Networking, lobbying and bargaining for pensions: trade union power in the Norwegian pension reform

Abstract Norway reformed its pension system in 2011, introducing a Swedish-style, NDC system. Contrary to expectations, the reform was largely supported by the dominant confederation of trade unions, the LO. In this article, we look at LO involvement in the process at different stages. Through qualitative interviews with key reform architects, we have traced the process between 2005 and 2008, emphasising actors, meeting places and interests. Starting from the insight that unions can influence through lobbying, bargaining and (the threat of) mobilising, we suggest that lobbying can be a mutual process, where parties and unions move each other’s positions. In addition, bargaining can take the form of behind-the-scenes cooperation, as well as of negotiations in the classic, Nordic-style industrial relations sense. Expanding on this framework, we suggest that the literature on pension reforms should pay more attention to negotiated and voluntary labour market occupational schemes, and to the importance of expertise and networks.


Introduction
What is the role of trade unions in policy processes leading up to welfare state reforms? In discussing this topic, the literature has moved beyond the view of trade unions as driving forces for welfare state expansion (as in the power resources literature; see for instance Korpi 1978), and of weakening trade unions as one of the preconditions for retrenchment (as in the new politics literature; see for instance Pierson 1996). Instead, there is a growing debate on the variation in trade unions' preferences and priorities, their strategies and their persistent potential to have an impact on policies and to sustain and develop social partnerships (e.g. Schludi 2005;Marier 2008;Hassel 2009;Häusermann 2010;Culpepper and Regan 2014). One insight that has come out of this literature is that trade unions' ability to influence is highly contextual, depending on historical traditions, as well as political and institutional factors (Anderson and Meyer 2003;Trampusch 2006;Andersen et al. 2014). Therefore, there are important lessons to be learned from an accumulation of case studies of single countries. In this article, we aim to contribute to this cumulative literature through a case study of the reform of old age pension in Norway.
The Norwegian pension reform, which was implemented in 2011, has a lot in common with the previous Swedish reform: both countries introduced NDC systems with flexible pensioning ages on actuarially neutral terms. However, the Swedish pension reform was passed practically without union involvement (Anderson and Meyer 2003;Schludi 2005;Marier 2008). In Norway, by contrast, the dominant union the Norwegian Confederation of Trade Unions, LOwas at the table, participating actively in designing the new system. Why would they do that? Why would a trade union confederation actively promote and defend a reform that significantly reduced the pensions of large member groups, in a situation where there were no external pressuressuch as a financial crisis, like in Swedento do so? Answering these questions, we suggest, will shed light on the politics of pension reform and the role of trade unions in advanced capitalist societies, in particular on the complexity of union preferences and the various power strategies modern unions have.
Pension reforms are special cases of welfare reforms: not only because they are complicated and have long-term effects, but also because they are found at the intersection of welfare policies and industrial relations (IR). An important institution for IR in the Nordic countries is the coordinated and relatively centralised systems of wage bargaining, combined with an extensive system of tripartite dialogue. This has resulted in relatively high and compressed wages, but also allowed for wage moderation, negotiated through centralised bargaining, as a way to ease effects of external shocks (Andersen et al. 2014). However, the Nordic countries differ in a number of areas in their IR tradition. Important for our purposes here is a tradition for state intervention in stalled conflicts between the social partners: this is common in Norway (and Denmark), but unheard of in Sweden (Andersen et al. 2014, 17). Other examples of Norwegian exceptionalism within the Nordic region include the traditionally strong ties between the Labour party and the LO (Allern et al. 2007): The LO chair is still elected onto the Labour party's executive committee in Norway, and collective membership was phased out only in 1997 (Allern et al. 2007). A third particular institutional feature of Norwegian wage negotiations is that negotiated results are subjected to balloting among the rank-and-file. This creates a need to secure broad member support, making the Norwegian case particularly interesting when it comes to the role of unions in IR and in policy reforms (Hippe and Nergaard 2016).
These exceptional features of the Norwegian modelthe tradition for state intervention in wage bargaining, the close party-union relationship and the balloting of negotiated resultshas implications for the dialogue on pensions. The tradition for tripartite dialogue has created platforms for interaction, as well as networks of trust and shared understandings that have been important also in pension processes. Moreover, the tripartite cooperation did at one point give birth to a pension scheme in which the social partners and the state are interweaved: the negotiated early retirement scheme, AFP. This was negotiated by the social partners in 1988 and later expanded. To break what looked like a deadlock in 1988, the state intervened and committed to fund 1/3 of the expenses to AFP and also to take on administration of the scheme. Elements of the pension system are based on collective agreements in all the Nordic countries (Kangas et al. 2010), but this heavy state involvement in a negotiated scheme is unusual. Even so, the relationship between the party and the union is not all harmony. LO has repeatedly shown that it is capable of blocking unpopular reforms, as it has done for decades with regard to the sickness benefit scheme (Hagelund and Pedersen 2015).
In the next section, we present theoretical perspectives and previous research on the role of trade unions in welfare reforms generally, and old age pension reforms particularly. We move on to outline the Norwegian pension system and the main contents of the reform, before presenting our data and analysis. We follow the timeline of the reform from the launching of the Pension Commission's report in 2004 to the wage negotiations in 2008. Analytically, we focus on three distinct events in order to show the arenas and mechanisms for union involvement: the LO congress in 2005, the Parliamentary settlement in 2007 and the wage bargaining in 2008. We shall argue that this was a carefully orchestrated process, and that different strategies were important at different stages. Nevertheless, as we shall show, the outcome of this lengthy process was remarkably similar to what the Pension Commission proposed at the outset, and thus far from the original status quo.
In this article, we limit our attention to the process in the private sector. Following the pension reform, negotiations on occupational pensions in the public sector were initiated in 2009 (Hagelund and Grødem 2017), but the negotiations failed to fully produce the desired outcomes, and the process is still ongoing by 2018. However, the problems of reforming public sector occupational pensions had no repercussions on the finalising of the reform process of state pensions. White-collar unions play a crucial role in wage negotiations first of all in the public sector. In the private sector, LO is dominant. A total of 52% of organised privatesector workers are LO members (Nergaard 2016), and LO is by far the largest organisation in the export industries, which set the pace for wage negotiations. This implies that we are mainly dealing with the male-dominated, blue-collar sections of LO, and that other unions outside the LO play a limited role in this story.

Perspectives on pension reform and the role of trade unions
While there is a booming literature on welfare state change and retrenchment, the role of trade unions plays a limited role in these contributions (Starke 2006). When the welfare state retrenchment literature brings in trade unions, it typically highlights how trade unions are weakened in terms of membership and also increasingly fragmented, and therefore not the impressive unified force it was under "old" welfare policies (e.g. Häusermann 2010, 70ff;Culpepper and Regan 2014). In addition, new interest groups have emerged, sometimes organised around welfare state arrangements. Unions thus increasingly appear as a bundle of interest groups among many when welfare state reforms are on the table.
Other contributions assign a more active role to trade unions, even under new circumstances. The literature on social pacts, for instance, obviously highlights trade unions as key actors (Pochet et al. 2010;Avdagic et al. 2011). A growing number of studies discuss how unions matter in advanced capitalist societies with different social policy traditions, often with an emphasis on their complex priorities and various power strategies. To understand a sustained social partnership in Europe, Hassel (2009) differentiates between policy interests and power interests. All actors have both policy interests and power interests, and differences in social partnerships are shaped by the interest constellations of actors as much as by institutional features. Trampusch (2006Trampusch ( , 122-123, see also 2007 argues that retrenchment policies could be supported by collective actors offering an opportunity of welfare delegation into occupational programmes, and hence that research should take into consideration the IR system and the institutionalisation of labour relations. Along the same lines, Andersen et al. (2014, 95) see Nordic unions as willing and able to combine conflict and cooperation and to mobilise both power and responsibility, taking part not only in crises resolution in the face of external shocks but also in government reform proposals. In an influential account on pension reforms in "Bismarckian" systems, Schludi (2005, 222) notes that governments are typically the "first movers" in reform processes, so that trade unions mainly play a reactive role. Trade unions typically prefer an outcome closer to the status quo than the government does, and thus it will be crucial to find an agreement pointa compromise that can satisfy both parties. Moreover, trade unions typically are more sensitive to the interests of older workers than the government is (Anderson and Meyer 2003, 24;Schludi 2005, 80). As strategies to move the governments' position, unions can use lobbying, bargaining and mobilising (Schludi 2005). Lobbying implies to try to influence reform contents through party channels, hoping to change the minds of a critical number of the ruling party's supporters. Bargaining implies to change reform outcomes by securing a package deal with the government, often "trading" support for concessions or side payments. Mobilising entails direct action, including public declarations, demonstrations or strikes.
The three strategies are not mutually exclusive, and their implementation may depend on political and institutional factors (Schludi 2005, 231). Lobbying may be particularly successful in situations where the government is ideologically fragmented, and/or the unions control a sizeable share of key positions within the party apparatus. The success of mobilising strategies, on the other hand, depends heavily on rank-and-file involvement. Bargaining may be most efficient when unions have credible arenas to exercise veto power. Several authors have suggested that countries where the reform process has been basically parliamentary are able to introduce more far-reaching and radical reform than countries where the social partners are integrated into the pension system (e.g. Anderson and Meyer 2003;Marier 2008).
Trade unions have historically had the opportunity to establish welfare arrangements on their own, supplementing or competing with state schemes. This includes occupational pensions, which play an increasingly important role in many countries (Natali 2017;Natali and Pavolini 2017). However, the turn to contractual pensions plays a limited role in the analyses of pension reforms quoted so far, yet establishing and controlling such schemes are obvious potential sources of power for trade unions (Trampusch 2007;Grødem et al. 2018). One important reason for this is that occupational pensions interact with state pensions, sometimes in such ways that they can neutralise desired effects of state pension reforms. Increased pensioning age or stronger built in incentives to work longer in state pensions will, for instance, be rendered relatively irrelevant if most workers can retire early with a negotiated pension. Therefore, contractual occupational pensions, based on collective agreements, give trade unions a potential trump card in pension reform processes.
Another effect of occupational pensions is that they require trade unions to build competence in the pension area. If they are going to negotiate pensions, and make reasonable demands on pensions and on wages, they need their own expertise. This is another potential power source, given the importance of expert networks and "epistemic communities" (Maher 2017;Bridgen and Meyer 2018) in pension reforms. Drawing on the tradition from Heclo (1978), these recent contributions argue that the complexity of the pension issue is understood by relatively few, so that in-depth discussions typically take place within limited networks. A link can be made to recent work by Stefan Svallfors, who argues that "new" politics are increasingly relying on nonelected experts (Svallfors 2016a(Svallfors , 2016b. While Svallfors' mainly refers to paid political employees, he acknowledges that "trade union and business association experts" can play this role (Svallfors 2016b, 4). If the trade unions are thrown out the main door in policymaking through the erosion of corporate channels, then, it is possible that they are let back in through the back door in roles as policy experts.
In what follows, we build on these insights regarding the importance of the institutional set-up, expert networks and trade unions' power strategies, including ownership of negotiated pensions. We suggest that while Norway in many respects is similar to Sweden in the sense that the pension reform process was anchored in Parliament, trade unions were brought into the process in a way they were not in Sweden. The reason for this was partly that the contractual early retirement scheme (AFP), which was governed by the social partners and which had the potential to undermine the reform, created a stumbling block. Politicians thus had important reasons to cooperate with the unions also in the process of reforming the state system, in a way that was not necessary in Sweden. Second, we adopt Schludi's (2005) concepts of lobbying and mobilisation, but we suggest an adaptation of his concept of bargaining: bargaining can refer both to the formalised and institutionalised negotiation processes, and to less formal, behind-the-door cooperation in what can be understood as an expert network. Hence, we will build on the literature on expert networks, but placed in a Nordic setting.

The Norwegian pension reform
The Norwegian pension reform was initiated by the appointment of the Pension Commission in 2001. The Pension Commission came as one in several initiatives, launched by the first government of Prime Minister Jens Stoltenberg, to bring state expenses under control. The Pension Commission was made up of all parties in Parliament plus a few experts, while the social partners were not represented. The leader of the Commission was a prominent Labour politician, former Minister of Finance Sigbjørn Johnsen. The Commission reported in 2004 (NOU 2004:1) and proposed a Swedish-style NDC scheme where all years of pension accrual up to the age of 75 count equally, and a flexible pensioning age (62-75) on actuarially neutral terms. These measures were intended to promote longer working careers. In addition, the reform contained two pure retrenchment measures: longevity adjustment, and indexing of pensions at 0.75% below the annual wage increase (Pedersen 2014). All political parties in the Commission, with the exception of the two representatives from left-wing Socialist Party and the right wing Progress Party, stood united behind the proposal.
This consensus was a good starting point, but the Commission's proposal was still just that: a commission's proposal. Successive governments emphasised the need to reach binding cross-party consensus in order to secure a sustainable reform for the future. The Parliament therefore voted twice on the key principles of the new pension system in what was called parliamentary settlements in 2005 and 2007. A settlement is basically an extended parliamentary process on a white paper, and is seen as a recent parliamentary innovation that contributes to binding political consensus in larger reforms (Narud et al. 2014). The settlement in spring 2005 (Finansdepartementet 2004(Finansdepartementet -2005  The trade unions, most notably LO, had pursued their own pension policies well before the Commission was appointed. The unions had long required an early retirement scheme, yet in the course of the 1980s they realised this demand was not going to be met in the National Insurance. In 1988, as mentioned, they took matters into their own hands and negotiated their ownthe AFP scheme. This was an occupational scheme negotiated nationally and made available to all employees in workplaces covered by collective agreements. The state agreed to administer the scheme, and also to subsidise the scheme by 1/3 of the costs, in order to help land the 1988 wage negotiations. This was one of the state interventions that are common in Norway, but unusual elsewhere (Andersen et al. 2014). The scheme was expanded during the 1990s, and since 1998 it had allowed for pensioning at the age of 62 on the same terms as if the claimant had worked until the age of 67. The AFP scheme was one of LO's crown jewels. The Pension Commission, however, understandably deemed the AFP scheme as incompatible with the new principles of life-long pension accrual and flexible pensioning on actuarially neutral terms, and proposed to end the state funding. This move would kill the scheme, which was a worst-case scenario for LO.
The Pension Commission, then, laid the basis for parliamentary consensus, while leaving the largest trade union fuming on the side-lines. The remainder of this article outlines how LO was brought into the overall consensus and became an important supporter of the pension reform.

Data and methods
Our analysis relies on qualitative interviews with informants who were central in the pension reform process. We interviewed senior officials in the Ministry of Labour and Social Inclusion, 1 LO officials and staff (including former leaders Valla and Flåthen), key politicians (including former Prime Ministers Stoltenberg and Bondevik, former Minister of Labour and Social Inclusion Hansen, former Ministers of Finance Johnsen and Schjødt-Pedersen), as well as central players in unions outside LO, and in the Confederation of Norwegian Enterprise (NHO). We also talked to a few independent experts. All in all, we carried out 14 interviews with 18 informants between November 2015 and September 2016. Only one of the respondents asked to participate did not respond to the invitation. All the interviews were recorded and fully transcribed. A complete list of informants is provided in Table A1 (Appendix).
Inspired by the process tracing method (Falleti 2016;Trampusch and Palier 2016), we made it a key aim of the interviews to pinpoint the process in as much detail as possible: who talked to whom when, under what circumstances and institutional conditions, promoting which interests and so on. Both authors of this article participated in almost all interviews, and when that was not possible we made contact shortly afterwards to discuss the new information. Most of our informants had been close to some parts of the process while absent or excluded from other parts of it, and thus their narratives only partly overlapped. An important part of the analysis was thus the ongoing dialogue between us, where we aimed to establish an overview of the process, weighing the input from different informants against each otherand also discussing what we needed to clarify in subsequent interviews with new informants. By the end of the data collection, therefore, we were confident we had a good overview of both meeting places, actors and the interests the actors sought to promote in different settings. In the analytical phase, we went back and forth between our memories of the interviews, our interpretations and the interview transcripts, in a hermeneutic manner that deepened and clarified our understanding of the process.
We have also consulted some of the key documents produced during the process. In this article, however, these documents are referred to mainly to fill out the information given in the interviews, and do not play an independent role as sources of information.
2005: getting the LO on boardmutual lobbying All the larger parties in Parliament were committed to overarching principles of the pension reform, and thus, in 2004, the Norwegian reform process was set to mimic the almost purely parliamentary process than had taken place in Sweden ten years earlier (Schludi 2005;Marier 2008). However, this is where the Norwegian story turns interesting, as seen from the vantage point of the pension literature. Unlike Swedish unions, LO did not offer "tacit acceptance" (Anderson and Meyer 2003) for the new reform principles. LO had its own processes running with regard to pensions, and our interviews indicate that they had not expected the 2004 Commission to be a game-changer. When the Commission reported, LO-leader Gerd-Liv Valla was furiousin the interview we did with her, she called the Commission's proposal "unacceptable". LO was particularly enraged by the proposals to end public funding for AFP and to weaken occupational pensions in the public sector, and also demanded a better "social profile" in the new pension system. Moreover, LO was sceptical about the two most important cutslongevity adjustment and indexing.
Theoretically, given the solid backing for the reform in Parliament, LO could have been left on the side-lines. However, Jens Stoltenbergleader of the Labour party, with clear ambitions to become Prime Minister after the 2005 electionhad other plans. Stoltenberg was a main proponent of the reform, and getting LO on board was a strategic priority for him. He launched a one-man campaign to turn LO around, starting with Gerd-Liv Valla. He did this, he said in his interview, because he recognised LO's "Veritas function" 2 : if LO launched a campaign saying the reform had negative social consequences, his voters would believe it. LO might not have much to play on with the Parliamentary opposition, but there was an opposition to the pension reform within his own party. If LO joined this opposition, it would not only be harder to mobilise support and win the election, his future job as Prime Minister would also become infinitely more difficult. In short, securing support from a broad alliance in Parliament would simply not be enough. Stoltenberg therefore initiated and maintained close informal contact with Gerd-Liv Valla in 2004/2005, and the two discussed pensions almost incessantly for several months (as confirmed in both their interviews). Eventually they did reach a consensus: Gerd-Liv Valla said she was honestly convinced by Stoltenberg's argument that when expected longevity increased it was only fair that people should work a little longer. Stoltenberg, on the other hand, said that he too was convinced about something important in these long and informal discussions: he realised that distribution effects mattered, that he could not "sell" the reform solely as a reform to safeguard public finances. Stoltenberg's rhetoric thus changed between 2001 and 2005 (as also noted by Ervik and Lindén 2015).
Labour's policies on pensions in 2005 came to deviate somewhat from the Pension Commission's proposal, as the party and the union found common ground. The Labour party by 2005 promoted a model of pension accrual that was more beneficial to the lowest paid, and promised to maintain AFP and public sector occupational pensions. Stoltenberg and Valla presented a joint proposal to the Labour party's executive committee along those lines, and as the two were perceived as leaders of opposing fractions, the rest of the committee supported their compromise.
The upcoming LO Congress in May 2005 was, however, the main stumbling block. The Congress is LO's main political assembly, held every 4 years with 300 elected delegates from all over the country. The LO leaders and the Labour party had found a compromise, but this was of limited interest if the LO Congress voted it down. At the Congress, LO leaders stressed repeatedly that the Congress was not invited to vote on the Pension Commission's proposal, but rather on the Labour/LO model. Jens Stoltenberg was invited to the Congress, and held a carefully prepared address where he defended the main principles of the reform (LO 2005). He went far in threatening the Congress: if the delegates voted the Labour/LO model down, it would cause disruption and weaken Labour's chances of winning the upcoming election. The passing of the reform would then be left to the parties on the right, who would ride roughshod over LO's interests.
The (left-wing) opposition at the LO Congress initially raised three demands that diverged from the LO-Labour consensus, but allowed two of them to fall during the debates. However, the final demand stood: no decision on adjustments for longevity or indexation at this point in time (LO 2005). The opposition argued that one did not know how long anyone would live in the future, and thus there could be no need to rush into a decision on longevity (LO 2005, 320). This demand had solid backing. Several of our informants who were present at the Congress talked how Gerd-Liv Valla personally initiated a frantic, last-minute campaign the night before the final vote in order to secure support for the LO executive proposal. According to one informant, Valla and (then) deputy leader Roar Flåthen put all their authority behind the proposal, and the last-minute campaign worked: the majority proposal passed, with 189 votes against 121. The Labour party thus got the mandate it hoped for in the negotiations with the government, and a parliamentary settlement on pensions was formalised on 26 May 2005 (Stortinget 2004(Stortinget -2005. The 2005 compromise left a lot of work, and a lot of hard negotiations, to be done, but it got the reform over the first hurdle. As Stoltenberg and others had hoped, old age pensions and the pension reform were mainly kept out of the 2005 election campaign. Labour did well in the election, and Jens Stoltenberg became Prime Minister for a coalition government made up of Labour, the Socialist Party and the Centre party.
This story from 2005, we suggest, was a story of mutual lobbying and exchange: Stoltenberg lobbied LO, in a long campaign that culminated with his speech to the Congress. Simultaneously, LO lobbied the Labour leadershipof which the LO leader was a partand convinced the party to drop the proposal to end state funding of AFP and to improve the social profile of the reformed state pensions. All in all, this made for a successful "party-to-union" lobbying campaign, in an exchange where both sides won some and lost some. The Stoltenberg strategy was, however, risky, as the issue of the contractual AFP programme was swept under the carpet and few, if any, saw how an early retirement scheme could be reformed and coexist with a DC-styled old age pensions scheme. In addition, the more precise distributional profile of pension accumulation in the new system was still to be decided.

2007: towards a settlementtrust and expertise
The 2005 Stoltenberg government started immediately to prepare for a second broad pension settlement in the Parliament. A new White Paper was presented on 20 October 2006, almost exactly one year after the government took office (Arbeids-og inkluderingsdepartementet 2006(Arbeids-og inkluderingsdepartementet -2007. The laboratory for working out the White Paper, we learned through the interviews, was the Ministry of Labour and Social Inclusion. All our informants hinted at this processmost of them had been directly involved in itbut they differed in how concrete information they gave. One of the most candid accounts came from Bjarne Håkon Hansen, who was Minister of Labour and Social Inclusion at the time: There were two types of meetings. In other words, the government worked parallel with developing the pension model and anchoring it among the coalition partners. This was a two-step process: the features of the model were worked out in a group made up by Labour members of government, senior bureaucrats in the Ministry of Labour and Social Inclusion and central LO representatives. This group met frequently and worked intensely. Then there were the other kind of meetings, where the aim was to get the support of the government coalition. The differences between the parties were too big to be handled at the level of political advisors, and thus the party leaders were brought in. However, the agenda in these meetings was not to open up for the coalition partners to make their mark on the reform, but rather to convince them that the choices made in the working group were the best ones.
Interestingly, the Ministry of Finance was barely involved in this process. Jens Stoltenberg explained why: It is quite remarkable that Finance was not involved in that process. And that was because Kristin Halvorsen was there, and even though I trusted Kristin, I did not trust the Socialist party. And there were so many political advisors from the Socialist Party in that Ministry, it would just have slipped and made it difficult for the Socialist Party to get on board. […] Even secret to large sections of the LO executive! Secret to two-thirds of the government! Stoltenberg openly acknowledged that in terms of democracy and ideals of open procedures this was a dilemma. His justification was that this process was so complex, and it was vital to maintain trust and create a space where ideas could be tried out without the press and the opposition breathing down their necks. In the absence of this closed space, he argued that positions would freeze too early, participants would not be able to move from established positions without losing face and the process would have imploded.
The process towards the second parliamentary settlement illustrates very clearly the importance of networks in pension policies (Bridgen and Meyer 2018), and also point to the role of expertise and trust relationships in modern politics more generally (Svallfors 2016a(Svallfors , 2016b. Many of our informants remarked on the sense of trust that developed between the members of this group, noting in particular how trade union executives and ministry advisors found common ground and worked as a team. An important aspect of this trust relationship was the confidence that no one would leak to the press, so that the group could work in peace. This trust was not extended to other actors, not even advisors in the Minister of Finance. In this process, LO appears to have been seen by leading Labour politicians as fully loyal collaborators, and such relationships can be crucial in "the hard and unforgiving environment" that modern politics, including the relationship with the press, entails (Svallfors 2016a, 3). In addition, expertise in the complex pension area is hard to come by, but it can be found in the LO secretariat. LO could thus provide both loyalty and expertise.
The discussions leading up to the second White Paper (Arbeids-og inkluderingsdepartementet [2006][2007] and the following parliamentary settlement were about details of the new system rather than principles. None of the parties came into the discussion with a blueprint for their preferred outcome, and the setting was informal and dialogue-oriented. We suggest that the Norwegian set-up, with long traditions for tripartite dialogue, provided the ground for these proceedings. Ironically, however, the tripartite cooperation had also produced results that made close contact between the union and the Ministry necessary: the social partners and the government were locked together in the AFP scheme, which was owned by the partners but administered and partially funded by the state. If the work incentives inherent in the new system were to have any meaning, the contractual early retirement schemewhich offered 62-year-olds the opportunity to retire on the same terms as 67-year-oldshad to be significantly remodelled.
2008: tripartite dialogue and an expensive occupational scheme AFP was addressed in the wage negotiations in the private sector in 2008. The stakes were high: large sections of LO had accepted the reform of state pensions partly because they expected to compensate for any losses through contractual schemes. The year 2008 was the first time this assumption was explicitly confronted. Also, obviously, because the scheme was contractual, the action now shifted to a new arena: in 2006-2007, LOs were guest actors in a play that mainly took place on the political arena. In 2007-2008, the main arena was the bipartite private sector wage negotiations, where politicians and bureaucrats participated only to the extent the social partners would let them.
The negotiations were primarily between LO and NHO (the Confederation of Norwegian Enterprise). NHO welcomed the reform in state pensions, and were willing to make concessions in the AFP negotiations in order to safeguard the gains in the pension reform. Our interviews show that NHO's primary interest was to keep costs stable for employers; they had no strong interests in the distribution profile of the AFP scheme. The main discussions on AFP, therefore, were between LO and the government, as the government partly funded the scheme.
The 2008 process started more or less where the settlement in the Parliament left off. Only two months after the 2007 settlement, the Minister of Labour and Social Inclusion established a committee to prepare the negotiations. This committee included many of the same players who had been instrumental before the settlement, with the important exception that the NHO now was fully included. The committee issued a report that worked as a background document for the upcoming negotiations, but did notand did not aspire tocome up with a proposal that could be sanctioned by the partners.
The document laid the groundworks for an elite compromise. In this round, however, elite compromise would not do: the social partners were negotiating, and the final result would be balloted. Among the LO rank-and-file, there was considerable scepticism against touching the AFP, and the left-wing opposition were rattling their sabres. Thinking ahead, Roar Flåthennow LO leader and a main player in the negotiationscalled a meeting with two key opposition leaders early in the negotiation process and outlined his preferred outcome. He sketched a solution where AFP was remodelled into a lifelong addition to state pensions, to be drawn on actuarially neutral terms, but where the initial level was so generous that the oldest cohorts would obtain the same replacement level at 62 as they did under the prevailing scheme when combining AFP and state pensions. In his own words -"those two gentlemen though it over fairly quickly, then said yes, if you can get that, Roar, we're in" (R. Flåthen, interview November 2015). Then Flåthen went back to the negotiating table and announced that if he got his way, he could almost guarantee LO support.
Flåthen's strategy was effective, and the negotiations were summed up as an astonishing victory for LO. The contractual AFP scheme survived as a top-up to the state pension scheme, offering life-long pensions to employees covered. The end result was that the government put about NOK 50 billion on the table to ease the transition from the "old" to the "new" AFP. The Ministry, however, also felt victorious: it got an AFP that complimented the National Insurance pension instead of undermining it. The 50 billion was an estimate of accumulated sums over a long period, and could be seen as a long-term investment in the overall reform. Similarly, NHO was satisfied: they achieved system change without increasing employer costs.
Among LO rank-and-file, however, not everybody was convinced. LO launched an information campaign immediately after the conclusion of the negotiations to promote the result. The efforts that had been made beforehand to bring the left-wing opposition leaders on board now paid off: those were strong union leaders with considerable credibility on the LO left, and they campaigned actively together with LO executives to ensure support. In the end, the ballots had an exceptionally high turn-out, and 76% voted in favour of the negotiated result. This concluded the pension reform process in the private sector and the final bill on the new state pension could be passed through the parliament.
The 2008 process was bargaining in the real sense. However, it was a set-up that would not have been possible in countries with a weaker tradition for tripartite dialogue: the social partners negotiated, but the state was only a phone-call away, willing to smooth the negotiations with literally billions of kroner. The process thus drew heavily on the traditions for tripartite cooperation. The outcome would probably not have been possible, had it not been for LO's involvement in the process leading to the 2007 settlement in parliament. This process fostered relationships of trust, as well as mutual understandings of the issues. In the end, the chain of processes led to a result where all the key actors could claim victory: Labour politicians got their reform, complete with harmonised contractual pensions, and the trade unions got an exceptionally generous transition period.

Discussion
This analysis leaves no doubt that the pension reform was a carefully orchestrated process. Jens Stoltenberg, as Labour leader throughout this period and as Prime Minister from 2005, was in a position to determine the timing of events, and he managed this role carefully. In the interview we did with him, he freely admitted that he did not envision at the beginning of the process how this was going to succeed, he only knew that it would have to be done one step at the time. The thorniest issues had to be postponed, as the aim early in the process was to secure broad support for the basic features of the new national pension model. The relationship with LO moved from one of "mutual lobbying" to one of networking and bargaining. Indeed, many of our informants from LO talked about this as a continuous process: this was the period when they buzzed in and out of the government's offices and discussed pensions, the distinction between the National Insurance pension and the AFP is blurred in their narratives. Nevertheless, the reform architects made sure that the National Insurance process was finalised before the real debate on the contractual early retirement scheme started, so that the arguably less predictable AFP process was not allowed to interfere with the debate over the National Insurance. The timing and sequencing of events undoubtedly mattered.
Following from this, this study confirms what has been pointed out by others (Allern et al. 2007)that the relationship between the LO and the Labour party in Norway was very close even by the early 2000s, and perhaps closer than discussed in most IR and welfare state literature. In the process described here, we have seen that the Labour leader was given ample space to address the LO Congress in 2005 to argue for the party's politics, LO was invited in to design welfare policy schemes and the Labour-led government followed the 2008 negotiations between the social partners like a shadow. The areas of combat changed from LO's home field (the Congress) to Labour's (governmental offices) back to LO's again (the negotiations), but the key figures in the party and the union maintained close dialogue irrespective of where the formal action was. In addition, while LO was actively involved, other potentially important players were side-lined. This included the coalition parties in the 2005-2013 governments and, even more strikingly, the Ministry of Finance.
We have suggested that the first phase considered here can be described as "mutual lobbying". The leader of the Labour party and the leader of LO discussed the principles of the pension reform, frequently and informally, and in the process both adapted their policies and rhetoric. Their compromise was later accepted by the Labour party's executive committee, LO's executive committee and, finally, the LO Congress.
The second phase was one of expert networking. We suggest that the most striking feature of this stage of the process is the inclusion and exclusion of actors: when the newly elected Labour-led coalition government was to fine-tune its pension policy, Labour took the lead and invited LO into the process. Not the government coalition partners, not the Ministry of Finance, not the finance industry -LO. Also, this took place without the usual formal hearings and underway reports to the parliament. This is indicative, we have argued, of the importance of expert networks in pension policies, and also of the potential importance of durable networks where mutual trust develops over time. LO and the Labour party have, historically, had this close relationship in Norway. Also, we have suggested that the existence of occupational pensions and a contractual AFP scheme made it necessary for LO to build and maintain competence on pensions within its own secretariat. The historical relationship between the party and the trade union, the Nordic-style IR regime and the pension nexus in Norway made LO's role in this phase possible.
The shared understandings developed in the second phase were carried over to the third phase, which was bargaining in the classical sense. LO had the upper hand in this round, as it claimed ownership to the contractual scheme being negotiated, and the Labour-led government only got its desired result after putting considerable amounts of money on the table. The outcome was seen as a major victory for LO, who finally got paid for years of loyal support to the Labour-party-initiated reform.

Conclusion
What light can this case study shed on the role of trade unions in welfare retrenchment processes more generally? When seen in the light of key contributions like Schludi (2005) and Marier (2008), the Norwegian case is an anomaly. Schludi (2005) makes the case that reform architects must secure the support of either the opposition in Parliament or of the trade unions, and provides a nuanced discussion of the conditions under which politicians prefer to pursue each option. Moreover, unions are seen to push the reform compromise as close to status quo as possible. From this vantage point, one must wonder at the Norwegian reform architects' double effort: the Pension Commission, in which all parties in parliament were represented, had laid the ground for a solid "Swedish-style" parliamentary consensus. However, Stoltenberg referred to LO's "Veritas position": given LO's credibility on social issues, intensive LO lobbying against his own party would have made his life as Prime Minister difficult. Moreover, he knew from the start that AFP had to be dealt with at some point. He also knew that an open confrontation with LO on AFP would be traumatising on both sides, and should be avoided at any cost. This is perhaps the most important reason why Stoltenberg and his fellow Labour leaders worked so hard to anchor the key principles of the new pension system within LO from 2004 onwards: if LO adopted these principles, they might agree to redesign AFP along the same lines later. As we have seen, this is precisely what happened.
Against this background, we have suggested that contractual pensionsespecially when they are designed so that they can neutralise or undermine reforms in state pensionshave potential to represent a formidable source of power for trade unions. This point has received little attention in previous literature on unions' power strategies in welfare reforms.
A second point that should be made is that the Norwegian IR regime, combined with the historically close ties between the party and the union, has laid the ground for a deep sense of trust between the parties. The difficult bargaining on the future pension system was done "in the cup-board"to paraphrase Stoltenbergbetween bureaucrats, the responsible Minister and LO officials. This select group built on the informal channels nurtured in the tripartite dialogue and the tradition for confidentiality.
A third point that comes out of this analysis is that when studying lobbying and bargaining processes, we should pay more attention to the role of expertise. Modern politics are complicated, and often even decisionmakers struggle to make sense of the issues (Svallfors 2016a). Nonelected experts, then, be they political advisors, political secretaries, think-tank employees or trade union/business association professionals, can therefore play a major role. This is not least true in pension politics (Bridgen and Meyer 2018). Lobbying and bargaining thus not only implies promoting one's point of view, but can also mean participation in mutual processes of problem construction, shared understandings and a joint search for acceptable solutions. This was LO's role in Norway in 2006/2007, and understandings developed in this phase were then carried over to the 2008 formal negotiations.
In one aspect, however, LO's role in the Norwegian reform was textbook: LO very clearly prioritised the interests of older workers over younger. The entire AFP settlement was about securing the same level of pension for ageing workers about to retire. Younger cohorts will not be sheltered from the retrenchment elements in the reform in the same way (as shown by Fredriksen and Stølen 2015).
In the introduction, we highlighted the need for an accumulation of case studies to shed light on the complex interplay between social policy and IR in the field of pension reform. We have analysed a country where trade unions remain powerful, where the relationship between the largest trade union and the Labour party is close and where there is a strong traditioneven by Nordic standardsfor tripartite cooperation. We suggest that the observations we have made about the role of contractual pensions, trust between the parties, as well as of expert networks, will have validity even beyond this immediate context, although exactly how this plays out will of course vary between different settings. Still, the institutional settings are changing even in Norway. It remains to be seen whether the 2011 pension reform will go down in history as the pinnacle of the Nordic-style tripartite cooperation model, or whether future reforms again will be able to draw on the same mechanisms of mutual lobbying and networking combined with hard power.