A special fund for gender equality? Institutional constraints and gendered consequences in Swedish collective bargaining

Funding information Forskningsrådet om Hälsa, Arbetsliv och Välfärd, Grant/Award Number: 2018‐00467 Abstract Sweden is often described as one of the world's most gender‐equal societies, but the gender pay gap nevertheless remains large. In 2007, a special gender‐equality fund that targeted women workers was successfully implemented in the collective bargaining agreement. Although it decreased the gender pay gap, it was controversial and has never been employed since. The aim of this article is to increase knowledge concerning the conditions for such “gender‐equality bargaining” through a detailed empirical case study. An in‐depth study of the bargaining process casts light on how the institutional features of bargaining have changed after 2007 in a way that significantly restricts further attempts to make relative wage changes in gender‐segregated labor markets. The findings indicate that increased employer coordination and a strong industry norm appear to hinder gender‐equality initiatives.


| INTRODUCTION
Gender inequalities continue to permeate labor markets worldwide. The gender pay gap across the world was about 20% in 2018, and it has only changed marginally in recent decades (ILO, 2019). In OECD countries, the median full-time female worker earns on average almost 15% less than her male counterpart-a difference that has barely changed since 2010 (OECD, 2017).
In order to foster the incorporation of gender-equality concerns into collective bargaining, previous research has addressed so-called gender-equality bargaining as an efficient means for targeting gender-related inequalities in the labor market (Briskin, 2006;Colling & Dickens, 1998;Gregory & Milner, 2009;Williamson & Baird, 2014).
The aim of this article is to expand knowledge of the conditions for gender-equality bargaining by means of a case study of gender-equality bargaining in Sweden.
In 2007, the Swedish Trade Union Confederation (LO) placed gender equality on the agenda in the collective bargaining process through the introduction of a gender-equality fund termed the "Gender Equality Pot" (GEP), which was to provide a means for equalizing wage differences between men and women. The idea was that there would be an extra wage pot calculated by means of a mathematical formula based upon the proportion of women in each sector with wages below 20,000 SEK (∼2000 EUR). In this way, women with low salaries would receive a higher wage increase than other groups, which would decrease the gender pay gap across sectors.
This measure was successfully implemented and had the intended effect. The unions state that womendominated sectors received wage increases of 4.3% to 4.8%, while male-dominated sectors received 3.4% (SvR, 080530). Although several trade unions have argued that a GEP is the most efficient means for targeting the gender pay gap, it has never been used again. Moreover, it was subject to criticism from the very beginning, not only because it favored low-wage women-dominated sectors at the expense of other groups, but also because of its technical construction, which provided no room for negotiation (interviews 16, 10, 2). The adoption of the GEP comprised a critical case of gender-equality bargaining in the sense that it recognized gender as a relevant category in collective bargaining, thereby challenging class-based interests as the primary concern.
Swedish industrial relations have changed significantly in recent decades toward a neoliberal trajectory that is marked by increased employer discretion, decentralized wage-setting, a high degree of individualization, and labor market flexibility (Baccaro & Howell, 2017, p. 144;Howell & Givan, 2011, p. 237). This shift has taken place to a large extent within the existing collective bargaining institutions by means of changes in institutional practices and functioning (Baccaro & Howell, 2017, p. 144). This article seeks to understand the role of the institutional framework in gender-equality bargaining and cast light on how recent shifts toward greater employer discretion restrict possibilities for implementing relative wage changes in general and gender-equality investments in particular. More specifically, the present article aims to establish how it was possible to adopt a gender-equality pot in 2007, but impossible to do so afterward. The following questions will be addressed: (1) How was the bargaining process of 2007 conditioned by institutional features? (2) Which institutional features made possible the introduction of a gender-equality pot, and have they changed? (3) What are the gendered consequences of institutional changes?
The point of departure for this article is the idea that setting pay levels must be viewed as a political rather than a technical issue, and that gaining an understanding of this question requires "an analysis of divergent interests" of the various actors (Rubery & Grimshaw, 2014, p. 339). Koskinen Sandberg (2018, p. 36) argues in this regard that a sociological perspective which includes social structures, institutions, and actors is better able to study the dynamics of gendered pay gaps than an economic approach, which views wages as reflecting productivity and determined solely by market forces.
The analysis establishes that institutional features have changed after 2007 in a way that significantly restricts further attempts to make relative wage changes between sectors dominated by men and women. The findings indicate that a strong wage norm strictly set by competitive export-oriented industry appears to hinder initiatives that recognize gender as a relevant category in collective bargaining. In addition, a greater coordination among 2 -ERIKSON employers has led to a shift in the balance of power between unions and employers toward the latter, with negative implications for gender-equality concerns. The findings show the usefulness of a gender perspective for challenging the "apparent neutrality of markets" (Rubery & Hebson, 2018, p. 416) and highlighting the need for collective and political efforts in collective bargaining. The article also demonstrates how endogenous institutional change-due to new practices and interpretations of institutional functioning-have implicit, albeit direct consequences for genderequality investments.

| Gender-equality bargaining
The importance of collective bargaining as a means for improving gender equality in the labor market has engaged researchers in the field of gender and organization studies. On a macrolevel, scholars have found that the level and scope of collective bargaining are substantially more important than specific equal-pay policies (Rubery, 1992). In addition, centralized wage setting and high levels of government employment appear conducive to women receiving relatively high salaries (Whitehouse, 1992). Studies comparing union and nonunion workplaces further underscore the importance of collective bargaining. Not only do women employees who are represented by a union enjoy better terms and conditions than those who are not, they also have better job protection (Booth, 1995;Martin & Roberts, 1984;Millward & Woodland, 1995). Colling and Dickens (1998) developed the concept of equality bargaining in order to capture bargaining that targets gender inequalities. They remark that the concept of equality bargaining encompasses the collective negotiation of provisions that are of particular interest or benefit to women and/or are likely to facilitate gender equality ('special measures'); equality awareness on the part of negotiators in handling commonplace bargaining agenda items such as pay and pay opportunities ('gender-proofing'), and the injection of an equality dimension (specifically, addressing gender disadvantage) to the negotiation of change, for example reforming a grading structure (Colling & Dickens, 1998, p. 390).
Although equality bargaining includes bargaining on a range of social inequalities, a focus on gender has come to predominate (Williamson & Baird, 2014). Later work in the field primarily utilizes the concept of gender-equality bargaining (Gregory & Milner, 2009;Williamson & Baird, 2014).
Empirical studies of gender-equality bargaining have sought to identify factors that are favorable for successful bargaining in this regard. In one of the most extensive studies, Heery (2006) investigated equality bargaining concerning equal pay through a survey of paid union officers in British trade unions. He found that while unions have raised procedural issues with the greatest frequency, substantive concessions, such as the extension of pay supplements or benefits to women, are less frequently reported (Heery, 2006, p. 527). Heery concludes that equality bargaining "is a function of women's voice within unions, the characteristics and preferences of bargainers themselves and of a favorable public policy environment" (Heery, 2006, p. 522). Other factors noted in the literature as favorable for promoting equality bargaining are the gender identity of negotiators (Dickens, 2000), the presence of women in union structures (Gregory & Milner, 2009, p. 142), and the formation of coalitions and alliances, both within unions and across sectors (Briskin, 2006). Briskin (2014) argues, however, that the emergence of individual champions of equality is not sufficient to guarantee change, noting that there is also a need for "collective mechanisms which ensure that women's interests are represented," such as constituency committees (Briskin, 2014, pp. 209, 217). While these factors concern conditions within unions, Gregory and Milner (2009) point out the importance of the external structures of opportunity-including policies, projects, and organizational initiatives on the national and European levels-for union engagement with such gender-equality issues as worklife-balance (Gregory & Milner, 2009, pp. 141 and 142). Milner and Gregory (2014) remark that the existence of ERIKSON -3 adequate legal enforcement is a necessary condition for gender-equality bargaining to take place (Milner & Gregory, 2014, p. 259).
Certain studies have highlighted the obstacles to successful gender-equality bargaining. For example, Colling and Dickens (1998) find in their case study of gender-equality bargaining involving British Gas that although there was an ambitious agreement in place, its implementation was flawed. They argue that state involvement and collective bargaining, particularly universal bargaining across several sectors, possess the potential to ameliorate gender inequalities on the labor market, while deregulation and privatization work to diminish gender-equality initiatives (Colling & Dickens, 1998). Milner and Gregory (2014) point out that gender equality may be "crowded out" on the bargaining agenda in times of recession (Milner & Gregory, 2014, p. 258). There is also a risk that classbased claims may be prioritized over gender-based claims (Dawson, 2014).
The importance of institutionally entrenched barriers is demonstrated in a recent study of a gender-equal pay initiative in Finland targeting nurses (Saari et al., 2019). They conclude that while corporatism in practice is flexible, it is nevertheless highly gendered, stating that Institutions can be stretched just far enough to aim at reaching better, fairer, more gender-equal outcomes, yet still resist until the rubber band returns to its original equilibria, one that serves to illustrate the stability of inequality (Saari et al., 2019, p. 21).
The present discussion cast light on the institutional factors surrounding collective bargaining in Sweden and reveal the flexibility of the institutional framework in the sense that its practical functioning can either promote or hinder gender-equality priorities.

| The Swedish case: Industrial relations and gender
While Sweden is characterized in many spheres of society by a high level of gender equality, the picture is mixed in working-life. On one hand, Sweden has extensive day care provisions, a high percentage of women in the workforce (SOU, 2015:86, p. 151), and a high level of unionized women (SCB). On the other, women take more sick leave than men (SOU, 2015:86, p. 299), work part-time to a greater degree (LO, 2018), and still face a gender wage gapwomen earned on average 89.3% of men's salaries in 2018 (NMO, 2018). Furthermore, the labor market is highly gender segregated (SOU, 2015:86, p. 168), and the average wage covaries strongly with the level of women in a given sector. Women-dominated sectors have significantly lower wages on average than male-dominated sectors (LO Fakta).
Wage bargaining in Sweden has traditionally been characterized by "strong, highly centralized, nationwide confederations for both employers and employees, which cover all sectors of the economy. So-called 'peak-level bargaining'" (Pontusson & Swenson, 1996). In contrast to many other countries where trade unions only bargain directly at the industry level, union shop-floor negotiations thus take place in Sweden within frameworks established by industry and nationwide agreements (Thörnquist & Thörnqvist, 2018). This regime was severely challenged by the end of the 1980s due to inflation, high unemployment, and high strike levels. Moreover, the employer confederation decided at that time to withdraw from centralized bargaining (Baccaro & Howell, 2017, p. 152). The 1997 Industrial Agreement (rectified in 2010) between unions and employers in industry re-established a regime of coordinated, multisectoral bargaining in the private manufacturing sector (Elvander, 2002). This agreement relied upon the "technocratic construction of a wage norm" (Baccaro & Howell, 2017, p. 155) that was calculated on the basis of wages in competitive, export-oriented industry. It was followed by similar agreements in the public sector which accepted that wage increases would follow the lead of the industrial sector (Howell & Givan, 2011, p. 239).
While the new regime in large part re-established the traditional model of coordinated and centralized collective bargaining, it has also entailed a shift toward greater marketization and employer discretion that is evident in 4 -ERIKSON decentralization and in the individualization of bargaining on the company level (Baccaro & Howell, 2017;Howell & Givan, 2011). By consequence, the new regime "privilege wage flexibility and wage restraint over wage leveling" (Baccaro & Howell, 2017, p. 170).
Another significant change is that trade unions have seen a dramatic decline in union density during recent decades that negatively affects their bargaining capacity. While the density in Sweden is still high in international comparisons, it has declined from a peak of 85% in 1993 to 67% in 2019 (Arena Idé report, 2019; Kjellberg, 2015).
It should be noted that Swedish trade unions have maintained their distinct organization of separate nationwide trade union confederations for blue and white-collar workers (Thörnquist & Thörnqvist, 2018). LO, the Swedish Trade Union Confederation, is the largest central organization, organizing blue-collar workers, while salaried employees are members of TCO, the Swedish Confederation of Professional Employees. A range of academic professions are represented by SACO, the Swedish Confederation of Professional Associations.
Previous research has described the trade union movement as a male bastion dominated by the norm of a male family provider and a masculine-coded ideal of leadership (Ledwith et al., 1990). The relation between class and gender has historically been a sensitive issue within the Swedish labor movement (Curtin & Higgins, 1998).
From the founding of LO at the end of the 1800s until 1947, when LO established a women's council, the issue of women's conditions was completely subordinated to male working-class interests (Qvist, 1974;Waldemarson, 2000). The situation changed somewhat during the 1970s in light of the increasing level of women's participation in LO unions and the growing gender-equality debate in society, with gender-equality issues acquiring a somewhat more prominent place on LO's agenda (Bergqvist, 1994). In general, however, gender-equality issues have continued to take a back seat to male-coded class interests in recent years (Ekström, 2012).
This article focuses on wage negotiations in the blue-collar sectors. LO organizes 14 affiliated unions with almost 1.5 million members, including such classic working occupations as construction workers, janitors, cooks, and shop assistants, as well as public service workers in the fields of child and elderly care. Most of these unions have either a female or a male majority, which reflects the existence of a very gender-segregated labor market. For example, approximately 78% of the membership of IF Metall, the Swedish Metal Workers' union, are men, while women comprise about 80% of Kommunal, the Swedish Municipal Workers' Union. LO's counterpart among employers is the Confederation of Swedish Enterprise (SN), which is Sweden's largest and most influential business federation, representing 49 member organizations and 60,000 member companies with over 1.6 million employees (see SN homepage).

| An institutional perspective on wage bargaining
Wage bargaining may be conceptualized as an institution consisting of various rules, norms, and practices which regulates the process of negotiation that establishes wage agreements (Howell & Givan, 2011;Pontusson & Swenson, 1996). While institutions can be either formal or informal to varying degrees, their function is to define the rules of the game, that is, how to behave and act in a particular setting. Disobedience generates sanctions that may be official or social in character (Lowndes & Roberts, 2013). Collective bargaining is one of the "institutional constants of economic life" (Flanagan, 1999(Flanagan, , p. 1151, and it includes a mix of formal rules and more or less voluntary agreements and norms. However, the specific characteristics of collective bargaining systems vary substantially among countries and over time (Flanagan, 1999(Flanagan, , p. 1151. Howell and Givan (2011) note that there is a high degree of "institutional plasticity" in industrial relations in Sweden and elsewhere, which is to say that existing institutions can take on new functions and meanings and produce different practices and consequences (Howell & Givan, 2011, p. 235). It is thus not enough to simply focus upon institutional forms since "a set of institutions can appear largely unchanged but in fact come to perform in quite different ways than before" (Baccaro & Howell, 2017, p. 15).
The Co-Determination in the Workplace Act (SFS 1976:580) is the most important formal rule in Sweden that governs relations between social partners, including such issues as the right to join a union, when a strike is legitimate,  and how to proceed when there is a conflict. Although collective agreements do not possess the status of law, they are also formally codified and contain both procedural and substantive regulations (Traxler & Mermet, 2003). In addition to these formal rules, there are also many informal rules of importance for collective bargaining, such as peak-level agreements, which have traditionally played a significant role in Swedish collective bargaining. Although they are not legally binding, they are based upon mutual trust and backed up by strong organizations willing to undertake sympathy actions in case of conflict (Pontusson & Swenson, 1996). Further examples include the "wage solidarity norm," which traditionally is an important component of the Swedish model of wage negotiations (Pontusson & Swenson, 1996), and the "industry norm," which is widely shared by social partners beyond actors in the industrial sector and functions as a "coordinating tool across industries and bargaining levels" (Kjellberg, 2019, p. 591).
The present article adopts an institutional perspective concerning the collective bargaining process that concluded with the GEP agreement in 2007, the aim being to identify the specific rules and practices that made this outcome possible. The analysis also intends to describe how rules and practices have changed over time in order to gain an understanding of the institutional conditions for gender-equality bargaining more generally.

| MATERIAL AND METHOD
The first step of the analysis involves an in-depth exploration of the process of the 2007 bargaining round (cf., Bennett & Checkel, 2014) in order to describe how events in that process unfolded over the various stages of the negotiations. It also serves to identify the institutional features that conditioned the chain of events. More specifically, the analysis seeks to pinpoint how the GEP was moved forward through each step of the bargaining process. The second step of the analysis focuses on the institutional features identified as having made the GEP possible in 2007, and it examines how they have subsequently changed in a manner that now obstructs similar agreements. An issue of particular interest concerns how these institutional features are gendered in a way that restricts gender-equality initiatives regardless of the macroeconomic conditions. The case study is based upon interviews, reports, and news articles. A total of 18 interviews were conducted during 2019-2020 with representatives of trade unions, employer organizations, and the National Mediation Office (NMO). The interviewees, several of whom were directly involved in the 2007 bargaining process, have been assigned numbers between 1 and 18 that will be used in the empricial analysis for the sake of anonymity (see the Appendix in Supporting information Material for more information). The interview questions specifically revolved around the 2007 collective bargaining round, including the introduction of the GEP, but also more generally addressed gender equality in the labor market and gender-equality work in collective bargaining. The reports in question documented the collective bargaining process, such as the annual report from the NMO on collective bargaining and wage formation along with other reports produced at the time. The media have also comprised a useful source of contemporary information concerning the main actors, including their arguments and actions, insofar as this bargaining process received a great deal of media attention. These written sources have been helpful in countering difficulties related to respondents' memory lapses, such as by providing cues and correcting misrepresentations of the events. Important information has been verified through triangulation with several sources to ensure that the findings are trustworthy (Schwartz-Shea & Yanow, 2013).
The proposal of a gender-equality pot is not only interesting in itself, it also comprises a suitable case for gaining further knowledge of the conditions for gender-equality bargaining in general-because of its explicit focus on gender it brings to the fore both favorable factors and obstacles.

| The backdrop for the gender equality pot in 2007
In 2007, Sweden was enjoying a period of economic growth, high investment, and low unemployment (NMO, 2007, pp. 15) that provided favorable conditions for wage negotiations. The bargaining process at that time affected 6 -ERIKSON almost three million employees in blue-collar, white-collar, and academic professions, with 500 of the labor market's 600 collective agreements being up for negotiation. While almost all unions and employer organizations were involved in the bargaining process, the present article focuses on the blue-collar sectors.
The question of gender equality had already been on the bargaining agenda for some time. The agreements of 2001 and 2003 in the blue-collar sectors had included explicit intentions to equalize wage levels through targeted investments in low-wage sectors, which included many women-dominated sectors (Interview 12, 6, 16). However, these agreements involved general low-wage investments with no specific gender focus, and the increases earmarked were limited in comparison with the later GEP (Handels rapporter, 2019: p. 3, Interview 16).
The view within LO at the time was that such investments had not been sufficient, and that there was a need for something more radical to target the persistent gender pay gap. A model of a Gender Equality Pot was launched within this context as a means for explicitly targeting gender inequalities (Interview 12). Several respondents maintained that the idea originated directly from the LO board and was developed with the assistance of a statistician (Interview 6, 9, 12). The GEP was thus constructed as a mathematical formula that calculated an extra wage increase for low-wage and women-dominated sectors on top of the general wage increases-the higher the percentages of women, the greater the amount of extra money. For instance, a sector consisting of 100% women with salaries below 2000 EUR/month would generate approximately 20 EUR/month extra per women employee, while a sector with no women would receive no extra funding (NMO, 2007, p. 113). A list of abbreviations and information concerning the various actors involved is provided in the Appendix in the Supporting information Material.

| Analyzing gender-equality bargaining in 2007
A collective bargaining process in the blue-collar sectors proceeds through a number of stages characterized by both horizontal and vertical coordination (Traxler & Mermet, 2003), that is, coordination between unions across sectors as well as coordination within a union between its central and local sections. Each stage is conditioned by formal and informal rules and practices. The following sections describe the stages of the process, and identify the most significant elements of the institutional framework that made the introduction of the GEP possible.

| Union coordination and employer coordination
Sweden has a system of multiemployer bargaining such that both unions and employer associations coordinate their own positions and activities on a central level (Traxler & Mermet, 2003, p. 233). In practice, the first step in a bargaining round is that the 14 LO unions attempt to establish what is termed "the LO coordination." Within this coordination, the unions agree on a list of joint demands to put forward in the sectorial negotiations that will follow, and they commit themselves to support each other in cases of conflict (2). The top priority is normally the level of wage increases, but additional issues regarding working conditions may also be included, such as pensions and additional parental leave subsidies (2). It is important to note that this coordination is voluntary in character-each union handles the negotiations within their sector and they have the discretion to make their own deal. The confederation thus has no coercive power in respect to the bargaining process (2). The LO coordination has historically been very important in bargaining, and LO unions have typically managed to coordinate themselves with a list of joint demands (2). Informal meant that each union could decide on their own whether and how to undertake sympathy actions in support of other unions (LO-Tidningen, 061201). A male respondent from a union in favor of the GEP maintained that "They [IF Metall] were afraid that they would have to engage in sympathy actions to enforce the demand of extra money for women" (9). An LO representative stated that another factor which helped persuade IF Metall to accept the agreement was that they were not obliged to apply the GEP in their own agreements, but could instead use the estimated level of the wage increase thereby created for other investments (16). The final list of demands put forward by the LO coordination included a 1-year agreement with a general wage increase of 3.9%, or 825 kr (85 EUR), and a GEP calculated on the basis of the percentage of women with low wages in each agreement. The expressed intention of the GEP was that it should be used to reduce the gender-equality pay gap ( NMO, 2007, p. 113). The fact that the GEP was included in the joint list of the LO coordination was important for the further process in spite of the internal disagreements noted above.
In parallel with the LO coordination, employer associations within SN undertake their own coordination prior to a bargaining process. Although they do not draw up a list of substantial common demands-that is the responsibility of each sector-the employer coordination establishes the overarching goals and directions for bargaining and specifies the forms of coordination among employers. The 2007 SN statute clearly states that although each branch organization handles their own negotiations, they should "consult the confederation in respect to essential issues brought up in the bargaining process" (NMO, 2007, p. 110). SN coordination may be described as more reactive than proactive, and it plays a more important role during the actual bargaining process than prior to its beginning.

| Bargaining begins
In the second stage, when the actual bargaining begins, each union negotiates their agreements with their employer counterpart. The 14 unions in the LO confederation encompass not only different labor market sectors, but also differing agreements within each sector. This step of the process is defined by guidelines for bargaining that can be of either a procedural or substantive character (Traxler & Mermet, 2003, p. 234). As mentioned before, there is a strong procedural norm in Sweden after the Industrial Agreement of 1997 that the industrial sector is the first to reach an agreement, which establishes the level of the "cost norm" for the entire labor market on the basis of productivity growth. Not only should the costs of wages be accounted for within the norm, but also increased costs due to changes in working conditions, insurance, and so forth (Elvander, 2002;MI, 2007, p. 117).
In line with this model, the industrial sector reached the first agreement by the middle of March 2007-the total value of the 3-year agreement was calculated to be 10.2% ( NMO, 2007, p. 117). This was intended to serve as the "mark" for the rest of the labor market. The industrial unions had decided not to apply the GEP, but its estimated cost was included in their agreement and designated to be used for an extra increase in the lowest wages (Handels rapporter, 2019:03, p. 24). Shortly after this agreement was reached, the bargaining process continued in the commercial sector between the Union of Commercial Employees (Handels) and the employers' association, the Swedish Trade Federation (Svensk Handel). Insofar as this was a women-dominated and low-income sector, the collective bargaining process in this sector came to be decisive for how to interpret and apply the GEP in the negotiations that followed in other sectors. The parties in the trade sector managed to reach a 3-year agreement rather smoothly, with a total wage increase of 12.6% and the GEP.
But the bargaining process took an unexpected turn when the agreement was about to be formally signed. SN rejected the informal agreement between the sectorial parties, and Svensk Handel was not permitted to sign (DN, 070326). Not only did SN maintain that the wages were too high and not in line with the "mark," representatives from employer organizations confirm that SN also disliked the GEP and similar investments for designated groups (OpO, 2007, p. 31;Interviews 8, 18, 17). SN sought to force Svensk Handel to cancel the deal, but they refused.
Handels declared at this point that they would go on strike if the deal was not accepted. They described their dialog with Svensk Handel as good, and regretted that they would nevertheless have to declare a strike because of 8 -ERIKSON SN's actions (DN, 070326). There was obviously disagreement among employers concerning how to interpret the industry norm, and the fragmentation became evident when Svensk Handel's managing director publicly defended the informal agreement, putting pressure on SN to approve it. He stated that If you relate the amount of money that industrial workers will get to what employees in the commercial sector will receive, the latter will receive a higher percentage increase, but not a larger amount of money to spend [due to higher industrial wages]. I have tried to explain how we see this, and there is no problem at all within the federation I represent [Svensk Handel]. But there is a fear that the commercial sector will set the "mark" for wages instead of the industrial sector (DN, 070326).
He also maintained that there was no reason to continue the bargaining process since an agreement had already been accepted by both parties.
The actions of Svensk Handel supposedly violated the norms and practices for coordination among employers, and representatives from unions as well as employers maintain that they were threatened with fines and criticized for making the conflict public (14, 7, 17). However, several factors specific to the trade sector help explain why Svensk Handel more easily supported the GEP than other employers. For example, extra funds were available for wages due to a reorganization of occupational pensions and decreased demands for full-time employment, but it was also necessary to avoid a strike during Easter. Respondents from both unions and employers remarked that the latter would have caused substantial damage (2, 14, 16).
When the risk of conflict became a fact, LO's board discussed the contents of the agreement and found that it was in line with the joint demands of the LO coordination. In contrast to SN, they also found that it was in line with the industry norm (NMO, 2007, p. 118). As a respondent from LO noted, Handels could consequently count on the support of the other LO unions (12), several of whom publically declared their willingness to declare sympathy strikes if necessary. For instance, the representative of the Swedish Transport Workers' Union stated that "Handels can count on us if there is a strike" (DN, 070327).
While the unions managed to find ways to maintain coordination during the bargaining process, the employers were fragmented.

| Declaration of conflict-Calling in the NMO
The institutional framework for mediation is important in shaping bargaining processes and their outcomes (Ibsen, 2016). In Sweden, the NMO was established in 2000 as a public agency tasked with mediating conflicts, fostering sound wage formation, and collecting wage statistics (NMO, homepage document). The NMO must be informed in writing at least 7 days prior to a strike declaration, and it then appoints independent mediators for the purpose of finding a solution to the conflict. The mediators should not only seek to avoid a conflict, they are to identify a solution that is "in line with the norm set by the competitive sector" (NMO, webpage document, p. 7), which is interpreted in light of the prevailing norms and practices. The mediation procedure normally involves both parties being called to additional negotiations, after which the mediators submit a bid in an effort to resolve the conflict. It should be noted that the mediators cannot force the parties to sign an agreement.
One of the union representatives described the conflict that arose during the 2007 negotiations as very unusual.
We accepted the engagement of mediators, but I said "I want to know how you're supposed to mediate when we already have reached an agreement?" It was a very short meeting. The mediators returned after an hour or so and suggested an alternative agreement, which we turned down. After that we signed the [initial] deal (7). ERIKSON -9 The strike was thereby avoided and the initial agreement signed in spite of the disagreement regarding its relation to the industry norm. SN thus found themselves defeated, and they had to accept an agreement in line with the initial proposal in order to avoid conflict.

| Adoption of the GEP in all women-dominated sectors
The contract for retail workers, which is a women-dominated sector, resulted in a 13% wage increase because of the adoption of the GEP (NMO, 2007, p. 248). This agreement was of great significance for the rest of the bargaining process, particularly for other women-dominated areas, which also adopted agreements that included the GEP (OpO 2007, p. 33). After the employers' initial defeat, the rest of the bargaining thus proceeded without further conflicts related to the GEP.
The NMO concluded in their 2007 report that the competitive sector's normative role for wages was complemented in sectors with many low-income workers, particularly in sectors with many women with wages below 20,000 kr/month. The wage increase for a three-year period has been about 13% in sectors with a large percentage of women with low wages (NMO, 2007, p. 119).
The NMO also maintained that it is difficult to estimate the exact "cost" of an agreement and determine whether or not it is within the industry norm, although their report contains an underlying critique of excessive wage agreements (NMO, 2007, p. 119). Svensk Handel officially stated their motivation for wage increases above the industry norm through reference to changes in how working time was calculated of (NMO, 2007, p. 248).
The 2007 bargaining process was described as a victory by representatives from women-dominated unions.
They remarked that "This means a shift towards gender-equal salaries" (GP, 070331), hailing it as "An important victory in the struggle towards higher real wages and more gender-equal wages" (LO, 070525). It was also criticized, however. A few months after the conclusion of the bargaining process, before the agreements were fully implemented and evaluated, the chair of IF Metall declared that they would not support the GEP again. He stated that "For us there will be no more LO coordination based on the GEP" (LO-Tidningen, 070926). This opposition resided upon criticism of the notion that there was a general value discrimination against women, with low-income investment being viewed as "more fair" than a gender-based model (SD 070926; Kommunalarbetaren 070926).

| Four key institutional elements in change
The analysis of the 2007 bargaining process demonstrates how the institutional setting at the time was sufficiently flexible to allow for the adoption of the GEP. In the next step, the analysis focuses on the four institutional elements identified as particularly important.  (2), with disagreements concerning the GEP being a 10 -ERIKSON contributing factor. As noted above, industry unions disliked the GEP from the beginning, and although they had accepted it in 2007, they declared that they would not support coordination based upon it again. They found the construction of the GEP problematic because they regarded it as unfavorable to low-paid men, to women in maledominated sectors, and to women in higher income groups. They also questioned the assumption that women's lower wages in general are a result of value discrimination, stating that We cannot be sure what the lower salary is due to in individual cases. It may be the number of working years, education, or skill. However, we know that since a smaller number of women working in industry have more highly qualified jobs, women generally have lower wages (Handelsnytt,080711).

| Trade union coordination
In contrast, unions in women-dominated sectors were satisfied with the outcome and strongly supported similar investments in future bargaining processes.
A central LO representative argues that the gender issue, which was brought to the fore in 2007, is an important reason for the recent difficulties with coordination.
It was easier before we started working from the perspective that women's salaries were wrongly valued. As long as we talked only about special low-salary increases, everybody supported us (6). Other macroeconomic conditions that have rendered coordination particularly difficult after 2007 are declining union density, the economic crisis of 2008, and the turbulence in the industrial sector prior to unions and employers signing a new Industrial Agreement in 2011. Previous research has also noted the tendency toward growing difficulties in coordinating between unions. Baccaro and Howell (2017) conclude in this regard that "finding common interest within each union confederation, let alone across the entire labor movement, has been far more difficult in recent years" (p. 157).
In summary, new practices marked by more fragmented and less coordinated bargaining on the part of the unions have constrained gender-equality bargaining.

| Employer coordination
The low level of coordination between employer organizations in 2007 was another contributing factor in the successful GEP implementation. This was evident in the differing interpretations of the industry norm and in the lack of SN sanctions that could have forced Svensk Handel to toe the line with the other employer associations. SN in fact found themselves forced to accept the informal agreement after a failed mediation. Several of the respondents from the unions have the impression that the employer organizations had not expected this type of model and were probably taken by surprise (2, 4, 16). Although a statute was in place at the time that required SN member organizations to consult with the employers' confederation before a deal was reached, it apparently did not work in  practice. When Svensk Handel informed the confederation about the deal, an informal agreement was already in place and they did not wish to revise it, regardless of SN's opposition.
Not only was this situation unusual, it constituted a tremendous defeat for SN that one of the respondents from the confederation described as "humiliating" (16). Respondents from both employer organizations and unions have remarked that changes were made after 2007 to prevent similar situations-coordination has become stronger among the SN member organizations (8,16,12,4,2,14). A representative of an employers' organization recalls that We had long discussions on the employers' side that this cannot happen again. We need to have an order that we stick to, that everyone accepts, which maintains the "mark." They [employer organizations] were supposed to toe the line (8).
A number of union representatives remarked that employer organizations have been much more coordinated in bargaining processes after 2007, and that new practices are in place through which SN controls their member organizations. One explicitly stated that "No employer would make a preliminary deal [today]. They go to SN to get their approval" (16). The new practice is that a confederation council thoroughly examines and approves member organizations' agreements before they receive permission to sign. Respondents from both LO and SN confirm that employers who do not comply with these procedures risk economic sanctions-unless the agreement has been approved in advance, they have no SN financial support in cases of conflict (16, 3).
Such changes mean that collective bargaining has become more controlled and regulated on the part of employers, with individual employer organizations having far fewer opportunities to act without SN support.
One union representative remarked that there have been occasions when they reached an agreement with an employer organization, but SN did not permit the employers to sign the deal (4). Not only would the GEP probably not have been implemented if this practice had been in place before 2007, one employer representative stated that similar investments have in fact subsequently been significantly restricted (17).
This is yet another manifestation of increased employer discretion in recent decades (Baccaro & Howell, 2017).

| Industry norm
The importance and interpretation of the Industrial Agreement were brought to the fore in the 2007 bargaining. A generous interpretation of the Agreement was in fact decisive for the adoption of the GEP.
As noted before, the purpose of the Industrial Agreement was to let industry set the norm for the level of the wage formation in the labor market as a whole (Elvander, 2002). The analysis demonstrates that the 2007 negotiations were controversial for the employer organizations, with the outcome contributing to the decision by industrial employers to leave the Industrial Agreement in 2010. Employers maintained that "it became clear after the last negotiations [2007] that several actors in the labor market did not accept industry's function as a norm-setter" (Lag och Avtal, 100429). They argued that the Agreement had played out its usefulness and refused to accept it as setting the "floor" in wage bargaining. The unions in the industry sector shared this view, even if they did not take the initiative to leave the Agreement. A new Industrial Agreement was reached in 2011 in which both employers and unions in the industrial sector emphasized the normative function of industry, stating that the Agreement takes precedence over other agreements and collaborations (Arbetet, 110509).
Several LO unions, such as those in the construction sector (Byggnads, 110519), immediately criticized industry's strengthened position in the new Agreement. A number of respondents maintain that the new Industrial Agreement has in practice established a much stricter interpretation of industry as the norm. This has 12 -ERIKSON been noted by both LO and employer organizations, and criticism of the norm has recently increased. One recurrent issue is that the cost norm has become too strict, and that it hinders relative wage changes and perpetuates wage gaps (2,6,16,15,17,12,9). One LO representative commented that "Se previously said that the industrial sector sets the norm, but that is a different thing than deciding the exact 'mark'" (6). Several respondents from unions that are in favor of the GEP maintain that the interpretation in place in 2007 allowed more room for adjustments and relative salary changes, such as the GEP (6,16,7,4). Even representatives from employer organizations are concerned that the model might "implode" if it is too strictly interpreted (17).
Unions also raise the issue that the new Industrial Agreement has a negative impact upon their ability to coordinate themselves (4, 16), with one remarking that "The Industrial Agreement has become superior to all other collaborations" (4).
SN claims that the 2007 negotiations exceeded the norm. Those supporting the GEP counter that the supposed excess was within the interpretation of the norm, and that it was necessary in order to achieve a relative wage change in a gender-segregated labor market. Although it is theoretically possible to make such investments as the GEP within the industry cost norm, it is difficult in practice to combine them with a strict interpretation of the norm since certain groups would then have to accept wage increases below the "mark" in order to compensate for increased costs.
Briefly stated, the GEP would not have been possible if the industry norm had been strictly interpreted in 2007, and similar investments are now restricted by the new and more strict norm.

| The approach of the NMO to conflict
The fourth institutional aspect of significance for wage negotiations concerns the procedures to be followed in cases of conflict, particularly the role played by the NMO. The NMO's interpretation of their assignment in 2007 was important for implementing the GEP, and a different understanding on their part might well have made it more difficult to reach an agreement.
The NMO was established in order to foster a "wage formation in balance with the national economy" (Government Bill 1999/00:32, p. 1), but the precise meaning of that goal is not a given-various approaches were in fact mentioned in the preparatory work. A respondent from the NMO stated that several definitions are of importance in this regard, namely to promote real wage increases, increase employment, limit the number of conflicts, make possible relative wage changes, and maintain the competitive sector's wage-normative role (13).
The 2007 negotiations between Handels and Svensk Handel became a case of conflict due to SN's refusal to sign the agreement that had been reached. The fact that the mediators only briefly engaged with the case, and their only suggestion was turned down, indicates that they were willing to accept the initial agreement even though it included the GEP and, according to SN, exceeded the industry norm in women-dominated sectors. This event became symbolically important for the rest of the bargaining process, and other agreements that included the GEP were adopted with neither conflict, nor the NMO's involvement.
Several respondents from unions that support gender-equality initiatives maintain that there has been a change in how the NMO interprets their role after the 2007 bargaining round. In line with the new Industrial Agreement, they have come to interpret their task of working for a well-functioning wage formation in a much stricter way-it has in fact become synonymous with ensuring that the industry norm is followed (9,16,4). the way in which the Office has chosen to interpret their assignment to protect the wage-normative role of the competitive sector, and the consequences this has for opportunities to gradually break the value discrimination that exists between women and male-dominated branches of the Swedish labor market (NMO, 2011, pp. 135-136).
This more strict interpretation is also referenced in the NMO's own report concerning the 2012 negotiations in their statement that "there was no help to get from the mediators for those who wanted to set agreements above the norm" (NMO, 2012, p. 140). This new practice meant that a bid from the NMO would not exceed the norm, even in the event of a strike. This change has been a factor in why it has become very difficult to reach agreements with relative wage changes, such as the GEP. As one representative from LO remarks The problem is that when we try to accomplish these changes-either low-income investments or a gender equality pot-then the NMO states that it conflicts with their assignment, which is to mediate on the basis of the norm set by industry (16).
The design and mandate of the NMO have implications for collective bargaining in that although the NMO cannot directly affect the ratification procedures of parties engaged in bargaining, it now serves to "effectively guard the manufacturing norm" (Ibsen, 2016, p. 298).
In short, the analysis of the four institutional features reveals the emergence of new practices and new interpretations of norms that together significantly restrict gender-equality bargaining after 2007. Gains and Lowndes (2014) have observed that rules and practices can be gendered in different ways. There are in fact several examples of neutral rules with gendered consequences in the collective bargaining case examined (Gains & Lowndes, 2014). Stated otherwise, although the institutional framework that conditions bargaining is not gendered-it neither explicitly rules out gender, nor treats men and women differently-the present analysis demonstrates how institutional features indirectly influence in a negative way the possibility of targeting gender inequality. While certain conditions made the GEP possible in 2007, in spite of various obstacles and strong resistance, the institutional framework has subsequently changed in a way that restricts similar investments. The period in question has also been marked by declining union density, economic crisis, and increasing employer discretion (Baccaro & Howell, 2017). Previous research has shown that this not only comprises a generally unfavorable context for unions, it also gives rise to the risk that gender-equality claims will be "crowded out" by other issues (Milner & Gregory, 2014).

| The gendered consequences of institutional shifts
In respect to the practice of coordination among unions, several respondents have noted the importance of unity in the struggle for gender equality as well as the difficulty of forging unity in support of gendered claims. Even as unions agree on the importance of reducing the gender pay gap, they have very different views concerning how best to attain that goal (1, 2, 6). For example, there is a shared understanding among unions from both industry and women-dominated sectors that unions in the industrial sector strongly oppose any targeted investment based on gender, while women-dominated unions, such as Handels and Kommunal, argue for the necessity of targeting women-dominated sectors with special investments (7,16,9,4,10). LO coordination has broken down several times since 2007 because of such disagreements.
Although there have been various types of low-wage investments, unions have not been able to unite again behind a joint investment that explicitly targets the gender pay gap. Although the investment in assistant nurses in 2016-initiated by Kommunal and supported by IF Metall-did target a women-dominated sector, it was criticized by other women-dominated unions for its limited focus on one profession. LO representatives maintain that this lead 14 -ERIKSON to a subsequent breakdown in coordination that year (16, 6). While the formation of coalitions among unions is a factor mentioned as favorable for gender-equality bargaining (Briskin, 2006), this case reveals how an uncoordinated bargaining process restricts the possibilities for making gender-equality investments in collective bargaining.
There are two reasons for this. First, the issue of gender equality is less likely to be on the bargaining agenda if unions cannot unite around a shared understanding insofar as employers in general do not undertake initiatives to equalize wages. Second, even if individual unions do bring up gender equality, their leverage in the bargaining process is limited when they lack support from other unions in cases of conflict.
Increased coordination among employer organizations has proceeded in parallel with the fragmentation of union coordination. For example, the institutionalization of an SN council with expanded gatekeeping functions is in itself a gender-neutral procedure, but it nonetheless further restricts the possibility of making gender-based claims.
SN has openly opposed the GEP and similar investments, and they have been very outspoken in their criticism of collective agreements that reflect gender-based demands. The interviews indicate that even if certain employer organizations within SN have a more positive approach to gender-equality measures, employer organizations rarely or never bring up gender equality in collective bargaining. Several sources from employers as well as unions confirm that they in fact tend to dislike all group-based demands (14,1,7,4,15). SN's more strict gatekeeping function and the fragmentation of union coordination are practices that indicate a shift of power from unions to employers in the bargaining process that restricts the inclusion of gender-based demands in collective agreements.
The more recent interpretation of the industry norm concerning the level of wages and the NMO's new interpretation of their assignment-which serves to legitimize the industry norm-have also had negative implications for gender-based investments. While these institutional elements are gender-neutral in their construction, they have clearly gendered consequences in that they maintain the status quo and hinder relative changes in wages between groups, including men and women. As noted, the more strict application of the industry norm in all individual sectors tends to rule out such agreements as the GEP, which are specifically aimed at shrinking the gender pay gap between groups and sectors. It is worth noting that although the industry norm has the same negative impact upon all agreements that include higher wages for targeted groups, its effects appear to be more restrictive for gender-based demands than for those based upon other social categories. There are in fact examples of recent agreements that exceed the norm, such as the 2016 agreement targeting assistant nurses. Several respondents from LO and women-dominated unions argue that class-based claims-those that target low wages in general or a particular underpaid profession-are more easily accepted than gender-based claims (6, 2, 4). A representative of Kommunal related the following experience in this regard.
If I go to SN or to employers and talk about problems related to value discrimination and gender equality, they never admit that it exists. But if I say that we have to maintain welfare levels and attract qualified and educated workers, then it is much easier to get support, although not from a genderequality perspective (4).
The more strict application of the industry norm has thus had negative consequences, particularly for genderbased claims. This finding indicates that class-based interests are still prioritized as the primary concern in collective bargaining (cf., Ekström, 2012). The NMO's stricter interpretation of the norm further underscores the restrictive function of the "mark" upon gender-based claims. If the NMO would utilize a less rigid interpretation of the industry norm and, for instance, encourage the adoption of gender-equality measures, then trade unions with gender-equality ambitions could push more strongly for gender-based claims in collective bargaining and risk the emergence of conflict. However, knowing that the NMO would not support any agreement that exceeds the industry norm makes it more difficult to do so.
The NMO's restrictive role is paradoxical in that it has had a specific gender-equality assignment from the government since 2015, whereby it should not only analyze wage formation from a gender perspective, but also foster gender equality in bargaining processes. More specifically, the NMO should "draw attention to the  importance of designing the central collective agreements so that they promote the local parties' work on wages from an equality perspective" (NMO's homepage). It is worth noting that certain respondents express a slightly positive attitude concerning the NMO's role in the promotion of gender equality in the near future because of the appointment of a new woman director with a social democratic background and an apparently more pragmatic approach to mediation (16,4). It is an open question, however, whether there is room for agency in the interpretation of the NMO's role, and whether the new director will champion gender equality (cf., Dickens, 2000).

| CONCLUSION
The main empirical finding of this study is that the adoption of the GEP was far from a smooth and easy process. On the contrary, the 2007 agreement was a rather remarkable exception that was implemented in spite of many obstacles. While there was a favorable economic situation in society at the time, along with an element of surprise that had a positive impact, the article also demonstrates that the institutional framework was important. The rules and practices that condition collective bargaining in Sweden were perhaps not favorable for gender-equality bargaining, but at least they were flexible enough to make such bargaining possible. Norms and practices were also open for certain degree of interpretation, and actors supporting the GEP could consequently take advantage of these loopholes to push through the agreement. Baccaro and Howell (2017) argue that industrial relations in Sweden have changed in recent decades in a neoliberal direction characterized by increased employer discretion in collective bargaining. In spite of the fact that many institutions that regulate collective bargaining remain intact, their practice and functioning have changed to include more decentralized wage-setting, a high degree of individualization, and labor market flexibility (Baccaro & Howell, 2017, p. 144). The present article has demonstrated that the institutional framework for collective bargaining has changed in a manner that efficiently restricts gender-equality investments. The GEP would in fact have been impossible within the new framework. This is an example of "institutional plasticity" when "existing institutions take on new functions and meanings and produce different practices and consequences" (Howell & Givan, 2011, p. 235). In the Swedish case, norms and practices have become more formalized and thereby less flexible and less open to interpretation, and the strong industry norm in particular appears to hinder genderequality initiatives. Women-dominated sectors as well as the domestic construction sector have in recent years intensified their criticism of industry being the norm setter for the wage level of the entire labor market.
The increased discretion of employer organizations noted by Baccaro and Howell is also evident in the increased coordination among employers at the central level after 2007. This shift of power toward employers at the expense of unions, which is accentuated by decreased union density, is detrimental to gender-equality concerns. Not only does the weakening of union organizations have a particularly strong impact upon low-wage (women-dominated) sectors, the remaining possibilities unions have to place gender equality on the bargaining agenda are also undercut. As Milner and Gregory (2014) observe, the differential capacity of trade unions and employers to set the agenda is crucial for gaining an understanding of the conditions for gender-equality bargaining (Milner & Gregory, 2014).
The restrictive effect of the new institutional framework can be noted in the fact that union coordination broke down in 2010, 2016, and 2019, with disagreements over whether and how to target women-dominated sectors contributing significantly to the breakdowns. The perceived conflict between class-based low wage investments and gender-based wage investments was evident in IF Metall's opposition, and it remains relevant today in LO (see Erikson forthcoming for a further discussion of this point).
These findings support the argument of Saari et al. (2019) that research should direct greater attention both to the rules and practices that condition bargaining, and also to how seemingly neutral practices can have negative consequences for gender-equality bargaining.

| Interviews
A total of 18 interviews were conducted, with respondents being assigned ID numbers from 1 to 18 for the sake of anonymity. See the Appendix in the Supporting Information Material for the complete list.

DATA AVAILABILITY STATEMENT
Data available on request due to privacy/ethical restrictions (interviews).