The Wage Impact of Being a Works Council Representative in Germany

(1)

The Wage Impact of Being a Works Council Representative in Germany

A Case of Strategic Discrimination?

Brébion, Clément

Document Version Final published version

Published in:

Industrial Relations

DOI:

10.1111/irel.12307

Publication date:

2022

License CC BY-NC-ND

Citation for published version (APA):

Brébion, C. (2022). The Wage Impact of Being a Works Council Representative in Germany: A Case of Strategic Discrimination? Industrial Relations, 61(4), 418-455. https://doi.org/10.1111/irel.12307

Link to publication in CBS Research Portal

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

Take down policy

If you believe that this document breaches copyright please contact us (research.lib@cbs.dk) providing details, and we will remove access to the work immediately and investigate your claim.

Download date: 20. Oct. 2022

(2)

Industrial Relations. 2022;00:1–38. wileyonlinelibrary.com/journal/irel | 1 DOI: 10.1111/irel.12307

O R I G I N A L A R T I C L E

The wage impact of being a works council representative in Germany: A case of strategic discrimination?

Clément Brébion

This is an open access article under the terms of the Creative Commons Attribution- NonCommercial- NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is noncommercial and no modifications or adaptations are made.

Copenhagen Business School (CBS) and Centre d’Etudes de l’Emploi et du Travail (CEET), Frederiksberg, Denmark Correspondence

Clément Brébion, Copenhagen Business School (CBS) and Centre d’Etudes de l’Emploi et du Travail (CEET), Frederiksberg, Denmark.

Email: cbr.msc@cbs.dk

Abstract

Works councils provide an essential mechanism for worker participation in decision- making. While the literature has extensively explored their impact on worker and establishment outcomes, the negotiation process between works council representatives and their employer has remained largely unexplored. This article contributes to filling this gap by investigating wage discrimination towards works councilors in Germany. Fixed effects models leveraging panel data show that councilors receive a wage premium that positively correlates with the sectoral coverage of collective bargaining. In the manufacturing sector, where the tradition of bargaining is heavily entrenched, employers positively discriminate councilors. In contrast, in the service sector, where the culture of bargaining is weak, employers penalize works councilors. In both sectors, partisan and unionized works councilors are the most affected. The most likely hypothesis to explain these results is that employers strategically discriminate these councilors in order to bypass the traditional constraints of establishment- level participation. This article therefore questions the quality of industrial democracy in Germany.

(3)

I N TRODUCTION

Works councils are a key instrument for nonunion worker participation in establishment- level decision- making. Amidst productivity slowdown, falling labor shares, and a reduction in union density, this institution has enjoyed a renewed interest both in Europe and the United States thanks to its potential for improving working conditions and establishment performance (Jäger et al., in press; Mueller & Stegmaier, 2020). Against this background, most of the literature has focused on measuring the effects of works councils on worker and establishment outcomes (for reviews, see Addison, 2009; or Jirjahn & Smith, 2018). The negotiation process between works council representatives and their employer, however, has remained largely unexplored. This article contributes to filling this gap. It shows that focusing on the power relation between the actors of codetermination casts a revealing light on both the quality of industrial democracy and its outcomes.

Works council representatives play two separate roles with respect to their employer: as dependent employees, they work under orders; as councilors, they are their equal discussion partners.¹ While ostensibly independent, these roles may interfere with each other because the hierarchical subordination implied by the employment relation gives employers leeway to pressure the works councilor during negotiations. Employers may seek to cow or bribe works councilors through (negative or positive) wage premiums to influence their behavior in the negotiation process. Such wage discriminations, if uncovered, would greatly undermine industrial democracy: it would likely both distort the pool of workers willing to run for elections and warp negotiation outcomes.

This article contributes to the works council literature by precisely investigating wage discrimination towards works councilors in Germany. The German case is chosen because of the strong legal rights afforded to works council representatives there. Although German works councils are released from bargaining over the most divisive distributional questions that usually take place at the industry level between unions and employers’ associations, they benefit from extensive codetermination rights and veto powers on a wide variety of social matters.

In principle, the German Works Constitution Act (WCA) rules out the prospect of any (positive or negative) wage discrimination targeted at works councilors. By law, works councilors and employers must strive to cooperate “in a spirit of mutual trust […] for the good of the employees and of the establishment,” and councilor remuneration should evolve along the same path as comparable workers of the establishment.² In practice, however, the cooperative and trustful negotiations promoted by the WCA may not always apply. Councilors might for example use their extensive rights to force employers to strike informal deals on the distribution of rent, and the latter could seek to limit such moves (Jirjahn & Smith, 2006, 2018). Other employers may value their entrepreneurial freedom so highly that they become reluctant to share decision power with councilors despite the potential benefits accrued from truthful negotiations (Jirjahn & Mohrenweiser, 2016). These incentives may tempt employers to bribe or punish councilors to influence them, in violation of the WCA. Case studies have highlighted such behavior in Germany. They have shown how some employers implement a variety of strategies to both eschew the establishment of works councils, and later control or sideline the councilors—

with the ultimate goal of neutering the institution (Arora, 2017; Artus, 2013; Bormann, 2007;

Köhnen, 2006; Royle, 1998). This article exploits a large- scale dataset to investigate whether such behavior is widespread.

The empirical analysis builds on the estimation of fixed effects models using the German Socio- Economic Panel (GSOEP) between 2001 and 2015. The main sample is restricted to employees working in establishments with a works council. In the absence of an

1The terms “works council representatives”, “works council members” and “works councilors” are synonyms.

2The WCA sets a prison sentence and fines should councilors be unduly favoured or penalized because of their position.

(4)

establishment identifier, I restrict each individual of the sample to the longest unbroken working spell within an establishment in their employment history so that the variation of interest— becoming a councilor or leaving the council— is measured within establishment and within individual.³

This paper yields three main results. First, I show that the wage impact of being a works council representative positively correlates with the sectoral coverage of collective bargaining.

In particular, I find a positive wage premium of 4.6% in the manufacturing sector, where the support for collective bargaining is strong and historically rooted. Conversely, a penalty of 3.7% is exhibited in the service sector, where the culture of bargaining is weak. Second, I show that in the manufacturing sector, the career trajectories of future worker representatives had generally evolved more slowly before they entered the council than those of workers with similar characteristics. When taking into account this negative selection, it appears that councilor earnings in the sector are 13.3% higher than what they would have been had they followed their pre- election trend. I further show that conditions imposed on employers by the Works Constitution Act are insufficient to explain this very large effect. Taking these conditions into account and clearing the negative selection bias leaves us with a residual wage effect of being a works councilor of +5.9%. Such a trend analysis is not possible for the service sector due to data limitations.⁴ Third, I provide statistical evidence showing that, in both sectors, being a works council representative primarily affects the earnings of partisan and unionized individuals. The preferences of these representatives, I explain, are central to the demands of works councils. Employers therefore have an incentive to treat them specially, either positively or negatively depending on the strength of collective bargaining in the sector.

Strategic discrimination from rational employers towards labor representatives is the most cogent hypothesis to explain the different signs of the wage impact of being a works councilor across sectors. Specifically, I argue that in the service sector, where the tradition of establishment- level participation and sectoral collective bargaining is weak, negative discrimination is a workable strategy to undermine opposition and to avoid sharing decision- making power. In contrast, the strong institutionalization of codetermination and collective bargaining in German manufacturing makes positive discrimination an attractive option for employers to bypass the constraints of the traditional model of industrial relations.

This article makes several contributions to the industrial relation literature. First, building on a stream of theoretical works that describe works councils as both rent- seeking and rent- generating organizations (Freeman & Lazear, 1995; Hübler & Jirjahn, 2003; Jirjahn, 2017), the works council empirical literature has chiefly tried to determine which of the two facets of councils dominate. A very large stream of papers has estimated the impact of works councils on working conditions, wages, productivity, or profits, mostly in Germany (e.g., Addison et al.

2010; Jirjahn & Mueller, 2014; Mueller & Neuschaeffer, 2021; Mueller & Stegmaier, 2017).

Overall, the literature presents a cautiously positive picture of the economic impact of works councils (Mueller & Stegmaier, 2020).⁵ This article complements that facet of the literature on the negotiation outcomes of works councils by shedding light on the black box of the negotiation process between employers and works councilors.

3In other words, the wage impact of being a works councilor is identified by: (i) individuals observed both before they begin a mandate as works councilor in an establishment, and during their mandate there, and (ii) individuals observed both during their mandate as works councilor and after their mandate is over, in the same establishment. Notably, these sources of identification in a sample of workers employed in establishments that have a works council ensure that the nonrandomness of the emergence of works councils across establishments does not affect the internal validity of the results.

4I can therefore not formally rule out the possibility that reverse causality explains the negative works councilor premium in this sector. However, the explanation I propose satisfactorily explains employers’ behaviour in the service sector as well.

5Of particular note for this paper, past research has revealed a rather similar impact of works councils on labour productivity in both manufacturing and service sectors (Frick and Möller, 2003; Mueller, 2015; Brändle, 2017). If anything, the effect is slightly larger in the latter (Mueller, 2015). This precludes sharing of the rent created by the works council as an explanation for my results.

(5)

Second, a few papers have attempted to reconcile the overall positive economic effect of councils with employers’ avoidance strategies found in case studies. They can be cleaved into two main arguments. First, the economic effects of works councils on the performance of establishments are heterogeneous across a vast number of characteristics, e.g., the type of own- ership of the establishment (Jirjahn & Mueller, 2014), the age of the works council (Mueller &

Stegmaier, 2017) and coverage by sectoral collective bargaining agreements (CBAs) (Hübler &

Jirjahn, 2003; Mueller, 2011). Works councils may therefore have only small or even negative impacts on profits in some establishments, which may try to avoid them (Mueller & Stegmaier, 2020). Secondly, some employers may oppose establishment- level participation because they ascribe high utility to their entrepreneurial freedom, including in cases where it would bring profits to the establishment (Jirjahn & Mohrenweiser, 2016). This article contributes to this literature by stressing how employers’ avoidance strategies negatively affect works councilor earnings in the service sector. It also highlights the existence of positive discrimination towards works councilors in the manufacturing sector.

This article is close in spirit to Breda (2011, 2014) and Bourdieu and Breda (2016) who, to my knowledge, are the only ones to have worked on the wage impact of being a shop- floor representative. One difference with these papers is my focus on works councilors in Germany where Bourdieu and Breda observe union delegates in France. A second difference is that I exploit panel data to capture both observed and unobserved time- invariant individual characteristics.

In contrast, Bourdieu and Breda use cross- sectional data. The two authors find that exerting mandates of union delegates in France is associated with a wage penalty of approximately 10% on average, which can reach 20% for delegates belonging to the most pressing unions. My results on works councilors in Germany and Bourdieu and Breda's results on union delegates in France therefore have some commonalities in that negative strategic discrimination is found where the institutionalization of employer- employee negotiations is weak (in France overall and in the German servicesector).

The article is organized as follows: In the next two sections, I successively describe the institutional context and construct testable hypotheses in relation to past literature. Sections 4 and 5 exhibit the data and some descriptive statistics. I then use regressions and robustness checks to show that the wage impact of being a works councilor differs according to the sector.

In the last section, I present a heterogeneity analysis, which contributes to the argument that this wage impact is the result of strategic discrimination.

I NSTIT U TIONA L CON TEXT

Industrial relations in Germany are structured around two pillars that, respectively, involve unions and works councils. First, collective bargaining agreements (CBAs) are usually negotiated at the industry level between unions and employers’ associations. They set the main rules regulating the distribution of rent, including matters of wages, working conditions, working time and job classification. Such agreements are legally binding for member firms of the signa- tory employers’ associations but are rarely extended to the whole sector.⁶

Working conditions can also be negotiated at the establishment level. Here, works councils are a key mechanism for employees’ participation in decision- making. Per the Works Constitution Act (WCA), their rights comprise: (i) extensive information rights on establishments’ strategic orientations; (ii) codetermination rights on the organization of working time (including overtime and plans for reduced working time), on the manner of remuneration, on the monitoring of employee behavior and performance, and on the fixing of job and bonus

6Firm- level agreements also exist but cover a very small share of establishments.

(6)

rates; and (iii) veto powers over management initiatives regarding certain individual staff movements. On the other hand, works councils are forbidden from striking deals on issues that are normally addressed by collective agreements, including wage bargaining. In short, councilors are tasked with less divisive personal and social matters.

The WCA also lays out the functions of works councils, which are distinct from those of unions. Works councils represent all employees of the establishment, irrespective of their union status, and their role is to ease relations between labor and employers at the establishment level and to increase the joint surplus. Thus, the WCA explicitly states that works councils and management must cooperate “in a spirit of mutual trust […] for the good of the employees and of the establishment” (WCA, section 2). As a result, works councilors must refrain from activities that “imperil the peace in the establishment” (WCA, section 74) and may not call for strikes.

Work councils may be formed in establishments of at least five permanent workers. They are not automatic: employers must facilitate the constitution of a works council and bear its costs only when referred to by workers. Professional elections take place every 4 years.⁷ The number of works councilors per establishment and the number to receive full release depend on the establishment size (see Table A1, in Appendix 1). Councilors not fully released are expected to decide on the time needed to accomplish their duties and inform their employer in a timely manner.⁸

Let us stress here that in legal terms, a works councilorship is an honorary position.

Councilors continue to work on the same contract they had before their election and the WCA explicitly states that councilor remuneration should evolve along the same path as comparable workers of the establishment.⁹ Breaking these terms— by either penalizing or favoring councilors— carries a prison penalty and fines. Thus, the WCA explicitly outlaws premiums for works councilors.

BACKGROU N D DISCUS SION

Employer incentives to influence the behavior of works councilors

In practice, it is unrealistic to expect the WCA to fully specify and determine the relations between management and councilors. In particular cases, employers have incentives to influence works councilors’ behavior against the “spirit of mutual trust” promoted by the WCA.

I discuss these in the next paragraphs before turning to the potential implications for works councilor wage premiums.

First, works councils can use their comprehensive rights on social and personnel matters to constrain employers to informally negotiate issues over which they normally have no intervention power (Jirjahn & Smith, 2006, 2018). For instance, councilors may trade their veto over working time reorganization for an informal commitment from employers to increase wages. In such instances, works councils exceed their legal prerogatives by trying to shape business decisions with a direct impact on the distribution of rent. Furthermore, by crystallizing conflicts between the workforce and management, such rent- seeking claims may also indirectly reduce the positive

7Since 2000, they should have taken place between 1 March and 31 May in 2002, 2006, 2010, 2014 and 2018 in all establishments.

Poll records cannot be combined with individual data.

8Representation hours must be spent during normal working hours. When this is not possible, overtime spent on a mandate should be converted to time off in the following 30 days and, as a last resort, be paid out (WCA, section 37– 2).

9Section 37– 4 of the WCA states that “during his term of office and for one year thereafter the remuneration of a member of the works council shall not be fixed at a lower rate than the remuneration paid to workers in a comparable position who have followed the career that is usual in the establishment. The same shall apply to general benefits granted by the employer”.

(7)

returns of works councils that ought to stem from improved information flows and increased mutual trust in the workplace when the most divisive distributional questions are discussed at the industry level (Freeman & Lazear, 1995). As a result, one may expect the overall returns of establishment- level participation on establishment profitability to be negatively correlated with the intensity of councilors’ rent- seeking behavior. Employers may therefore be tempted to sway councilor behavior away from such demands.

Employers can also initiate negotiations on questions involving essential business decisions to increase their rent, with the same effect of placing councilors in pivotal positions unwar- ranted by the WCA. First, employers may seek approval by the works council to ease the ac- ceptability of change on matters that normally fall under their own prerogatives but towards which the workforce shows reservations (Royle, 1998). Second, facing downward pressure on their membership rates, employers’ associations and unions have introduced innovative institutional designs enabling firms to deviate from CBAs under particular circumstances. The most common types are opt- out clauses, which generally require employers to obtain the support of the works council (Haipeter, 2011). In these two cases, works councilors may be reluctant to approve employer initiatives, especially if workers express concerns. Employers may therefore try to influence their behavior to secure their consent.

Finally, employers may try to inhibit the independence of works councilors because establishment- level participation reduces their entrepreneurial freedom. The behavioral literature has revealed individuals’ taste for discretionary power, including in cases when sharing powers would serve their material interests (Bartling et al. 2014; Fehr et al. 2013). Specifically for our case, Harcourt et al. (2020) argue that German employers view codetermination as a departure from the default situation in which entrepreneurial prerogative applies and, for this reason, they underestimate the gains yielded by codetermination. This applies especially to owner- managers who seemingly ascribe a high utility to being the “ultimate boss” within their establishment (Jirjahn & Mohrenweiser, 2016). Overall, employers ascribing high utility to their entrepreneurial freedom may strive to avoid sharing power with works councilors and, when they cannot avoid it, may try to sway the decisions of the works council despite the potential positive returns of truthful negotiations.

Strategies to control or weaken the works council

Breda (2011, 2014) and Bourdieu and Breda (2016) provide a theoretical framework rationalizing the strategies that employers may develop to influence councilor behavior. The authors explain that shop floor representatives play two negotiation games with their employer at the same time: one through their mandate, on behalf of their colleagues, and another concerning their own career evolution (promotions, working conditions, etc.) similar to any other employee.

Two Nash equilibria can result from these negotiation games in the presence of incentives for the employer to influence councilors’ behavior. In the first equilibrium, the representative is pliable and trades laxity in her demands as an elected delegate for particular improvements in her working conditions relative to her colleagues. A second, different equilibrium occurs if the representative is inflexible and strongly negotiates on behalf of her colleagues. In this case, the employer could ensure that the delegate's career stagnates to deter further activism in the establishment. This framework therefore predicts that in the presence of incentives to influence councilors’ behavior, rational employers strategically discriminate councilors by offering them negative or positive wage premiums depending on the flexibility and the strength of their claims. Notably,the weight of works councilors in the establishment or in the works council is likely to play a mitigating role. For instance, employers are expected to pay stronger attention to the demands of the heads of works councils than to those of standard councilors. The former prediction is therefore likely to best apply to the most influential councilors.

(8)

We have seen that by law, such strategic discriminations are forbidden under the WCA (see Section 2). In practice, however, managers have some scope in setting the wages of works councilors differently from those of comparable workers with low risks of being reported. Overpaid councilors are unlikely to report this bypassing of the law. The opposite scenario— career stagnation of councilors— can take different forms. The pool of workers deemed comparable to a given councilor before her election to the works council may first benefit from increases in wage rates for tasks that the councilor no longer performs. Likewise, the employer may pro- mote some individuals from the pool of comparable workers and justify this by emphasizing specific benchmarks that councilors are not able to hit because of their mandate. In both instances, the case is not easy to capture and demonstrate. Councilors may not feel comfortable claiming these compensations and may receive little support from their colleagues when doing so. Overall, the risk that employers run with such behaviors may therefore be limited.

German case studies have shed light on such cases of strategic discrimination. These studies are not systematic but have the virtue of proving that monetary HR policies targeted at councilors do take place in Germany. For instance, Royle (1998) shows that the management at McDonald's seeks to co- opt the most moderate and opportunistic councilors by offering them cash compensations or promotions, and to discourage the most inflexible ones by refusing to compensate them for the time spent on counselling duty. Likewise, interviews with works councilors at H&M suggest that some councilors’ careers stagnate because of their role as representatives (Köhnen, 2006). More recently, a jail sentence handed to the head of Volkswagen's works council, and senior managers has shed light on a case of long- lasting collusion. In 2005, managers unsatisfied with their profit margins at a time of high labor costs and low productivity in the firm had bribed the head of the works council with pay raises and in- kind benefits to convince the rest of the works council and the workforce of the need for uncompensated increases in working hours in order to save jobs (Abdelal et al., 2008; Arora, 2017) .

Based on the theoretical arguments of Breda (2011, 2014) and Bourdieu and Breda (2016) and on the anecdotal evidence of these case studies, one can construct the following hypothesis.

Hypothesis 1 The works councilor premium is most sizeable for influential councilors and nega- tively correlates with the strength and inflexibility of their demands.

The tradition of establishment- level participation and collective bargaining in the sector may also shape employers’ strategies in the presence of incentives to influence councilors’

behavior. Consider a sector with low unionization rates, low coverage of sectoral CBAs and a weak culture of worker participation in decision- making. By definition, workplaces of this sector share strong norms of entrepreneurial freedom. Therefore, as previously discussed, employers are expected to oppose establishment- level participation as a whole more often than elsewhere despite the potential positive impact of works councils on profitability. In the presence of a works council, such employers are particularly likely to penalize councilors to deter activism and, ultimately, to achieve the suppression of the council. Such strategies can prove costly as works councilors may appeal to labor courts. However, the risk is more limited in these sectors than elsewhere due to relatively low collective support for codetermination in the workplace and beyond. In particular, employers with strong preferences against works councils and collective bargaining can choose to hire workers who are less likely to mobilize to support councilors and can decide not to renew the fixed- term contracts of workers who show interest in labor representation. Such HR policies are well described in Royle’s (1998) typology of employer avoidance strategies applied against works councils and, more recently, by Artus (2013) in reference to the service sector. Managers in such cases rely on nonunionized outsiders with relatively poor knowledge of their rights and on short- term contract workers with little interest in the long- run performance of the establishment. Beyond the workplace, unions’ interest in engaging in long and costly fights

(9)

against employers may also be limited if union density is low. While formally independent, works councils effectively rely on union expertise and judiciary counselling. Councilors’

abilities to engage in lawsuits against employers in cases of negative discrimination are therefore reduced if unions only offer them weak support (Artus, 2013). Therefore, in expec- tations, a cost- benefit analysis should drive employers to use punishment strategies more often in sectors with a weak tradition of establishment- level participation and sectoral collective bargaining. Hence, I propose the following hypothesis:

Hypothesis 2 The weaker the institutionalization of establishment- level participation and sectoral collective bargaining, the greater the risk of negative discrimination towards councilors.

In Germany, the culture of worker participation in decision- making and collective bargaining is much weaker in the service sector than in the manufacturing sector. Thus, in 2011, respectively, 34% and 66% of employees in those sectors worked in establishments with a works council (see Table 1). There is also a clear negative trend in the service sector, which is not found in the manufacturing sector. Likewise, in manufacturing, CBAs covered approximately 20% more workers in 2011 and union density is approximately 250% greater (Visser, 2019). These figures reflect a stronger norm in favor of entrepreneurial decisions in the service sector. Thus, a survey shows that managers classify works agreements as profit- able overall and as decreasing rigidity in the establishment much less often in services than in manufacturing (Nienhueser, 2009).¹⁰ Likewise, in line with H2, Behrens and Dribbusch (2020) show that union representatives from the service sector notice more actions made against existing works councils than their counterparts from the manufacturing sector.

Interestingly, these differences across sectors correlate poorly with the actual effect of works councils as the literature has found that they enhance productivity in both sectors to a rather similar extent (Brändle, 2017; Frick & Möller, 2003; Mueller, 2015). If anything, the positive impact of councils on the value- added per worker is even stronger in the service sector (Mueller, 2015). From this I derive H2.1:

Hypothesis 2.1 The wage impact of being a works councilor is more often positive in the manu- facturing sector than in the service sector in Germany.

DATA

The three hypotheses that the previous section develops are tested using data from the German Socio- Economic Panel (GSOEP), a yearly survey representative at the household and individual levels (Haisken- DeNew & Frick, 2005). To my knowledge, this database is the only one that combines information on wages and works council membership in Germany. Both variables are available for waves 2001, 2003, 2006, 2007, 2011, and 2015, which I therefore use for the main analysis.

Employees working in establishments with a works council differ from the rest of the workforce in many observable characteristics (Table A2.1 in Appendix 2) and, we can expect, in a number of unobservable features. Workers from establishments with no works council should not be included in the sample of analysis; however, information on whether establishments have a works council is available only for 2001, 2006, 2011, and 2016 (Table 2). Works council

10Note that owner- managers, who share strong norms for employers’ right to manage (Jirjahn and Mohrenweiser, 2016), are also more often found in services than in manufacturing industries (Gottschalk et al. 2017).

(10)

elections normally take place once every 4 years; assuming that an establishment with no creation or destruction of a works council over a 5- year span has experienced no variation on the matter in the intervening time, I can determine the presence of a works council in the establishment of 40% of the respondents for the waves 2003, 2007, and 2017. I drop the remaining individuals in these years.¹¹ Some robustness checks are performed in Section 6 using an alternative sample built without these recoding assumptions.

Works councils depend effectively on unions, both for expertise and because around two- thirds of works councilors are also union members. The impacts of the two variables on wages should therefore be distinguished. Information on union status is given for all years of interest except for 2006 (Table 2). I approximate the 2006 status of respondents who answered in both 2006 and 2007 without changing establishment with their status in 2007.¹² The other observations for 2006 are dropped.

I further restrict the sample to full- time workers (i.e., between 30 and 60 working hours per week) aged 20– 64 and employed on open- ended contracts in firms with more than 5 employees. Civil servants are dropped, as are voluntary workers, members of the military and workers in the agricultural sector or in the extractive industries.

There is no establishment identifier in the panel. To capture the establishment dimension, I restrict each individual of the sample to the longest unbroken working spell within an establishment in her employment history. As the main model of identification is a fixed effects model, the estimations are free of establishment and individual characteristics that are con- stant over time. This sampling strategy, combined with the fact that individuals are observed in an establishment with a works council, ensures that the nonrandomness in the emergence of a works council does not affect the internal validity of the results.

11In practice, I approximate for 2003 the works council status (presence or absence of a works council in the establishment) by the status applicable in 2001 and 2006 if the establishment did not change status and the worker did not change establishments. For all three waves, I then drop observations for establishments with no works council or for which the works council status could not be recovered.

12The method produces limited noise: the yearly variation in union membership affects approximately 5% of the sample in the other waves.

TA B L E 1 Share of German employees working in establishments…

… with a works council … covered by sectoral CBAs

Whole

economy Service sector

Manufacturing sector

Whole

economy Service sector

Manufacturing sector

2000 47.6% 41.4% 65.4% 59.1% 54.3% 63.9%

2011 42.4% 34.5% 65.9% 47.8% 43.4% 51.8%

Source: Addison et al. (2017). The data were drawn from the IAB establishment panel. Establishments with fewer than 5 employees or from the agriculture sector, extractive industries, and public corporations are excluded.

TA B L E 2 Availability of the main variables of interest in the German Socio- Economic Panel by wave

2001 2003 2006 2007 2011 2015 2016

Presence of a works council in the

establishment ✓ ✓ ✓ ✓

Works council membership ✓ ✓ ✓ ✓ ✓ ✓

Union membership ✓ ✓ ✓ ✓ ✓

Source: German Socio- Economic Panel.

(11)

To limit the risk of outliers driving the main results, I further trim the bottom and top 1% of the yearly distribution of the gross hourly wage. I finally remove individuals who are observed only once. The final unbalanced panel includes 8323 observations from 2530 respondents. On average, a respondent is observed 3.3 times over 6.7 years (corresponding to the time span between the first and last observations).

The main dependent variable is the log- value of the gross hourly wage.¹³ All fixed- effect estimations— except Equation (3)— include the following categorical control variables: age (4 categories), seniority (4), SES (9), firm size (6), the month of the interview (12), and a dummy to control for whether the number of weekly hours is specified in the working contract. A year fixed effect is also included.

DE SCR I PTI V E STATISTICS

Table 3 shows the share of the works council and union memberships in the final sample. As mentioned above, two- thirds of works councilors are unionized in the main sample. In total, works council members account for 7.9% of the sample.¹⁴ Overall, one- third of the sample includes members of a trade union.

Table A2.1 in Appendix 2 displays the average values of different variables for the main sample according to works council membership status (columns (3) and (5)). T- tests for mean difference are also provided in the last column. Works councilors earn, on average, approximately 0.8 euro per hour less than their colleagues— significant at the 1% level. This value accounts for approximately one- eighth of a standard deviation in the gross hourly wage distribution, or 4.4% of the average. The difference stems mostly from monthly earnings, although councilors report working approximately 25 min less per week (Table A2.2).

Interestingly, Table A2.3 and A2.4 show that these relations do not apply similarly across sectors for the main sample. Works councilors actually earn approximately 0.50 euro per hour more than their colleagues in the private manufacturing sector and 1.90 euro per hour less than their colleagues in the private service sector. For the former, both monthly earnings and actual working hours drive this difference, whereas for the latter, monthly earnings are the sole driver.

13It is computed using the answers to the questions: “How high was your income from employment last month?” and “How many hours [per week] do your actual working hours consist of, including possible overtime?”.

14This figure is higher than listed by legal requirements (Table A1 in Appendix 1). The criteria applied to build the main sample do over- select works councilors: e.g., because they are rarely on short- term contracts.

TA B L E 3 Incidence of works council and union memberships in the final sample

Works councilor

Member of a trade union

No Yes Total

No 5408 2260 7668

65.0% 27.2% 92.1%

Yes 218 437 655

2.6% 5.3% 7.9%

Total 5626 2697 8323

67.6% 32.4% 100.0%

Source: German Socio- Economic Panel, own calculations.

(12)

More generally, works council members are on average older, have more seniority, are less well educated, and are more often males than their colleagues. They also more frequently work in smaller firms, an expected consequence of the WCA (Table A1 in Appendix 1). Statistics for individuals who identify the main effect of the fixed effects model are also displayed (column (4) of Table A2.1). These individuals do not differ much from the average works councilor. If anything, their numbers of weekly hours are closer to those of nonelected workers. To provide more information, additional columns show the same statistics for workers in the GSOEP before the main selection procedures were applied (column (1)) and for workers of establishments with no works council (column (2)).

E STI M ATIONS Baseline model

Column (1) of Table 4 displays the estimation of the baseline fixed effects model— described in Equation (1)— over the whole sample.

Time and individual fixed effects are, respectively, measured by c_t and α_i. WoCo_M_i,t is a dummy variable taking the value 1 if the individual i is a works councilor in year t. The coefficient of interest, β, gives us the size of the wage impact of being a works councilor. It is identified by: (i) individuals observed both before they begin a mandate as works councilor in an establishment, and during their mandate there, and (ii) individuals observed both during their mandate as works councilor and after their mandate is over, in the same establishment.

Let us stress that creation and destruction of works councils do not play any role here given the sample selection described in Section 4. I control for union membership (U_M_i,t) and for the list of covariates (X_i,t) detailed in Section 4. Column (1) of Table 4 shows that, overall, no association between works council mandates and wages can be found in Germany.

I then proceed to test whether this result hides a link between the sectoral degree of institutionalization of collective bargaining and the works councilor premium, as hypothesized in H2. For this, I focus on the private sector and I use yearly data measuring the share of employees working in establishments that follow CBAs or are formally covered by CBAs.¹⁵ These yearly data were computed at the sector x region level by Ellguth and Kohaut using the IAB Establishment Panel (e.g., see Ellguth and Kohaut (2008) for the 2007 data).¹⁶ For each individual in my dataset, I compute the average value of this index across the years when they appear in my sample— using the share in the sector and region of relevance for their establishment.

After centering and standardizing the variable, I interact it with the works councilor dummy in an equation otherwise similar to Equation (1). Column (2) of Table 4 shows the result. It corroborates H2: a one standard deviation increase in the share of employees working in establishments following or covered by CBAs is associated with an increase of the works councilor premium of 3.3 percentage points.

(1) ln(w_i,t)=𝛽. WoCo_M_i,t+𝛾. U _M_i,t+X_i,t.Γ + 𝛼_i+c_t+𝜖_i,t

15Many German establishments are not formally covered by collective agreements but still follow the agreement applying in their industry. Considering only the share of workers in establishments formally covered by CBAs would fail to properly account for the degree of institutionalisation of collective bargaining in the sector.

16In these IAB data, sectors are 1- digit NACE industries and regions are West or East Germany. These information are lacking for a few observations in the SOEP, hence a number of observations in column (2) slightly smaller than the sum of columns (3) to (5).

(13)

TABLE 4Effect of works council and union memberships on the log gross hourly wage by sector (baseline model) (1)(2)(3)(4)(5)(6)(7) All sectors Private sector Public sector (not incl. civil servants)All private sectorsManufacturing sectorConstruction sectorService sector (incl. banking and insurance)

Service sector (not incl. banking or insurance) Works councilor0.0020.0100.046**−0.020−0.021−0.037**−0.014 (0.010)(0.012)(0.018)(0.029)(0.018)(0.019)(0.016) Works councilor * coverage of CBAs0.033** (0.015) Member of a trade union−0.016*−0.021*−0.0110.001−0.063***−0.067***0.000 (0.009)(0.011)(0.016)(0.023)(0.018)(0.020)(0.015) Individual fixed effectYesYesYesYesYesYesYes Time fixed effectYesYesYesYesYesYesYes Observations8323568024321241203315532617 Adjusted R2 0.8540.8610.8350.8260.8990.9010.828 Note: Fixed effects model. CBAs stands for collective bargaining agreements. Coverage of CBAs is computed at the sectoral level and gives account of the share of employees working in establishments that voluntarily follow or are formally covered by CBAs in the sector. This variable is centered and standardized. Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Source: German Socio- Economic Panel, own calculations.

(14)

In a third step, to clarify this result, I repeat the estimation of Equation (1) in each sector. In the private manufacturing sector, being a works councilor is associated with a positive wage gain of 4.6%. As expected, this sector shows a pattern different from that of the rest of the private sector: a nonsignificant penalty of approximately 2% is found in both the construction¹⁷ and service sectors. Among private services, it is known that industrial relations in the financial sector are unique due to very high coverage of both sectoral CBAs and works councils (respectively, 82.6% and 78.3% of employees in 2011 (Addison et al. 2017)). Separate estimations for this sector are not robust due to the sample size and are therefore not displayed here— if anything, they show a nonsignificant positive premium.¹⁸ When limiting the estimation to the remainder of the services industry, the wage impact of being a works councilor is negative (−3.7%) and statistically significant at the 5% level (column (6)). Finally, no works councilor premium is found in the public sector.

The impacts of all control variables for columns (1), (3), and (6) are displayed in Table A3.1 in Appendix 3. On a side note, the estimates provided in Table 4 are identified by workers changing status (i.e., observed both in- mandate and out of mandate). Useful information on their number by specification is given in the last appendix (Tables A6.1, A6.2, A6.3). The rest of this paper separately focuses on the manufacturing and the service sectors— excluding the financial sector because of its specificity.¹⁹

Overall, the works councilor premium therefore increases with the sectoral degree of institutionalization of collective bargaining. At the core of the Germany traditional model of industrial relations, in the manufacturing sector, the premium is positive. Conversely, at its periphery, in the service sector, the works councilor premium is negative. These results are consistent with both hypotheses H2 and H2.1 elaborated in Section 3. It is interesting to compare the size of these effects to those found by Breda (2014) and Bourdieu and Breda (2016) in France. These authors find that, on average, French union delegates receive a 10% wage penalty. Therefore, overall, employers in Germany value works councilors more than employers in France value union delegates. However, the penalty in the service sector is in fact close to what the researchers found in France overall.

Estimates of the relation between works council membership and wages in Equation (1) are the resulting effect of changes in works council membership occurring when voted into and out of the works council. Table A3.2 in Appendix 3 distinguishes between the two, pre- senting results separately for the private manufacturing and the service sectors (not including banking and insurance). To simplify, I separate agents who switch status at least once (“switchers”) into two groups: respondents whose first change in status involves becoming a works councilor and respondents whose first change in status involves leaving the works council. For both groups, I consider only observations that precede a second change in status if applicable. I re- estimate Equation (1) after successively dropping each of the two groups. Thus, in columns (1) and (3), the coefficient of interest measures the impact of becoming a works councilor, whereas, in columns (2) and (4), it measures the impact of leaving the works council. For both sectors, no significant difference can be observed between the two types of estimations, and the point estimates are similar to those given in Table 4. In other words, in both sectors, the association between works council membership and wages

17Industrial relations in the construction sector are not easy to assess. As in the manufacturing sector, it exhibits high coverage by sectoral CBAs. However, as in the service sector, councilors have limited relations with their sectoral union, and the share of establishments with a works council has been very low and exhibited a strong negative trend since the early 2000s. For this reason, the construction sector is set aside in the rest of the analysis.

18The sector notably differs from the manufacturing sector, where union density is much higher, the relationship between unions and works councilors is stronger and rights to open clauses and firm agreements are more widespread. For this reason, I do not merge the two sectors in the analyses hereafter.

19Therefore, in the remainder of the article, the service sector is composed as follows: 33% of the individuals work in trades, 21%

work in transport, 10% work in services to industries and 36% work in personal services.

(15)

seems to occur throughout the mandate period and to vanish when individuals are voted out of the organization.

For union membership, Table 4 shows an overall wage penalty of 1.6%. This result may be surprising: in Germany, employers covered by collective bargaining must pay the negotiated wage rate to both union members and nonmembers. However, this result converges with some anecdotal evidence on union- busting found in Germany (Behrens & Dribbusch, 2013).

Furthermore, the premium clearly stems from the service sector where bargaining coverage is the lowest, and it does not appear in large firms where bargaining coverage is the highest (Table A4.1). Finally, past research has already found union premiums in institutional settings that formally prevent them. For instance, Bourdieu and Breda (2016) measure a negative union premium in France (−3.5%) whereas industry- level agreements are often extended to the whole sector and, like in Germany, union contracts should apply to all workers, independently of their union membership (Breda, 2015).

A works councilor premium despite unchanged working hours

In Tables 5, I carry out regressions in a similar spirit as the baseline regressions, but I do so using first the actual number of weekly working hours, then the log of the monthly gross wage as a dependent variable. It appears that works council and union memberships affect income rather than working hours. Columns (1) and (2) of Table 5 show significant results of very similar size to the results of Table 4 for both manufacturing and service sectors.

Note that the negative impact of works council membership on monthly incomes in the service sector should not be read as a wage drop in nominal terms. Since earnings are computed from the following question: “How high was your income from employment last month?” respondents are expected to include additional income from sources such as bonuses or premiums in providing their answers. A decline in such earnings is the first possible explanation.

A differentiated rate of promotion between councilors and their comparable colleagues is also possible, as the combination of an average yearly wage growth of 1.25% and flat wage

TA B L E 5 What drives the baseline results? Estimations of alternative dependent variables: the log monthly gross wage and the number of actual working hours

Dependent variable: log of the monthly gross wage

Dependent variable: number of actual working hours

(1) (2) (3) (4)

Manufacturing

sector Service sector

Manufacturing

sector Service sector

Works councilor 0.044** −0.031* −0.119 0.312

(0.018) (0.018) (0.432) (0.412)

Member of a trade union −0.014 −0.062*** −0.116 0.191

(0.016) (0.020) (0.319) (0.551)

Individual fixed effect Yes Yes Yes Yes

Time fixed effect Yes Yes Yes Yes

Observations 2432 1553 2432 1553

Adjusted R² 0.877 0.926 0.661 0.694

Note: Fixed effects model.

Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.

Source: German Socio- Economic Panel, own calculations.

(16)

evolutions for works councilors would be sufficient to generate an average disadvantage of 3.1% over a mandate.

Finally, Table 5 exhibits no significant correlation between works council or union memberships and working hours— although, if anything, the correlation tends to inflate the effects on councilors’ incomes in both sectors. This result may seem unexpected. It is sometimes argued that time releases granted to councilors are not sufficient to fulfill their role and, more generally, that works councilors and union members spend additional time in meetings or organiz- ing the labor force. The present estimation shows us that either (i) these presuppositions are wrong or that (ii) works councilors already managed work- related issues outside of working hours before being elected to the works council.

Works councilor premium versus works council premium

It is interesting to compare the works councilor premium with the works council premium, that is, the wage impact of being a representative of the works council (as measured in Tables 4 and 5) with the wage impact of the presence of a works council in the establishment. To proceed, I add to the main sample observations from workers in establishments with no works councils—

but that otherwise respect the restrictions specified in Section 4. The OLS model in Table A3.3 in Appendix 3 shows that workers in establishments with a works council earn 12% more than the rest once taken into account the standard covariates.²⁰ This is very close to estimates previously measured by the literature (e.g., Addison et al. 2010; Kraft & Lang, 2008; Mueller &

Neuschaeffer, 2021). However, many factors may bias these estimates. Thus, Columns (4) to (6) show that the works council premium becomes nonsignificant when a fixed effects model is used, with a point estimate in the manufacturing sector of +2.6% (resp. −3.7% in the service sector). These latter results are similar to previous articles using time variations as a source of identification (Grund & Schmitt, 2013; Guertzgen, 2009; Kraft & Lang, 2008; Mueller &

Neuschaeffer, 2021).²¹ It is worth adding that, in both the manufacturing and the service sectors, the point estimates of the works council and works councilor premiums are not far from each other.

Robustness checks

In this section, I perform robustness checks. They are first motivated by the risks of meas- urement error stemming, on the one hand, from the self- declared feature of the presence of a works council in the establishment and, on the other hand, from the recoding procedure described in Section 4. To prevent these pitfalls, I build an alternative sample that includes workers from firms with more than 200 employees, independently on whether or not they declared the presence of a works council in their establishment. It is known that there is a strong positive correlation between the presence of a works council in an establishment and its size.

According to Addison et al. (2017), approximately 90% of establishments with more than 250 workers have a works council. The same restrictions as those applied to the main sample apply otherwise.

20Namely: age, seniority, SES, firm size, the month of the interview, and a dummy controlling for whether the number of weekly hours is specified in the working contract.

21All these articles but one finds no effect of works councils on wages. Guertzgen (2009) finds a works council premium of 3% in the manufacturing and mining sector, which is close to my point estimate for manufacturing. Interestingly, other articles that use a Heckman selection model do find a significant works council premium. For instance, Addison et al. (2010) find a positive relation between works council coverage and wages that is larger in the service sector than in the manufacturing sector (resp. 6.8% and 4.5% in the absence of CBA and 12.2% and 9.9% in the presence of CBAs).

(17)

This alternative sample includes 8885 observations. Table A4.1 in Appendix 4 shows that all conclusions previously mentioned regarding the association between works council membership and wages still apply to this sample with no strong change in nature or magnitude.

Conversely, correlations between wages and union status described in the baseline analysis are not found in this sample of large firms as the effect loses its significance in the service sector.

The larger coverage of CBAs may explain this evolution.

Furthermore, I argue that the opposite sign of the works councilor premiums found in the manufacturing and service sectors are not driven by differences in the productivity, location, or size of firms or in the level of rent creation to expect from works councils.

Gross hourly wages, which capture the productivity dimension, typically differ between the manufacturing and service sectors, which may limit the comparability of the results.

However, the difference vanishes when trade industries are removed from the service sample (see Table A4.2) and Table A4.3 shows that the baseline results persist when the analysis is led on the manufacturing sample and on this comparable subsample. Likewise, Table A4.4 shows that the results do not pertain to manufacturing establishments being more often situated in West Germany (where the tradition of collective bargaining is stronger).

Finally, the literature has shown that works councils enhance performance in both sectors to a rather similar extent (Brändle, 2017; Frick & Möller, 2003; Mueller, 2015). If anything, the positive impact of works councils on the value- added per worker is even stronger in the service sector (Mueller, 2015). The works councilor premium does therefore not result from the sharing of the rent created by the works council.

Taking into account individual- specific wage trends

The baseline model does not control for the possibility of wage trend differences between switchers— who switch status from noncouncilor to councilor or the reverse, within an establishment that has a works council— and nonswitchers in the sample. The estimates given in Table 4 could therefore simply be the result of a positive (respectively, negative) association between wage trends and the probability of becoming a works councilor in the manufacturing (resp. service) sector. I provide different arguments in Appendix 5 that point together against the possibility of a correlation between an individual's probability of becoming a works councilor and the evolution of the average wage in their establishment.

There remains, however, a risk that individual- specific variations in wages correlate with the probability of becoming a works councilor. What follows tests for this risk by compar- ing wage trends between “councilors- to- be” and workers who will never hold office and then prevents it. Due to data limitations, the procedure is only performed for the manufacturing sector.²²

To use wage trends of sufficient length, I build a new yearly sample (hereafter called the

“updated sample”). I assume that none of the respondents entered or left the works council twice between two waves of the main sample. This is a safe assumption, as there is normally one election at most between them (see note 7). Based on this assumption, I can impute works council memberships for waves with no information on industrial relations as long as it remained the same in the two surrounding waves of the main sample for the individual. For example, if an individual was out of office in 2001, 2003, 2006, and 2007 and then observed in office in 2011 and 2015, I assume that he was also out of office in 2002, 2004, and 2005 and in office between 2011 and 2015; observations between 2007 and 2011 are not used. The other restrictions specified in Section 4 apply. The “updated sample” is the restriction of this dataset

22Too few individuals are observed over a sufficiently long time as a noncouncilor before becoming a works councilor in services.

The equivalent of the number in Table A6.3 for the service sector is only 17.