included in the analyses in a few studies. Aiming to contribute to this literature, this article attempts to analyse the current membership and previous membership separately in multiple countries by constructing logistic regression models with the same explanatory variables based on more up-to-date data for 15 European countries drawing primarily on the supply–demand theory. In this context, the following research questions will be analysed in this article:

• ▪ Are personal, occupational and workplace characteristics of European employees effective on the probability of being a current trade union member? If they are, are they consistent among European countries?

• ▪ Are personal, occupational and workplace characteristics of European employees effective on the probability of being a previous trade union member? If they are, are they consistent among European countries?

• ▪ Do the personal, occupational and workplace characteristics affect previous membership as much as current membership in Europe?

Certain personal, occupational and workplace characteristics of employees, determined considering the literature in this field, will be used as proxy variables in this article as the costs and benefits of membership (or their perception) would differ depending on them. In this context, the following hypotheses will be tested especially within the scope of the first two research questions:

• ◦ The probability of membership follows an inverted U-shaped pattern as the age increase,

• ◦ The probability of membership is higher for males than for females,

• ◦ The probability of membership decreases as the level of education increases,

• ◦ The probability of membership is higher for natives than foreign-born employees,

• ◦ The probability of membership decreases as the qualification of employee increases,

• ◦ The probability of membership is higher for contracted employees than the ones with limited contracts and no contracts,

• ◦ The probability of membership increases as the size of the workplace increase,

• ◦ The probability of membership is higher for employees in the public sector than the ones in the private sector.

Methodology

The research questions primarily focusing on the effects of personal characteristics as well as workplace and occupational characteristics on current membership and previous membership in multiple countries necessitate drawing on multinational data including information about the union membership status of employees. In this respect, the empirical analysis within the scope of this article is conducted based on the secondary data belonging to the ninth round of the ESS fielded in 2018 which uses cross-sectional, probability samples representative of all persons aged 15 and over resident within private households in each country. In this article, the sample is further narrowed to current employees (with different union membership statuses) corresponding to a total of 14702 current employees from EU-15 countries, referring to the EU member states before the enlargement in 2004. It should be noted that Norway and Switzerland are included in the study considering their tight bonds with the EU instead of Greece (in which the ninth round of the ESS was not conducted) and Luxemburg. Accordingly, the countries included in this study are Austria, Belgium, Denmark, Finland, France, Germany, Great Britain, Ireland, Italy, the Netherlands, Norway, Portugal Spain, Sweden and Switzerland.

The ESS provides a vital opportunity for researchers willing to conduct a cross-national analysis of trade union membership as it includes the following question: “Are you or have you ever been a member of a trade union or similar organisation? If yes, is that currently or previously?” In this context, the pattern of the survey allows the researcher to distinguish the current members and previous members as well. This article will also benefit from this distinction to further analyse the topic by incorporating the previous membership aspect as well.

There are some weaknesses in the data. First of all, it is not clear what the question means by “similar organisation”. For the purpose of this research, all responses are assumed to refer to trade unions. Furthermore, the information requested about the trade unions is confined to the membership statuses of the respondents in the ninth round of ESS. As underlined in the literature review, the determinants of membership are usually analysed through proxy variables. However, some variables with less proxy nature, such as the presence of a union in the workplace or the attitudes towards unions could be used. But the ESS does not include such questions directly related to the unions. Furthermore, considering that the primary reasons for being a previous trade union member are associated with changing jobs and being made redundant, etc., the lack of questions reflecting these kinds of changes in the employment circumstances of an employee in ESS limits especially the explanation of previous membership.

Individual-level determinants of membership are usually analysed through logistic regression analysis depending on the dichotomous nature of the membership as “member” and “non-member”. In this article, two logistic regression models with the same explanatory variables, one for current membership and another for previous membership, will be established for each country as well. Firstly, in the models established for the current membership, the options of the response variables will be “never-member” and “current member”. In this respect, they can be considered as specifically focusing on the determinants of the likelihood of being a current member for the first time in his/her employment lifetime in comparison with being a never-member. Secondly, the options will be “current member” and “previous member” in the models established for the previous membership. Accordingly, they can be evaluated as analysing the determinants of the likelihood of being a previous member in comparison with being a current member. In this respect, it should be noted that the interpretation of current membership should not be considered as joining the trade union or the interpretation of previous membership should not be considered as leaving the trade union since the empirical analysis is based on cross-sectional data rather than tracing the joining and leaving decisions of the same groups of people.

The method applied based on repetitive logistic regression models with the same explanatory variables, instead of a single cross-sectional model including all countries, will be applied for ensuring a clearer comparison of countries. Moreover, it will provide a simpler interpretation compared to multilevel logistic regression models considering that multiple countries are incorporated. Furthermore, the concept of the average marginal effects, referring to the average effect of changes in explanatory variables on the change in the probability of outcomes and calculated by taking the average of the changes in probability for each observation, is utilised in this article for the interpretation of the logistic regression models to ensure clarity. Moreover, the use of marginal effects is more appropriate when it comes to group comparisons in logistic regression as they are not scaled by unobserved heterogeneity (Long & Mustillo, 2021). Besides, statistical analysis within the scope of this article is conducted through R version 4.2.0.

The explanatory variables used in this article are determined based on the supply–demand theory in which the willingness of employees to be members and the willingness of unions to supply services are considered as demand and supply curves and the proxy variables are utilised to determine the factors shifting the demand and supply curves. Considering that the expectation from a trade union membership would differ depending on certain characteristics of employees, eight covariates having the potential to affect the willingness of employees to be members and the unions to supply services are used in this article which can be grouped as personal characteristics as well as workplace and occupational characteristics.

Starting with the ones within the scope of personal characteristics, the first covariate within the scope of personal characteristics is age. Considering the literature, age squared, in addition to age, is also included in this article in order to detect possible non-linear relationships, as the probability of current membership (or previous membership) does not have to increase (or decrease) at a constant slope. Furthermore, expecting to reveal the differences between men and women in terms of trade union membership, the second covariate is determined as gender. Moreover, the third covariate is about being born in the country analysed. Even though, there is not a clear theoretical explanation for the inclusion of this variable, the difference (or indifference) in the probability of trade union membership of natives and foreigners may provide insight into the trade union membership of foreign-born employees. Also, the fourth covariate and the only numerical one other than age, is the number of years in education. Similar to the age variable, the square of the years in education is also taken into consideration in order to detect a potential non-linear relationship.

The remaining four covariates reflect the workplace and occupational characteristics. In this respect, the fifth covariate is the size of the establishment, which refers to the number of people employed at the place where the respondent works. There are 5 categories for this variable including “under 10”, “10 to 24”, “25 to 99”, “100 to 499” and “500 or more”. Then, the sixth covariate is the type of organisation. Within the scope of this covariate, the design of the ESS’s relevant question provides 6 categories including “central or local government”, “other public sector (such as education and health)”, “state-owned enterprise”, “private firm”, “self-employed” and “other”. In order to identify the public-private distinction; “central or local government”, “other public sector (such as education and health)” and “state-owned enterprise” categories are merged under the “public” category in this study. Furthermore, “self-employed” and “other” categories are removed.

The seventh covariate is the qualification of the employee. The response categories in the ESS are given based on the International Standard Classification of Occupations known as ISCO08. In order to simplify the responses comprised of 590 categories; managers, professionals, technicians and associate professionals are coded as “high-skilled white-collar”, clerical support workers and service and sales workers are coded as “low-skilled white-collar”, skilled agricultural, forestry and fishery workers as well as craft and related trades workers are coded as “high-skilled blue-collar” and, plant and machine operators and assemblers along with elementary occupations are coded as “low-skilled blue-collar”. Moreover, the eighth covariate is the contract type with categories including “unlimited”, “limited” and “no contract”.

Analysis of findings

Descriptive statistics

According to the calculations made based on the design weights given in the survey, 57.80 per cent of the sample are found to be never-members while 30.97 per cent are current members and 11.23 per cent are previous members. Besides, the majority of the employees in the sample, 52.09 per cent, are females and the average age of the sample, in general, is 42.42. Furthermore, the average number of years spent in education is 14.60 while the vast majority of the sample, 85.21 per cent, were born in the country of their employment.

When it comes to the type of organisations, 31.39 per cent of the employees work in the public sector. Analysing the data considering the size of the workplaces shows that 22.59 per cent of the employees work in workplaces employing less than 10 employees while 30.03 per cent work in workplaces with a total number of employees between 10 and 24. Moreover, the largest portion, 25.72 per cent, work in workplaces with a total number of employees between 25 and 99. Furthermore, 17.41 per cent work in workplaces employing between 100 and 499 employees. Lastly, the group comprised of employees working in workplaces employing more than 500 constitutes the smallest share, namely 14.24 per cent.

Regarding the qualification of the employees, more than 15 per cent are low-skilled blue-collar, while less than 10 per cent are high-skilled blue-collar. Also, 28.76 per cent are low-skilled white-collar and 45.78 per cent are high-skilled white-collar. Besides, the majority of employees, namely 76.86 per cent, have unlimited (open) employment contracts while 17.90 per cent have limited (with a specific end-date) ones. On the other hand, more than 5 per cent do not have any contract in return for their employment.

Determinants of current trade union membership and previous trade union membership

Logistic regression models with the same explanatory variables are built in order to comprehend the determinants of current membership and previous membership in 15 European countries. Within the scope of the models built for current membership, the areas under ROC curves indicating goodness of fit differ between 68.36 per cent in Switzerland and 88.3 per cent in Denmark while the average is 78.85 per cent. Regarding the models built for previous membership, they vary between 64.93 per cent in Sweden and 79.73 per cent in Portugal while the average is 72.27 per cent. Calculated based on logit models for current membership and previous membership, Tables 1 and 2, respectively, give the average marginal effects along with the p-values in parenthesis while the base categories for the categorical independent variables are “born in the country: yes”, “establishment size: under 10”, “gender: male”, “qualification: low-skilled blue-collar”, “type of organisation: private” and “contract type: unlimited”.

Table 1. Average marginal effects of determinants of current trade union membership.

Note: Signif. codes: ***p≤0.001; **p≤0.01; *p≤0.05; p≤0.1.

^† Number of cases include the current employees who are previous members and current members.

Source: Author’s calculations based on the ESS Round-9 dataset.

Table 2. Average marginal effects of determinants of previous trade union membership.

Signif. codes: ***p≤0.001; **p≤0.01; *p≤0.05; p≤0.1.

^† Number of cases include the current employees who are previous members and current members.

Source: Author’s calculations based on the ESS Round-9 dataset.

As can be seen in Table 1, inverted U-shapes, indicating an increasing tendency for being a current member with age at a decreasing rate and falling at higher years of age, are prominent in 13 of 15 countries while more than half of them are statistically significant at α = 0.05. This finding, in line with Blanchflower (2007) as well as Ebbinghaus, Göbel, and Koos’s (2011), implies that younger and older employees need union benefits less than middle-aged employees. Blanchflower (2007) associated this pattern with the cohort effect even though it remains after the removal of cohort effects. In this context, one might expect a U-shaped (or at least positive) relationship between being a previous member and age at the first glance. However, as can be seen in Table 2, there is not a prevalent pattern found in this article. Rather, the results in different groups of countries are even in contrast with each other while very few of these relationships are found to be statistically significant.

Regarding gender as a determinant of current membership, the findings are in favour of males although the directions of the relationships differ between Scandinavian counties along with Great Britain and the rest. Namely, the probabilities of being a current member are higher for males in 10 countries while the differences in probabilities are statistically significant in 4 of them at α = 0.05. On the other hand, the probabilities of being a current member are observed to be higher for females in Scandinavian countries and Great Britain. Among those, the only statistically significant probability difference belongs to Finland at α = 0.05. When it comes to the effect of gender on being a previous member, it is not as obvious as it is on being a current member. Namely, the probabilities of being a previous member are higher for males compared to females in nine countries while none of them are statistically significant at α = 0.05.

The lower unionisation of women is usually explained by women’s weaker attachment to the labour market due to care responsibilities and their lower average income (Fitzenberger & Beck, 2003). Furthermore, Visser (2003, p. 397) explains this as a result of women’s higher engagement in part-time jobs. Besides, the results of this article give hints of narrowing the gender gap in unionisation as the differences in many countries are not statistically significant and there are countries in which being female increases the probability of current membership. This can be interpreted as a reflection of the rise of female union membership along with the decline of industrialisation that began around 1970 and occurred nearly everywhere, with some countries in northern Europe and the Anglo-Saxon world taking the lead (Visser, 2019, p. 29).

The interpretation of the effects of gender on previous membership becomes more meaningful when the effects of gender on current membership are also taken into consideration. For instance, the directions of the effects of gender on current membership are converse to the ones on previous membership in 11 countries. Namely, in Belgium, France, Germany, Ireland, the Netherlands and Portugal, where the probabilities of current membership are higher for males, the probabilities of previous membership are higher for females. It can be inferred that the higher attachment of males to the labour market in these countries lowers their tendency to be a previous member. Also, in Denmark, Finland, Norway, Sweden and Great Britain, where the probabilities of current membership are higher for females, the probabilities of previous membership are higher for males. In this context, it can be inferred that the higher attachment of females to the labour market especially in Scandinavian countries is apparent within the scope of previous membership as well.

The findings of this article in terms of the relationship between the level of education and being a current member are in line with Schnabel and Wagner (2007), who observed that the probabilities of membership decline in higher and lower-educated employees compared to medium-level educated employees. Namely, concave relationships between the years spent in education and being a current member are observed in the majority of the countries. This means that, as the years spent in education increase, the probabilities of current membership increase but the effects of years spent in education decrease. Regarding the effect of education level on being a previous member, in comparison with being a current member, the results are more fragmented among countries. Namely, there are several types of relationships including convex, concave and positive linear. However, the effects of education on being a previous member are not statistically significant in any country. In this respect, it can be inferred that the level of education does not constitute an important or consistent factor for previous membership.

The relationships between being foreign-born and current membership are found to be negative in 12 countries in this article unlike Schnabel and Wagner (2007) which obtained varying signs for this variable in different countries. Furthermore, most of the negative relationships are found to be statistically significant. On the other hand, the effects of being born in another country than the country of employment on the probabilities of previous membership are found to be negative in eight countries and positive in seven countries while only one of them is statistically significant at α = 0.05. In this context, it can be interpreted that being native is not as influential in previous membership as it is in current membership. Moreover, the slight effects it has do not have a prevalent direction in Europe.

The results of this article regarding the effect of being native, especially on current membership, are in line with the expectation of less unionisation of foreign-born employees compared to native employees.¹ According to Visser (2019, pp. 32–33), it is not clear if the lower unionisation of migrant employees is associated with the lower propensity of migrant employees to join or the lack of unions targeting migrant employees. On the other hand, Gorodzeisky and Richards (2013) and Kranendonk and de Beer (2016) conclude that the lower unionisation of migrant workers in Europe cannot, or can only to a very small degree, be explained by the individual characteristics of migrants.

The interpretation of the qualification of employees is more complicated compared to other variables. First of all, the probabilities of current membership are higher for high-skilled blue-collar compared to low-skilled blue-collar in 12 countries. However, when we compare the low-skilled blue-collar with a one-step more qualified employee group, namely the low-skilled white-collar, the results get somewhat mixed. Namely, in eight countries, the probabilities of current membership are lower for low-skilled white-collar. Furthermore, in 11 countries, the probabilities of current membership are observed to be lower for high-skilled white-collar compared to low-skilled blue-collar. In this respect, it can be interpreted that the groups comprised of high-skilled blue-collar have the highest probability of current membership in the majority of the countries. On the other hand, the most qualified groups, namely the high-skilled white-collar, are usually the ones with the lowest probabilities of current membership in the majority of the countries. Regarding the effect of qualification of employees on previous membership, similar to current membership, the most salient finding is about the high-skilled blue-collar employees. This group constitutes the employees with the lowest probability of previous membership in nine countries. Other than this, high-skilled white-collar employees are the ones with the highest probability of previous membership in seven countries. The findings concerning the effect of the qualification of employees are in line with the traditional explanations within the scope of the supply–demand theory. For instance, according to Schnabel (2002, p. 19), white-collar workers are often observed to be less likely to be union members than blue-collar employees because of more homogeneous preferences and working conditions of blue-collar employees which make them more likely to organise.

Regarding the contract types, the probabilities of current membership are found to be higher for unlimited contracted employees than for limited contracted employees in all countries. Besides, the probabilities of being a current member are higher for unlimited contracted employees than the employees with no contract in 11 countries. Furthermore, the probabilities of being a current member are higher for employees with limited employment contracts than the ones with no contract in 10 countries. When it comes to previous membership, the probabilities of being a previous member are higher for employees with limited contracts compared to the ones with unlimited contracts in almost all countries. Moreover, in nine countries, the probabilities of being a previous member are higher for employees with no contract than the ones with unlimited contracts. Briefly, the probabilities of being a previous member are higher for the employees with no contracts and limited contracts compared to the ones with unlimited contracts in the majority of the countries. The findings regarding the type of contract can be interpreted in line with the higher extent of labour force attachment of employees with unlimited contracts. Furthermore, from the point of view of trade unions, atypically employed ones are more difficult to recruit (Ebbinghaus, Göbel, & Koos, 2011, p. 112).

Concerning the effect of establishment size, in 12 countries, the probabilities of being a current member are lowest for the employees working in workplaces with less than 10 employees, which constitute the smallest workplace group. Furthermore, in most cases, the probabilities of current membership increase as the size of the establishment increase. On the other hand, the effect of this variable on the previous membership is not as definite as it is on the current membership. Namely, it can be concluded that the probability of being a previous member is higher among the employees working in smaller size workplaces but there is not a clear distinction between medium-sized and large workplaces. Within the scope of the supply–demand theory, the explanation of the higher probability of being a member among employees working in larger workplaces can be made associated with lower organising costs for unions in larger units. Also, union services may be valued more in large bureaucratic organisations where employees are likely to be treated impersonally and feel a greater need for representation and protection (Schnabel & Wagner, 2007, p. 22). On the other hand, in Denmark, the Netherlands, Portugal, Sweden and Switzerland, there is not any statistically significant difference in probabilities of current membership between employees in different sizes of establishments. Among these countries, in Denmark and Sweden, the smaller effect of workplace size might be associated with the Ghent system employed in these countries which helps organising employees in small workplaces as well (Ebbinghaus, Göbel, & Koos, 2011, pp. 120–121).

Lastly, the type of organisation can be considered as the most effective variable for both current membership and previous membership. To be more precise, the probabilities of being a current member are higher for employees in the public sector compared to the private sector in all countries while almost all of the probability differences are statistically significant at α = 0.05. Besides, in all of the countries, the probabilities of being a previous member are higher in the private sector while the probability differences are statistically significant in nine countries at α = 0.05. Similar to this article, Antos, Chandler, and Mellow (1980), Visser (2002, p. 417), and Windolf and Haas (1989) found that union membership is more likely for employees in the public sector than the ones in the private sector (as cited in Schnabel, 2002). Resembling the establishment size, higher current membership and lower previous membership inclination in the public sector might be associated with the higher tendency of union recruitment along with easier and less costly retention in large, homogeneous organisations with a bureaucratic nature and a low turnover rate (Schnabel, 2002, p. 19).

Why don’t the personal, occupational and workplace characteristics affect previous membership as much as current membership?

All of the variables within the scope of the personal characteristics incorporated in this article are found to be influential in the current membership. Namely, the probability differences are found to be statistically significant in many cases, while their patterns are consistent among European countries. On the other hand, the effects of personal characteristics on previous membership are observed to be neither statistically significant in most cases nor consistent among countries. Namely, the effects of age, gender and being born in the country of employment on previous membership are not as definite as they are on current membership. Similar to personal characteristics, occupational and workplace characteristics are also observed to be influential in being a current member. Their effects are frequently significant and their patterns are common in most European countries. When it comes to previous membership, the effect of occupational and workplace characteristics are not as commonly obvious as they are on current membership even though they are more apparent compared to personal characteristics.

Taking into account that the supply–demand theory depicts the union membership based on employees seeking to maximise their utilities, one might expect individual-level characteristics determined based on this theory, to affect the current membership and previous membership in inverse directions. For instance, Vaona (2010, p. 1093) claims that the workers having a smaller probability of joining tend to coincide with those having a higher probability of leaving and he associates this with labour market segmentation. It can be argued that this article revealed findings in line with this expectation in general. Namely, in the majority of the countries studied, the effect of being female has inverse directions for the probabilities of current membership and previous membership. Furthermore, high-skilled blue-collar employees are found to have the highest current membership and lowest previous membership probabilities in most countries. On the other hand, high-skilled white collars are found to have the lowest current membership and highest previous membership probabilities in the majority of the countries. When it comes to the contract type, unlimited contracted employees are found to have the highest current membership and lowest previous membership probabilities in most cases. Moreover, the employees working in workplaces employing less than 10 employees are observed to be least likely to be a current member while being most likely to be a previous member in the majority of the countries. Furthermore, employees in the public sector are revealed to have the highest probabilities of current membership in all countries while having the lowest probabilities of previous membership in most countries. Briefly, it can be argued that various variables, especially the ones within the scope of occupational and workplace characteristics, or certain categories of these variables, have effects with inverse directions on current membership and previous membership.

Even though the findings of this empirical article are considered to be in line with the literature, it is remarkable that the effects of personal characteristics and occupational and workplace characteristics on previous membership are usually more indistinctive in comparison with their effects on current membership. This relative indistinctiveness might be associated with several reasons. Firstly, the logic behind being a previous member does not have to be simply the inverse of the one behind being a current member. For instance, once become a member, union-related reasons may be more influential in previous membership. In other words, the level of satisfaction (or dissatisfaction) with the union is of vital importance to decide whether to retain or leave (Waddington, 2006, pp. 22–26). A member might be dissatisfied with the union because of several reasons such as its inefficiency, incompatible political approach between membership and union leadership, high costs, industrial actions, limited information provided, attitude towards the workplace, and so forth However, variables reflecting the attitude of a member towards the union are not included in the analyses in this article as the data regarding the trade union in ESS is only limited to membership statuses.

Another reason for less obvious effects on previous membership than current membership might be related to the bounded rationality of the employees once they become members.² Namely, union members tend not to conduct cost–benefit analyses for their membership anymore after they become members. Usually, other than union-related reasons, substantial changes in the career such as retirement, being made abundant or changing jobs constitute the basis for re-evaluating their union membership-related position. As this article focuses on the current employees, the most potential career change-related re-evaluation of membership might be associated with changing jobs. However, depending on the data limitation, variables reflecting job change are not incorporated into this article.

Lastly, the secondary nature of being a previous member dependent on being a current member might be influential in the less explicit effects of the variables on previous membership compared to the current membership. In other words, the fact that being a previous member is a later process in terms of the sequence in an employee’s lifetime as it necessitates being a current member first may lead the effect to be less distinctive. For instance, being a previous member at an older age might also be associated with being a member at an older age as well. However, the duration of membership, which would be a more convenient variable from this point of view, is not included in this article because of data limitations. Besides, it should be noted that the fact that the previous membership is subsequent to the current membership does not only influence the effect of age (or time-related variables) on previous membership. Rather, this might also influence the effect of other variables since being a current member functions as a preselection. For instance, being born in another country than the country of employment might affect previous membership less significantly, as this variable already affected current members at the very beginning. As foreign-born employees who are relatively not sensitive to the negative effect of being foreign-born tend to be union members, the effect of this variable on previous membership might not be as strong (or consistent) as it is on current membership. In this respect, previous membership’s being dependent on the current membership might be influential in the relatively more indistinctive effect of most of the individual-level characteristics on previous membership.

Conclusion

This article primarily aims to reveal the individual-level determinants of current trade union membership and previous trade union membership in 15 European countries. It focuses on personal, occupational and workplace characteristics considering the supply–demand theory. In this respect, it provides an opportunity to get a comprehensive insight into current membership and previous membership at the same time, in addition to allowing for detecting the determinants not affected by country-specific factors through incorporating various countries with different union policy settings and unionisation levels.

Analyses conducted in this article provide a wide range of notable findings. Firstly, inverted U-shaped relationships are found between age and being a current member in almost all countries. Furthermore, the probabilities of current membership are revealed to be higher for males in most countries. Besides, the probabilities of being a current member are found to increase at decreasing rates as the years spent in education increase in the majority of the countries. Moreover, being foreign-born is observed to decline the probability of current membership in almost all countries. Also, high-skilled blue-collar employees are found to be the group with the highest probability of current membership in the majority of the countries while the same group is found to have the lowest probability of previous membership. On the other hand, high-skilled white-collar employees are observed to have the lowest probability of current membership in most countries, while they are the most common group having the highest probabilities of previous membership in the majority of the countries. Furthermore, the group comprised of employees working for workplaces employing less than 10 employees constitutes the group with the lowest probability of being a current member and the highest probability of being a previous member in most cases. Lastly, a positive effect of being employed in the public sector is revealed. Namely, the probabilities of current membership are observed to be higher for employees in the public sector than the ones in the private sector in all countries while the probabilities of previous membership are observed to be lower in all countries.

The variables examined are found to affect current membership and previous membership in inverse directions in most cases. This is reasonable as the supply–demand theory, which constitutes the basis of variable selection in this article, considers the employees as seeking to maximise their utilisation through cost and benefit analysis. However, the effects of these variables on being a current member are more apparent compared to being a previous member. Namely, the effects of the variables on the probabilities of current membership are found to be statistically significant in more cases while their patterns tend to be more consistent among European countries in comparison with their effects on the probabilities of previous membership. This finding might be associated with several reasons. Firstly, the level of satisfaction (or dissatisfaction) with the union is of vital importance for previous membership. Accordingly, the effects of the personal, occupational and workplace characteristics on previous membership become less obvious as they do not reflect employees’ perceptions of unions. The second reason might be related to the tendency of union members not to conduct cost–benefit analyses for their memberships anymore once they become members unless facing important career-related changes. Lastly, the fact that being a previous member necessitates being a current member initially may lead the effects of the variables on previous membership to be less distinctive. This reason affects the time-related variables (such as age) directly as being a previous member is also dependent on the time of being a current member. However, more remarkably, being a current member initially may play a filter role for being a previous member later. Accordingly, the sensitivity regarding certain characteristics might be different for the never-members and the current members as the current members are comprised of employees who have already made cost and benefit analyses and given a certain decision.

This article provides comprehensive information about the member profile of unions. Remarkably, understanding the current profile of the trade union membership is critical as extending the member profile is necessary for revitalisation in the era of de-unionisation. Besides, the benefits of this article can be enhanced by a future article extending its scope to include social customs theory while maintaining its focus on current membership and previous membership in multiple countries.