Accessible Published by De Gruyter Oldenbourg July 2, 2016

A Sequential Decomposition of the Drop in Collective Bargaining Coverage

Bernd Fitzenberger and Katrin Sommerfeld

Abstract

Union representation has been in strong decline in most OECD countries with potentially important consequences for wages. What drives this decline? We try to answer this question by developing and implementing a detailed Fairlie decomposition approach. Using linked employer-employee data from the German Structure of Earnings Survey for 2001 and 2006, we document a sharp drop in collective bargaining coverage that amounts to 17 percentage points for males and 20 percentage points for females in West, and 8 and 14 percentage points, respectively, in East Germany. We find that neither changes in the characteristics nor changes in the coefficients associated with the characteristics as a whole provide an explanation for the drop in collective bargaining coverage. The drop in coverage is the result of an unexplained time trend.

JEL: C21; J51; J52

1 Introduction

Union representation has been in strong decline in most OECD countries (Visser 2006; Lesch 2004; OECD 2004; Card et al. 2003; Schnabel 2013). It is interesting to see that the same trend affects countries with very different institutional set-ups such as the USA, Canada, the UK, and Germany. [1] This trend may have important consequences. In particular, it may result in lower wage levels and growing wage inequality because unions tend to compress the wage distribution from below. [2] Germany is an interesting case to study because it underwent a remarkable transition from being considered as the “sick man” of Europe to experiencing a “jobs miracle” (European Commission 2002; Krugman 2009; Möller 2010; Dustmann et al. 2014). In Germany, collective agreements dominate the entire process of setting working conditions – not only for union members.

Collective wage agreements in Germany define minimum working standards not only for union members but typically for all employees of a covered establishment. Therefore, coverage rates are much higher than membership rates and more relevant for assessing the range of union representation in the labour market. Just as union membership has dropped, also the share of employees covered by a collective bargaining agreement has dropped sharply. [3] According to Ellguth and Kohaut (2004, 2014) this share dropped from 70 % in 2003 to 60 % in 2013 (also see Addison et al. 2010, Antonczyk et al. 2010). The strongest decline in coverage on a year-to-year basis seems to have occurred during the first half of the 2000s which is the period we investigate in this paper. Our study analyses the trend during the early 2000s by means of a statistical decomposition analysis. Can changes in the composition of the workforce explain this trend (i. e. a characteristics effect)? Or were there large changes in the propensity to work in a covered establishment for certain employee groups (i. e. a coefficients effect)? Or is this a trend that affects all groups of employees equally?

A large part of the variation in collective bargaining coverage is explained by sector affiliation (Fitzenberger et al. 2011; Antonczyk et al. 2010). Thus, the currently observed sectoral shift from manufacturing to services could potentially drive collective bargaining coverage down. Moreover, collective bargaining coverage varies substantially with firm size (Biebeler and Lesch 2007; Fitzenberger et al. 2011). Furthermore, age and tenure of an individual employee are positively related with the likelihood of working in a covered establishment. Educational upskilling could reduce union representation over time if higher skilled employees tend to work in non-covered establishments. Further determinants of union membership include risk aversion (Goerke and Pannenberg 2012) and political attitudes (Biebeler and Lesch 2007; Fitzenberger et al. 2011; Schnabel and Wagner 2007).

A few studies have decomposed the drop in union representation in Germany over time, as measured by the drop in union membership (Fitzenberger et al. 2011; Addison et al. 2011; Schnabel and Wagner 2007). Following the decomposition approach for limited dependent variables introduced by Fairlie (2005), they conclude that changes in the composition of the workforce are not the main driving force of the drop in membership. Rather, the residual effect looms large which is due to the changing associations between the covariates and collective bargaining coverage as well as the impact of unobservables.

Our analysis tries to determine the mechanism through which the reduction in coverage operates in a statistical sense. This paper adds to the literature by developing a detailed decomposition approach and applying it to decompose changes in collective bargaining coverage rather than in union membership over time. This approach allows separating out the effects of first individual characteristics, second firm characteristics and third industry branch. Moreover, within the residual effect the method allows separating the effect due to changes in the different coefficients from unexplained changes over time. This approach is based on Fairlie (2005) and extends it in order to sequentially distinguish the contribution of certain groups of characteristics. This is similar to the approach of Antonczyk et al. 2010, 2009) or more generally of DiNardo et al. (1996) for the case of continuous dependent variables. To the best of our knowledge, our study is the first to sequentially decompose the drop in union representation.

The second contribution lies in the analysis of union coverage instead of union membership. We use the German Structure of Earnings Survey (GSES), a large and reliable linked employer-employee data set provided by the Research Data Centres of the German Statistical Office. Previous studies on Germany have often used either ALLBUS or SOEP data for union membership or the IAB establishment panel for firm-level analyses. The ALLBUS and the SOEP do not provide any information on union coverage at the individual or firm level. The IAB establishment panel involves union coverage at the establishment level, information which is self-reported by a firm representative and which has been used in a number of studies for Germany. The GSES data also provides information on coverage at the establishment level, which is arguably more precise than the information contained in the IAB establishment panel, because the GSES is based on personnel records of the establishment sent directly to the Federal Statistical Office. In this study, almost 1,500,000 full-time employees will be analysed for the years 2001 and 2006. We argue that this time frame is interesting because it is the time during which the German labour market underwent a strong transformation and labour market conditions improved considerably. During the same time, wage inequality increased strongly and the drop in coverage was particularly strong (as measured on an annual basis). Moreover, considering the time period from 2001 to 2006 leaves out potential effects from the great recession starting in 2007.

Admittedly, the GSES comprises less information at the individual level or at the establishment level compared to some of the aforementioned data sets, respectively. Union membership may be associated with employee characteristics (Fitzenberger et al. 2011; Addison et al. 2011; Schnabel and Wagner 2007), which are not recorded in the GSES. The decline in union membership precedes the decline in union coverage, however, there may be a relation between both. For instance, unions may find it harder to organize new establishments if membership is low. Thus, a decline in membership may result in a decline of coverage subsequently. Establishments under distress may opt to drop out of collective bargaining agreements (Addison et al. 2010; Dustmann et al. 2014). Thus, a drop in coverage may be associated with indicators of the development of the establishment over time (e. g. profits, firm growth), and a weak labour market (e. g. as measured by the local unemployment rate) may be a general indicator of distress (Gürtzgen 2016). Thus, an analysis based on the GSES should be viewed as a complement rather than a substitute, on the one hand, to the analysis of union membership based on the SOEP or ALLBUS and, on the other hand, to a panel analysis of union coverage based on the IAB establishment panel (possibly augmented by information on individual employees from social security records). The advantages of our analysis based on the GSES are the large sample size and the highly reliable information provided which are based on personnel records of the establishments. The large sample size of the GSES allows for a detailed sequential decomposition analysis.

Our results based on the GSES show that collective bargaining dropped sharply over the period from 2001 to 2006. While for male employees in West Germany, the drop in the share of employees who work in a covered establishment amounts to about 17 percentage points (ppoints), for West German females the drop amounts to nearly 20 ppoints! For East Germany, the drop in collective bargaining amounts to 8 ppoints for male and 14 ppoints for female employees. The decomposition results clearly show that only a minor part of the drop in collective bargaining coverage can be explained by the characteristics or their corresponding coefficients, in both West and East Germany. We interpret the complex pattern in East Germany as the results of the ongoing structural adjustment process on-site. Both for West and East Germany, the drop in coverage is the result of an unexplained time trend. This means that the drop in collective bargaining coverage is not confined to certain industry sectors, to firms of a certain size nor to certain educational groups.

The rest of this article is structured as follows: The next section briefly explains the German institutions concerning collective bargaining. Next, Section 3 develops the methodology starting out from the existing decomposition for limited dependent variables and extends it to a sequential decomposition for our case. Then, Section 4 describes the data used and descriptive statistics. Section 5 presents the results. Finally, Section 6 concludes. An additional online appendix provides further details on the empirical analysis. [4]

2 Institutional background and literature

Collective bargaining agreements in Germany are generally negotiated between an employers’ association and a union. As an alternative to these forms of collective negotiation, employers and employees can negotiate individual contracts. When a collective agreement is achieved, it applies to all firms that form part of the corresponding employers’ association and that operate in the relevant sector and region. On the side of the employees, legally, the collective contract only needs to be applied to union members. However, it is very common that employers pay all employees according to the collective contract. [5] This is because employers want to reduce negotiation costs and to reduce the incentive to become a union member (Fitzenberger et al. 2011). For this reason, collective bargaining coverage is much higher and more relevant than union membership in Germany (Fitzenberger et al. 2013).

The bargaining process can take place at the sectoral or at the firm level, so as to reach more or less centralised results. Sectoral agreements apply to all establishments in the corresponding sector and region and may have to accommodate very different firms in terms of e. g. size and profitability. Meanwhile firm-level agreements can be adjusted much more specifically to the single employer. For this reason, it was expected that the firms’ need for flexibility might lead to a situation in which the drop in sectoral bargaining is accompanied by a rise in firm bargaining. However, this could not be confirmed empirically (Antonczyk et al. 2010). Rather, the specific literature finds a stagnation or only a small drop in firm-level bargaining, for both West and East Germany (Ellguth/Kohaut 2004, 2005, 2007, 2011, 2012, 2013, 2014). Therefore, and because both types of collective bargaining are rather similar, we will not further differentiate between these two types of agreements.

As a matter of course, employers are always free to pay higher wages or premia than the collective agreement requires them to (“favourability principle” or “Günstigkeitsprinzip” Bosch 2004). But they may not undercut the collective agreement. [6] In this sense, collective agreements define minimum working standards for all employees working in a covered establishment (also see Fitzenberger et al. 2013). For this reason, our measure of collective bargaining coverage will reflect whether or not an employee works in an establishment that is covered by collective bargaining (as in Antonczyk et al. 2010). [7]

Finally, the contracts of individually contracted employees often explicitly or implicitly refer to a collective agreement. In other words, some firms that are not part of an employers’ association and for which the application of a collective agreement is not binding may still use a collective agreement as a benchmark in their wage setting (“Bezugnahme-Klausel” Hold 2003: 478). Although the application of a collective contract comes into effect voluntarily from the employer’s side, the collective contract might turn to be legally binding under certain conditions (Hold 2003). Regarding those employees who work in establishments that are not directly covered by collective bargaining, about half of them work in establishments that still use the collective contract for orientation (Ellguth/Kohaut 2004, 2005, 2007, 2011, 2012, 2013, 2014; Addison et al. 2015). For the reasons discussed here, coverage by collective agreements is likely to exceed membership in an employers’ association (Schnabel and Wagner 1996).

While this “orientation” towards a collective agreement is of large relevance among German employers, it is naturally difficult to find any precise numbers on its distribution. These numbers are not recorded in the data set that we use. What we will focus on later in this study is whether or not an employee works in an establishment that is covered by collective bargaining.

What is the extent of union coverage in Germany? Ellguth and Kohaut (2004, 2007, 2014) report a share of covered employees of 70 % for West Germany for the year 2003 which dropped to 65 % in 2006 and further declined to 60 % in 2013. For East Germany, the coverage rate on the employee level was at 54 % in 2003 and 47 % in 2013 (Ellguth and Kohaut 2004, 2014). Their results are based on the IAB establishment panel and differences to the results obtained from the GSES data are probably due to different response behaviour or different data selection. Based on the same data set, Addison et al. (2010) report a drop of coverage at the employee level from 64.1 % in 2000 to 55.8 % in 2008 for the whole of Germany.

Naturally, data on coverage at the establishment level reveals very different shares because coverage is strongly related to the size of the establishment. Coverage at the establishment level was reported to be at 48.1 % in West Germany in 2000 (Schnabel et al. 2006). When establishments with only ten or more employees are considered, this share increases to 61.7 % (ibid.). Meanwhile, Addison et al. (2011) report a drop from 62.5 % in 1998 to 51.1 % in 2004, whereas Addison et al. (2010) report coverage shares of 49.9 % in 2000 and 38.1 % in 2008 for entire Germany. Once again, it seems as if different data sets obtain different coverage shares but they uniformly describe a clear drop. This very sharp drop is to be explained by the following decomposition approach.

A similar picture emerges based on union membership rates which are at a much lower level due to the institutional set-up in Germany as explained above. Fitzenberger et al. (2011) report membership rates of 29.9 % in 1985, 26.7 % (37.3 %) in 1993 and 20.0 % (17.5 %) in 2003 for West (East) Germany. Based on a different data base, Schnabel and Wagner (2007) reports 32.7 % in 1980, 28.7 % in 1992, and 21.7 % in 2004 for West Germany. For Germany as a whole, Schnabel (2013) reports 24.6 % for 2000 and 18.6 % for 2010. Irrespective of the data source, all studies confirm a strong drop in union membership.

What explains the drop in union representation? The reduction in the employment rates of full time employees, males, and blue collar workers was expected to lead to a reduction in union membership (Schnabel and Wagner 2007). Based on a similar composition argument, the shift in the industry structure towards the service sector was expected to go along with lower union representation because establishments in the service sector are less frequently covered by collective bargaining than in the manufacturing sector (Hassel 2007). Moreover, firms which become more exposed to international competition may have a growing need for flexibility and might therefore leave the system of collective bargaining (Kohaut and Bellmann 1997; Bosch 2004). However, in contrast to these considerations, recent empirical studies do not confirm the expected role of changes in the composition of the workforce (Schnabel/Wagner 2007; Fitzenberger et al. 2011; Addison et al. 2011). The two decomposition analyses by Fitzenberger et al. (2011) and Schnabel and Wagner (2007) show that changes in the composition of the workforce – as captured by the characteristics effect in a decomposition following Fairlie (2005) – explain little or hardly anything about the drop in union membership. Instead, Schnabel and Wagner (2007) attribute more than 90 % of the drop in union membership to a residual effect which they do not interpret any further. We are aware of only one study that decomposes changes in collective bargaining coverage in Germany which is carried out by Addison et al. (2011). However, this study uses an Oaxaca (1973) and Blinder (1973) type of decomposition which ignores the non-linearity of the dependent variable. The authors find that changes in the coefficients fully explain the drop in collective bargaining coverage and interpret these as behavioural changes.

The present study adds to this literature in two ways. First, we analyse union coverage rather than union membership because this may be the more relevant measure for outcomes like wages. Second, we extend the decomposition approach by Fairlie (2005) to consider in detail the separate contributions of different sets of covariates. In this way, we address the research question to which extent changes in the composition of first individual characteristics, second firm characteristics and third industry branch have affected the drop in collective bargaining coverage. At the same time, we separate the effects of changes in the three corresponding sets of coefficients from each other and from the residual effect. We will focus on this methodology in the next chapter.

3 Methodology

Several decomposition procedures have been developed in order to decompose changes in some dependent variable into parts that are attributable to changes in characteristics or in coefficients. The original approach by Oaxaca (1973) and Blinder (1973) applies to the linear regression case. However, when studying changes in collective bargaining coverage, the dependent variable is binary and thus a non-linear parametric model is required. For this case Fairlie (1999, 2005) develops a decomposition approach on which this study is based.

We want to decompose changes in collective bargaining coverage over time. Adapting Fairlie’s method to our application, the decomposition reads:

[1]Y¯2006Y¯2001=[i=1N06F(X06β^06)N06i=1N06F(X06β^01)N06]Residual+i=1N06F(X06β^01)N06j=1N01F(X01β^01)N01Characteristics

where X is the covariates matrix and β is the coefficients vector. In this case, the function F corresponds to the standard normal cumulative density function, corresponding to a probit model. The shorthand notation 01 refers to the year 2001 and likewise 06 to the year 2006. N06 and N01 denote the sample sizes of the two years. Hats refer to estimated values. The hypothetical value F(X06β^01) estimates the propensity of being covered by collective bargaining for individuals with characteristics from 2006 as if they had lived in the labour market of 2001. We estimate all the decompositions separately for males and females and for East and West Germany. For all covariates we take differences to their 2001 means within the corresponding subsample (males/females; East/West). This allows us to interpret changes in the constant as changes over time.

The second term in eq. [1] is called the “characteristics effect” as it represents differences in the outcome variable that occur due to the differences in the distributions of X (Fairlie 2005: 307). The first term in brackets captures those differences that occur due to changes in the coefficients and in the constant. In case there were relevant factors which are unobserved to the researcher, the constant would be affected. In this case, also the coefficients could be biased in case the unobservables correlate with the covariates. For this reason, the corresponding first term of the decomposition is usually labelled “residual” term or “unexplained” part (Fairlie 2005: 307; Schnabel and Wagner 2007).

The coefficients βˆ are obtained from probit regressions of a collective bargaining dummy on a set of covariates. The covariates can be grouped into three subgroups of interest:

  • P: Personal characteristics of the employee, i. e. age, tenure and education.

  • F: Firm characteristics of the job match, i. e. firm size, region and share of male employees.

  • S: Sector of the firm, i. e. industry branch. [8]

Next, we extend the decomposition approach in order to consider the contributions of these different sets of characteristics separately. This step requires a matching of the observations for the construction of a hypothetical counterfactual combination. [9]

Fairlie (2005: 308) suggests matching the observations based on the ranks of the fitted values of the estimated nonlinear functions. In case both subgroups are not of the same size, he further suggests using a random subsample of the larger group. However, this approach does not explicitly take account of the correlations between the covariates and therefore we further develop the approach at this point (similar to Antonczyk et al. 2010). The following approach is based on the sequential decomposition suggested by DiNardo et al. (1996) and further developed by Chernozhukov et al. (2013) and by Antonczyk et al. (2009, 2010). While all these approaches apply to the case of a continuous dependent variable, we will now translate them to the case of a limited dependent variable based on Fairlie (2005).

Thus, we want to decompose:

[2]Yˉ06Yˉ01=FβP06,βF06,βS06,β006,XS06,XF06,XP06FβP01,βF06,βS01,β001,XS01,XF01,XP01

where XP, XF and XS denote sets of personal and firm characteristics and the industry sector respectively, and βP, βF and βS the corresponding coefficients. β0 denotes the constants obtained from the two underlying probit regressions for 2001 and 2006.

We will analyse the contribution of each of the components separately by changing them step by step as denoted by the following sequence of effects:

[3]Δ1=F(βP06,βF06,βS06,β006,XS06,XF06,XP06)F(βP01,βF06,βS06,β006,XS06,XF06,XP06)Δ2=F(βP01,βF06,βS06,β006,XS06,XF06,XP06)F(βP01,βF01,βS06,β006,XS06,XF06,XP06)Δ3=F(βP01,βF01,βS06,β006,XS06,XF06,XP06)F(βP01,βF01,βS01,β006,XS06,XF06,XP06)Δ4=F(βP01,βF01,βS01,β006,XS06,XF06,XP06)F(βP01,βF01,βS01,β001,XS06,XF06,XP06)Δ5=F(βP01,βF01,βS01,β001,XS06,XF06,XP06)F(βP01,βF01,βS01,β001,XS01,XF06,XP06)Δ6=F(βP01,βF01,βS01,β001,XS01,XF06,XP06)F(βP01,βF01,βS01,β001,XS01,XF01,XP06)Δ7=F(βP01,βF01,βS01,β001,XS01,XF01,XP06)F(βP01,βF01,βS01,β001,XS01,XF01,XP01)

The choice of a sequence is not innocuous because the order matters in any sequential decomposition, i. e. they are path-dependent. [10] We choose this specific sequence of counterfactuals because it reflects the idea that we transfer the individuals from 2006 ‘back in time’ to the year 2001. We argue that this is meaningful because in this way the first step reflects in what sense the changing labour market remunerations (i. e. coefficients) contributed to the drop in coverage, given the individual characteristics of 2006. Only then, we change the characteristics. The complete sequential decomposition of changes in collective bargaining coverage from 2001 to 2006 can be summarised as:

Yˉ2006Yˉ2001=Δ1Personal+Δ2Firm+Δ3SectorCoefficients+Δ4Residual+Δ5Sector+Δ6Firm+Δ7PersonalCharacteristics

The first term of this detailed decomposition, Δ1, reflects changes in the propensity to work under collective bargaining that occur due to changes in the coefficients which correspond to personal characteristics. For example, for certain educational groups, if the probability of working under collective contracts changes over time relative to other educational groups, this is reflected in the first component.

The second term of the detailed decomposition, Δ2, captures changes in the coefficients which correspond to firm characteristics. For example, for employees working in small firms, if the probability of working under collective contracts changes over time relative to large firms, this is reflected in the second component.

The third term, Δ3, captures changes in the coefficients which correspond to the industry sector. For example, for certain industries, if collective bargaining coverage changes more strongly over time than for other industries, this is reflected in the third component.

The fourth term, Δ4, captures changes in the constant of the regression model over time. This includes an average time shift that applies to all industries, all firms and all employees. Further, a change in the constant includes changes in all unobservable variables. Therefore, the fourth component reflects all residual factors.

The fifth component, Δ5, captures changes in the industry composition of the workforce. For example, if there is a trend towards tertiarisation and collective bargaining coverage in the service sector differs from the one in the manufacturing sector, this compositional effect is reflected in the fifth component.

The sixth component, Δ6, captures changes in the composition of firms. For example, if there is a trend towards larger firms and if larger firms have different propensities to be covered by collective bargaining than smaller firms, then this is reflected in the sixth component.

The seventh component, Δ7, captures changes in the composition of employees. For example, if there is a trend towards educational upskilling and if highly educated employees display lower probabilities of collective bargaining than lower educated employees, then this is reflected in the seventh component.

All seven components add up to the total change in collective bargaining coverage over time as given by the difference between the average predicted values from the conditional models (see eq. [3]). [11]

Up until step 4, it is sufficient for the implementation of the procedure to plug in certain coefficients from 2001 together with the individual observations from 2006. Then, for the fifth step it is necessary to simulate in which industry sectors the individuals from 2006, who work in firms in 2006, would have worked in 2001. This is implemented by kernel matching based on the normal Gaussian kernel. Similarly, for the sixth step it is necessary to match the individual employees from 2006 to some firms and industry sectors in 2001. Again, this is implemented by Gaussian kernel matching.

The crucial assumption that underlies the estimation of a hypothetical counterfactual distribution is that a change in the covariates X does not affect the parameters of the conditional distribution model given X (e. g. Chernozhukov et al. 2013; Antonczyk et al. 2010). In other words, the decomposition approach ignores general equilibrium effects. This is similar to other decomposition methods in the literature (e. g. DiNardo et al. 1996). [12] This means that if changes in the characteristics cause the coefficients to change or vice versa, this interrelation cannot be detected by the decomposition approach. [13]

Another caveat to the standard Fairlie method refers to the fact that the residual effect does not further differentiate between the impact of coefficients and of the constant (Schnabel and Wagner 2007). This point is also addressed by our approach because changes in the coefficients are separated from changes in the constant.

Finally, as explained above, sequential decompositions are path-dependent. Therefore, the order will be reversed later in order to test for robustness (see the online appendix for details).

4 Data and descriptive statistics

Our analysis uses the 2001 and 2006 waves of the German Structure of Earnings Survey (GSES, “Verdienststrukturerhebung”). [14] This is a large mandatory linked employer-employee data set (LEED) consisting of repeated cross-sections. There are comparable data sets in other EU countries, such as the Spanish Earnings Structure Survey used e. g. by de la Rica et al. (2015). As data are reported by the personnel departments of the establishments, such variables like coverage by collective bargaining, industry sector, firm size, wage payments, and hours worked are very reliable. [15] The data involve a random draw among all establishments with at least ten employees. The advantage of linked employer-employee data is the joint availability of highly reliable firm-level and individual-specific data. [16] The disadvantage of this specific data set is that it consists of repeated cross-sections rather than a panel. We still use it due to its way of measuring collective bargaining coverage.

We limit our sample to establishments in the private sector which operate in those industries that are available in both years. [17] We limit our sample to prime aged employees in Germany who work full time. [18] For 2001, the final sample includes more than 400,000 employees in West Germany and about 125,000 employees in East Germany (see Table 1). In 2006, there are more than 700,000 employees in West Germany and 210,000 in East Germany. All observations are weighted by the inverse sampling probability reported with the data. Table 2 provides definitions of the variables available in our data.

In the data, for every individual employee it is reported whether one is covered by sectoral level bargaining, firm level bargaining or individual bargaining. We combine sectoral and firm level bargaining into just one category of collective bargaining (see Section 2). Next, we define an employee as covered by collective bargaining as soon as one works in a covered establishment, i. e. an establishment with a minimum number of individually-covered employees. [19] This definition takes account of the fact that collective bargaining coverage defines minimum standards for all employees working in a covered establishment in general. Note that we still analyse shares of covered individuals and not of covered firms (as done in Addison et al. 2010, 2011; Schnabel et al. 2006). This allows filtering out effects of changes in individual-specific characteristics, such as increases of the educational attainment. Table 1 reports the shares of covered employees. The figures show the sharp drop in coverage over the course of only five years. This applies to both, males and females. While for males, coverage dropped from 70.3 % (43.1 %) in 2001 to 53.4 % (35.5 %) in 2006 in West (East) Germany, for females the shares are 65.9 % (49.8 %) in 2001 and 45.9 % (35.8 %) in 2006. Thus, coverage dropped by 8–17 (ppoints) for males and by 14–20 ppoints for females.

Compared to the literature on Germany, the coverage reported in the GSES is lower. A different level of coverage can be partly explained by a different data selection. We exclude not only the public sector but also the entire health and education sector. Furthermore, the literature reports a smaller drop in coverage over time. We think that the coverage variable in the GSES is more reliable than self-reported coverage in non-mandatory surveys. Despite these differences, there is a consensus in the literature that a sizeable decline in coverage by collective bargaining occurred during the 2000s. We try to explain this drop by the following decomposition approach.

Table A.1 and A.4 in the online appendix display further descriptive statistics. On average, females are younger, have less tenure, and work in smaller establishments, compared to males. Tables 3 and 6 show the differences in coverage between different groups. While in West Germany, the medium-skilled show the highest coverage in 2001 and also for males in 2006, coverage in East Germany is highest for the low-skilled in 2001 and also for low-skilled males in 2006. In contrast, coverage in 2006 in both West and East Germany is highest for high-skilled females. The federal states North Rhine-Westphalia (NRW), Bavaria, Saxony-Anhalt, and Berlin start out with a high level of coverage in 2001 and experience a very sharp drop in 2006 by 17 ppoints and more. As expected, coverage generally increases with establishment size (with the exception that, in 2006, the largest coverage rate among females in West Germany is found for the second largest group of establishments because the largest establishments show a very large drop). Regarding the differences in coverage across industries, coverage is large for electricity, gas, water (in West Germany), mining and quarrying, manufacturing of transport equipment, post and telecommunications, finance and insurance, and coke, chemicals (in East Germany). The industries with the lowest coverage are data processing (in West Germany), real estate (in East Germany), and research, other services. The general descriptive findings are in line with the results reported by Ellguth and Kohaut (2011) for 2010. There are some minor differences in the ranking of industries between West and East Germany. Furthermore, the drop in coverage is much more uniform across industries in West Germany compared to East Germany. The sector of data processing in fact experienced an increase in coverage in both East and West Germany. For East Germany, there are a number of further cases where coverage increases over time (electricity, gas, water shows the strongest increase and the level in 2006 in East Germany is higher than in West Germany). The largest drop in both West and East Germany occurs in the post and telecommunications sector (from more than 90 % down to somewhere around 50 %) which is most likely related to the liberalisation and privatisation in these industries.

5 Results

First, we discuss the estimated probit regressions of coverage by collective bargaining on the observed covariates of the employees. Second, we present and discuss the detailed decomposition results which are based on the estimated probit regressions.

5.1 Probit results

We estimate flexible probit regressions by gender, year, and region. The detailed coefficient estimates and average marginal effects are reported in the additional appendix online. Most results are fairly similar for the different genders, years, and regions. Our discussion will describe the qualitative nature of the results and highlight some important differences.

While education plays only a small role in West Germany, coverage is U-shaped in education in East Germany for 2001, where the high-skilled actually show a higher coverage than the medium-skilled. In 2006, coverage falls with higher education in East Germany. [20]

The association of age with coverage is significantly negative or insignificant, and the association of tenure with coverage is positive. This is consistent with coverage being higher among older workers. Similarly, the literature reports higher levels of union membership among older employees (Fitzenberger et al. 2011) and higher levels of membership in employers’ associations among older firms (Schnabel and Wagner 1996).

The association of firm size with collective bargaining coverage is strongly positive and significant, both statistically and economically. The effect is similar to membership in employers’ associations being higher among larger firms (Schnabel and Wagner 1996). In our analysis, there are strong differences between small and large firms for all cases in 2001. For example, employees working in establishments with 10–99 employees in West Germany are about 45 ppoints less likely to be covered by collective bargaining in 2001 than employees working in establishments with 2,000 employees or more (the reference group). However, these differences are reduced over time and we observe even some changes in the ranks as firms with 1,000–1,999 employees show the highest coverage in 2006.

For West Germany in both years and for East Germany in 2006, the share of male employees shows a strong positive association with coverage. However, this does not apply to East Germany in 2001. The finding may indicate that collective bargaining is more strongly male dominated in West Germany (see discussion in Fitzenberger et al. 2011) and that there is a convergence of East Germany to West German patterns.

There are some differences in coverage across German regions and the ranking of regions changes over time. While North Rhine-Westphalia (NRW) has a high coverage level in 2001, coverage falls below the level of all other West German states in 2006. The same applies to Berlin in the East German sample. We cannot explain these changes.

There are also noticeable differences across industries (note that the omitted category refers to manufacturing of metals). The finance and insurance industry shows very high coverage in all cases, in the order of 20–45 ppoints above the omitted category. Also, manufacturing of coke and chemicals shows high coverage in all cases. There are also some remarkable differences. For instance, real estate shows a very low coverage in all regions in 2001 but only for West Germany in 2006. Turning to the changes over time, the ranking of industries changes little for West Germany and much more so for East Germany. A noticeable change in both West and East Germany involves the strong increase in coverage for data processing and information systems. In East Germany, coverage in real estate increases strongly over time.

Summing up, there are some noticeable changes in the coefficients over time regarding firm size and industries. These changes could drive the observed decline in aggregate coverage to some extent, an issue which would show up in the coefficients effects. Therefore, we now turn to the detailed decomposition analysis.

5.2 Sequential decomposition results

The benchmark decomposition results are reported in Tables 4 and 7. The upper parts of the tables involve a simple decomposition into only two components, the residual (coefficients) and the characteristics effect. The lower parts involve the detailed decomposition into seven components, as described in Section 3. The overall decline in coverage, that is to be decomposed, amounts to −16.9 (−8.4) ppoints for males in West (East) Germany and to −19.9 (−13.9) ppoints for females in West (East) Germany. [21]

For West Germany, the simple decomposition shows that more than 90 % of the decline are explained by the residual effect. For East Germany, the residual effect amounts to more than 100 % for both, females and males. The residual effect includes changes in the coefficients and in the intercept. For West Germany, only 8 % (males) to 9 % (females) are explained by changes in the composition of the workforce (characteristics effect). For East Germany, the characteristics effect amounts to −20 % for females and −24 % for males, meaning that the changing characteristics worked against a drop in collective bargaining coverage in the East. For West Germany these results are in line with the recent literature on union coverage which documents a minor role of the characteristics effect (Addison et al. 2011; Antonczyk et al. 2011; Fitzenberger et al. 2011; Schnabel and Wagner 2007).

The sequential decomposition further decomposes the residual effect and the characteristics effect, leaving the size of these total effects unchanged. The first three components correspond to the coefficients for the personal characteristics (Δ1), firm characteristics (Δ2), and for industry affiliation (Δ3). For West Germany, all of these contribute very little to the drop in coverage (the effects contribute at most 2.6 % and for males the industry coefficients would even have implied an increase in coverage). For East Germany, Δ1 and Δ2 also do not contribute in a sizable way to the drop in coverage while the industry coefficients effect Δ3 would have implied a notable increase in coverage.

Turning to the three different characteristics effects, these also contribute in a minor way to the drop in coverage for West Germany. Most importantly, changes in the industry structure, Δ5, explain about 11 % of the drop in coverage. Changes in the firm characteristics, Δ6, contribute with a smaller share of 5 % for males and 2 % for females. Changes in personal characteristics, Δ7, counteract the decline in coverage to a small extent. Once again, for East Germany, the results are more complex and the contribution of some of the sequential effects is quite large. Here, changes in the industry structure, explain a major part of the drop in coverage (Δ5: 23 % for males and 6 % for females) while changes in firm characteristics (Δ6) would have implied a sizable increase in coverage (54 % for males and 29 % for females). Changes in personal characteristics (Δ7) contribute little. We argue that the compensating firm and industry characteristics effects in East Germany still reflect the ongoing structural adjustment process of the East German economy. On the one hand, firm characteristics change in a way that would increase coverage. On the other hand, industry changes counteract the firm characteristics effect in a way to reduce coverage in East Germany. Both together may reflect strong structural changes in the East German economy.

The component of the residual effect associated with the change in the intercept (constant, Δ4) contributes primarily to the aggregate drop in coverage. This effect reflects the unexplained time trend, i. e. the change in coverage that affects all groups of workers and firms alike, including changes in unobserved covariates. The residual component Δ4 amounts to −16 (14) ppoints or 95 % (168 %) for males in West (East) Germany and to −17 (19) ppoints or 85 % (136 %) for females in West (East) Germany. Thus, despite sizable changes in slope coefficients and characteristics, the overall drop in coverage remains unexplained reflecting changes in the intercept. For instance, this means that the notable differences in the change in coverage across industries (or other subgroups) discussed in Section 4 do not drive the aggregate drop in coverage. The drop in coverage has basically affected all groups of employees, albeit to a varying degree. Put differently, the change in coverage for the employee with average characteristics in 2001 (this is what Δ4 estimates) almost coincides with (even exceeds) the overall drop in coverage for West Germany (East Germany), i. e. the noticeable differences of the change in coverage (coefficients effect) and the noticeable changes in characteristics almost cancel each other out. For instance, note that not all industries experience a decline in coverage relative to the omitted category. In other words, different industry coefficients in West Germany compensate each other, resulting in an industry coefficients effect that is close to zero. This finding is in contrast to the notion that the drop in coverage is associated with a decline of employment in high-coverage industries or with certain large industries dropping out of collective bargaining.

These results are in line with the literature on coverage (mostly for West Germany) which also documents the dominating role of the residual effect (Addison et al. 2011; Antonczyk et al. 2011). [22] Recent decomposition analysis of the decline of union membership have also found a minor role of the characteristics effect (Fitzenberger et al. 2011; Schnabel and Wagner 2007). In contrast, earlier studies find a noticeable role of changes in characteristics (Fitzenberger et al. 1999; Beck and Fitzenberger 2004). Taking an intermediate position, Biebeler and Lesch (2007) argue that workers’ preferences play an important role in explaining the drop in union membership. There is more of a consensus for coverage implying that changes in the composition of the workforce are not the main factor of the drop in coverage. Our sequential approach adds to this literature by showing that not the slope coefficients per se but changes in the constant or in unobservables can fully rationalise the drop in coverage.

Which factors could be driving the strong unexplained reduction in coverage (Δ4) which is very similar for both males and females and even higher in East Germany than in West Germany? Dustmann et al. (2014) argue that Germany was increasing the flexibility of its labour market in the late 1990s and the early 2000s in order to cope with the high unemployment rate and to improve its competitiveness. During this time period, wage inequality increased strongly (see also Antonczyk 2011) and a sizable share of firms opted out of collective bargaining with the goal to increase wage flexibility at the firm level. Furthermore, wage inequality increased more strongly among covered firms during the early 2000. Thus wage setting under collective bargaining became much more responsive to the conditions at the firm level (Dustmann et al. 2014).

5.3 Robustness check

In order to check for robustness of our results, we now reverse the direction of the decomposition. The simple decomposition into only two effects reads now:

[4]Y¯2006Y¯2001=[i=1N06F(X06β^06)N06i=1N01F(X01β^06)N06]Characteristics+i=1N01F(X01β^06)N06j=1N01F(X01β^01)N01Residual

For the counterfactual used for the decomposition in the previous section, the characteristics X from 2006 were evaluated at the coefficients from 2001, so far. Now, for the counterfactual the characteristics X from 2001 are evaluated at the coefficients from 2006, see eq. [4]. The detailed sequential decomposition starts now from individuals in 2006 evaluated at coefficients from 2006, and then first changes the personal characteristics (matching firm and industry from 2006), next the firm characteristics (still matching the industry from 2006), and finally the industry. After that, we transfer the hypothetical observations from 2001 “back in time” by using the 2001 coefficients. The additional appendix shows the detailed sequential decomposition in reversed order.

The results of this reversed decomposition analysis are reported in Tables 5 and 8. It is astonishing that the results of this simple decomposition remain nearly identical when compared to the original ordering for West Germany. For East Germany, the residual component becomes even much stronger but some of the other sequential effects change. The residual effect (constant, Δrev4) fully explains the drop in coverage. In all cases, except for females in West Germany, this unexplained time trend would have predicted an even higher drop in coverage than observed because the contribution in these cases is above 100 % (it is 107 % (195 %) for males in West (East) Germany and 151 % for females in East Germany). Here, the higher contribution of the unexplained time trend compared to Section 5.2 (see eq. [3]) implies that either changes in characteristics or changes in coefficients from 2001 to 2006 have partly worked against the drop in coverage. Furthermore, our findings for the coefficients effects typically show a small contribution for West Germany but a noticeable contribution for East Germany in the case of firm coefficients, Δrev6. In East Germany, the firm coefficients contribute to a rising tendency in coverage. As for the characteristics effects, all three are almost negligible for West Germany. In contrast, for East Germany we find compensating effects between strong positive personal and industry characteristics (Δrev1 and Δrev3, personal characteristics and industries change in a way that would predict an increase in coverage) and negative firm characteristics (Δrev2, firms change in a way that would predict a drop in coverage). These strong and compensating characteristics suggest that, evaluated at 2006 coefficients, the changes in characteristics imply a stronger impact on coverage compared to Section 5.2. Recall that in all cases the characteristics effect as a whole does not provide an important contribution to explain the drop in coverage. Thus, despite the differences in the detailed decomposition results, the robustness analysis confirms the main finding above, namely, that the unexplained time trend basically fully explains the drop in coverage.

6 Conclusions

In Germany, as in most industrialised countries, union representation has been in strong decline. This is reflected not only in the drop in union membership but also in the drop in coverage by collective bargaining. This trend is important as it is likely to translate into higher wage inequality (Card 2001; Card et al. 2003; Antonczyk et al. 2010; Dustmann et al. 2014; Addison et al. 2014).

This study investigates the drop in coverage for the case of Germany during the early 2000s. We develop and implement a sequential decomposition approach that extends upon Fairlie (2005) by using a well defined sequence of counterfactuals. This allows gaining further insight on which of the covariates from the individual or firm level dominate. Our study is the first one developing and applying a sequential decomposition approach to the drop in collective bargaining coverage.

The empirical analysis uses linked employer-employee data from the German Structure of Earnings Survey (GSES, “Verdienststrukturerhebung”) comparing 2001 and 2006. The advantages of the GSES are the large sample size and the highly reliable information provided which are based on personnel records of the establishments. Our data show that the share of male employees working in covered firms dropped from 70 % (44 %) in 2001 to about 53 % (36 %) in 2006 for West (East) Germany. Even more so, coverage among female employees dropped from about 66 % (50 %) in 2001 to 46 % (36 %) in 2006 for West (East) Germany.

The decomposition results clearly show that only a minor part of the drop in collective bargaining coverage can be explained by changes in the characteristics or their corresponding slope coefficients. While the development of collective bargaining over time varies substantially between different industries and groups of firms, these differences do not fully explain the drop in coverage because relative gains and losses cancel out. Instead, unexplained changes over time drive the result. This means that the drop in collective bargaining coverage is not confined to certain industries, to establishments of a certain size, nor to certain educational groups of employees. Nevertheless, we find strong evidence for different trends in coverage across industries and groups of firms. These differences are particularly strong in the case of East Germany where firm characteristics change in a way that reduces coverage, whereas changes in industries and personal characteristics would have implied an increasing coverage. Altogether, these effects may reflect strong ongoing structural changes in the East German economy. These findings would be compatible with growing heterogeneity of industries and firms (Card et al. 2013).

There are a number of caveats regarding our analysis. First, it should be kept in mind that we only observe whether firms explicitly state that they adhere to a collective contract. This may include firms who are not legally required to do so and this may exclude firms that are merely “oriented” towards a collective contract. Thus, a natural next step for research would be to analyse whether selectivity into collective bargaining coverage has changed over time (Gürtzgen 2016). Second, a limitation of any decomposition method is that it does not allow causal conclusions and that it ignores general equilibrium effects. Instead, the method points to those factors which are considered important in association with the drop in coverage by collective bargaining – a contribution that we consider as important. Third, the GSES data we use involve less information at the individual level or at the establishment level compared to some of the data sets used in the literature. Thus, our analysis based on the GSES should be viewed as a complement rather than a substitute, on the one hand, to the analysis of union membership based on the SOEP or ALLBUS and, on the other hand, to a panel analysis of union coverage based on the IAB establishment panel. Furthermore, the GSES data allow for a distinction between firm-level and sectoral-level bargaining contracts which may be explored in future work.

The findings presented here call for further research. What are the unobserved drivers and mechanisms which lead to the drop in coverage? Because a drop in coverage is observed in many countries (e. g. Visser 2006) and because we find a strong unexplained time trend, could there be a universal explanation, e. g. a response to the globalisation of economic activity? As there is only a small and loosely related literature on this hypothesis (Neumayer/de Soysa 2006), we leave this question to future research.

Acknowledgement

We thank the editor and three anonymous referees for helpful comments. Moreover, we thank David Card and participants at the Workshop on the Analysis of Administrative Data (WAAD) at IAB Nuremberg in December 2012. Furthermore, we thank Marina Furdas for help with the programming. We also thank the Research Data Centre (FDZ) at the Statistical Office of Hesse and in particular Manuel Boos, Hans-Peter Hafner, Alexander Richter and Florian Fischer for support with the data. All errors remain our own responsibility. An additional online appendix for this paper is available at www.jbnst.de/en.

References

Addison, J.T., A. Bryson, P. Teixeira, A. Pahnke (2011), Slip Sliding Away: Further Union Decline in Germany and Britain. Scottish Journal of Political Economy 58 (4): 490–518. Search in Google Scholar

Addison, J.T., A. Bryson, P. Teixeira, A. Pahnke, L. Bellmann (2010), The State of Collective Bargaining and Worker Representation in Germany: The Erosion Continues. IZA Discussion Paper, 5030. Search in Google Scholar

Addison, J.T., A. Kölling, P. Teixeira (2014), Changes in Bargaining Status and Intra-Plant Wage Dispersion in Germany: A Case of (Almost) Plus Ça Change? IZA Discussion Paper, 8359. Search in Google Scholar

Addison, J.T., P. Teixeira, K. Evers, L. Bellmann (2015), Is the Erosion Thesis Overblown? Alignment From Without in Germany. Industrial Relations,Forthcoming. Search in Google Scholar

Antonczyk, D. (2011), Using Social Norms to Estimate the Effect of Collective Bargaining on the Wage Structure. Mimeograph University of Freiburg. Search in Google Scholar

Antonczyk, D., B. Fitzenberger, K. Sommerfeld (2010), Rising Wage Inequality, The Decline of Collective Bargaining, and the Gender Wage Gap. Labour Economics 17 (5): 835–847. Search in Google Scholar

Antonczyk, D., B. Fitzenberger, K. Sommerfeld (2011), Anstieg der Lohnungleichheit, Rückgang der Tarifbindung und Polarisierung. Zeitschrift für Arbeitsmarkt Forschung – Journal for Labor Market Research (ZAF) 44 (1–2): 15–27. Search in Google Scholar

Antonczyk, D., F. Fitzenberger, U. Leuschner (2009), Can a Task-Based Approach Explain the Recent Changes in the German Wage Structure? Journal of Economics and Statistics (Jahrbücher für Nationalökonomie und Statistik) 229 (2–3): 214–238. Search in Google Scholar

Beck, M., B. Fitzenberger (2004), Changes in Union Membership Over Time: A Panel Analysis for West Germany. Labour 18 (3): 329–362. Search in Google Scholar

Biebeler, H., H. Lesch (2007), Zwischen Mitgliedererosion und Ansehensverlust: Die deutschen Gewerkschaften im Umbruch. Industrielle Beziehungen 14 (2): 133–153. Search in Google Scholar

Blinder, A.S. (1973), Wage Discrimination: Reduced Forms and Structural Estimates. Journal of Human Resources 8 (4): 436–455. Search in Google Scholar

Bosch, G. (2004), The Changing Nature of Collective Bargaining in Germany. Coordinated Decentralization. 84–118 in: H.C. Katz, W. Lee, J. Lee (editors), The New Structure of Labor Relations: Tripartism and Decentralization, Chapter 4. Ithaca, Cornell University Press. Search in Google Scholar

Burda, M., B. Fitzenberger, A.C. Lembcke, T. Vogel (2008), Unionization, Stochastic Dominance, and Compression of the Wage Distribution: Evidence from Germany. SFB 649 Discussion Papers, HU Berlin. Search in Google Scholar

Card, D. (1996), The Effect of Unions on the Structure of Wages: A Longitudinal Analysis. Econometrica 64 (4): 957–979. Search in Google Scholar

Card, D. (2001), The Effect of Unions on Wage Inequality in the U.S. Labor Market. Industrial and Labor Relations Review 54 (2): 296–315. Search in Google Scholar

Card, D., J. Heining, P. Kline (2013), Workplace Heterogeneity and the Rise of West German Wage Inequality. The Quarterly Journal of Economics 128 (3): 967–1015. Search in Google Scholar

Card, D., T. Lemieux, C. Riddell (2004), Unions and Wage Inequality. Journal of Labor Research 25 (4): 519–562. Search in Google Scholar

Card, D., T. Lemieux, W.C. Riddell (2003), Unions and the Wage Structure, Chapter 8, pages 246–292. Edward Elgar. Search in Google Scholar

Chernozhukov, V., I. Fernández-Val, B. Melly (2013), Inference on Counterfactual Distributions. Econometrica 81 (6): 2205–2268. Search in Google Scholar

de la Rica, S., J.J. Dolado, R. Vegas (2015), Gender Gaps in Performance Pay: New Evidence from Spain. Annals of Economics and Statistics 117–118, 41–59. Search in Google Scholar

DiNardo, J., N.M. Fortin, T. Lemieux (1996), Labor Market Institutions and the Distribution of Wages, 1973–1992: A Semiparametric Approach. Econometrica 64 (5): 1001–1044. Search in Google Scholar

Dustmann, C., B. Fitzenberger, U. Schönberg, A. Spitz-Oener (2014), From Sick Man of Europe to Economic Superstar: Germany’s Resurgent Economy. The Journal of Economic Perspectives 28 (1): 167–188. Search in Google Scholar

Ellguth, P., S. Kohaut (2004), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2003. WSI-Mitteilungen 57 (8): 450–460. Search in Google Scholar

Ellguth, P., S. Kohaut (2005), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel. WSI-Mitteilungen 58 (7): 398–406. Search in Google Scholar

Ellguth, P., S. Kohaut (2007), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2006. WSI-Mitteilungen 60 (9): 511–514. Search in Google Scholar

Ellguth, P., S. Kohaut (2011), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2010. WSI-Mitteilungen 64 (5): 242–247. Search in Google Scholar

Ellguth, P., S. Kohaut (2012), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2011. WSI-Mitteilungen 65 (4): 297–305. Search in Google Scholar

Ellguth, P., S. Kohaut (2013), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2012. WSI-Mitteilungen 66 (4): 281–288. Search in Google Scholar

Ellguth, P., S. Kohaut (2014), Tarifbindung und betriebliche Interessenvertretung: Aktuelle Ergebnisse aus dem IAB-Betriebspanel 2013. WSI-Mitteilungen 67 (4): 286–295. Search in Google Scholar

European Commission (2002), Germany’s growth performance in the 1990’s. Economic Papers, 170. Search in Google Scholar

Fairlie, R.W. (1999), The Absence of the African-American Owned Business: An Analysis of the Dynamics of Self-Employment. Journal of Labor Economics 17 (1): 80–108. Search in Google Scholar

Fairlie, R.W. (2005), An Extension of the Blinder-Oaxaca Decomposition Technique to Logit and Probit Models. Journal of Economic and Social Measurement 3 (4): 305–316. Search in Google Scholar

Fitzenberger, B., I. Haggeney, M. Ernst (1999), Wer ist noch Mitglied in Gewerkschaften? Eine Panelanalyse für Westdeutschland. Schmollers Jahrbuch (Journal for Economics and Social Sciences) 119 (2): 223–263. Search in Google Scholar

Fitzenberger, B., K. Kohn (2005), Gleicher Lohn fuör gleiche Arbeit? zum Zusammenhang zwischen Gewerkschaftsmitgliedschaft und Lohnstruktur in Westdeutschland 1985–1997. Zeitschrift für ArbeitsmarktForschung 38 (2–3): 125–146. Search in Google Scholar

Fitzenberger, B., K. Kohn, A.C. Lembcke (2013), Union Density and Varieties of Coverage: The Anatomy of Union Wage Effects in Germany. The Industrial & Labor Relations Review 66 (1): 169–197. Search in Google Scholar

Fitzenberger, B., K. Kohn, Q. Wang (2011), The Erosion of Union Membership in Germany: Determinants, Densities, Decompositions. Journal of Population Economics 24 (1): 141–165. Search in Google Scholar

Fortin, N., T. Lemieux, S. Firpo (2011), Decomposition Methods in Economics. 1–102 in: O. Ashenfelter and D. Card (editors), Handbook of Labor Economics, Volume 4A, Chapter 1, North Holland. Search in Google Scholar

Garloff, A., N. Gürtzgen (2011), Flexibilität durch Öffnungsklauseln: Wenn weniger mehr ist. IAB Forum 1: 26–29. Search in Google Scholar

Gerlach, K., G. Stephan (2006), Bargaining Regimes and Wage Dispersion. Jahrbücher für Nationalökonomie und Statistik (Journal of Economics and Statistics) 226 (6): 629–645. Search in Google Scholar

Goerke, L., M. Pannenberg (2012), Risk Aversion, Collective Bargaining, and Wages in Germany. Labour 26 (2): 156–173. Search in Google Scholar

Gürtzgen, N. (2016), Estimating the Wage Premium of Collective Wage Contracts – Evidence From Longitudinal Linked Employer-Employee Data. Industrial Relations, forthcoming. Search in Google Scholar

Hassel, A. (2007), The Curse of Institutional Security: The Erosion of German Trade Unionism. Industrielle Beziehungen 14 (2): 176–191. Search in Google Scholar

Heinbach, W.D. (2005), Ausmaß und Grad der tarifvertraglichen Öffnung. IAW-Report 2: 49–68. Search in Google Scholar

Heinbach, W.D. (2006), Bargained Wages in Decentralized Wage-Setting Regimes. IAW-Diskussionspapier, 26. Search in Google Scholar

Heinbach, W.D., S. Schröpfer (2007), Typisierung der Tarifvertragslandschaft: Eine Clusteranalyse der tarifvertraglichen Öffnungsklauseln. Jahrbücher für Nationalökonomie und Statistik (Journal of Economics and Statistics) 227 (3): 219–235. Search in Google Scholar

Hold, D. (2003), Flucht aus dem Tarifvertrag – Möglichkeiten und Grenzen. Betrieb und Wirtschaft 11: 476–481. Search in Google Scholar

Kohaut, S., L. Bellmann (1997), Betriebliche Determinanten der Tarifbindung: Eine empirische Analyse auf der Basis des IAB-Betriebspanels 1995. Industrielle Beziehungen 4 (4): 317–334. Search in Google Scholar

Krugman, P. (Nov 12, 2009), Free to Lose. The New York Times. Available at: http://www.nytimes.com/2009/11/13/opinion/13krugman.html (Published: 12th Nov. 2009, last accessed 12th June 2015). Search in Google Scholar

Lesch, H. (2004), Trade Union Density in International Comparison. CESifo Forum 4: 12–18. Search in Google Scholar

Möller, J. (2010), The German Labor Market Response in the World Recession – De-Mystifying a Miracle. Zeitschrift für ArbeitsmarktForschung42 (4): 325–336. Search in Google Scholar

Neumayer, E., I. de Soysa (2006), Globalization and the Right to Free Association and Collective Bargaining: An Empirical Analysis. World Development 34 (1): 31–49. Search in Google Scholar

Oaxaca, R. (1973), Male-Female Wage Differentials in Urban Labor Markets. International Economic Review 14 (3): 693–709. Search in Google Scholar

OECD (2004), OECD employment outlook. Chapter 3, Wage-Setting Institutions and Outcomes. Organisation for Economic Co-operation and Development, Paris. Search in Google Scholar

Schnabel, C. (2013), Union Membership and Density: Some (not so) Stylized Facts and Challenges. European Journal of Industrial Relations 19 (3): 255–272. Search in Google Scholar

Schnabel, C., J. Wagner (1996), Ausmaß und Bestimmungsgründe der Mitgliedschaft in Arbeitgeberverbänden: Eine empirische Untersuchung mit Firmendaten. Industrielle Beziehungen (The German Journal of Industrial Relations) 3 (4): 293–306. Search in Google Scholar

Schnabel, C., J. Wagner (2007), The Persistent Decline in Unionization in West and East Germany, 1980–2004: What Can We Learn From a Decomposition Analysis? Industrielle Beziehungen 14 (2): 118–132. Search in Google Scholar

Schnabel, C., S. Zagelmeyer, S. Kohaut (2006), Collective Bargaining Structure and Its Determinants. An Empirical Analysis with British and German Establishment Data. European Journal of Industrial Relations 12 (2): 165–188. Search in Google Scholar

Schwiebert, J. (2015), A Detailed Decomposition for Nonlinear Econometric Models. The Journal of Economic Inequality 13 (1): 53–67. Search in Google Scholar

Visser, J. (2006), Union Membership Statistics in 24 Countries. Monthly Labor Review 129 (1): 38–49. Search in Google Scholar

Appendix

Table 1:

Shares of employees regarding coverage by collective bargaining.

YearMalesFemales
2001200620012006
West Germany
Covered70.29 %53.42 %65.85 %45.91 %
Not covered29.71 %46.58 %34.15 %54.09 %
No. of observations311,054517,969101,992184,247
East Germany
Covered43.84 %35.51 %49.76 %35.79 %
Not covered56.16 %64.49 %50.24 %64.21 %
No. of observations84,186140,37441,76272,658
Table 2:

Definition of variables.

LabelDescription
Individual level
Low educationLow level of education: no training beyond a school degree
Medium educationIntermediate Level of education: vocational training
High educationHigh level of education: university or university of applied sciences
Education n/aMissing information on the education level
AgeAge in years
TenureTenure in years
Firm level
10–99 employeesFirm has between 10 and 99 employees
100–199 employeesFirm has between 100 and 199 employees
200–999 employeesFirm has between 200 and 999 employees
1,000–1,999 employeesFirm has between 1,000 and 1,999 employees
More than 2,000 employeesFirm has more than 2,000 employees
Share of male employeesShare of male employees, ranges between 0 and 1
West:
Schleswig-Holstein, HHFirm is located in Schleswig Holstein or Hamburg
Lower Saxony, BremenFirm is located in Lower Saxony or Bremen
NRWFirm is located in North Rhine-Westphalia
HesseFirm is located in Hesse
RLP, SaarlandFirm is located in Rhineland-Palatinate or Saarland
Baden-WürttembergFirm is located in Baden-Württemberg
BavariaFirm is located in Bavaria
East:
BerlinFirm is located in Berlin
Brandenburg, Meck-PomFirm is located in Brandenburg or Mecklenburg-West Pomerania
SaxonyFirm is located in Saxony
Saxony-AnhaltFirm is located in Saxony-Anhalt
ThuringiaFirm is located in Thuringia
Industry
Mining, quarryingMining and quarrying
Manufact: FoodManufacture of food products, beverages and tobacco
Manufact: TextilesManufacture of textile and textile products, leather and leather products
Manufact: WoodManufacture of wood and wood products
Publishing, printingPublishing, printing and reproduction of recorded media
Manufact: Coke, chemicalsManufacture of coke, refined petroleum products and nuclear fuel; chemicals and chemical products
Manufact: Rubber, plasticManufacture of rubber and plastic products
Manufact: Non-metallicManufacture of other non-metallic mineral products
Manufact: MetalsManufacture of basic metals; fabricated metal products, except from machinery and equipment
Manufact: MachineryManufacture of machinery and equipment
Manufact: Electr. machineryManufacture of electrical machinery and apparatus
Manufact: Electr. equipmentManufacture of electrical & optical equipment; radio, TV, & communication equipment & apparatus
Manufact: InstrumentsManufacture of medical, precision and optical instruments, watches and clocks
Manufact: TransportManufacture of transport equipment
Manufact: n.e.c.Manufacture not elsewhere classified
Electricity, gas, waterElectricity, gas and water supply
ConstructionConstruction
Auto sales, repairSale, maintenance and repair of motor vehicles and motorcycles; retail sale of automotive fuel
Wholesale tradeWholesale trade and commission trade except for motor vehicles and motorcycles
Retail tradeRetail trade, except from motor vehicles and motorcycles; repair of personal and household goods
Hotels, restaurantsHotels and restaurants
TransportLand, water and air transport
Auxiliary transportSupporting and auxiliary transport activities; activities of travel agencies
Post, telecommunicationsPost and telecommunications
Finance, insuranceFinancial intermediation, insurance and pension funding, except compulsory social security
Real estateReal estate activities; renting of machinery and equipment without operator
Data processingData processing and information systems
Research, other servicesResearch and development and other services
Table 3:

Descriptive statistics: collective bargaining coverage in subgroups, West.

MalesFemales
200120062006–2001200120062006–2001
ShareShareΔ ShareShareShareΔ Share
Individual level
Low education0.7150.550−0.1650.7050.524−0.181
Medium education (Reference)0.7260.571−0.1550.6760.485−0.192
High education0.7070.568−0.1390.6360.485−0.151
Education n/a0.4410.244−0.1970.4330.235−0.198
Firm level
10–99 employees0.4560.292−0.2460.4280.248−0.180
100–199 employees0.6680.467−0.2010.6040.431−0.173
200–999 employees0.8080.614−0.1940.7590.551−0.208
1,000–1,999 employees0.9000.788−0.1120.8840.735−0.149
More than 2,000 employees (Reference)0.9640.810−0.1540.9430.670−0.273
Schleswig-Holstein, HH0.5820.478−0.1040.5710.433−0.138
Lower Saxony, Bremen0.7480.611−0.1370.7080.502−0.206
NRW (Reference)0.7350.528−0.2070.7040.448−0.256
Hessen0.6560.506−0.1500.6020.442−0.160
RLP, Saarland0.7180.590−0.1280.6880.534−0.154
Baden-Württemberg0.6610.498−0.1630.6110.460−0.151
Bavaria0.7180.546−0.1720.6680.449−0.219
Industry
Mining, quarrying0.9470.775−0.1720.9250.627−0.298
Manufact: Food0.6710.390−0.2810.5600.331−0.229
Manufact: Textiles0.7840.494−0.2900.7120.390−0.322
Manufact: Wood0.6990.495−0.2040.6660.490−0.176
Publishing, printing0.7170.509−0.2080.6320.503−0.129
Manufact: Coke, chemicals0.8960.835−0.0610.7730.732−0.041
Manufact: Rubber, plastic0.6640.432−0.2320.5780.412−0.166
Manufact: Non-metallic0.8340.502−0.3320.7760.477−0.299
Manufact: Metals (Reference)0.6950.501−0.1940.6480.525−0.123
Manufact: Machinery0.7380.652−0.0860.7550.599−0.156
Manufact: Electr. machinery0.7890.603−0.1860.7200.550−0.170
Manufact: Electr. equipment0.7930.463−0.3300.7300.383−0.347
Manufact: Instruments0.6200.505−0.1150.5500.452−0.098
Manufact: Transport0.9380.820−0.1180.9220.751−0.171
Manufact: n.e.c.0.6870.441−0.2460.5940.371−0.223
Electricity, gas, water0.9600.886−0.0740.9190.881−0.038
Construction0.6960.511−0.1850.6630.505−0.158
Auto sales, repair0.6890.445−0.2440.6550.406−0.249
Wholesale trade0.5000.311−0.1890.4960.279−0.217
Retail trade0.6440.382−0.2620.7540.379−0.375
Hotels, restaurants0.5870.458−0.1290.5690.468−0.101
Transport0.5510.377−0.1740.6000.473−0.127
Auxiliary transport0.6100.301−0.3090.6620.366−0.294
Post, telecommunications0.9520.497−0.4550.9060.429−0.477
Finance, insurance0.9070.802−0.1050.9180.804−0.114
Real estate0.5790.263−0.3160.5020.241−0.261
Data processing0.3000.4240.1240.2770.3490.072
Research, other services0.3390.4300.0910.3430.322−0.021
Overall0.7030.534−0.1690.6590.459−0.199
No. of observations311,054517,969101,992184,247
Table 4:

Decomposition results, West.

ComponentMalesFemales
ValuePercentValuePercent
Difference in coverage 2001–2006−0.169100−0.199100
Residual effect (incl. coeff. effect)−0.14887.8−0.17889.3
Characteristics effect−0.02112.2−0.02110.7
Δ1Personal coefficients−0.0021.0−0.0021.2
Δ2Firm coefficients−0.0010.5−0.0052.6
Δ3Industry coefficients0.014−8.2−0.0021.0
Δ4Constant/Residual effect−0.16094.5−0.16884.5
Δ5Industry (characteristics)−0.01911.0−0.02211.1
Δ6Firm characteristics−0.0095.2−0.0052.3
Δ7Personal characteristics0.007−4.00.005−2.7
No. of observations829,023286,239
Table 5:

Sensitivity check: decomposition in reversed order, West.

ComponentMalesFemales
ValuePercentValuePercent
Difference in coverage 2001–2006−0.169100−0.199100
Residual effect (incl. coeff. effect)−0.16295.6−0.18291.8
Characteristics effect−0.0074.4−0.0168.2
Δrev1Personal characteristics−0.0095.10.003−1.3
Δrev2Firm characteristics0.001−0.6−0.0031.8
Δrev3Industry (characteristics)0.000−0.2−0.0157.7
Δrev4Constant/Residual effect−0.192107.6−0.18995.2
Δrev5Industry coefficients0.005−2.8−0.0062.9
Δrev6Firm coefficients0.013−7.80.012−5.9
Δrev7Personal coefficients0.003−1.40.001−0.4
No. of observations829,023286,239
Table 6:

Descriptive statistics: collective bargaining coverage in subgroups, East.

LabelMalesFemales
200120062006–2001200120062006–2001
ShareShareΔ ShareShareShareΔ Share
Individual level
Low education0.6010.425−0.1770.5830.389−0.194
Medium education (Reference)0.4340.364−0.0700.5020.367−0.135
High education0.5200.397−0.1230.5560.423−0.133
Education n/a0.3140.244−0.0700.3850.256−0.129
Firm level
10–99 employees0.2420.188−0.0530.2690.163−0.106
100–199 employees0.4890.331−0.1490.5240.330−0.194
200–999 employees0.7030.520−0.1840.7210.502−0.219
1000–1999 employees0.8720.720−0.1520.8790.675−0.204
More than 2000 employees (Reference)0.9050.692−0.2140.8950.652−0.243
Berlin0.6000.378−0.2220.6440.370−0.274
Brandenburg, Meck-Pom (Reference)0.3930.341−0.0520.4930.358−0.135
Saxony0.4160.384−0.0320.4380.385−0.053
Saxony-Anhalt0.4080.322−0.0860.5210.352−0.170
Thuringia0.3830.323−0.0590.3960.295−0.101
Industry
Mining, quarrying0.8400.745−0.0950.8430.8990.056
Manufact: Food0.4990.332−0.1680.4040.239−0.165
Manufact: Textiles0.4680.242−0.2260.3380.109−0.229
Manufact: Wood0.3580.324−0.0340.3470.323−0.023
Publishing, printing0.5840.330−0.2540.5660.304−0.262
Manufact: Coke, chemicals0.8380.638−0.2000.7690.602−0.167
Manufact: Rubber, plastic0.3820.209−0.1730.2920.181−0.112
Manufact: Non-metallic0.4150.415−0.0000.6180.365−0.253
Manufact: Metals (Reference)0.3420.239−0.1030.3080.223−0.086
Manufact: Machinery0.4540.225−0.2290.4380.247−0.191
Manufact: Electr. machinery0.6230.465−0.1570.5080.410−0.098
Manufact: Electr. equipment0.5740.169−0.4050.5210.180−0.340
Manufact: Instruments0.4110.253−0.1580.2570.184−0.073
Manufact: Transport0.8000.653−0.1470.7360.589−0.147
Manufact: n.e.c.0.3510.117−0.2350.3330.080−0.253
Electricity, gas, water0.7360.9070.1710.7290.9210.192
Construction0.3210.306−0.0150.3340.3370.003
Auto sales, repair0.3010.083−0.2180.2410.062−0.179
Wholesale trade0.4440.274−0.1700.5120.250−0.262
Retail trade0.5030.354−0.1490.5920.339−0.253
Hotels, restaurants0.4450.284−0.1610.4540.341−0.112
Transport0.4220.371−0.0520.5620.564−0.001
Auxiliary transport0.3760.169−0.2060.5520.274−0.278
Post, telecommunications0.9460.574−0.3720.9480.528−0.420
Finance, insurance0.9030.736−0.1680.9450.783−0.162
Real estate0.2400.3940.1550.2450.2480.003
Data processing0.4500.4030.0470.4320.399−0.032
Research, other services0.2660.4290.1630.3700.308−0.062
Overall0.4380.355−0.0830.5000.358−0.140
No. of observations84,186140,37441,76272,658
Table 7:

Decomposition results, East.

ComponentMalesFemales
ValuePercentValuePercent
Difference in coverage 2001–2006−0.084100−0.139100
Residual effect (incl. coeff. effect)−0.104123.8−0.167119.9
Characteristics effect0.020−23.80.028−19.9
Δ1Personal coefficients0.001−1.00.001−0.5
Δ2Firm coefficients0.008−9.4−0.0010.8
Δ3Industry coefficients0.029−34.00.023−16.2
Δ4Constant/Residual effect−0.142168.2−0.189135.8
Δ5Industry (characteristics)−0.02023.3−0.0086.1
Δ6Firm characteristics0.046−54.30.041−29.3
Δ7Personal characteristics−0.0067.2−0.0053.2
No. of observations224,560114,420
Table 8:

Sensitivity check: decomposition in reversed order, East.

ComponentMalesFemales
ValuePercentValuePercent
Difference in coverage 2001–2006−0.084100−0.139100
Residual effect (incl. coeff. effect)−0.112132.7−0.178127.9
Characteristics effect0.028−32.70.039−27.9
Δrev1Personal characteristics0.039−46.30.036−26.0
Δrev2Firm characteristics−0.02833.4−0.02719.4
Δrev3Industry (characteristics)0.017−19.80.029−21.2
Δrev4Constant/Residual effect−0.164194.5−0.210151.4
Δrev5Industry coefficients0.013−15.80.010−7.4
Δrev6Firm coefficients0.038−45.60.023−16.3
Δrev7Personal coefficients0.000−0.3−0.0000.2
No. of observations224,560114,420
Received: 2014-6-6
Revised: 2015-2-27
Accepted: 2015-6-2
Published Online: 2016-7-2
Published in Print: 2016-2-1

©2016 by De Gruyter Mouton