Union Density and European Integration: Diverging Convergence

Transcription

1 European Journal of Industrial Relations 2000 SAGE (London, Thousand Oaks, CA and New Delhi) ISSN Volume 6 Number 2 pp Sabine Blaschke Universität Wien, AUSTRIA Union Density and European Integration: Diverging Convergence ABSTRACT This article uses multivariate regression analysis to investigate trends in union density in western Europe in the period While countries with union-managed unemployment insurance schemes (Ghent system) saw density rates increase or at least remain stable, the others typically experienced a decline in unionization. Economic factors (shifts in the structure of employment, and also the business cycle and internationalization of trade) exert an influence, though less substantial, on changes in union density. The evidence is ambiguous as to the impact of European economic integration as such on unionization. Introduction European integration exerts pressure for convergence in many areas of economic policy, changing the environment for industrial relations actors at national level. What impact does this have on the interest organizations themselves? This article examines trends in trade union density in all European Union (EU) member states with the exception of Greece and Luxembourg, together with Norway and Switzerland, to establish whether there is evidence of convergence. Union density (membership as a proportion of wage and salary earners) is one indicator of union strength. Others include mobilizing capability, political influence, bargaining power, organizational stability and legal protection. Union density is however the most readily available indicator, often used in judging claims of representativeness, as well as in assessing unions economic impact (Ebbinghaus and Visser, 1999: 135 6). But one must bear in mind that the meaning of union density and its relationship to union power may vary cross-nationally. In France, for example, unions have traditionally put less emphasis on recruiting a stable mass membership, which is one of the explanations for the low rate of unionization (Goetschy, 1998: ). It must also be noted that generating union density data that are comparable over time as well as across [2000/07]6:2; ;013392

2 European Journal of Industrial Relations 6(2) countries rests in part on estimates which may be inaccurate (Visser, 1991: 98 9). The data used in this article are from the Internationalization, Labour Relations, Competitiveness (ILC) project, directed by Franz Traxler, which collected a wide variety of data on industrial relations on 20 OECD countries between 1970 and Specialists from each country completed a standardized questionnaire, which was validated by comparison with existing literature; ambiguities and inconsistencies were discussed with the specialists (Traxler et al., 2000). Economic indicators and employment data were taken from the OECD (Economic Outlook and Labour Force Statistics). Union membership data are based on responses by the country specialists and by the work of Jelle Visser (Ebbinghaus and Visser 1997; Visser 1991, 1999). Convergence and Divergence in Union Density The overall mean of union density in Europe rose in the 1970s and declined during the 1980s. Measures of variation show that there is no increasing similarity, but growing diversity. Previous research has concluded that the presence or absence of a union-managed unemployment scheme (the Ghent system) is the most important determinant of crossnational differences in levels and trends in union density (Ebbinghaus and Visser, 1999; Freeman, 1990; Rothstein, 1992; Western, 1997). On account of this overwhelming evidence, the development of union density was examined separately for Ghent systems and those without (Table 1). While in the Ghent countries the mean of union density increased or remained more or less stable throughout the period covered, in the other countries it generally declined in the 1980s and first half of the 1990s. Within each sub-group, variation is considerably smaller than for all countries together. In the non-ghent countries, variation measured by the standard deviation (but not the variation coefficient) decreased after This decrease must be attributed to the inclusion in the later period of Spain and Portugal, which have to be considered as special cases (see later). Data are therefore presented both including and excluding these two countries (Table 2). Economic Influences Several economic factors are widely regarded as influences on union density and are commonly presented as explanations of the decline in unionization in the majority of European countries. First, growing economic internationalization may be a cause of decline. 218

3 Blaschke: Union Density and European Integration TABLE 1. Union Density a in European Countries Country Ghent Countries Sweden b Denmark b Finland b Belgium b Non-Ghent Countries Norway b Ireland b United Kingdom b Italy b Germany b Switzerland b Austria b Netherlands b France b Spain n.a. n.a b Portugal n.a b a Net density (employed union members as percentage of the labour force in dependent employment) except for Ireland, Portugal and Switzerland where only gross membership (all union members, including unemployed and retired) is available. b In the literature, it is often argued that increased international competition harms unionization by inhibiting wage increases, hardening management attitudes, and intensifying competition among employees, leading to the erosion of solidarity (Griffin et al., 1990: 177; Smith, 1999: 10). 2 High unemployment may reinforce the negative effects on union membership. The empirical evidence from previous multivariate analysis using foreign trade dependence as the indicator of internationalization is mixed (Griffin et al., 1990: 181; Wallerstein, 1989: 490 1; Western, 1997: 189). In the following analysis, three indicators are used: foreign trade dependence, liberalization of financial markets and foreign direct investment (FDI). Unemployment and inflation are often seen as the most important economic influences on union density. According to the business-cycle theory, unemployment has a negative effect on union growth because bargaining power is reduced (see for example Bain and Elsheikh, 1976). Inflation is assumed to exert a positive influence, as workers try to maintain their real wages and employers can afford concessions during boom periods. The evidence on the importance of these factors for explaining 219

4 220 TABLE 2. Indicators of Union Density in Europe All countries Non-Ghent Countries Ghent Countries incl. Spain, Portugal a excl. Spain, Portugal Year Mean SD b CV c Mean SD CV Mean SD CV Mean SD CV European Journal of Industrial Relations 6(2) a Portugal from 1975, Spain from b standard deviation. c coefficient of variation.

5 Blaschke: Union Density and European Integration cross-national differences in union density and growth is however ambiguous (Ebbinghaus and Visser, 1999: 148; Freeman, 1990: ; Griffin et al., 1990: 182 3; Western, 1997: ). Unemployment has risen in Europe, while inflation has declined since the early 1980s. According to the business-cycle theory, this means a decline in union density is to be expected. Union decline may also reflect changes in the structure of the economy. Unionization rates are usually higher in industry and the public sector, and lower in agriculture and private services. In addition, the unionization rates of men, full-time employees and blue-collar workers usually are higher than those of women, part-time employees and white-collar workers (Ebbinghaus and Visser, 1999: 141 2; Visser, 1991: ; Western, 1994: ). In Europe in recent decades, there has been a decline in employment in manufacturing and an increase in private services. The highly organized public sector also has expanded, but especially since 1980 this has not compensated for the decline in industry. To assess the effect on union density of structural change, the employment share of manufacturing and changes in this proportion were included in the analysis. Data and Methodology Yearly growth rates of union density could not be calculated because data are often available only for certain years. Therefore changes in union density were calculated at five-year-intervals. 3 Differences for the more slowly changing variables of employment also were calculated between those years, while those for more strongly fluctuating variables (strike activity, unemployment, inflation) are based on period averages over several years to correct for outliers over time. 4 For the examination of their effect on change in union density, liberalization of financial markets and foreign trade dependence were used as levels (and not as changes), because the degree of openness is hypothesized to have a negative influence on union density. 5 Multivariate regression models (pooled models with results checked in periodwise cross-section models) were calculated to examine the determinants of changes in union density. Most of the economic factors supposed to exert an influence on the development of union density moved (at least over several periods) in the same or in the opposite direction as union density. In a pooled regression model, this will blur the findings on the effects of the independent variables. Therefore period dummies were included, which reflect the overall development and neutralize common trends in dependent and independent variables. Thus the models examine whether the independent variables explain deviations from the period average of the dependent variable. For calculating standard errors, the Beck 221

6 European Journal of Industrial Relations 6(2) and Katz (1995) procedure for deriving panel-corrected standard error (PCSE) estimates was used to correct for the dependency of observation. As well as the specific variables examined, control variables were included in the models. It is not possible to control for all possible explanations in one model, so apart from the period dummies, only the control variables with a substantial influence were retained. In case of correlations between explanatory variables in a model, the coefficient estimate of one variable may be changed by the inclusion of another (correlated) variable. For reasons of space, a comprehensive discussion of shifting coefficients depending on the presence of other variables is impossible here; the most relevant changes will be noted. Inspection of relationships between variables showed that Spain and Portugal have to be considered special cases: after a late restoration of democracy they seem to have been in a transformation phase, with distinctive mechanisms at work. Because of this, dummies for these countries were included as control variables (and show a large effect). Given the evidence of the strong influence of the Ghent system, every model contained the dummy variable for this as a control. Results Using multivariate regression models, the effect of internationalization is tested in Table 3. Foreign trade dependence exerted a small negative TABLE 3. The Effect of Economic Factors on Change in Union Density in Europe (OLS-Regression) a (1) (2) (3) Foreign Trade Dependence b [ 2.30] [ 0.69] [ 2.79] Financial Deregulation c 0.07 [ 0.37] Change in Share of Employment in Manufacturing d [2.76] [2.91] [2.55] Change in Share of Employment in Public Sector d [2.79] [2.68] [3.10] Unemployment Rate e [ 2.41] [ 3.57] [ 2.17] Change in Unemployment Rate e [1.92] [2.32] [2.18] Change in Inflation f [3.08] [3.37] [3.68] Control variables: Ghent System [9.09] [10.04] [8.50] 222

7 Blaschke: Union Density and European Integration TABLE 3. Continued (1) (2) (3) Strikes g 2.91 [3.12] Real wage growth h 0.33 [1.68] Spain [4.83] [5.32] [5.66] Portugal [ 3.47] [ 3.92] [ 4.28] Period [0.48] [0.33] [1.33] Period [ 1.86] [ 1.83] [ 0.68] Period [ 1.18] [ 0.77] [ 0.59] Period [ 0.10] [0.65] [0.81] Constant [0.91] [0.53] [0.14] i (t 1) [ 0.48] [ 0.75] [ 0.17] adj. R N T a Changes are calculated as absolute differences between two values in 5-year intervals: , , , , (unless noted otherwise). Reference category for the period dummies is Data missing for Spain , , Portugal , the Netherlands b Foreign trade dependence: sum of exports and imports as percentage of GDP, period averages for period , for period , for period , for period , for period c Liberalization of financial markets: measure for legal restriction of payments and receipts of capital, goods and invisibles, restriction decreases with score, maximum score: 14 (data set provided by Dennis Quinn), period averages as foreign trade dependence. d Dependent employment only, missing: manufacturing for the Netherlands e Unemployment rate: period averages, as foreign trade dependence. Change in unemployment rate: absolute difference between period values. f Inflation: change of consumer price index, geometric mean for periods, as foreign trade dependence. Change in inflation: absolute difference between period values. g Working days lost per employee, period averages as foreign trade dependence. h Growth of wage rate for the business sector deflated with the deflator for private consumption, geometric mean for periods as foreign trade dependence; missing: Finland and Ireland for , Sweden for and i (t 1) : Coefficient of lagged residuals in regression of OLS residuals on their lags, controlling for all independent variables (Durbin s m test). 223

8 European Journal of Industrial Relations 6(2) influence on union density. The sign of the coefficient for liberalization of financial markets was in line with expectations, but the coefficient too small and insignificant to allow the conclusion that this exerts a negative influence on union density. 6 Figures for FDI (sum of in- and outflows) were available only from 1980 on, with a few countries missing. In the regression models calculated for the cases with available data (not shown in the table), foreign direct investment showed no systematic influence on union density. Unemployment and inflation were first included in the form of levels as well as changes. The level of inflation was omitted later because it had only a minor (positive) influence. The pronounced positive effect of the change in inflation corresponds to the hypothesis. As expected, the level of unemployment had a considerable negative influence, but the positive effect of change in unemployment is not in line with expectations. 7 The coefficient estimates express that persistent high unemployment rates have a stronger negative effect than unemployment rates that have risen recently. 8 However, periodwise regressions show that the sign of the coefficients for change in inflation and unemployment, and the unemployment level, are not stable over time. The coefficients for unemployment in the pooled model are influenced more by the 1980s and 1990s, when unemployment was much higher than in the 1970s. The coefficient for change in inflation is almost entirely influenced by the 1970s, when inflation was rising. When inflation was falling in the 1980s and 1990s, it had no influence on union density; thus increased inflation helps explain union growth in the 1970s, but its reduction is not part of the explanation of the later decline. Although there is some instability in coefficients for unemployment in the 1970s, the results allow the cautious conclusion that the higher unemployment rates of the 1980s and 1990s have contributed to the decline of unionization. It must be noted that the causal relationship between unionization on one hand, and inflation and unemployment on the other, is not clear. Inflation and unemployment may exert an influence on union density; but union density may in turn exert an influence on inflation and unemployment. 9 In line with expectations, decline in manufacturing employment exerted a negative influence on union density. The coefficient is large and significant. The control variable for change in public employment also has a large and significant influence on change in union density. The overall effect of the changes in the employment structure on change in union density was positive in the 1970s (especially between 1970 and 1975), and negative thereafter. Strikes are assumed to stimulate collective organization, therefore time lost per employee and changes in the level are included as control variables. The positive coefficients for both measures of strike activity correspond with the findings of Western (1997: 119) and Griffin et al. (1990: 224

9 Blaschke: Union Density and European Integration 183). The coefficient for change in strike activity is rather small and statistically insignificant, and is therefore omitted from the model shown in Table 3. There is a negative correlation between strike activity and foreign trade dependence, hence the inclusion of strike activity reduces the coefficient for foreign trade dependence. Another important variable is the growth in real wages: the reasoning is that unions that are effective in wage bargaining are more successful in attracting members (Bain and Elsheikh, 1976: 64). As predicted, the growth in the real wage rate exerts a positive but weak influence in the pooled model. 10 If only the 1980s and 1990s are considered, real wage growth has a more significant influence. Periodwise regression shows that this already was the case in the late 1970s. The regression coefficients of the other economic factors remain unchanged when strike activity or real wage growth is controlled for in the pooled model, except foreign trade dependence when strike activity is included. As with inflation and unemployment, a reciprocal action between strike activity or the change in the real wage rate and the change in union density may exist. The results from multivariate regression suggest that an important cause of the decline in unionization in the majority of west European countries in the 1980s and 1990s was the change in the employment structure. Higher unemployment and foreign trade dependence seem to have contributed further to this decrease. Institutions and Union Growth In the analysis of industrial relations systems, persisting institutional diversity across countries has shown that institutions, once established, develop according to their own logic, which in turn leads to different trajectories of industrial relations (Ferner and Hyman, 1992: xxxiii). The following paragraphs explore how far the institutional context of industrial relations and the characteristics of union organizations at national level shape changes in union density. The focus is on institutions that may exert influence on recruitment or retention of union members. First, selective incentives are important because they reduce the problem of free-riding. Such a selective incentive is union-administered unemployment insurance: the Ghent system. Usually it is not necessary to be a union member, but the connection between union membership and unemployment insurance, as well as the regular contact with the union, can be seen as strong incentives for employees to become union members and to stay in the union when unemployed (Scheuer, 1992: ). Another factor that could have a positive effect on unions is the institutionalization of employee representatives, which can provide union 225

10 European Journal of Industrial Relations 6(2) access to the workplace (Ebbinghaus and Visser, 1999: 143 4). Here, one can distinguish between works councils (including joint committees) established by law, and union structures. The latter may be established by collective agreements alone or be backed by legislation. In several countries, more than one system exists. In contrast to works councils and union bodies backed by legislation, the existence of other union bodies depends in most cases on the power of the union to authorize them through collective agreements or to make employers tolerate their activities. Strong unions may be a precondition for legislation on statutory employee representation; but once such laws are in force, they take effect independently of the power of the unions. Therefore to examine the effects of institutions on union membership, statutory bodies (works councils and union representative mechanisms) are more appropriate than union bodies established solely by collective agreement. Union security arrangements such as the closed shop are also assumed to foster unionization (Ebbinghaus and Visser, 1999: 147 8; Olson, 1965: 76 85). Again, a reciprocal relationship is possible: high union density may be necessary to include such arrangements in collective agreements. Another potential influence is the bargaining order (single-employer bargaining versus inclusive, multi-employer bargaining), which is commonly assumed to influence employers attitudes towards unionization. According to the theory of collective action (Olson, 1965), encompassing domains and high centralization should exert a negative influence on unionization. For the individual employee, the propensity to unionize should be higher when there is a union that is tailored to his/her situation, needs and attitudes. Hence small specialized unions are expected to have more success in attracting and retaining members than large unions. With growing centralization of internal decision-making, the individual member s influence decreases, which may prevent potential members from joining. Another line of reasoning however suggests that unified or centralized union movements are more capable of recruiting members because they can coordinate the organizing effort, and use resources more efficiently, while competition among unions is reduced (Chaison, 1997: 20 1; Western, 1997: 31). In the following analysis, the number of union confederations is used as the measure of union differentiation. The governability of the union confederations (formal power vis-à-vis direct affiliates) is the indicator for centralization. Centralization measures are available only for the largest union confederation, which is assumed to be representative of the other confederations. Some authors contend that union participation in socio-economic policy-making influences unionization positively (Armingeon, 1989: ). The argument is that a union that has more success in achieving its goals by exerting influence on public policy attracts more members. However, a counter-argument is that this influence is a public 226

11 Blaschke: Union Density and European Integration good providing no incentive to join (Hassel, 1999: 17 18). On the other hand, the union can generate selective incentives from its participation in public policy (e.g. establishing and sustaining a Ghent system) to its benefit. In order to examine this assumption, the activities of the largest union confederation in socio-economic policy-making are taken as a measure of union participation. Data and Methodology For the analysis of institutions and change in union density, the Ghent system, specific systems of employees workplace representation and closed-shop arrangements were again coded as dummy variables. Many institutional variables remain constant over the entire period of investigation. To take into account time lags in case of alterations, period averages (mean or mode according to the level of measurement) were used. The control variables were enlarged by adding those economic factors that proved to be most important. In order not to overburden the models, financial deregulation was not included because it contributed almost nothing to the explanation of variance. Furthermore, real wage growth was not included because several cases were missing. Strike activity was omitted because of collinearity with foreign trade dependence. Though it contributed more to the explanation of variance, strike activity was excluded instead of foreign trade dependence, because the latter is among the central variables of the analysis and therefore was more important to control for when examining the influence of institutions. Results As already mentioned, the Ghent system exerts a large positive influence on change in union density, which is confirmed by the results from the multivariate regression models (Tables 3 and 4). Over time, the Ghent system has retained its importance, so European integration does not diminish its differentiating influence. Surprisingly, the results from multivariate regression suggest that statutory employee representation does not exert a positive influence on union development. Indeed when union bodies backed by legislation are excluded, the existence of works councils (including joint committees) has a pronounced negative influence. 11 Contrary to the fact that national studies often view works councils as securing union access to the workplace and playing a key role in recruiting members, comparative analysis shows that their existence does not provide a recruiting advantage for national union movements (though in other respects they may enhance union strength). There are several possible explanations for this unexpected result. First, 227

12 European Journal of Industrial Relations 6(2) TABLE 4. The Effect of Institutions on Change in Union Density in Europe (OLS-Regression) a (1) (2) (3) (4) (5) (6) Ghent System [7.96] [8.82] [8.86] [7.46] [7.53] [6.56] Statutory Employee 0.63 Representation b [ 0.66] Statutory Works Council c 1.48 [ 2.49] Statutory Works Council 0.47 without Pervasive [ 0.58] Extension Practice d Statutory Works Council 1.99 plus Pervasive Extension [ 3.13] Practice d Closed Shop under Single Employer Bargaining e,f [2.39] Closed Shop under Multi Employer Bargaining e,f [0.64] Abolition of Closed Shop e 2.98 [ 2.03] Single Employer Bargaining 0.21 without Closed Shop f [0.10] Union characteristics: Number of Union 0.07 Confederations [0.23] Governability of the 0.21 Largest Union [ 0.85] Confederation g Control Variables: Foreign Trade Dependence [ 1.92] [ 0.99] [ 1.02] [ 2.12] [ 0.73] [ 0.74] Change in Share of Employment in [2.65] [2.16] [2.17] [2.53] [2.46] [2.69] Manufacturing Change in Share of Employment in Public [2.81] [2.69] [2.79] [2.20] [2.62] [2.97] Sector Unemployment Rate [ 2.31] [ 2.13] [ 1.78] [ 1.89] [ 2.65] [ 3.28] Change in Unemployment Rate [2.15] [2.13] [1.90] [1.31] [1.85] [2.08] Change in Inflation [3.55] [3.48] [3.30] [4.12] [3.37] [3.44] 228

13 Blaschke: Union Density and European Integration TABLE 4. Continued (1) (2) (3) (4) (5) (6) Spain [5.22] [5.97] [6.14] [4.99] [5.10] [5.46] Portugal [ 4.18] [ 3.74] [ 3.14] [ 3.87] [ 3.02] [ 3.89] Period [0.65] [0.38] [0.18] [0.98] [0.45] [0.56] Period [ 1.72] [ 2.09] [ 2.12] [ 1.28] [ 1.88] [ 1.75] Period [ 1.08] [ 1.32] [ 1.41] [ 0.64] [ 1.23] [ 1.00] Period [ 0.18] [ 0.46] [ 0.55] [0.45] [ 0.31] [0.01] Constant [1.24] [1.21] [1.15] [1.06] [0.56] [1.39] h (t 1) [ 0.65] [ 0.90] [ 1.20] [ 0.58] [ 0.43] [ 0.52] adj. R N T a Data and periodization as in Table 3. Regression coefficients from pooled OLS analysis, t-values in brackets, calculated with PCSE (Beck and Katz, 1995). b Statutory employee representation (dummy variable): Austria, Belgium, France, Germany, Italy, the Netherlands (from 1980), Portugal (from 1975), Sweden (from 1975), Finland, Spain (from 1980); in case of change it is assumed that there was a time lag in the effect on union density of at least two years, therefore Switzerland where statutory works councils exist from 1994 is coded 0 for the period c Statutory works council (dummy variable): Austria, Belgium, France, Germany, the Netherlands (from 1980), Portugal (from 1975), Spain (from 1980 on). d Pervasive extension practice (dummy variable): Austria, Belgium, Finland, France, Portugal (from 1975), Spain (from 1980). e Closed shop: Ireland, the United Kingdom until 1986 (dummy variable, coded 1 when closed shop or similar arrangements occur more than infrequently). Abolition of closed shop (dummy variable): the United Kingdom f Single/Multi-employer bargaining (dummy variable): single-employer bargaining coded 1 for the United Kingdom, Ireland (only ), other countries multiemployer bargaining coded 1 (includes combination of single-employer bargaining and industry bargaining with domains overlapping). Single-employer bargaining without closed shop (dummy variable): the United Kingdom g Governability of the largest union confederation: formal power of the largest confederation vis-à-vis its direct affiliates, scale from 0 7, governability increases with score. h (t 1) : Coefficient of lagged residuals in regression of OLS residuals on their lags, controlling for all independent variables (Durbin s m test). 229

14 European Journal of Industrial Relations 6(2) there are usually legal and, even more important, practical thresholds for employee workplace representation systems. Statutory (like voluntary) systems are more common in larger firms, in the public sector and in manufacturing (Ebbinghaus and Visser, 1999: 144). Structural changes in the economy (such as the decline of large industrial firms) reduce these areas, while in the growth sectors (such as private services) it may be difficult to establish works councils even if they are mandated by labour law. 12 Unions that have been able to rely on works councils may be less prepared for these shifts than others because they have neglected to develop their own recruitment structures. Second, it could be that works councils (which usually are at least formally independent from unions) may give employees the feeling that unions are not that necessary any more. This may be reinforced by the fact that in countries with works councils established by labour law, multi-employer bargaining is predominant and pervasive; or at least, extension practices exist. So there is less need for union activity at the plant level to make an employer comply with the collective agreement. Employees may even be unaware that unions are necessary to negotiate the collective agreements. To check for this assumption, works council legislation was differentiated by the existence of moderate or pervasive extension practices. The combination with pervasive extension practices was shown to exert a negative influence on the change in union density. There is some evidence that the negative influence of pervasive extension practices and works councils was greatest in the 1970s, rather small or even nil in the 1980s, and somewhat larger again in the early 1990s. In Europe, closed-shop clauses are infrequent except in Ireland and (until the late 1980s) the United Kingdom, where the closed-shop clauses were increasingly legally restricted from 1980 and prohibited in 1990 (for this analysis, it was decided to regard the United Kingdom between 1985 and 1990 as affected by the change in legislation). Contrary to expectations, the Irish case showed that closed-shop arrangements in a multiemployer bargaining environment need not favour union growth (see also Ebbinghaus and Visser, 1999: 153). For the United Kingdom, a positive effect of the closed shop can be observed between 1975 and 1980, while its abolition exerted a negative influence. The small number of cases prevents a general conclusion, but the results indicate that closed-shop practices do not prevent union decline. With the absence of control variables, union encompassingness exerted a positive influence, and centralization a negative influence. This is in line with the hypotheses derived from the theory of collective action. But as soon as the Ghent system is controlled for, the influences become smaller and rather insignificant. Because of multicollinearity, an assessment of the influence of union participation in public policy is not possible. Union participation in socio-economic policy-making does not exert an influence 230

15 Blaschke: Union Density and European Integration when change in public employment is controlled for, but exerts a positive influence in line with expectations when change in public employment is excluded. The size of the coefficient is increased when the variable for the Ghent system is also omitted from the model. Separate analyses for Ghent and non-ghent countries show that with exclusion of change in public employment, union participation had a positive influence only in those countries without the Ghent system in the 1970s. Among all institutional variables, the recruiting assistance derived from the Ghent system is the one that has a large and stable influence. Among all the determinants of union density examined, it contributes most to the explanation of variance. Surprisingly, statutory works councils and pervasive extension of collective agreements seem to have exerted a certain negative influence on union growth, which lessened during the 1980s. Conclusion: A Specific European Development? In Europe, the existence or non-existence of a union-administered unemployment scheme, the Ghent system, leads to two distinct paths of development in union density. While both (albeit to a different extent) showed increasing density in the 1970s, afterwards the non-ghent countries generally experienced a decrease in density, while the Ghent countries increased or at least stabilized their density rates. The Ghent system exerts a strong positive impact on union growth, and so far European integration has not weakened its influence on cross-national divergence. The findings from multivariate regression analysis show that in addition to the Ghent system, economic factors (especially shifts in the structure of employment, but also the business cycle and trade penetration) exert an influence on union density. The question now is whether these developments, and thus their influence on union density, can be attributed to European integration. A similar question arises in connection with the impact of the period dummies which were included in the regression models as control variables. Along with other factors, they may contain aspects of European integration that are not captured by other independent variables. European integration has to be seen as a continuous process. Nevertheless, some important steps can be identified: the Single European Act of 1987, the Treaty on European Union in and the completion of the Single Market in Thus for the periods from and , effects of European integration on union density should be most visible. The cross-country averages of the economic indicators with a negative influence on union density (decline of manufacturing, unemployment, foreign trade dependence) started to rise or fall in the period, if not earlier (for manufacturing). In addition, the change in inflation, which 231

16 European Journal of Industrial Relations 6(2) exerted a positive influence in the 1970s, started to become negative in Also in the period , the positive influence of public sector growth no longer compensated for the loss of employment in manufacturing (the stagnation of public employment growth started later). Thus most of the pronounced changes in the economic determinants of union density took place earlier than expected. This does not support a clear-cut causal relationship between European integration and these changes, and thus between European integration and the negative impact of economic factors on union density. Turning to the period dummies of the multivariate pooled regression models, it must be stated that their coefficients were positive in the 1970s, negative in the 1980s and (depending on the value of the constant of the various models) positive or negative in the early 1990s. As with the economic indicators, the chronological order impedes establishing a clear-cut relationship between elements of European integration possibly contained in the period dummies and the decline in union density. A comparison between European and non-european OECD countries could also shed some light on whether the findings on trends in union density and their determinants are specific for Europe and thus could be attributed to European integration. Analysis of five non-european countries in the ILC data set (Australia, Canada, Japan, New Zealand and the United States) show that among them, employment structure and inflation changes exerted an influence on union density as well, but foreign trade dependence and unemployment did not. There are some differences in the development of economic determinants of union density between European and non-european countries. On average, foreign trade dependence is much larger in Europe than in non-european countries, with the gap widening in the 1970s and 1980s. Further, the decline in manufacturing was on average somewhat larger in Europe in most periods particularly in and and unemployment was higher after 1980, particularly in the early 1990s. These differences, in combination with the non-relevance of some of the determinants in the group of non- European countries, would suggest that union decline caused by economic factors was larger in Europe than in the non-european countries. This could be partially compensated by the growth in public employment, which was larger in Europe, particularly from and The comparison of union growth (and decline) between European and non-european countries in the ILC data set shows no major differences. In the 1970s, European countries experienced higher union growth, but as soon as the Ghent system and the change in public employment are controlled for, most differences diminish: there is only a small, insignificant negative coefficient for Europe after Thus the differences in the negative economic influences on change in union density between European and non-european countries are not large enough to establish a distinct 232

17 Blaschke: Union Density and European Integration European pattern. Finally, the period dummies could contain some specific European developments, but in models calculated separately for European and non-european countries, the coefficients for the period dummies are very similar. It is possible that economic developments, as well as other developments contained in period dummies, are caused by different factors in European and non-european OECD countries, and therefore can be attributed to European integration. In sum, European integration could have exerted some negative influence on union density, but no definite conclusion can be derived from the results. ACKNOWLEDGEMENTS I thank Franz Traxler, Richard Hyman and Bernhard Ebbinghaus for useful comments on an earlier version of this article and Bernhard Kittel for useful comments as well as providing the procedure for calculating the PCSE of the pooled regression models. NOTES 1 Among the countries without a Ghent system, the following deviate from the average pattern of increase in the 1970s and subsequent decline: Austria (constant decline), Norway (stability after 1980), the Netherlands and Switzerland (stability after 1990), Spain (constant increase). Among the Ghent countries, Belgium (rise and decline) deviates from the pattern of increase and stability. 2 Note however that Wallerstein (1989: 490) argues that foreign trade dependence exerts a positive influence on union density. He derives this assumption from Katzenstein (1985: 32 3), who suggests that trade dependence encourages a corporatist institutional framework. 3 In most cases, differences were calculated for , , , , For Germany, differences were corrected to reflect the estimated effect of unification. As there are large differences between the country-specific levels of union density, percentage changes were considered; but this created extreme outliers, therefore absolute differences were used. 4 For 1970, the period averages were used; for 1975, the period averages ; for 1980, of ; for 1985, of ; for 1990, of ; and for 1995, of Alterations in liberalization of financial markets and foreign trade dependence do not have an immediate effect on union density, therefore period averages were used. 6 Because of collinearity between liberalization of financial markets and other variables, in models which do not control for the Ghent system and other economic factors, the negative regression coefficient for the liberalization of financial markets is significant. 7 In periodwise regression models without change in unemployment, the coefficient for the level of unemployment is smaller and less significant. 233

18 European Journal of Industrial Relations 6(2) 8 In the small subgroup of Ghent countries, the direction of the coefficients for unemployment and its change are the same as in the overall model. This is not in line with the results of Freeman (1990: ). The main reason for the difference in findings seems to be that Freeman uses gross densities for his calculations, while the results in this article are based on net densities. 9 On the assumption that effects lag at least one year, an examination of annual changes in union density and the accompanying lagged changes/levels of inflation and unemployment could shed some light on which comes first. In the ILC data set, this approach was not possible because yearly values of union density are available only for some countries. In a similar analysis of change in union density, Griffin et al. (1990: 183) could account for possible reciprocal action in their regression models. Their estimates suggested that even if changes in unemployment and inflation are both cause and consequence of changes in union density, that did not unduly bias their results, according to which change in unemployment and inflation exerted an effect on change in union density. 10 Because of collinearity, the effect of real wage growth becomes much larger and more significant when inflation is added to the pooled model. 11 In periodwise regressions, the coefficient for the works council variable sometimes changes its sign (depending on the other variables in the model); but when only the Ghent system, Spain and Portugal are controlled for it is always negative. 12 In the multivariate models, these sectoral changes should be captured to a certain extent by the variable for the change of the share of dependent employment in manufacturing. REFERENCES Armingeon, K. (1989) Arbeitsbeziehungen und Gewerkschaftsentwicklung in den achtziger Jahren: Ein Vergleich der OECD-Länder, Politische Vierteljahresschrift 30: Bain, G.S. and Elsheikh, F. (1976) Union Growth and the Business Cycle. An Econometric Analysis. Oxford: Blackwell. Beck, N. and Katz, J. (1995) What to Do (and Not to Do) with Time-Series Cross-Section Data, American Political Science Review 89: Chaison, G.N. (1997) Reforming and Rationalizing Union Structure: New Directions and Unanswered Questions, in M. Sverke (ed.) The Future of Trade Unionism: International Perspectives on Emerging Union Structures, pp Aldershot: Ashgate. Ebbinghaus, B. and Visser, J. (1997) When Labour Institutions Matter: Union Growth and Decline in Western Europe, , paper to the European Sociological Association Conference, 2 August. Ebbinghaus, B. and Visser, J. (1999) When Institutions Matter. Union Growth and Decline in Western Europe, , European Sociological Review 15: Ferner, A. and Hyman, R. (1992) Introduction: Industrial Relations in the New Europe, in A. Ferner and R. Hyman (eds) Industrial Relations in the New Europe, pp. xvi xlix. Oxford: Blackwell. 234

19 Blaschke: Union Density and European Integration Freeman, R. (1990) On the Divergence of Unionism among Developed Countries, in R. Brunetta and C. Dell Aringa (eds) Labour Relations and Economic Performance. Proceedings of a Conference held by the International Economic Association in Venice, Italy, pp New York: New York University Press. Goetschy, J. (1998) France: The Limits of Reform, in A. Ferner and R. Hyman (eds) Changing Industrial Relations in Europe, pp Oxford: Blackwell. Griffin, L.J., McCammon, H.J. and Botsko, C. (1990) The Unmaking of a Movement? The Crisis of US Trade Unions in Comparative Perspective, in M.T. Hallinan, D.M. Klein and J. Glass (eds) Change in Societal Institutions, pp New York: Plenum Press. Hassel, A. (1999) Gewerkschaften und sozialer Wandel. Mitgliederrekrutierung und Arbeitsbeziehungen in Deutschland und Großbritannien. Baden-Baden: Nomos. Katzenstein, P.J. (1985) Small States in World Markets. Industrial Policy in Europe. Ithaca, NY: Cornell University Press. Olson, M. (1965) The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press. Rothstein, B. (1992) Labor-Market Institutions and Working-Class Strength, in S. Steinmo, K. Thelen and F. Longstretch (eds) Structuring Politics: Historical Institutionalism in Comparative Analysis, pp Cambridge: Cambridge University Press. Scheuer, S. (1992) Denmark: Return to Decentralization, in A. Ferner and R. Hyman (eds) Industrial Relations in the New Europe, pp Oxford: Blackwell. Smith, R. (1999) The Convergence/Divergence Debate in Comparative Industrial Relations, in M. Rigby, R. Smith and T. Lawlor (eds) European Trade Unions. Change and Response, pp London: Routledge. Traxler, F. (1994) Collective Bargaining: Levels and Coverage, OECD Employment Outlook: Traxler, F., Blaschke, S. and Kittel, B. (2000) National Labour Relations in Internationalized Markets. A Cross-National Analysis of Institutional Patterns, Change and Performance. Oxford: Oxford University Press. Visser, J. (1991) Trends in Trade Union Membership, OECD Employment Outlook: Visser, J. (1999) Societal Support for Social Dialogue. Europe s Trade Unions and Employers Associations, in G. Huemer, M. Mesch and F. Traxler (eds) The Role of Employer Associations and Labour Unions in the EMU. Institutional Requirements for European Economic Policies, pp Aldershot: Ashgate. Wallerstein, M. (1989) Union Organization in Advanced Industrial Democracies, American Political Science Review 83: Western, B. (1994) Institutional Mechanisms for Unionization in Sixteen OECD Countries: An Analysis of Social Survey Data, Social Forces 73: Western, B. (1997) Between Class and Market: Postwar Unionization in the Capitalist Democracies. Princeton, NJ: Princeton University Press. 235

20 European Journal of Industrial Relations 6(2) SABINE BLASCHKE is a sociologist and currently lecturer and research fellow at the University of Vienna. ADDRESS: Betriebswirtschaftliches Zentrum, Universität Wien, Bruenner Straße 72, A-1210 Vienna, Austria. [ sabine.blaschke@univie.ac.at] 236