Organizational Structure of Labor Unions and Bounded Solidarity: Implications for Welfare State Building

Transcription

1 Organizational Structure of Labor Unions and Bounded Solidarity: Implications for Welfare State Building Jae-jin Yang (Yonsei University, and Hyeok Yong Kwon (Korea University, Paper presented at the 24th IPSA World Congress of Political Science 23 to 28 July, 2016 Poznań, Poland 1

2 I. Introduction Labor unions are the principal institution of workers in modern capitalist societies. Nobody would deny the importance of unions. Yet, there have been series of debates over the sociopolitical effects of labor unions. Many economists and the political right view unions largely as selfinterested interest groups who monopolize the labor market and raise members wages at the expense of unorganized people and of the efficient functions of the economy. In contrast, unions are also viewed by unionists and the political left as providing workers both with protection against arbitrary management decisions and with a voice in the political area for the ordinary working people (Freeman and Medoff, 1984). In fact, much of the literature on welfare state development regards unions as strong supporters and builders of the modern welfare state rather than a mere interest group. Unions are usually depicted not only as proponents of solidarity but also one of the most powerful prowelfare political forces since they provide a framework for the articulation and aggregation of social demands, leadership networks in and across industries, and more importantly the mobilization of workers in close alliance with social democratic parties. In a word, labor unions are viewed as natural protagonists of welfare state building (Korpi, 2006). What bands workers together as one is not just shared interests, but a feeling of solidarity. Unions usually promote the sense of solidarity among workers to secure organizational unity, a sine qua non for their collective action. Often value-bound emotional impulses may motivate industrial action whose expected material consequences are too small by themselves to have justified that action and its costs (Swenson, 1989: 14). We agree that unions tend to promote the sense of solidarity and to pursue a welfare state. But we challenge the taken-for-granted assumption that all unions seek to cultivate solidarity among the working class and support public welfare in most comparative study of welfare states, especially in the literature of the so-called 2

3 power-resources model (PRM). This paper argues that solidarity stays within the boundary of unions. The scope and degree of solidarity differ country by country according to the dominant form of labor unions: national vs. industrial vs. enterprise unions. In a country where labor movement is dominated by narrowly organized unions such as enterprise unions, the reach of solidarity is limited within their company so that club goods-style social policies are provided (e.g. corporate welfare). On the other hand, encompassing unions such as general unions at the national level or industrial unions at the industrial level tend to cultivate solidarity beyond corporate boundaries so that redistributive social policies are promoted (e.g. tax-based universal programs or at least industrywide social insurance programs). In order to examine whether and how dominant organizational form of unions affect the sense of solidarity among workers, we examine union members attitudes toward redistribution, using the International Social Survey Programme (ISSP 2009). The rest of the paper is organized into three sections. The first section is literature review on the heterogeneity of labor and solidarity as opposed to the basic assumption of the PRM. It also deals with different mode of unionism and its implication for the bounded solidarity in terms of social policy preferences. The next section discusses the data, variables, and empirical model specification, and then presents the results of analysis. The last section concludes with theoretical implications. II. Heterogeneity of Labor and Bounded Solidarity 1. Conventional Wisdom and Revisionist View on Union and Solidarity The power-resources model is the dominant paradigm in explaining variations in the welfare state development in the West. As a demand-side logic of politics, working class mobilization is 3

4 viewed as the key explanatory factor for early welfare state consolidation and subsequent expansion of public social security (Korpi, 1983; Hicks, 1999; Esping-Andersen and Korpi, 1984; Huber and Stephens, 2001; Castles and Mckinlay, 1979). For them, major differences in the welfare state development are determined by the relative success of organized labor and leftist political parties in the process of democratic class struggle (Korpi, 1983) or politics against markets (Esping-Andersen, 1985). For them, the balance of power between labor and capital is a major determinant of the character and extent of the welfare state. In the literature of power-resources model, the working class is the prime beneficiary of the social security programs, and thus trade unions are assumed to be at the center of welfare state building. Labor union density is regarded a key determinant of welfare state development. High union density gives unions more capacity to disrupt the economy, more resources for electoral campaigns and lobbying, and more power to extract pre-emptive concessions from the state and employers. In addition, legal extension of collective bargaining and institutionalization of social dialog among governments and peak organizations of employers and workers boost labor union power (McGuire, 1999). 1 If labor unions are closely allied with strong social democratic parties, union capacity to achieve welfare state development increases further still. In short, unions are viewed as both proponents and powerhouses of welfare state building. Also, unions are viewed as pursuing broad interests of the working class and thus have an interest in achieving greater public social welfare. The depiction of unions as representatives of the working class is by and large suited to the historical experiences of the European welfare state, especially Scandinavian countries. However, many revisionists pointed out, workers are not a homogeneous group with united interests in the introduction and expansion of public welfare programs (Oude Nijhuis, 2013; 1 Union participation in tripartite corporatism at either the national or administrative levels (e.g. the pension administrative board) provides them with important voice in the making of social welfare policy. 4

5 Becher and Pontusson, 2011; Hacker, 2002; Yang, 2013). As Pontusson(1992: 12) rightly pointed out, wage earners like any social class, and perhaps to a greater extent than other social classes, have multiple and partly contradictory interests. In fact, there is plenty of studies by comparative political economists that analyzed heterogeneity of labor and conflicts of interest among unions. The first line of studies emphasizes sectoral conflicts of interest between unions organizing sheltered sectors and those representing trade-exposed sectors. In the 1930s and 1950s in Sweden, conflict over centralization of wage bargaining between Metalworkers Union representing trade-exposed/low-paid workers and unions organizing domestic sheltered/highpaid industries such as construction is the good case in point (Swenson, 2002). Secondly, scholars of varieties of capitalism (VoC) demonstrate that workers with industry/firm-specific skills have different social policy preferences from those with general skills (Estévez-Abe, Iversen and Soskice, 2001). Thirdly, under the logic of insider-outsider, Rueda (2007 and 2006) shows that insiders prefer job protection to ALMPs(active labor market policies), and left parties are more likely to produce pro-insider policies rather than pro-outsider policies such as ALMPs since insiders constitute the core constituency of left parties. Fourthly, there is a growing body of literature that highlights the effect of organizational structure of labor movement on social policy preferences of workers. For example, Oude Nijhuis(2013) and Yang(2013) find that craft unions in the UK and enterprise unions in Korea did not pursue redistributive policies that would benefit lower-skilled and thus lower-paid risk-prone workers. As such, individual workers payoffs are diverse, depending on their sector, skill level, employment status and union structure. We cannot take for granted solidarity among workers and unions preference for redistributive social policies. Some unions often act like an interest group. They tend to represent clear occupational interests, seeking selective benefits for their members. This is why Milton Friedman (1962: 124) maintained that unions have not only harmed the public at large.. [t]hey have also made the incomes of the working class more unequal by reducing the opportunities available to the most disadvantaged workers and that 5

6 their effect has been to make high-paid workers higher paid at the expense of lower-paid workers. From this point of view, this paper question the conventional wisdom of the PRM that takes labor union s support for solidarity and redistribution for granted. There is no a priori reason to believe that organized labor pursues redistributive public social welfare other than corporate and private welfare. 2. Union Structure, Bounded Solidarity and Social Policy Preferences 1) Union Structure and Bounded Solidarity As Swenson (1989: 1-3) pointed out, the union leaders play two roles at the same time. One is to reconcile labor and capital to the former s advantage or often to their mutual advantage since workers prosperity depends on the performance of firms. Meanwhile, the union leaders try to manage conflicts of interest among workers so as to unify the working class organizationally and thereby enhance influence over employers and public policymaking. In the interest of material gains, unions employ collective bargaining and use the threat of strike action to extract higher wage from employers while pursuing more egalitarian wage policies within the unionized sector to maintain organizational solidarity. Therefore, the net effect of collective bargaining on wage inequality depends on the relative size of within sector effect (i.e. the effect of collective bargaining on decreasing pay differentials within unionized members) and between sector effect (i.e. the effect of collective bargaining on increasing the pay gap between unionized and non-unionized workers) (Hayter and Weinberg, 2011: 143). If unionized sector is small, then the between sector effect would overwhelm the within sector effect so that wage differentials will prevail. In contrast, if unionized sector is large, then the between sector effect would be outnumbered by within sector effect so that it would generate more egalitarian wage structure across the country. It is in this latter situation that union leaders need to play harder upon workers moral 6

7 sensibilities to regulate complaints and maintain membership support because egalitarian wage compression requires sacrifice of skilled workers at affluent sectors/firms who could be paid more. Unauthorized so-called wild cat strikes would occur when union leaders fail to regulate their complaints. If unions represent only privileged workers, then the need to rely on the sense of solidarity is not high. In a word, solidarity is bounded. A wide scope and high degree of solidarity is required to maintain encompassing unions whereas such scale and effort is not necessary for narrowly organized unions, including craft/enterprise unions. Then, can encompassing union leadership foster solidarity among workers? Ahlquist and Levi (2013) argues that union leaders possess useful knowledge about the state of the world that the average member does not have and can shape members attitudes on a policy. Members come to accept the leaders positions so long as the organization continues to deliver good basic outcomes. In a similar vein, Kim and Margalit verifies that unions are not merely a voice of workers preferences, but also an effective institution that is able to systematically shape and cohere that voice toward a given policy objective (2014: 37). That is why Mosiamann and Pontusson (2013: 1) can expect that a well-paid workers who belongs to an encompassing union, i.e. a union that organize workers across a wide range of the earnings distribution, will be more supportive of redistribution than her non-union counterpart because unions primarily promote solidarity among their members. (ibid. 7). All in all, we can conclude that basically union cultivate the feeling of solidarity among its members and more importantly that the scope and degree of cultivated solidarity depends on the degree of encompassing-ness of unions. 2) Union Structure and Social Policy Preferences This bounded solidarity also will affect policy preferences of workers. Most social welfare programs are designed to be public goods in the sense that social benefits go to every individual, whether or not the individual contributed to the costs of collective action to achieve them. 7

8 However, as Figure 1 indicates, social welfare has a variety of forms. Some are pure public goods for all citizens (e.g. tax-based universal schemes) and others are club goods for a particular group only (e.g. corporate welfare). 2 Some fall in-between (e.g. industry or income-based selective social insurance), achieving different levels of publicness. Also, private welfare could function as protection against social risks, taking a variety of forms of pure private goods such as higher wage, private insurances and savings. [Figure 1 about here] As discussed above, labor unions have diverse policy preferences in advancing workers material interests and protect them against labor market risks, depending on the income and risk profile of their members. Union leaders of national-level trade unions, which encompass not only skilled professionals but also unskilled manual workers, have interests in installing a nationwide redistributive social welfare system to enhance solidarity among workers. Relatively free from firm-specific or sector-specific interests, national union leaders are likely to be farsighted and to pursue long-term economic interests for the working class as a whole. If bargaining authority is centralized, then union leaders at the top will be better situated to overcoming the collective action problem and tendency toward segregation within and across companies and sectors. These characteristics will in turn increase their capabilities to achieve desired ends. In response to recent neoliberal pressure for labor market flexibility, national union leaders must contend with the potentially conflicting situation of low-paid temporary (or non-regular) workers. Coordinated flexibilization or flexicurity would be the favored alternative for leaders of encompassing unions. Although accommodating the inevitable need for flexibility, they would seek to weave comprehensive social safety nets for outsiders (the youth, the unemployed, the unskilled, etc) and provide them with more opportunities to upgrade their skills and build careers. 2 A characteristic of club goods is that exclusion is feasible, but the optimal size of the club is in general larger than one individual. In this sense, it is not a private good (Burton, 1978: 44). 8

9 Universal income-supporting programs with low entry barriers and extensive ALMPs would have policy priority over regulatory protections or segmented income maintenance programs for insiders in the labor market (Drøopping, Hvinden, and Vik, 2002). On the contrary, if workers are organized predominantly at the firm level and lack institutional means for comprehensive negotiations regarding distributional issues at the industrial or national level, distributional struggles will take place at the firm level and they will tend to be particularistic and temporally short-sighted. Under such conditions, necessity to cultivate solidarity is weak and voices for universal social welfare or industry-wide social welfare scheme are unlikely to prevail. Instead, enterprise unions would favor wage maximization and corporate welfare. In a word, club goods are favored. This likelihood will increase to the extent that enterprise union leadership is weak vis-à-vis their rank and file workers since union leaders will feel compelled to satisfy short-term demands of the rank and file. Therefore, other things being equal, these organizational arrangements are more likely to result in a under-developed small welfare state characterized by large wage differentials, a relatively high level of private welfare, and the overdevelopment of corporate welfare. This system will leave large segments of workers under-protected. In response to pressure for labor market liberalization, individual unions would prioritize job security over flexicurity, since losing a job means losing everything due to the weak social security system. Active labor market policies are a remote interest for individual enterprise unions. Therefore, a huge cleavage is likely to emerge between the internal and external labor markets, between large and small firms, and between organized and unorganized workers. In such settings, it is more likely that unorganized workers in the external labor market will be victims of labor market flexibility. In between, if labor movements are led by strong industrial unions, segregated social welfare schemes along industrial lines are likely to emerge. Nevertheless, industrial unions are more solidarity/redistribution-enhancing than enterprise unions since they represent relatively 9

10 low-paid/low skilled workers within the same industry. If they forge a close partnership with encompassing leftist political parties, their power may be strong enough to build a universal social security system although privileged workers within affluent sectors would be unwilling to sacrifice part of their share to benefit others unless they are ideologically determined to do so or governed by strong organizational discipline. 3 In response to pressure for labor market flexibility, regulatory employment protections and industry-wide unemployment insurance programs would be preferred by unions especially in affluent and powerful industries. This would help preserve specific job profiles and skills (re)employable for specific industries, as argued by VoC theorists (Estebez-Abe, Iversen, and Soskice, 2001). All in all, union leaders need to promote solidarity among workers to enhance collective power, and this tendency is strongest in the encompassing national unions. This cultivated solidarity can change the policy preferences of rank-and-file workers within the boundary of the union. Therefore, we present two hypotheses as follows. Hypothesis 1: Union members are more supportive of redistribution. Hypothesis 2: Encompassing union members are more supportive of redistribution than narrowly-organized union members (National union > Industrial Union > Enterprise Union). 3 If union leaders of affluent industries face leadership competition, they would have little room to pursue such a universal welfare state strategy. Competition to satisfy their constituencies is likely to result in industry-specific welfare programs. Competitive increases in wage and welfare between industries is also possible. 10

11 III. Empirical Analysis 1. Data and Variables Our empirical analysis makes use of individual-level ISSP 2009 data and country-level data. The countries included in analysis are Australia, Austria, Belgium, Bulgaria, Czech Republic, Denmark, Estonia, Finland, France, Germany, Iceland, Japan, South Korea, Latvia, New Zealand, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, UK, and USA. To measure redistribution preferences, we use the following question from ISSP: Please say to what extent you agree or disagree with the following statement: It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes." The score is on a metric of 1-5, where 1 represents agree strongly, 2 agree, 3 neither agree nor disagree,' 4 disagree, and 5 disagree strongly.' To make interpretation more intuitive, we converted the score so that 5 refers to agree strongly.' In addition, we measure unemployment policy preferences by using the following question: Please say to what extent you agree or disagree with the following statement: The government should provide a decent standard of living for the unemployed. We also converted the score so that 5 refers to strongly agree. For our present purpose, crucial questions are two-fold. First, are respondents who are union members more likely to support redistributive measures than respondents who are not union members? Second, how do different levels of union encompass-ness influence the effect of union membership? To address these questions, our analysis includes an individual-level variable Union member (coded 1 if the respondent is a union member, 0 otherwise), a country-level variable, encompass-ness of union structure, for which we operationalize by using Visser s indicator of the predominant level of collective bargaining in a country. We rely on ICTWSS (Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts in 51 countries in ). A level is predominant if it accounts for at least two- 11

12 thirds of the total bargaining coverage rate in a given year and country. Level 5 means that bargaing predominantly takes place at central or cross-indusry level and there are contrally determined binding norms to be respected by agreements negotiated at lower levels while level 1 indicates that bargainig predominantly takes place at the local or company level (Visser, 2013: 11) We include a multiplicative term to capture the cross-level interaction between these two variables (Union members Wage bargaining level). We estimate the effects of these variables on the preferences for redistribution while taking into account individual- and country-level variables that might influence attitudes towards redistribution. Our analysis includes Subjective income strata, survey respondents' selfplacement on a 10-scale poor-rich continuum, with higher scores indicating higher income group. Although this variable is a subjective self-placement measure, the data availability leaves us to use this measure because the ISSP data does not include a consistent and unified measure of income levels of respondents. Consistent with numerous political economy models, we expect subjective income strata to be negatively associated with support for redistribution. We include Left-Right ideology, a respondent's self-placement of their ideology on a 1-5 scale, where 1 denotes far Left and 5 extreme Right. Needless to say perhaps, we should expect a negative association between rightist ideology and redistributive preferences. We include Age (actual age), and dummy variables for Gender (female=1) and Unemployed in the analysis. As older people, women and the unemployed tend to rely more than their counterparts on the public provision of social welfare and redistribution, we expect these variables to have positive effects. We also include Education (1-5 scale). To take into account the argument about a substitutive relationship between religion and the welfare state (Scheve and Stasavage 2006), we include a dummy variable Religion (Christian=1). 4 4 Alesina et al. (2003) and Roemer and Van der Straeten (2006) focus on the racial conflict that could explain the refusal of redistribution, when individuals expect migrants to get all the benefits from it. It might have been better to include respondents view on immigration. However, the ISSP 12

13 To examine country-level contextual effects, we include Union density, Growth (GDP growth rate), and Electoral disproportionality. If the policy mood literature (Erikson et al. 2002) is correct, we should expect a positive association between macroeconomic conditions and redistributive preferences. Electoral institutional structures help shape party strategies as well as individual social policy preferences. Accordingly, we include the Gallagher index of electoral disproportionality. [Table 1 about here] Table 1 shows redistribution and unemployment preferences and wage bargaining level by country. Portugal and post-socialist countries like Slovenia, Bulgaria, and Latvia show higher support for redistribution, whereas the liberal market economies such as Australia, New Zealand, and the US show lower levels of support. Many European democracies, on the other hand, show that the support for unemployment insurance is relatively higher than the support for redistribution. [Figure 2 about here] Figure 2 shows the mean difference in redistribution preferences (1-5 scale) between union members and non-union members by country. A positive score denotes that union members are on average more supportive of redistributive measures than non-union members are. Australia shows the highest degree of differences (i.e., the distinctiveness of union member preferences), followed by Belgium, France, Slovakia, Sweden, and Norway. Other countries show that the difference in redistributive preferences is not statistically distinguishable between union members and non-union members (95% confidence interval includes 0). 2. Model Specification 2009 data does not include a question on individual attitude towards immigration. 13

14 To estimate the effects of individual-level as well as country-level covariates, it is necessary to take into account the multilevel structure of the data (Steenbergen and Jones 2002). Combining multiple levels of analysis in a single model, we can specify predictors of individual redistributive policy preferences at different levels. Accordingly, we employ a multilevel modeling strategy. The categorical responses yy ii,jj --redistributive policy preferences of individual in a country j -- may be modeled by a generalized linear mixed model with country-level covariates and cross-level interactions (Raudenbush and Bryk 2002). llllllllll Pr yy ii,jj > ss = XX TT ii,jj ZZ jj γγ + XX TT ii,jj δδ jj κκ ss where XX ii,jj is individual-level covariates, ZZ jj is country-level covariates, γγ is a vector of coefficients for fixed effect, and δδ jj is the random effects for country j. We estimate the 1-5 scale ordinal responses about the preferences for redistribution by a generalized linear mixed model (GLMM) with ordered logit link. IV. Results This section presents estimation results. Before we discuss the results, we first examine associations between average redistributive preferences and wage bargaining levels. How does the mass public's support for redistribution vary across different wage bargaining levels? Does union membership effect vary by different levels of collective bargaining? [Figure 3 about here] Figure 3 shows an association between wage bargaining levels and redistributive preferences. The left panel suggests that countries with higher levels of collective bargaining tend to have relatively higher levels of redistributive preferences. In contrast, countries where labor union is 14

15 organized at plant- or company-level such as US and New Zealand show relatively lower levels of the support for redistribution. It is noteworthy that new democracies of post-socialist countries like Slovenia, Latvia, Bulgaria, Estonia, Poland, and Slovakia as well as East Asian new democracy, South Korea show comparatively higher levels of redistribution preferences than old democracies. The right panel presents the relationship between wage bargaining levels and the distinctiveness of union members' redistributive preferences, measured as the difference between union members' and non-union members' preferences. A positive score denotes that union members are on average more supportive of redistribution than non-union members are. The right panel in Figure 3 suggests that higher levels of collective bargaining are associated with higher levels of distinctiveness of union members' preferences, suggesting that unions create and sustain solidarity to their members especially when collective bargaining takes place at higher levels. [Figure 4 about here] What is the relationship between collective bargaining levels and unemployment insurance preferences? We observe, from the left panel of Figure 4, that collective bargaining level is not significantly associated with unemployment insurance preferences. The right panel shows, however, that the union effect on unemployment insurance preferences is higher in countries with higher levels of collective bargaining. That is, union members' distinctiveness in the preferences for unemployment insurance policy expansion tends to be more pronounced and distinctive in countries with higher levels of collective bargaining such as France, Sweden, and Austria. [Table 2 about here] Table 2 presents the results of analysis. Model (1) is a random intercept model without contextual fixed effects and model (2) is a random coefficient model with country-level fixed effects as well as a cross-level interaction term. Consistent with Hypothesis 1, the results suggest that there is a union effect on redistribution preferences. Union members are more likely to support than non-union members for redistributive measures to reduce income gaps. There are two possible explanations of this union-membership group identity effect. One is that being a 15

16 union member alters the way individuals perceive their interests. Suppose that we observe two individuals in the same job and with the same skills and income, one being a union member, the other not: the argument here is that the union member will be more likely to look for collective/public means to improve his or her employment security and relative income. The other explanation is that union membership serves as a proxy for certain objective characteristics (occupation and skill levels as well as perhaps characteristics of the place of employment) that are not captured by our control variables. But, it should be noted that the union effect we find was obtained while after taking ideological self-placement and a proxy for skill levels (i.e., education) into account. The coefficients for most of the individual-level variables have expected signs and clear conventional thresholds of statistical significance. Not surprisingly, subjective income stratum has a negative and statistically significant effect on redistributive preferences. As income increases, the probability of supporting redistributive measures decreases. Ideologically conservative respondents are more likely than their counterparts to oppose to redistributive measures. Educational attainment, a proxy for skill level, has a negative effect on the preferences for redistribution. It may be the case that education is correlated with income. The results also conform to the existing studies in that female is more supportive of redistribution than their male counterparts. The unemployed is found to have a positive association with redistributive preferences, but it merely reaches a 90% confidence level. It may be the case that the unemployed are more supportive of the public provision of welfare in the form of passive and active labor market policies, but they are not necessarily supportive of redistribution. Like Scheve and Stasavage (2006), we find a negative and statistically significant effect of religion (Christian) on redistributive preferences. Turning to the effect of contextual factors, the results of model (2) suggest that collective bargaining level exerts a positive effect on the support for redistribution. Countries with higher levels of collective bargaining system tend to have a higher level of overall support for 16

17 redistribution. More importantly for our purpose, the cross-level interaction term is statistically significant and has an expected sign, suggesting that Hypothesis 2 is supported. That is, the union member effect is pronounced when collective bargaining level is higher. Conversely, the union member effect is moderated or even muted in a context where wage bargaining takes place at company level and union structure is decentralized. Union density turns out to be negatively associated with overall level of redistribution preferences. Although we do not have a clear explanation for this surprising finding, it may suggest that the sheer magnitude of union membership is less critical in shaping attitude and policy preferences of workers than the mode of labor movement. Consistent with the policy mood literature that argues that policy mood tends to move towards a liberal direction in economic booms and move towards conservative in economic bursts (Erikson et al. 2002), this result suggests that good economic conditions are associated with higher levels of support for redistribution. Electoral disproportionality exerts a positive effect on the support for redistribution. Disproportional electoral systems affect overall levels of support for redistribution. Put differently, majoritarian systems tend to be associated with higher levels of average redistributive preferences. [Table 3 about here] Table 3 shows the results of analysis of unemployment insurance preferences. The results are qualitatively similar to the ones presented in Table 2. We found the union member effect on unemployment insurance preferences. We also found the cross-level interaction effect of union members and collective bargaining levels. Not surprisingly, the unemployed shows, unlike the previous result, a statistically significant association with unemployment insurance preferences. Estimated coefficients for other variables turn out to be qualitatively similar to the results of redistribution preferences. [Table 4 about here] To get a sense of the substantive effect of union membership and collective bargaining levels, 17

18 Table 4 shows the conditional predicted probabilities of favoring redistribution (`agree' and `agree strongly') across collective bargaining levels by union membership. Table 3 shows that while preferences for redistribution rise as wage bargaining level increases, the collective bargaining level effect is stronger among union members than non-union members. We can also observe from Table 3 that the union member effect, the difference between union members and non-union members, is more pronounced where there is a higher level of wage bargaining system than in contexts with decentralized collective bargaining system. For instance, the estimated probability of union members support for redistribution is 0.6 in countries with decentralized collective bargaining system (such as US, UK, South Korea), whereas it increases to 0.8 in countries with the highest level of collective bargaining system (Belgium). Union members in the medium level of collective bargaining system (3; such as Sweden, Denmark, Germany) are predicted to favor redistribution by V. Conclusion To be completed References Ahlquist, S. John and Margaret Levi In the Interest of Others: Organizations and Social Activism. New Jersey: Princeton University Press. Berger, J. Chris, Craig A. Olson and John W. Roudreau Effects of Unions on Job Satisfaction: The Role of Work-Related Values and Perceived Rewards. Organizational Behavior and Human Performance. Vol.32: Brady, David Comparing European Workers Part B: Policies and Institutions. Research in the Sociology or work. Vol.22: 1-6. Castles, Francis G. and R. D. McKinlay Public Welfare Provision, Scandinavia, and the Sheer Futility of the Sociological Approach to Politics. British Journal of Political Science. Vol.9(2): Esping-Andersen, Gøsta and Walter Korpi Social Policy as Class Politics in Post-War Capitalism: Scandinavia, Austria, and Germany. in John H. Goldthorpe (ed.) Order and 18

19 Conflict in Contemporary Capitalism. London: Oxford University Press. Esping-Andersen, Gøsta Politics Against Markets: The Social Democratic Road to Power. New Jersey: Princeton University Press. Estévez-Abe, Margarita, Torben Iversen and David Soskice Social Protection and the Formation of Skills: A Reinterpretation of the Welfare State. in Peter A. Hall and David Soskice (ed.) Varieties of Capitalism. New York: Oxford University Press. Freedman, B. Richard and James L. Medoff What Do Unions Do?. New York: Basic Books, Inc., Publishers. Friedman, Milton Capitalism and Freedom. Chicago: The University of Chicago Press. Form, H. William Job vs. Political Unionism: A Cross-national Comparison. Industrial Relation. May Vol.12(2): Hacker, S. Jacob The Divided Welfare State: The Battle over Public and Private Social Benefits in the United States. New York: Cambridge University Press. Hayter, Susan and Bradley Weinberg Mind the gap: collective bargaining and wage inequality. in Susan Hayter (ed.) The Role of Collective Bargaining in the Global Economy: Negotiating for Social Justice. Cheltenham, UK: Edward Elgar. Huber, Everlyne and John D. Stephens Development and Crisis of the Welfare State. Chicago: University of Chicago Press. Iversen, Torben and David Soskice Information, Inequality, and Mass Polarization: Ideology in Advanced Democracies. Comparative Political Studies. Vol.48(13): Kim, Sung Eun and Yotam Margalit Informed Preferencs? The Impact of Unions on Workers Policy Views. Unpublished paper, Department of political science, Columbia University. Korpi, Walter The Democratic Class Struggle. London: Routledge & Kegan Pauol Pblication. Korpi, Walter Power Resources and Employer-Centered Approaches in Explanations of Welfare States and Varieties of Capitalism: Protagonists, Consenters, and Antagonists. World Politics. Vol.58: McGuire, James Labor Union Strength and Human Development in East Asia and Latin America. Studies in Coparative International Development. Vol. 33(4):2-33. Michael Becher and Jonas Pontusson Whose Interests Do Unions Represent? Unionization by Income in Western Europe. Comparing European Workers Part B: Policies and Institutions Research in the Sociology of work, Vol.22: Mosimann, Nadja and Jonas Pontusson Bounded Communities of Solidarity: Union Membership and Support for Redistribusion in Contemporary Europe. University of Geneva. Oude Nijhuis, Dennie Revisiting The Role of Labor: Worker Solidarity, Employer Opposition, and the Development of Old-Age Pensions, in the Netherlands and the United Kingdom. World Politics. Vol.61(2): Oude Nijhuis, Dennie Labor Divided in the Postwar European Welfare State. New York: Cambridege University Press. Pontusson Introduction: Organizational and Political-Economic Perspectives on Union Politics. in Miriam Golden and Jonas Pontusson (eds.) Bargaining for Change: Union Politics in North America and Europe. Ithaca: Cornell University Press. 19

20 Rueda, Daved Social Democracy and Acrive Labour-Market Policies: Insiders, Outsiders and the Politics of Employment Promotion. Cambridge University Press. Rueda, David Social Democracy Inside Out. Oxford University Press. Swenson, Peter Fair Shares: Unions, Pay, and Politics in Sweden and West Germany. New York: Cornell University Press. Swenson, Peter Fair Shares: Unions, Pay, and Politics in Sweden and West Germany. Ithaca: Cornell University Press. Taylor, J. Andrew Trade Unions and Politics: A comparative introduction. London: Macmillan Education LTD. Visser, Jelle Data Base on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts, (ICTWSS). Yang, Jae-jin, 2013, "Parochial Welfare Politics and the Small Welfare State in South Korea," Comparative Politics, Vol.45, No.4. 20

21 Figure 1: Policy Spectrum of Social Welfare Private Welfare private goods pure Social welfare programs *higher wage *intra- *savings family *private transfer insurances quasi *charity Social Welfare club goods public goods *corporate *selective *tax-based welfare social universal insurances schemes Redistribution none weak strong Solidarity none weak strong Note: modified from Yang (2006: 214). 21

22 Figure 2: Union effect on redistribution preferences Notes: Dot plots denote the mean difference of redistribution preferences between union members and non-union members and lines indicate 95% confidence interval. 22

23 Figure 3: Wage bargaining levels and the union effect on redistribution preferences Notes: Left panel denotes the mean redistribution preferences and right panel the mean differences of redistribution preferences between union members and non-union members. Data for wage bargaining level is from Visser (2016). 23

24 Figure 4: Wage bargaining levels and the union effect on unemployment insurance preferences Notes: Left panel denotes the mean redistribution preferences and right panel the mean differences of redistribution preferences between union members and non-union members. Data for wage bargaining level is from Visser (2016). 24

25 Table 1: Preferences and wage bargaining levels by country Country Redistribution Unemployment insurance Wage bargaining level Mean Proportion Mean Proportion Australia Austria Belgium Bulgaria Czech Rep Denmark Estonia Finland France Germany Iceland Japan South Korea Latvia New Zealand Norway Poland Portugal Slovakia Slovenia Spain Sweden Switzerland UK USA Notes: Entries are the mean preferences and the proportion of favoring redistribution and unemployment insurance policy (combining agree and strongly agree ), and wage bargaining level (Visser 2016). 25

26 Table 2: Estimated effects of union membership and union structure on redistribution preferences Covariates [1] [2] Individual level Union member 0.193*** (0.030) (0.091) Subjective social strata *** *** (0.009) (0.009) Left-Right ideology *** *** (0.013) (0.013) Education *** *** (0.010) (0.010) Gender (female) 0.231*** 0.227*** (0.027) (0.026) Age * (0.0008) (0.0008) Unemployed 0.114* 0.118* (0.064) (0.064) Religion (Christian) *** *** (0.029) (0.030) Country level Wage bargaining level 0.327*** (0.015) Union Wage bargaining level 0.053** (0.026) Union density ** (0.001) Growth 0.114*** (0.006) Log(electoral disproportionality) 0.331*** (0.022) Country-level variance (0.010) (0.012) Log-likelihood N of observations N of countries

27 Table 3: Estimated effects of union membership and union structure on social policy preferences Covariates [1] [2] Individual level Union member 0.139*** (0.032) (0.091) Subjective social strata *** *** (0.009) (0.009) Left-Right ideology *** *** (0.013) (0.013) Education *** *** (0.010) (0.010) Gender (female) 0.055** 0.060** (0.026) (0.026) Age 0.005*** 0.006*** (0.001) (0.001) Unemployed 0.668*** 0.651*** (0.066) (0.066) Religion (Christian) *** *** (0.029) (0.030) Country level Wage bargaining level (0.015) Union Wage bargaining level 0.065** (0.031) Union density (0.001) Growth ** (0.006) Log(electoral disproportionality) *** (0.022) Country-level variance (0.009) (0.013) Log-likelihood N of observations N of countries

28 Table 4: Predicted redistribution preferences Wage bargaining level Union members Non-Union members difference n/a n/a Notes: cell entries are predicted probabilities of answering agree and strongly agree to the question It is the responsibility of the government to reduce income differences. 28

29 Appendix. Summary Statistics Variables Mean StdDev. Min Max N Redistribution preferences (1-5 scale) Redistribution preferences (dichotomous) Unemployment policy preferences (1-5 scale) Unemployment policy preferences (dichotomous) Union member Subjective income strata Left-Right ideology Education Gender (female) Age Unemployed Religion (Christian) Wage bargaining level Union density Growth Electoral disproportionality