Everything we do know (and don’t know) about collective bargaining: The Zeitgeist in the academic and political debate on the role and effects of collective bargaining

This article provides an overview of academic and public policy debates on the role and effects of collective bargaining. The motivation behind this article is that the academic and political debate is – and ever was – characterized by many controversies. It is explained that these controversies often arise because of different disciplinary, theoretical and empirical approaches. It will also be outlined how the empirical and theoretical debates influenced the Zeitgeist in public policy making. Hence, the article provides an overview of the knowledge on the role and effects of collective bargaining as well as how this knowledge influenced and guided (or not) politically initiated institution building and reforms of collective bargaining systems.


Introduction
The determination of wages and working conditions via collective (wage) bargaining, i.e. via negotiations between trade unions and employers or their representatives is one important characteristic of today's industrialized democratic economies. Even though collective bargaining is a key element for the functioning of labour markets and takes place in all industrialized democratic countries, the way it is conducted and how it functions, as well as its role and impact for actual wages and working conditions of workers, differ substantially across the industrialized democratic world. More specifically, there is a large variation in the coverage of collective agreements as well as in which areas of working life are affected (European Commission, 2014;Hayter, 2011;OECD, 2018;Traxler et al., 2001). Apart from these differences, there is also a variation in how bargaining works, i.e. in how the bargaining process is structured, organized and institutionally embedded (Arnholtz et al., 2018;Bechter et al., 2012;Crouch, 1993;Jansen and Lehr, 2022;Larsson and Törnberg, 2021;Locke and Thelen, 1995).
In this article it will be argued and explained that differences in the characteristics of collective bargaining are the reasons why there are differences not only for workers themselves, but also for the impact collective bargaining has on the 'performance' of different firms, sectors, and even countries. It is also argued that these differences are at the centre of both academic and political debates on the role collective bargaining should have and whether collective bargaining systems in different countries, i.e. different institutional and organizational forms of collective bargaining, should be reformed or not. On that basis it will be explained that even though there is an agreement in the literature that these differences matter, there is substantial disagreement on what exactly the implications are and which characteristics or features of different industrial relations systems are having which effect.
Motivated by this disagreement in the literature, this article gives an overview of what is known and what is (still) not known in the academic and political debate on the performance of collective bargaining by focusing on academic literature that is incorporated in studies that aim to inform public policy making on reforms of collective bargaining systems. Furthermore, it is also explained how the academic and theoretical debate changed and evolved since the early 1980s as well as how this debate influenced (or not) policy makers. In this context the article argues that policy makers appear to follow a Zeitgeist in making decisions that is not necessarily in line with the latest academic knowledge. However, the article also explains the limitations policy makers have to face since the functioning of collective bargaining systems and their performance are highly contingent upon a number of contextual factors that are not in the hands of policy makers.
As regards the literature and debates covered in this work, not all available literature formed the basis but a selection and a focus on seminal and influential work has been made. As will be shown and argued, this work is predominantly based on the macro-level cross-country studies since the 1980s, even though the relevance of micro-level studies increased recently. Since cross-country macro-level studies provide policy makers with information of a wider spectrum of the performance of collective bargaining systems, such work was traditionally incorporated in studies by international organizations such as for example the European Union (EU), the International Labour Organization (ILO), the International Monetary Fund (IMF) and the Organisation for Economic Co-operation and Development (OECD). Hence, this article does not cover all literature that is available on the effects and performance of collective bargaining but rather concentrates on seminal and influential work that aims to inform policy makers. Although much of the theoretical literature that is used in the academic and political debate is generalizable and potentially relevant for policy making all over the world, the fact that empirical studies and relevant debates focus almost entirely on industrialized democratic economies only, all conclusions drawn in this article are also limited to this domain. From a temporal perspective, the article discusses the academic and political debates from the early 1980s onwards even though references to previous seminal work is made if necessary.
The next section provides an overview of the different assumptions, indicators and methodological approaches that are used in the academic and political debate on the performance of collective bargaining. On the basis of that the article describes how the academic and political debate evolved over time, which is followed by an overview of the status quo in the debate and how it came to this status quo. The article finishes by summarizing everything we now know about the impact of collective bargaining as well as what we still do not know. As regards the latter, the article argues that because there are still unknowns on what the impact of collective bargaining is, decision makers should be more careful than before in policy making.

Different assumptions, 'performance' indicators and methodological approaches
The academic and political debate on the role and effects of collective bargaining is -and ever was -characterized by a number of controversies. While the reasons behind some of these controversies are based on differences in empirical and methodological factors used in different analyses, others are based on differing theoretical and ideological assumptions on the functioning of markets and the role and effects of labour market institutions in general that lie underneath the debate.
More specifically, some controversies in the literature on the role and effects of collective bargaining arise over differences in the perception of issues such as how perfect (labour) markets work as well as what the effects of institutions are and if markets are in 'equilibrium' or not. Hence, controversies arise on 'old' and fundamental assumptions on how markets work (or not) and if market 'forces' are able to bring supply and demand into balance or not. With respect to collective bargaining, such assumptions concentrate on the question of whether trade unions are able to enact their power resources (Refslund and Arnholtz, 2021) and 'distort' the functioning and 'clearance' of markets, or if they simply rebalance power asymmetries (Korpi, 1978) and therefore correct labour market imbalances and help to overcome imperfect markets.
This also means that different assumptions are made on the question of whether the power of trade unions and their involvement in collective bargaining simply lead to a different, e.g. fairer, distribution of income and rents, or if market efficiency is impeded (Breda, 2015;Nickell and Layard, 1999). Moreover, these differences appear to be irreconcilable not least because of different philosophical or even ideological stances within different schools of thought, but also appear to persist over time. Furthermore, differences in the academic debate have also translated into different policy recommendations and actions on how collective bargaining systems should be reformed (Arpaia and Mourre, 2009;Blanchard et al., 2014;van Ours et al., 2016).
These different assumptions on the functioning of labour markets and the role of institutions not only influence (empirical) analyses as such, but also the interpretation of results and what the perception of 'well performing' collective bargaining is. As regards the latter, much of the literature discusses the effects collective bargaining and uses the term 'performance' of collective bargaining -which is a terminology that was coined by the influential work of Calmfors and Driffill (1988). Although the term performance is vague and subjective since it could be expressed by many indicators, the literature concentrates predominantly on three indicators in order to express the performance of collective bargaining, i.e. on wages, employment and inflation. As regards the last, because of the fact that inflation is caused by many factors, the literature on the performance of collective bargaining usually concentrates on wage inflation, which describes the relationship that wage increases (can) lead to increases in prices because of an increased demand for goods and services.
While high employment and low (wage) inflation are associated with a good performance, the question whether high or low wages are beneficial is debated differently. However, the focus on the first two indicators is expressed by the fact that they are the main (or often only) factors in the utility function of unions and employers in economic textbooks on collective bargaining (Booth, 1994;Doucouliagos et al., 2017). Furthermore, concentrating on these two indicators can also be explained by the fact that they are directly negotiated in the bargaining process. Nevertheless, the focus on these two indicators neglects the fact that collective bargaining does not focus exclusively on wages and employment and that these are the only components in the utility function. Therefore, it is important to keep in mind that other (non-wage) topics and issues can be even more important and conflict-ridden in the negotiation process (Braakmann and Brandl, 2021). Even though the focus on wages and employment is evident in academic and political debates on collective bargaining, other studies exist which take a more holistic perspective on the effects and performance of collective bargaining by using other indicators such as, for example, equality and social peace as well (Garnero, 2021;Traxler and Brandl, 2009). Nevertheless, since the main academic and political debates on the performance of collective bargaining concentrate on the indicators wages, employment and wage inflation, the following discussion will also use these three indicators as principal reference.
Furthermore, controversies are also based on differences in the methodological approach which is used. More specifically, there are differences between micro-and macro-approaches for the analysis of the impact of collective bargaining. Although micro-and macro-approaches differ in many ways, some controversies arise around the use of different indicators in empirical analyses. Common in micro-level analyses are indicators such as nominal wage increases, wage premia, wage dispersion, 'excess wages', labour costs and labour productivity in order to evaluate the impact of collective bargaining (Brändle and Goerke, 2018;Dahl et al., 2013;Daouli et al., 2013;Gartner et al., 2013). Typical for macro-level cross-country studies is the use of aggregate indicators such as (changes in) (nominal and real) wages, unit labour costs, (un)employment, inflation, taxation and (wage) inequality (Brandl, 2012;Calmfors and Driffill, 1988;Flanagan, 1999;Johnston and Hancké, 2009;Soskice, 1990;Traxler et al., 2001). This means that even though both micro-and macro-level analyses do not necessarily differ with respect to the concentration on wages and employment as well as the fact that there is clearly an overlap in the basic ideas of the indicators, the empirical operationalizations and conceptualization of these factors do differ and lead to different results and conclusions. In addition to that, the empirical results of both micro-and macro-level studies differ because in different studies different country samples are used and different periods of time are considered in the analysis. Furthermore, empirical restrictions because of the availability of a limited number of observations also lead to limitations in the empirical estimation strategy and model uncertainty. These different empirical approaches together with empirical limitations all explain why the results presented in empirical studies and the conclusions that draw upon these results on the performance of collective bargaining differ within the literature.
In addition to the aforementioned differences in the debates, the literature is also characterized by the presence of different theoretical approaches. As will be argued in the following, even though some of the different approaches are 'only' refinements of others, the implications on the impact of collective bargaining are substantial and therefore also lead to very different public policy conclusions. In addition, empirical research, which aims to test different theoretical approaches, is often not clear-cut or able to resolve controversies because it is faced with a number of limitations. This is especially so when it concerns clarifying different implications which are derived from different methodological analyses. Most importantly, typical micro-level analyses are predominantly based on within-country empirical observations or a small cross-country sample which are only able to capture a small variation in different collective bargaining systems and therefore are not generalizable for the performance of collective bargaining systems per se. Microstudies have the advantage that they are able to capture the small country sample context well, but since the context is different in other countries, these studies are not suitable to make any inferences and policy recommendations regarding other countries. Such small variation of different collective bargaining systems is especially evident for micro-level studies in economics (Addison et al., 2014;Barth et al., 2020;Gürtzgen, 2016), but is also characteristic for studies which are based on case studies and are rooted within other areas of social sciences (Crouch, 1993;Dore, 1973;Meardi, 2018), even though they tend to consider a wider range of different systems. Macro-level analyses (Aidt and Tzannatos, 2002;Brandl, 2012;Johnston and Hancké, 2009; OECD, 2019) capture a larger variation in collective bargaining systems across different countries thus allowing for a higher degree of generalizability, and are more suitable to make policy recommendations with respect to the performance of a wide spectrum of collective bargaining systems. However, the disadvantage of macro-level studies is that they are often constrained by a limited number of observations which challenges the empirical validity of the analyses.
Yet despite the differences and controversies on the impact of collective bargaining, there appears to be a Zeitgeist in the academic literature over recent decades in the sense that there are some theories and approaches that have dominated. However, this Zeitgeist that is argued to exist in the following certainly does not imply that the theories and approaches have been unrivalled either academically or politically. Against the background that industrial relations and collective bargaining are very different in different countries and sectors and their functioning highly dependent upon a number of contextual factors, policy makers have nevertheless generally (or blindly) followed the Zeitgeist. In this sense, the following concept of a Zeitgeist rather describes a set of perceived best knowledge and beliefs about the functioning and effects of collective bargaining that are (loosely) shared by the majority.

The evolution and Zeitgeist of the theoretical debate
Collective bargaining had its heyday in the 1970s when in almost all Western European countries strong trade unions bargained over wages and working conditions for a large share of workers (Crouch, 1993). Therefore, it is not surprising that at the same time theoretical discussions and empirical studies on the impact flourished. However, what is somewhat surprising is the fact that at that time, i.e. from the early 1970s until the 1980s, the impact collective bargaining had was either (largely) dismissed in mainstream (economics) literature or regarded as inefficient for the functioning of markets (Freeman, 2008). This is because collective bargaining, as one form of a non-market institution, was considered to be either not necessary or even disruptive for the efficacy of the labour market and therefore causing suboptimal performances.
This perspective rested on the critical assumption of perfect markets and was strongly challenged by empirical evidence and theoretical developments from within the wider field of industrial relations and institutional economics. As regards the latter, further research gained inspiration and momentum from the work of North (1990). In subsequent debates more attention was given to differences in the institutional structure and functioning of labour market institutions in general (Nickell and Layard, 1999) and in particular to different systems of collective bargaining, i.e. to differences in the level and degree of coordination of collective bargaining (Freeman, 2008;Hall and Soskice, 2001). More specifically, the academic literature shifted increasingly towards analyses of the impact of different organizational structures of collective bargaining in which specific configurations of power, interests and norms have formed (Refslund and Arnholtz, 2021) and arguments over their relative importance and implications on outcomes such as wages, employment and (wage) inflation.

Demand-and supply-side perspectives
As regards the impact of collective bargaining in general, the literature in the 1980s started more strongly to distinguish between two main kinds of effects of collective bargaining. First, collective bargaining has implications for the demand side of the economy. This is because wages as well as employment are directly or indirectly set or regulated in a negotiation process between representatives of workers and employers and the outcome of these negotiations, i.e. the level and changes in wages and employment, affects the purchasing power of the workers, and therefore of consumer demand and prices.
Second, the literature in the 1980s became increasingly aware that collective bargaining also has implications for the supply side, since bargaining outcomes affect labour costs, (labour) productivity and therefore the competitiveness of firms (Addison and Hirsch, 1989;Doucouliagos and Laroche, 2003). Hence, collective bargaining also affects the ability of companies to employ workers. Moreover, since collective bargaining may also cover many non-wage issues, additional supply-side effects including skill formation as a factor for competitiveness are affected. The latter impact of collective bargaining on skill formation, i.e. qualifications, (vocational) training and development of the workforce, should certainly not be underestimated as skill formation is associated with a fundamental collective action problem and therefore with the systematic risk of market failure (Traxler and Brandl, 2009).
Furthermore, the literature increasingly shifted interest in the fact that the outcome of the negotiations in collective bargaining is likely to have distributive effects not only between 'capital' and 'labour' (i.e. on the wage/profit share), but also among the workforce. Since then the academic literature (overwhelmingly) agrees that almost any system of collective bargaining tends to lead to a more egalitarian wage distribution within the workforce (Aidt and Tzannatos, 2002;Breda, 2015;Carlin and Soskice, 1990;Flanagan, 1999;Hayter, 2011;Traxler and Brandl, 2009). However, independently from within-group effects on the worker and employer side, there can also be a trade-off between demand-side and supply-side goals and interests. The question of whether the demand-side or supply-side effect has more weight and in particular which side should be prioritized is largely a question of economic and political perspective as well as one of power and strength between the two sides in the bargaining process, i.e. between trade unions on the one side and the employers on the other (Freeman and Medoff, 1984;Rychly, 2009). Ideally, if both sides are 'equal' partners and share mutual interests, collective bargaining can be a way to synchronize the interests of both parties in order to achieve mutually beneficial outcomes. In this sense, collective bargaining also balances the power and ability to represent the interests between capital and labour (Korpi, 1978).
Since collective bargaining is a means to synchronize demand-and supply-side interests it could enable wage setting in such a way that consumer demand is stimulated fostering economic growth while also (excessive) wage demands are moderated, i.e. wage inflation is mitigated, securing (long-term) employment as well as firm competitiveness. Thus, collective bargaining enables the worker and employer side to synchronize different demand-and supply-side interests and goals. Furthermore, if collective bargaining covers a large domain, i.e. a substantial number of workers in a sector or country as was the case in the 1980s in the industrialized democratic world, the implications for the outcomes of bargaining are, of course, of a wider, e.g. country wide, relevance. However, independently of the range of the domain, the ability of collective bargaining to synchronize demand-side and supply-side considerations is usually described as wage synchronization, or as the ability of the bargaining system to internalize the (negative) consequences/effects of the bargaining outcomes.
In the 1980s the academic literature became increasingly aware of the importance that via collective bargaining the demand and supply side can (and need to) be synchronized. Upon this awareness different theoretical approaches arose which share the assumption of collective action from Olson (1965) that different collective actors have differing abilities to maximize their (self)interests but differ upon what the consequences of these differences are. More specifically, the next generation of theoretical approaches which emerged in the 1980s have in common the idea that the ability of different bargaining systems to synchronize wages depends very much on differences in collective bargaining systems and on the different (collective) actors associated with different abilities of bargaining units to externalize negative costs upon others, i.e. upon the economy and society. However, the approaches differ fundamentally over the nature and functioning of the different systems.

The corporatist thesis
The first main theoretical approach which emerged in the early 1980s that explicitly investigated different systems of collective bargaining, i.e. different institutional and organizational structures of collective bargaining in which specific configurations of power, interests and norms have formed, is the so-called corporatist thesis. The corporatist thesis (Cameron, 1984) dominated the academic and political Zeitgeist of the 1980s and argues that a linear relationship exists between the degree of centralization of bargaining and its performance. The higher the degree of centralization, the better the performance of the bargaining system. More specifically, the corporatist thesis argues that the ability of bargaining actors to synchronize wage setting in such a way that real wages are moderated (i.e. not set 'excessively' high) so that wages and employment levels are set in a mutually beneficial way, increases with the degree of centralization. The higher the level at which collective bargaining takes place, the greater the ability to synchronize wages, i.e. to moderate wages, because at higher levels of centralization the more difficult it is for bargaining actors to pass negative wage setting externalities to others. For example, if (in a closed economy) collective bargaining is completely centralized and only one collective agreement covers all firms and workers within a country, both the worker side and the employer side need to take into account all negative externalities in setting wages and therefore can be expected to set wages in such a way that mutually beneficial outcomes for the whole economy are maximized.

The hump-shape thesis
According to the corporatist thesis, the performance of a sectoral or national level collective bargaining system is higher than a system in which collective bargaining takes place at company level. Unfortunately, this theoretical reasoning did not match empirical observations at the time. In fact, countries in which collective bargaining at company level is characteristic performed relatively well compared with countries in which higher level bargaining was characteristic. This mismatch between theoretical reasoning and empirical evidence lead to the development of the hump-shape thesis proposed by Calmfors and Driffill (1988) in the late 1980s.
Even though the hump-shape thesis builds upon the basic idea that the more centralized bargaining is, the higher the need to internalize any negative costs -on basis of the assumption of perfect competition in product markets -the thesis integrates the role of market forces. The implication of that is that the relationship between the degree of centralization and real wages is non-linear, i.e. has a hump shape. More specifically, both highly centralized and highly decentralized collective bargaining processes are associated with not excessively high wages, i.e. a higher economic performance in this interpretation, while all levels in between, i.e. sector-level bargaining, are associated with (potentially excessive) high wages and therefore a detrimental economic performance.
The reason why highly centralized collective bargaining is associated with a beneficial performance is the same as that underlying the corporatist thesis, i.e. bargaining parties are unable to externalize any negative costs of wage setting. However, the argument behind the beneficial effects of completely decentralized bargaining is that bargaining parties and units, i.e. company level trade unions and the management of companies, are unable to externalize any negative costs onto others because (perfect) competition in product markets does not allow them to set (excessively) high wages. In other words, if unions in a firm which is exposed to competition with other firms pushed to enforce excessively high wages, i.e. wages that are higher than wages of competitor firms, then everything else being equal, the firm would not be able to survive. This market pressure in wage setting is irrelevant if collective bargaining takes place at the sectoral level, as sector level collective agreements establish a sector-specific cartel. This means that a sectoral collective agreement establishes common wages within the sector so that competing firms are all affected in the same way. However, such a wage cartel would also allow bargaining parties to externalize negative costs to others, especially to firms and workers in other sectors as well as to the public. Under the assumption that sectoral actors aim to maximize their self-interest it can be expected that all sector actors try to pass on the costs to others and therefore cause detrimental effects to the whole economy. This detrimental effect can be expected to be high in a closed economy in which firms within sectoral cartels do not compete with firms in other sectors.
In the case of open economies, the fact that sector level collective bargaining forms cartels that allow the costs of wage setting to be passed onto others makes the externalization of costs especially attractive for firms in sectors that are not exposed to international competition. Until the 2008 crisis, the public sector was a good example of a sector sheltered from international competition and in which powerful trade unions would be able to push and enforce high wages (and good working conditions) at the expense of firms and workers in all other sectors (Bach and Bordogna, 2016;Garrett and Way, 2000). Sectors that are highly exposed to international competition, such as manufacturing, are especially vulnerable and would need to pay these costs since these sectors need to maintain their competitiveness in global markets.

The key relevance of the hump-shape thesis for policy making and institution building
The logic of the hump-shape thesis was found to be convincing both in the mainstream academic and by the policy making community and in the early 1990s it became the predominant theoretical backbone of almost all studies. Furthermore, the policy implications that can be derived from the hump-shape thesis have been highly influential among policy makers and this is well documented in a number of policy relevant reports by national and international organizations (IMF, 2016;OECD, 1994OECD, , 2019. Not least when labour market institutions were about to be established in many Central and Eastern European countries (CEECs) in the early 1990s it became clear that the hump-shape thesis expressed the Zeitgeist in all debates and policy advice (Meardi, 2013). More specifically, since the hump-shape thesis sees collective bargaining at the sectoral level to be associated with the worst performance of all available options it is therefore hardly a surprise that decentralized bargaining, rather than sectoral bargaining was introduced in most CEECs. Furthermore, decentralized bargaining was also the policy option that was easier to implement since no further institution and capacity building was needed. More specifically, there was no need to strengthen or create from scratch employers' organizations and trade unions at higher level.
However, one fundamental problem of the hump-shape thesis is that it does not match a number of empirical observations on the performance of the degree of centralization in collective bargaining systems. Most notably, the hump-shape thesis failed to explain why countries that are characterized by sectoral agreements perform exceptionally well. For example, countries like Austria, Denmark, Germany and Sweden, where encompassing sectoral agreements were -and in some countries still are -the norm, perform exceptionally well and research showed that the performance of countries with higher level collective bargaining systems can be even better. In fact, the empirical support for the hump-shape thesis and the relationship between performance and the degree of centralization is rather indeterminate (Calmfors, 1993;OECD, 2019). Against the background that the empirical support for the hump-shape thesis is not (and actually never was) convincing it comes as a surprise that it dominated the Zeitgeist for a long time.

Current challenges and the Zeitgeist in academic and policy making and debates
The fact that the hump-shape thesis was not able to explain the situation, i.e. the empirical facts, correctly has motivated further theoretical developments that all have in common that further factors intervene in the relationship between collective bargaining and economic performance. Nowadays, empirical studies and theoretical discussions are (usually) more complex and provide a nuanced picture (OECD, 2019). As will be shown in the following, subsequent theoretical developments from the late 1990s onwards considered a number of additional factors that were identified to be of key importance. On the one hand, this has led to a better and deeper understanding of the impact of collective bargaining, but on the other hand, has made theoretical and empirical analyses more complex and challenging. Furthermore, different theoretical pathways were developed which all had in common that they aimed to overcome three kinds of shortcomings of the hump-shape thesis.
The first shortcoming is that the hump-shape thesis only holds under the assumption of perfect markets and competition, which rarely applies in today's product and labour markets, which are often highly regulated. Second, one of its core arguments, that sectors form a cartel, only holds for a closed economy context in which the sector level bargaining parties face a demand curve, which is not very elastic. As outlined earlier, in an open economy only a few sectors are able to form a cartel and with the increased internationalization and globalization of economies, the ability of national sectors to form cartels became a (rare) exception. Third, and perhaps most importantly, the hump-shape thesis focuses only on the level at which collective bargaining takes place but ignores holistic differences in the organizational functioning of the bargaining system itself.
More specifically, as argued by Soskice (1990), since the hump-shape thesis only looks at the level of bargaining, it ignores the fact that the horizontal sectoral coordination of collective bargaining can work as a functional equivalent to centralized bargaining. This means that sector level bargaining if horizontally coordinated across all sectors in an economy can be expected to have (at least) the same beneficial performance effects as national level bargaining. If bargaining is horizontally coordinated and any negative costs are not externalized, i.e. bargaining units pursue a wage setting strategy that looks on the mutual interests of all units, all costs of wage setting have to be internalized. Moreover, as shown for example by Brandl (2012), if horizontal coordination is achieved via exposed sector pattern bargaining, i.e. the exposed sectors set the pattern for the bargaining process of all sector bargaining units (Traxler et al., 2001), then horizontally coordinated bargaining can even outperform other forms of bargaining.
As regards vertical coordination, in the case of multi-level collective bargaining systems, Traxler (1995) showed that the effects on wages and employment can be very different. More precisely, depending upon the ability of higher level bargaining units to govern the actions of lower level bargaining units, the effects of wage setting at different levels can be very different. Basically, Traxler (1995) argued that any strategy on wage setting by higher level collective bargaining units can only work if it is supported (or at least not contradicted) by lower level wage setters. For example, without any support from the rank-and-file, lower level bargaining units might bypass or even stand against any higher level wage setting strategy if the higher level strategy fails to reflect the narrow interests of the lower level bargaining parties. This means that lower level bargaining units might also be able to externalize their costs of wage setting to others and therefore a similar collective action problem exists as for horizontal coordination. As shown by Traxler et al. (2001), if bargaining is vertically coordinated, its performance can be economically beneficial, while it is detrimental if not vertically coordinated.

The complexity and reality of today's collective bargaining systems
According to the argumentation before, a completely centralized bargaining system can be considered to be a well-performing bargaining system even though coordinated collective bargaining can also be considered to perform well. Therefore, policy makers might consider strengthening or building up a centralized bargaining system in a country not least because this might be the easiest system to implement. Such a policy making perspective, however, ignores the fact that in all industrialized countries, the freedom of association, i.e. the right to form unions and bargain collectively, is a fundamental human right (Gernigon et al., 2008;Hayter, 2011). Therefore, different collective bargaining units can form and organize themselves at any preferred domain, i.e. at any level, and act independently from each other. As shown in recent literature (Brandl and Bechter, 2019;Eurofound, 2016;Marginson, 2015), in the past decades bargaining units were increasingly formed at lower levels in addition to higher levels and therefore completely centralized collective bargaining systems rarely exist (anymore) among industrialized democratic countries. Therefore, the important question for current research is to investigate and understand the functioning and performance effects of these hybrid, complex multi-level systems.

The contingency thesis: The conditional effects of centralization and coordination
Against the background that any collective agreement beyond the decentralized level can (or might) be confronted with any lower level collective agreement, the question arises whether the outcomes of the agreements conflict or complement each other. A clear conflict exists if one bargaining unit aims to free ride and the agreement tries to bypass the other one by trying to externalize any negative costs onto others. Such a conflict could exist if, for example, a sectoral or national collective agreement that set real wages according to the long-term mutual interests of all parties involved is rivalled by a company collective agreement, e.g. from a firm in a sector that is not exposed to international competition and that sets wages 'excessively' higher according to the self-interest of the company. Of course, the same incentives to free ride exist for companies and sectors to deviate and bypass higher level collective agreements. Thus, coordination between different collective bargaining units is of fundamental importance in order to avoid any free-riding by lower level bargaining units.
However, the problem of free-riding and therefore the performance effects of the level of centralization, on the one hand, and the importance of horizontal and vertical coordination, on the other hand, depend also on a number of factors including the level at which bargaining takes place, the sizes of the domain of the bargaining unit at different levels as well as how much overlap there is in the domain.
More specifically, as regards differences in the size of bargaining units and differences in the level, smaller units have more incentives to free ride because free-riding of smaller units causes lower costs for others than free-riding of larger units. This means that the costs of smaller units can be externalized much more easily than the costs of larger units. Also, the bigger the difference in size the more incentives for smaller units to free ride. Since bargaining units are smaller at the most decentralized level and the higher the level the larger the bargaining units, the larger the difference and the problem of free-riding if bargaining takes place at the lowest level and the highest. The implication of this is that the importance of vertical coordination for performance increases with the difference in levels at which bargaining takes place as well as the number of bargaining levels that cover the same domain.
This also means that if collective bargaining takes place at different levels, there is a negative relationship between economic performance and the level of bargaining. The reason for this is that potential free-riding that leads to a negative performance becomes more likely the higher the level of collective bargaining in addition to lower level bargaining. For example, if bargaining takes place at the highest level, i.e. at national level, the incentives for lower level units to free ride are highest. Hence, there is a negative relationship between the level of bargaining and potential negative performance. Consequently, the higher the level of bargaining the more damaging free-riding can be and therefore the more important vertical coordination and the need to govern lower level units. This negative relationship, however, does not exist for the need of horizontal coordination. In fact, the logic is different since coordination between units at the same level, i.e. horizontal coordination, becomes less important the higher the level of collective bargaining. The reason for this is that at a higher level the bargaining units are increasingly encompassing and therefore the less ability they have to externalize costs exists and they are thus able to free ride. In fact, at the highest level, i.e. at national level, horizontal coordination becomes obsolete since there is only one bargaining unit.
This implies that the logics of horizontal and vertical coordination of collective bargaining are different and therefore need to be investigated separately. These different logics are considered by the contingency thesis, which is based on the idea that the performance of the degree of centralization and horizontal coordination is contingent upon the degree of vertical coordination of collective bargaining. If the collective bargaining system is not vertically coordinated, the performance decreases with the level of centralization as well as with the degree of horizontal coordination. Vice versa, if the bargaining system is vertically coordinated the performance increases. Recent empirical research has shown that the contingency thesis can explain empirical observations better than the corporatist and hump-shape thesis (Braakmann and Brandl, 2021;Brandl, 2012;Traxler et al., 2001). Most notably, the contingency thesis is able to explain how coordinated sectoral collective bargaining systems as well as some coordinated multi-level systems can perform exceptionally as well. However, this research also shows that this beneficial performance of some bargaining systems is also contingent on the functioning of coordination, which is something that cannot be taken as granted. In fact, the coordination of collective bargaining is a complex process that is also highly contingent on the wider socio-economic and socio-political system in which it is embedded. The functioning is highly context-sensitive (Locke and Thelen, 1995) and research has shown that some forms of coordination are also fragile (Brandl and Ibsen, 2017;Roche and Gormley, 2020). Research has also shown that the coordination of collective bargaining is a feature of the bargaining system that usually evolved from the bottom up over (a long) time. Therefore, it cannot be simply imposed or centrally decided by 'decision makers' in a top-down way of policy making. Hence, even though the contingency thesis clearly offers well-performing policy options, the implementation of these options is politically difficult and therefore disadvantageous.
Conclusions: What do we know about the performance of collective bargaining?
. . . there are things we know that we know. There are known unknowns. That is to say, there are things that we now know we don't know. But there are also unknown unknowns. There are things we do not know we don't know . . . 1 This article provided an overview of not only the main theoretical concepts that influenced academic and policy making discussion on collective bargaining, but also how debates evolved and changed over time. As explained, all theories share some common grounds which are based on collective action theory and the ability of bargaining parties to align, i.e. synchronize, mutual long-term interests of both sides. More precisely, all approaches share the idea that well-performing collective bargaining systems need to be able to balance the power and interests between labour and capital, i.e. synchronize demand-side and supply-side considerations, and must internalize any (negative) consequences/effects of their bargaining outcomes. The outcome or consequence of the alignment of the mutual long-term interests is that wages are set in a way that employment is high and wage-induced inflation is low.
However, the theoretical concepts differ with respect to which characteristics of the bargaining system matter for the ability to synchronize wages and to internalize costs and benefits of bargaining outcomes. More precisely, the theoretical concepts differ with respect to the relevance of achieving a superior performance of the level of centralization as well as with respect to horizontal and vertical coordination. Furthermore, since research on the performance of collective bargaining has to face a number of empirical problems empirical tests are not clear-cut but only provide researchers and policy makers with some evidence that favours different approaches. Thus, there is still some uncertainty in the debates, which causes academic and political controversies on the performance of collective bargaining.
Nevertheless, there are things that we, i.e. the academic community, know even though there is some uncertainty in the literature and without any doubt, there are things we do not know. Also, in analogy to the famous quote, this article showed that there are also things we thought that we knew but now we know that we were wrong. In other words, we thought that we knew what the performance of collective bargaining is but since subsequent research challenged this knowledge, we later found out that we did not know earlier. This means that although we learnt that we did not know earlier and therefore know more today, we still need to be aware that there are, of course, unknown unknowns in the debate on the performance of collective bargaining.
To summarize what we know: we definitely know that collective bargaining systems differ and that differences matter with respect to performance. We also know that these variations are mainly based on differences in the level at which bargaining takes place and in particular on the coordination of bargaining.
To summarize what we thought we knew but now we know that we did not know it before: against the background that the hump-shape thesis dominated academic and policy relevant debates and it was thought that sectoral collective bargaining was associated with an inferior performance, we (now) know that sectoral collective bargaining is not necessarily associated with a detrimental performance. In fact, we now know that the impact, i.e. performance, of sectoral collective bargaining is highly contingent upon horizontal and vertical bargaining. Moreover, as subsequent literature shows, if sectoral collective bargaining is coordinated, its performance can even be superior to any other form of collective bargaining. However, the literature also shows that its performance can be inferior if not coordinated.
Hence, the coordination of higher, i.e. sector, level collective bargaining appears to be a very important element for the performance of collective bargaining systems. Against the background that single-level bargaining systems were increasingly replaced in recent decades by multi-level bargaining systems, the importance of vertical coordination even increases. Furthermore, we now also know that encompassing higher level coordinated collective bargaining can exert positive performance effects because a cartelizing effect is generated. More specifically, since encompassing higher level collective agreements standardize wages and working conditions for all firms and workers within a domain, firms are unable to compete and increase their productiveness via lowering, i.e. dumping, of wages and working conditions. This standardization has the effect that unproductive firms are pushed out of the market and their market share is taken over by firms with higher productivity. Hence, encompassing higher level collective agreements also exert incentives to firms to improve their competitiveness by using (innovative) technologies rather than by the use of cheaper labour. In this sense, the logic of these performance effects is similar to the logic of Streeck's (1997) beneficial constraints where institutional and political constraints lead to a superior performance. In any case, the fact that encompassing higher level collective agreements tend to push less competitive firms out of the market and more productive firms take over their market share, means that the aggregate productivity and competitiveness of sectors or countries increase. Thus, collective bargaining is also an important 'tool' for economic policy making.
However, uncoordinated collective bargaining systems and lower, i.e. firm level bargaining, are unable to guarantee the positive cartelizing effects as higher level agreements might be bypassed or undermined by other units. Also, any other options that allow firms to deviate or opt out of encompassing higher level collective agreements might also undermine the cartelizing effect as 'loopholes' are generated.
Hence, the trend of decentralization of collective bargaining systems and the increase of opt-out options which were introduced in almost all industrialized democratic countries in recent decades (Katz, 1993) needs to be put in perspective. Especially the approach of disorganized decentralization (Traxler, 1995) in which there is no coordination and division of duties between bargaining units at different levels needs to be put in perspective. On the one hand, the disorganized decentralization process and the introduction of opt-out options introduced the necessary flexibility to individual firms to react individually and quickly to changing labour and product market situations. However, this flexibility often came at the expense of workers as it frequently led to a decrease in wages and working conditions. On the other hand, however, this flexibility favoured the supply side, and less attention was paid to the demand side and the effects of lower wages on the purchasing power of the workers. Furthermore, the trend of decentralization undermined the cartelizing effect and protected less productive firms from other firms that were more competitive because of their use of more productive technologies and allowed firm survival through lower wages and poorer working conditions. Unfortunately, which effects on both sides dominate is not definitively known, but the Zeitgeist of disorganized decentralization in the recent past suggests that the flexibility effect was favoured.
However, finally, and to conclude with more known unknowns, i.e. things we do know we do not know: we certainly do not know which collective bargaining systems work best for which country. The latter known unknown thing is especially important as policy advice on the reforms of collective bargaining systems is made based on our academic knowledge of the impact of collective bargaining. Given that the functioning of collective bargaining is dependent upon the context and bargaining systems are very different in different countries, any advice on reforms of these systems should be based not only on consideration of everything we know, but also on what we do not know. Therefore, it might not always be the best option to judge the performance of different collective bargaining systems on the (changing) Zeitgeist and, in particular, it might be ill-informed to derive policy making recommendations on the basis of the same blueprints for each and every country on the basis of the predominant Zeitgeist (as happened previously). It might be more useful instead to look at the nitty gritty complexity of the institutional structure and organizational functioning of collective bargaining in different national contexts and aim to understand their distinct efficacy and traditions. In doing so, some unknown unknowns might be revealed.

Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/ or publication of this article.

Funding
The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD
Bernd Brandl https://orcid.org/0000-0002-3264-0454 Note comparative employment relations and international HRM. He has managed a number of funded research projects and published in top-tier academic journals. He also worked as a policy advisor, expert and consultant to many international organizations including the European Commission and the ILO. Recent work includes the prerequisites and nature of the cooperation between trade unions, employers' organizations and the government in order to address the health-related and economic consequences of the COVID-19 pandemic.