Abstract
Introduction
In recent years, the discussion in political economy has turned from the rather static varieties of capitalism approach to a more dynamic view of growth models, where the wage setting of the labour market actors plays a key role for how growth is kept on track. According to Baccaro et al. (2022), most capitalist countries can be understood as belonging to a specific ‘growth model’ of economic and material improvement, with the models defined by distinct labour market policies, institutions, and relationships between the social partners. As economic growth is often essential to both material welfare and political legitimacy, understanding the mechanisms and outcomes of these growth models is of considerable research interest, especially as economic growth has stalled in many countries (Baccaro et al., 2022). In this paper we will evaluate the role of the wage-setting system in Sweden as a critical case of the ‘balanced growth model’ (Baccaro and Pontusson, 2023), which claims to deliver both economic growth as well as wage equality, and the extent to which the latter claim is true.
According to the growth model approach, most industrialised countries can currently be characterised as belonging to one out of two ideal-typical models of growth – the consumption-led or the export-led growth model. Both models contrast to previous Fordist era wage-led growth model, in which relatively high and uniform wages and real wage increases created aggregate growth through rising demand. Common components of this model were collective bargaining and the indexing of real wages to productivity growth, which helped keep wages at a high level. The model was eventually abandoned, commonly described as a result of wage militancy undermining the rate of profit and wage-induced inflation, lowering growth as a result (Baccaro et al., 2022).
The consumption-led model relies on domestic consumption and credit/debt-financed growth to generate aggregate demand and economic growth, exemplified by countries such as the USA and the UK. Noteworthy components of this model are the rising wages of high-income groups and the expansion of credit and debt for those with lower income to maintain a standard of living in a context of increasing inequality (Baccaro et al., 2022; Reisenbichler and Wiedemann, 2022). The second is an export-led growth model where profits from exports, most commonly manufacturing, are the source of aggregate economic growth, with Germany being a case in point. A relatively low-wage share, especially in sectors outside manufacturing, help keep down domestic demand, with a low national interest rate as a result, which retains prices of export goods at a competitive level (Baccaro et al., 2022).
However, there is also a third model of balanced growth with Sweden being the ‘ideal-typical’ case of this growth model (Baccaro et al., 2022: 42). In this balanced model, exports are still essential, but the role of the wage-setting system is central. On the one hand, the wage-setting system keeps down real wage growth and inflation, which keep export prices competitive, fostering profit-led export growth. On the other, and unlike the export-led growth model, the wage floor is also kept at a relatively high level in sectors outside manufacturing that encourage a domestic wage-led growth. The balanced model seems to offer a seemingly beneficial growth model, offering both economic growth as well as wage equality through a relatively compressed wage dispersion.
In this paper we will evaluate to what extent this latter claim is true. If excessive wage dispersion is present in the ideal-typical case of the balanced growth model represented by Sweden, especially if the dispersion strengthens structures of inequality, this will undermine the claims of the balanced growth model. If the balanced growth model cannot deliver on both goals, it will call into question if it is possible to combine economic growth with wage equality or if all Western models of capitalism tend to promote the former to the detriment of the latter.
We will in this paper explore the issue of wage equality, using the Swedish wage-setting system as a case in point, with a more nuanced measure of employee wage equality that goes beyond sector comparisons. Most studies of Sweden have focused on the rather stable period since the deep crises in the 1990s, when unemployment as well as public debt skyrocketed. Thereafter, the Swedish labour market model stabilised and wage earners could count on real wage increases for most years. Even the financial crisis of the late 2000s only partly affected the Swedish model, although Erixon and Pontusson (2022) found signs of a shift in the direction of finance and consumption-led growth model due to strongly increased private debt. However, in this paper we will also analyse how the even greater challenge of the COVID pandemic and the following high-inflation period has affected the wage-setting model.
Since 1998, Sweden has adopted a wage-setting model known as the Industrial Agreement (Industriavtalet, IA), which aims to foster international competitiveness by preventing excessive wage dispersion and inflation. The IA applies a form of pattern bargaining that requires export industries to determine the scope for wage increases among all sectors (Müller et al., 2018). In subsequent years, the IA model successfully became the overall norm for wage growth and was further institutionalised with the establishment of the Swedish National Mediation Office (SNMO) in 2001, which enforces wage increases in the export industries as benchmarks for the rest of the labour market (Elvander, 2003).
In recent research, the IA's wage-setting system was described as an important explanation for the balanced growth in Sweden, with the effect of diffusing wage growth from the manufacturing to other sectors and sheltered industries (Erixon and Pontusson, 2022). According to Erixon and Pontusson (2022), this pattern is to a high degree explained by the wage-setting model in combination with strong unions and a large public sector (which is also a union stronghold). Wage growth in the sheltered sectors could therefore keep pace with that in the export-dominated sector. Relatively high and uniform wages encouraged domestic consumption, which added to the aggregate growth.
Overall, the Swedish model and economy seemed to be functioning well before the COVID crises, with real wage increases being on average 2.1% annually throughout the 2000s and 1.4% during the 2010s compared with 0.24% and 0.42%, respectively, in the Eurozone (Swedish National Mediation Office, 2022). Regarding equality, the literature indicates that the IA helped to counteract wage inequalities between occupations, with wage differences being relatively stable since the 2000s (Alfonsson et al., 2024; Marginson and Dølvik, 2020). Karimi et al. (2024) find that the 90:10, 50:10, and 99:50 ratios of low-, medium-, high-, and very high-wage earners are essentially stable from 2010 to 2021. In addition, there were no indications of a growing low-wage sector in Sweden (Alfonsson et al., 2024). At the aggregate level, wage growth in Sweden seems to keep to a strong solidaristic pattern in the sense that the highest paid occupational categories have not moved away from the lowest paid categories.
However, there are also indications that the IA model is facing challenges regarding its promotion of wage equality. Granqvist and Regnér (2016) found noticeable differences in wage increases on occupational levels. These results are in line with a contrasting account of the Swedish wage-setting model that sees the dissolution of solidaristic traits with larger disparities between occupational groups using their power resources to influence wage increases. Arguably, Sweden has started to move along a neoliberal trajectory with increasing inequalities in the labour market (Baccaro and Howell, 2017). Combining these observations with the discovered shift in the direction to a finance and consumption-led growth (Erixon and Pontusson, 2022), the question arises as to how the wage-setting model performs in a more hostile environment of high inflation: Does it withhold wage increases for all, or is inequality on the rise? Two aspects of wage equality will thus be in focus, the presence of a uniform and compressed wage growth, and the reduction of the wage gap between high and low earners.
The paper analyses the wage-setting period from 2014 to 2023, which is divided into the pre- and post-COVID-19 pandemic periods. The years heading up to the COVID-19 pandemic were characterised by extremely low inflation, while the years since the outbreak have been shaped by massive inflation in addition to geopolitical turbulence and insecurity, with both periods having substantial consequences for the wage-setting model.
The paper is structured as follows. We begin by outlining the background and development of the IA as the central component of the Swedish balanced growth model, including its designated benchmarks for wage and cost increases during the specified period. This is followed by a review of the relevant literature, focusing on earlier research related to the Swedish wage-setting model and the structure of wage increases. Next, the data and method section presents the datasets and analytical approaches employed. The empirical results are then presented and discussed, and the paper concludes with a comprehensive discussion of the findings.
Background to the Swedish wage-setting model and the IA
Three distinct wage-setting models have been applied in Sweden since the 1950s. The first period from 1956 to 1982 was characterised by a centralised solidaristic wage policy (Erixon, 2010). The Swedish Trade Union Confederation (Landsorganisationen i Sverige, LO) and the Swedish Employers’ Confederation (Svenska Arbetsgivareföreningen, SAF) centrally determined the pace of wage increases, which were subsequently negotiated through collective and local agreements. This regime included mutual acceptance of wage restraints to enhance competitiveness in Swedish industries. Wage equalisation helped maintain lower wages in high-wage industries and raised them in low-wage industries. Pontusson and Swenson (1996) argued that this model not only equalised wages within occupational groups across sectors but also compressed the overall wage structure by decreasing wage inequalities between positions and occupations. From 1960 to 1982, real wages increased by an average of 2.3% per year (Swedish National Mediation Office, 2022). This follows what Baccaro et al. (2022) define as a wage-led growth model.
The second period (1983–1997) was characterised by a decentralised wage-setting model among the social partners and increased government intervention in wage negotiations. This period began with the withdrawal of the Swedish Engineering Employers’ Association (Verkstadsföreningen) and the Swedish metal workers’ trade union IF Metall from centralised bargaining. Subsequently, in 1984, the SAF also disengaged from centralised bargaining, which led to inflationary pressures and government attempts to steer the concerned parties back towards centralisation. In 1990, the government-appointed negotiation committee, known as the Rehnberg Group, brokered a 2-year agreement, which included a wage freeze for the first 6 months of 1991, a ban on strikes, and small wage increases in the second year (Elvander, 1997, 2003). In 1993, the social partners regained autonomy in negotiations, but decentralisation resurfaced, leading to a lack of consensus from both unions and employers on the duration and level of wage increases. No central agreement was reached, resulting in varying wage increases and duration of agreements across sectors (Elvander, 1997). The average real wage increase per year during this period was 0.93%.
The third period, beginning in 1998, is defined by the IA and marked a phase of centralisation aimed at curbing inflation and aligning profit levels with international standards. Following a period of low real wage increases and high inflation, the Social Democratic government in the mid-1990s began pressuring the social partners to find a wage-setting model that aligned with the European norm and curbed inflation. The government urged all social partners to participate in reforming the wage bargaining model to ensure it promoted economic growth and competitiveness (Albåge and Larsson, 2003). The IA was signed in 1997 (Elvander, 2003). To date, 11 agreements have been made under the IA model, with an average real wage growth of 1.32% per year between 1998 and 2023. The IA is in line with the balanced growth model, as it aims for a complementary relation between profit-led export sectors and wage-led domestic sectors.
In the IA, the manufacturing and export industry conditions serve as the foundation for negotiations, with unions and employers required to consider the impacts of wage decisions on inflation, employment, and competitiveness. The IA strives to curb inflation through moderate nominal wage increases, thereby achieving higher real wages without adverse effects on interest rates or competitiveness. Therefore, it acts in tandem with the central bank's fiscal policies and the bank is expected to keep inflation at 2% (Bender, 2024). These agreements are made with a high level of centralisation while actual wage negotiations are mainly made at the sectoral or local levels (e.g. at individual workplaces and by individuals) without central coordination. However, most local wage agreements are in line with the levels set by the IA (Bender, 2024; Swedish National Mediation Office, 2021; Ulfsdotter Eriksson et al., 2021).
Industrial agreements for the period from 2014 to 2023
Five agreements have been made within the IA framework during the period 2014–2023. The first agreement spans from April 2013 to March 2016, the second from April 2016 to March 2017, the third from April 2017 to March 2020, the fourth from November 2020 to March 2023 (with the COVID-19 pandemic delaying the start of the agreement), and the fifth from April 2023 to March 2025. The stated goals in the agreements include: ‘Sound and sustainable wage formation presupposes that the union agreements in industry constitute the cost norm and guidance for labour costs in the Swedish labour market’ (The Industrial Agreement, 2016: 19) and ‘The industrial partners undertake, individually and jointly, to work to ensure that the ‘cost benchmark’ in industry is the norm within which the other parties to the labour market should adhere’ (The Industrial Agreement, 2016: 19).
The IA does not specify what is considered an unacceptable deviation from the cost benchmark or wage increase norm, only that the negotiating partners in the IA or affected partners must come to an understanding on whether an agreement can be considered the norm (cf. Industrial Agreement, 2016). The cost benchmark is expected to cover not only wage increases but also all labour force costs, such as pensions, skills development, and paternity leave (Bender, 2024). The agreed nominal annualised cost increase is similar for the first three agreements (2.3% for the 2013 agreement, 2.2% for the 2016–2017 and 2017–2020 agreements) lower for the 2020–2023 agreement at 1.8% and higher for the 2023–2025 agreement at 3.7% (Swedish National Mediation Office, 2024).
Some exceptions to the norm have occurred during this period. From 2017 to 2018, the LO unions agreed that low-wage earners belonging to their unions could have proportionally higher-wage increases than would otherwise be allowed by the IA to promote economic equality (Edholm and Julius, 2019), which effectively raised health-care workers’ wages. In 2016, the Swedish government initiated the so-called ‘lärarlyftet’ (i.e. teachers’ raise), whereby the wages of teachers, especially competent teachers, should be raised.
The period of analysis in this paper covers dramatic changes in inflation, which are also reflected in the much higher-wage norm levels in the IA from 2023 compared with previous years. The start of the 2014–2023 period was one of relative economic and geopolitical stability, with Sweden having recovered from both the financial and Euro crises. The pre-pandemic years were also a period of extremely low inflation and the central bank tried to move the economy in the direction of the 2% inflation target by pushing down the interest rate, resulting in rising housing prices and very high levels of private debt. Erixon and Pontusson (2022) described this situation as the beginning of the balanced growth model tipping in the direction of a domestic consumption-dominated growth model driven by private lending and cheap money. However, this low-inflation period suddenly ended with the outbreak of the COVID-19 pandemic in addition to geopolitical turbulence and insecurity. Sweden started to reach inflation figures not seen since the 1980s. It is noteworthy that the rise in inflation is not wage-led, but to a large extent the result of firms’ rising mark-ups, effectively raising prices and the profit share, while real wages stagnate (Mastromatteo and Rossi, 2024). This runs counter to the logic of the IA, which is constructed on the assumption that inflation will be wage-led.
Literature review
Two aspects of the issue of wage equality are commonly addressed in the literature. The first is whether the IA can avoid deviations from the manufacturing norm, which could widen wage increase dispersion. The IA's pattern-bargaining model has been described as competitive corporatism (Bengtsson and Ryner, 2018), which is characterised by the interplay between centralised and decentralised negotiations (Thelen, 2001). This competitive corporatism increases the space for local flexibility, where the wages can be set individually within some sectors and for some employees, even though they usually match the central norm, while cross-sectoral coordination with the manufacturing norm on competitiveness-oriented upper wages has tightened (IER, 2017; Müller et al., 2018; Ulfsdotter Eriksson et al., 2021).
The trade union for managers (Ledarna) criticised the manufactural norm as acting as a ceiling, which excessively limits the possibility of setting individual wages, and believes that the manufacturing norm should be removed completely (Hermansson, 2019). The engineers’ and metal workers’ trade unions in addition to employer organisations see strict adherence to norms as detrimental, as it acts as both a ceiling and a floor in practice, resulting in too much centralisation (Hermansson, 2019). Several blue-collar unions in construction and maintenance stress that they, who frequently have centralised agreements with explicit wage growth figures, fall behind in wage increases compared with white-collar workers, as well as administrative regions and municipalities who more commonly have numberless agreements, which allows for local agreements with increases beyond the norm (Hermansson, 2019). Both the SNMO, which is tasked with mediating labour disputes to maintain the IA (2015, 2018) and Granqvist and Regnér (2016) found that comparing localised to more centralised agreements yield minor differences in wage increases but noticeably larger changes between occupations and positions within the same type of agreement. The Swedish National Institute of Economic Research (2018) found that industries with high labour shortages have greater wage increases compared to industries with low labour shortages, indicating that wage increases can be adjusted to match market needs.
The second aspect of wage equality concerns to what extent existing wage inequalities can be addressed through relative wage increase changes within the IA framework. Unions for construction and maintenance, as well as health workers, have pointed out that by defining wage increases in percentages, the IA does little to address gender wage inequality (Hermansson, 2019). The IER (2017) commented that there is often too much faith in central agreements to address these differences and that gender wage differences are often influenced by other factors, such as education and sector. As mentioned above, some exceptions to the norm have been implemented (Edholm and Julius, 2019; Hermansson, 2019). The IA has also been amended with stricter coordination procedures to make exceptions to the norm harder to implement (Müller et al., 2018). The role of the state should also be addressed, where the SNMO (2024) is tasked with promoting both the wage-setting role of the internationally competitive sector and enabling relative wage changes to promote well-functioning wage formation, which means allowing for employee groups to negotiate higher-wage increases than other groups, if the social partners consider it necessary (Bender, 2024). In addition, the increase in teachers’ wages was noteworthy because the government initiated this process instead of the social partners (Bender, 2024; Hermansson, 2019).
Overall, the literature shows that promoting wage equality through relative wage changes in the IA was contested among the social partners, who have different interpretations and implementations of the IA to achieve their own goals. There were also indications that while cross-sectoral coordination of wage setting is strict, there was more room for variation at the local level, which could affect the overall wage equality. Crucially, there is a lack of research that decomposes aggregate measures for occupations and wage increases to study the variation in wage changes between occupational groups, in particular considering structures other than the sector level. Earlier research indicates that detailed analyses of both occupational positions and wage levels are of interest as potential sources of wage inequality. Therefore, these analyses will be centred in our empirical analysis. The dramatic inflation changes during the 2014–2023 period also challenge the functioning of the IA and wage equality, which calls for further analyses of both low- and high-inflation years. These issues will be empirically analysed in the next section considering the state of wage equality, the function of the IA, and the implications of the Swedish balanced growth model.
Data and method
This paper will focus on wage growth as a potential source of wage inequality with a focus on occupational position and occupational wage. Two different data sources are used to analyse wage growth: the Swedish Wage Structure Statistics (Lönestrukturstatistiken, WSS) and the Swedish Labour Force Survey (LFS). The WSS is collected by Statistics Sweden on behalf of the SNMO, which provides the official data for occupational wages. This register collects wages, including variable pay supplements such as bonuses and overtime, recalculated to full-time monthly equivalences and includes individuals aged between 18 and 64. The register includes all wage-earners working in the public sector, or those employed in private companies with 500 employees or more. Information on the rest of the private sector is collected by sampling.
The LFS includes detailed information about employees’ labour market situation, such as contract types and working time, and is based on quarterly interviews with respondents, which means that the same individual may appear in annual data up to four times. To avoid duplicating cases in the LFS, we have restricted the analyses of particular years to respondents interviewed during the October–December quarter.
The occupational classifications used throughout this paper are based on the 2012 Swedish Standard Classification of Occupations (‘SSYK 2012’), which is in turn based on the 2008 International Standard Classifications of Occupations (‘ISCO-08’). The SSYK is a four-level hierarchical classification, and this paper focuses on the third (i.e. occupation) and first (i.e. major groups) levels. The SSYK attaches no or little weight to individual workers’ formal education, employment status (i.e. employed/self-employed), industry, or size of the company where the work is conducted. Instead, the SSYK focuses on the qualifications and specialisation that workers require to perform their tasks and duties within their occupation (Statistics Sweden, 2012). The major groups included in the SSYK are shown in Figure 1.
Overview of wage growth per occupation at the three-digit level. Annual real wage growth in per cent per occupation (adjusted for CPI, average percentage for the 2014–2018 (a) and 2020–2023 (b) periods, colour-coded according to the SSYK occupational structure). CPI: consumer price index; SSYK: Swedish Standard Classification of Occupations.
In the LFS dataset, 2014 is the first year with data collected according to the SSYK 2012 standard. Moreover, 2019 was excluded because of poor LFS data quality in that year. 1 The choice of time period is also due to the relative stability of the period and the three IAs for the 2014–2018 period having similar levels of agreed cost increases among industries (Swedish National Institute of Economic Research, 2024). In addition, 2020 was the first year of the COVID-19 pandemic, which led to high inflation. The period 2020–2023 includes three IAs with the industries’ annual cost increase set to 2.2% for the first two agreements and the third set to 3.7%, starting in April 2023 (Swedish National Institute of Economic Research, 2024). Hence, 2018 and 2023 were selected as sample years to represent these low- and high-inflation periods in the LFS. The year 2018 was chosen as it is the most recent year from the generally stable low-inflation period. The year 2023 was chosen as it is the single the year that best captures both the high inflation and an IA that responds to a labour market affected by high inflation.
In addition to the occupational category, the industry category is used in some of the following analyses. As industry is the main category through which wage growth is regulated in the IA, it acts as a central control variable as well as the standard measure of wage equality. The industry category is based on the Swedish Standard Industrial Classification (‘SNI 2007’), which is in turn based on the European Union’s (EU) recommended standard (NACE Rev. 2). Some industries are either merged or omitted due to few employees present in the dataset, for a total of 16 different industries. See Table A4 in the Appendix for an overview of industry categories, including mergers and omissions.
This analysis is conducted in two steps. First, a descriptive analysis at the aggregate level of the wage growth per occupation at the three-digit level is conducted for the 2014–2018 and 2020–2023 periods using the WSS data. This analysis focuses on the distribution of wage increases between occupations. Occupational wage growth is defined according to the average per occupation calculated by the annual wage growth for the periods 2014–2018 and 2020–2023, respectively, and divided by the number of years. The purpose is to avoid unrepresentative outliers in the wage changes that might occur for only 1 year. Wage refers to the WSS occupational wage recorded by the SNMO and Statistics Sweden.
Second, we use the individual-level LFS data and multinomial logistic regressions to study the likelihood for workers in particular occupations, industries, and wage levels to work in occupations characterised by different wage growth brackets (as defined in the first analysis). These probabilities are presented as marginal effects using the mean of the included control variables and show the difference in probability between the reference category (e.g. a specific occupational group) and another category (e.g. another occupational group) of having low-, medium-, or high-wage growth. To avoid an overtly complex model that needs to control for time effects and to facilitate a more accessible interpretation of results, the logistic regression is based on the representative single years for the two time periods, that is, 2018 and 2023, which will be the focus of analysis and comparison. The sample years of 2018 and 2023 have been compared with the other years for respective time period on these key variables and we find no results that indicate that 2018 or 2023 are outliers.
To control for the compositional effects in the key independent variables above, we must control for several other variables, including gender, age, country or region of birth (divided into Sweden, the EU-27/28 region, European countries not included in the EU, and non-European countries), level of education (i.e. primary to tertiary), union membership, where LO is the main trade union confederation for blue-collar workers, the Swedish Confederation of Professional Employees for professionals, and the Central Organization of Professional Employees for professionals with a university degree, working time (divided into under 25 h, 26–34 h, and 35 h or more per week), and employment type (i.e. permanent or temporary employment). Gender and union membership are notable because both were the focus of debates and actual implementations of relative wage changes. By including control variables, individual-level effects, that is, the relative distribution of people with distinct characteristics within occupations on these occupations’ wage growth, are controlled for.
Results
Descriptive analysis
Figure 1 provides an overview of the occupational wage growth for the 2014–2018 period (a) and for the 2020–2023 period (b) using occupations as described by the three-digit SSYK codes. Figure 1(a) shows the real wage growth (i.e. annual average change for the 2014–2018 period net of the consumer price index inflation rate) for the 121 occupations included in this analysis, ranked from the occupation with the highest real wage growth to the lowest and colour-coded according to the SSYK standard at the one-digit level (i.e. major occupational categories). The overall real wage change for all occupations is an annual increase of 1.18%. The average inflation for the same period is 1.20%, which makes the average nominal wage increase 2.38%. The wage norm for the specific period in the IA can be calculated as 2.24%.
As can be seen in Figure 1(a), there is a noticeable variety in the occupational levels regarding real wage growth during the 2014–2018 period. Compared with the mean level of 1.18%, some occupations show wage increases more than double the average level. The occupations with the highest wage increases are dominated by managers and those requiring advanced education. The specific occupations with the largest wage increases are retail and wholesale trade managers (3.24%), followed by administration and planning managers (3.16%), and financial and insurance managers (3.09%). 2
However, the 2% wage increase seems to be a break-off point. Above that level, managers, occupations requiring advanced education, and higher education to some extent, dominate. Below that level, a greater representation of service and manufacturing occupations can be seen. Towards the average, there is an assortment of occupations, with some elementary occupations present, but fewer service occupations. Closest to the average are building caretakers and related workers at 1.18%. Below the average, there is a great mixture of occupational groups, but with few occupations requiring higher education. At the bottom, some occupations display negative real wage growth, such as croupiers, debt collectors, and related workers (−1.59%), veterinarians (−0.97%), and cabin crew, guides, and related workers (−0.72%). The average annual inflation for the period is 1.2%, which means that occupations with nominal wage increases below this turned negative when deflated to real wage changes.
Figure 1(b) presents the corresponding overview for the 2020–2023 period, where the annual average nominal wage increase for the period is 3.03% and average inflation is 6.37%. The IA wage norm for the period is 2.48. Seven occupations are not included in the 2020–2023 WSS analysis due to missing data. There is a noticeable variety at the occupational level regarding real wage growth during this period. The average real wage growth for the 2020–2023 period is −3.3%, with all but one of the occupations showing negative development. Managers and occupations requiring advanced education are clearly present in both the highest (i.e. with the lowest amount of real wage reduction) and lowest third of the distribution, with a smaller share in the middle. At the top, we find managerial positions, such as real estate and head of administration managers (−1.53%) and managing directors and chief executives (−1.54%), and occupations requiring advanced education, such as medical doctors (−1.73%), accountants, financial analysts, and fund managers (−2.20%). At the bottom are legislators and senior officials (−6.20%), supply, logistics, and transport managers (−5.71%), authors, journalists, and linguists (−5.01%), and university and higher education teachers (−4.43%). Occupations requiring higher education, such as physical and engineering science technicians (−2.73%) are mostly found in the higher half of the distribution, while manufacturing occupations, such as electrical equipment installers and repairers (−3.49%) can be found mostly in the middle and lower third of the distribution.
A noticeable difference between the two periods is that a mixture of occupational groups comprises the occupations at the very top of the wage growth distribution in 2020–2023, whereas managerial occupations almost exclusively make up the top in 2014–2018, while managerial and higher education occupations have a greater presence at the bottom of the distribution in 2020–2023.
Wage growth, occupational wage levels, and occupational groups
As a key purpose of the IA is to set wage norms to avoid excessive wage dispersion, we analyse the structural patterns found in the different parts of the wage change distribution. The variation in wage growth has been divided into three brackets to capture low, average, and high real wage changes. The wage brackets for 2018 and 2023 are shown in Table 1. The specific cut-off points for the wage brackets are based on notable breaks in the 2014–2018 occupational wage growth structure. The share of respondents making up these brackets has then been maintained for the 2020–2023 period to the extent made possible by the data structure to allow for a meaningful comparison.
Wage growth brackets.
WSS: Wage Structure Statistics; LFS: Labour Force Survey.
The average annual wage change is based on yearly occupational wage change averages for the periods 2014–2018 and 2020–2023, respectively. Occupations with WSS wage data for each year for the respective time period are included.
Table 2 cross-tabulates the share of employees in the SSYK occupational categories with these three wage growth brackets and reveals several notable patterns in 2018. Managers, those with higher education, and mechanical manufacturing all have large shares in the high-wage growth bracket. While those with higher education have a small share in the low growth bracket, managers and mechanical manufacturing show a more polarised pattern with high shares in the low growth bracket. Service workers are close to the average, while all other groups have a large or very large share in the low growth bracket and a small or no share in the high growth bracket.
Real wage growth and occupational categories (share in percentages).
In 2023, these patterns changed somewhat. The low share increased for those with advanced education, while mechanical manufacturing is now mainly found in the medium growth bracket, which is also the case for building and manufacturing. Elementary occupations show a great increase in the high growth bracket. Administration and agriculture show a drastic swing towards the high growth bracket, but the latter group is small, making fluctuations more likely. The overall descriptive changes show that those higher in the SSYK hierarchy are still more likely to have a high-wage growth, or more adequate, not as poor a development as other lower groups, but the difference has decreased compared to 2018.
The occupational wage level (Table 3) expresses the nominal average monthly wage in SEK in the occupation years (i.e. 2018 and 2023). The average wage increase for all occupations between 2018 and 2023 was 5300 SEK and was added when calculating income groups in 2023 to facilitate comparative wage groups between the years.
Real wage growth and occupational wage level (share in percentage).
In 2018, a pattern can be seen where those earning 30,000–40,000 SEK have a small share in the low-wage growth bracket while those earning <30,000 SEK have a noticeably larger share. The 30,000–40,000 SEK group stands out by having a very high share in the high bracket, while it is almost zero in the lowest income group. Participants with the highest wages reported high levels in both the high- and low-wage growth brackets, which indicates a dual tendency within this group. The general trend for 2018 shows a pattern where those with lower wages also have lower-wage growth, compared with higher-wage groups. The patterns changed somewhat in 2023. Those earning ≥45,301 SEK have noticeably higher shares in the high growth bracket compared with 2018, similarly to those with the lowest wage. The 30,301–35,300 SEK group show a higher share in the middle growth bracket and a decrease in both the low and high growth brackets, while those earning 35,301–45,000 SEK have markedly poorer wage growth compared to 2018. The general change compared to 2018 is that those with high wages have even better wage growth, which is also the case for those with the lowest wages, although on a more modest level, while those in the middle wage group have fared worse.
Regression analysis
Figure 2 presents the marginal effects for occupational major groups in SSYK for 2018 and 2023, 3 which are derived from a multinomial regression including controls for industry and several variables for individual characteristics in the labour market (see Tables A9 and A5 (2018) and Tables A11 and A7 (2023) in the Appendix). The marginal effects are the statistical differences in the likelihood of occupational groups being in the low-, medium-, or high-wage growth categories compared with the reference group. The effects are expressed as probabilities with both possible positive values (i.e. higher probability) and negative values (i.e. lower probability) of belonging to a certain wage growth category. We present the probabilities as percentages (i.e. value × 100). Due to its small size and lack of wage growth variation in 2018, agriculture was removed from the analysis. Likewise due to the lack of wage growth variation in 2018, the service category was merged with administration. Besides SSYK, we also study wage groups and industry in later figures. The marginal effects are calculated when the controls are at their mean. For tables presenting the marginal effects, including all control variables, see Tables A9–A12 in the Appendix.
Marginal effects for probability of being in low-, medium-, or high-wage increase group for occupational groups in 2018 and 2023. Note: Marginal effects express higher (positive value) or lower (negative value) probability of being in a particular wage increase group compared to the reference category. The marginal effects are expressed in per cent (e.g. 0.2 = 20% higher probability of belonging to a wage increase group compared to reference category). Agriculture has been omitted. Service and administration have been merged into one category. Analysis includes controls for industry and individual characteristics (not shown in the figure).
The SSYK variable in Figure 2 for 2018 generally shows large marginal effect sizes, indicating that individual positions in the SSYK hierarchy have large effects on their predicted wage growth. Administration and service care (9%), building and manufacturing (27%), and elementary occupations (44%) all show large positive marginal effects in the low-wage growth category compared with the reference category of occupations requiring advanced education. Moreover, the same occupational groups have large negative probabilities for high-wage growth. Thus, most workers in occupations in the lower half of the occupation hierarchy had a higher risk of poor wage growth in 2018. The exception is mechanical manufacturing, a key occupational group for the IA wage norm, as a large share of those occupations is found in export industries. Workers in this occupational group have a higher probability to be found in the high-wage growth category compared with workers in occupations that require advanced education. The same pattern is also found for occupations requiring higher education and especially managers. In sum, those higher-up in the occupational hierarchy and workers with manufacturing occupations have a higher probability to work in occupations with better wage growth compared with workers lower in the hierarchy. This pattern emerges with all other variables held constant, that is, individual-level variables that could affect aggregate wage growth outcomes. Analysing occupational positions without controlling for industry (not shown) reveals that controlling for industry reinforces the differences between the occupational groups in SSYK, indicating that differences in occupational groups at the industry level contain larger variation and that patterns in occupational groups become more pronounced once we hold industry constant.
Turning to the results for 2023 most of the other occupational groups have a higher probability for medium-wage growth and lower probabilities of being in the low- and high-wage categories compared with the reference category (occupations requiring advanced education). Administration and service care workers, as well as building and manufacturing workers show a much lower chance for high-wage growth (−39% and −15%, respectively), while managers have a relatively high probability of being in the poor wage growth category. Mechanical manufacturing and transportation workers also showed a very high probability for the medium-wage increase category. The result indicates that there is a larger dispersion in wage growth for occupations requiring advanced education in the turbulent 2020–2023 period compared with 2014–2018. The greater differences (larger total marginal effect sizes), with mechanical manufacturing and transport in 2023 indicate greater wage dispersion from the IA benchmark, indicating that the effect of occupational position on wage dispersion has increased. The larger effect sizes compared with higher education and administration and services in 2023 (both large occupational groups), also indicate a greater occupational wage dispersion span in 2023.
Turning to industry in Figure 3 (controlling for SSYK and individual characteristics in the labour market, see Tables A9 and A5 (2018), and Tables A11 and A7 (2023) in the Appendix), manufacturing is now the reference category. Regarding the results in 2018, workers in several industries, such as financial and insurance activities and real estate, have a higher probability than the reference category of being in the medium-wage growth category, but a lower probability for low- and high-wage growth. This finding indicates that the growth for workers in the manufacturing industry is more dispersed, with higher shares in both the low- and high-wage growth categories, while wage growth is more centred to medium-wage increases in many other industries. However, workers in three industries stand out by having a higher probability than the reference to be found in the high-wage growth category. Health care and social work activities, and education were both the subject of notable wage increases outside the norm of the IA during the 2014–2018 period, which may explain many of these effects. Both made use of numberless agreements which might also have had an effect. The positive result of food services is less clear. Controlling for occupational position has the general effect of increasing the probability of industries being in the high-wage group compared with manufacturing, indicating that wage change variation due to occupational positions within the different industries can hide more distinct industry differences, unless controlled for.
Marginal effects for probability of being in low-, medium-, or high-wage increase group for industries in 2018 and 2023. Note: Marginal effects express higher (positive value) or lower (negative value) probability of being in a particular wage increase group compared to the reference category. The marginal effects are expressed in per cent (e.g. 0.2 = 20% higher probability of belonging to a wage increase group compared to reference category). Analysis include control for SSYK and individual characteristics (not shown in the figure). SSYK: Swedish Standard Classification of Occupations.
The results for 2023 show a stark trend, with almost all industries performing poorer than the manufacturing industry. Considering the function of the IA and the wage norm, it is noteworthy that manufacturing differs so much from all other industries. Part of the explanation can possibly be that unlike during the 2008 financial crisis, the export industries were only marginally affected by the COVID-19 pandemic. Education stands out with a very high risk for poor wage growth. Rising costs for municipalities from inflation combined with the fact that numberless agreements do not include a lower limit for wage growth are likely partial explanations. Workers in accommodation and food services have a very high probability for positive wage growth, especially as this industry experienced extensive layoffs due to the COVID-19 pandemic, particularly temporary or part-time workers. After the COVID-19 pandemic ended, this industry has struggled to fill vacancies, which forced employers to raise wages. Both these tendencies have affected wage growth (Swedish Agency for Growth Analysis, 2021). Comparing the results from 2018 with those from 2023 we do not see the same change in marginal effect sizes for industry as for occupational hierarchy. Manufacture seems to hold a more beneficial wage growth position in 2023 but there are no clear indications of an increasing wage dispersion span over time between industries compared to occupational hierarchy, which indicates that different mechanisms affect wage dispersion.
Figure 4 shows the effect of different occupational wage levels for the outcome in occupational wage growth (controlling for industry and individual characteristics, see Tables A10 and A6 (2018), and Tables A12 and A8 (2023) in the Appendix). Regarding the results in 2018, those with wages lower than the reference group (i.e. 30,001–40,000 SEK monthly) have a higher probability to be in the poor wage growth category. The 40,000–50,000 SEK group indicates a more similar distribution in wage increase probabilities compared to the reference group, although with a 16% lower probability to be found in the high increase category. The highest wage group reports an 8% higher probability for the low-wage growth category. In summary, those in the reference category have the most favourable wage growth, with a notably poorer wage growth among those with low wages. Entering the control variables separately (not shown) still showed the same wage growth for those with the lowest wages, indicating that wages in themselves have a very strong effect on wage growth for those with low wages compared to medium wages.
Marginal effects for probability of being in low-, medium-, or high-wage increase group for wage groups in 2018 and 2023. Note: Marginal effects express higher (positive value) or lower (negative value) probability of being in a particular wage increase group compared to the reference category. The marginal effects are expressed in per cent (e.g. 0.2 = 20% higher probability of belonging to a wage increase group compared to reference category). Analysis include control for industry and individual characteristics (not shown in the figure).
Turning to the results in 2023 the pattern is now inverted, with the two highest wage groups having the largest probability for positive wage growth. The second to lowest group performed somewhat worse than the reference group, while there were small and mostly insignificant improvements in the lowest wage group. A summary of the wage variable for both periods is that while the low-wage groups were the losers and the medium-wage group the winner in 2018 regarding wage growth, the high-wage groups are clearly the winners in 2023. The effect sizes are overall relatively stable over time, indicating that the wage dispersion span has not increased, even though the pattern is inverted.
Concluding discussion
This paper analyses how well the balanced growth model manages to uphold wage equality, using Sweden as a critical case. With Sweden being the ideal-typical example of a balanced growth model, which is believed to offer both economic growth and reasonable wage increases for all workers, the accuracy of the wage-growth norm can have far-reaching consequences for the model's legitimacy. The period covered by the analysis captures a move from low to high inflation, adding a particular challenge to the wage-setting model.
Analysing the development from the conventional industry perspective, the IA model appears to work as intended, even during this tumultuous period. The results indicate that the manufacturing sector has a generally higher probability of good wage growth in both 2018 and 2023 compared with other industries, where the few industries that outperformed manufacturing in 2018 can mostly be credited to explicit (and contested) exceptions to the IA model, such as the wage increases in the health care and educational sectors in the pre-pandemic period. We see no growing wage dispersion span over time, which would indicate a generally increase in wage inequality. While it is debatable whether the manufacturing industry having better wage growth than other industries is ideal, it seems to act as a benchmark that other industries do not move beyond.
However, when analysing the wage-setting norm through the occupational hierarchy and wage level the picture changes. We find large variations in wage increases between occupations. The occupational hierarchy in 2018 showed that those at the top, such as managers and occupations requiring higher education or more, had higher probabilities of better wage growth than those at the bottom. The exception being mechanical manufacturing, which is a group close to the centre of the IA wage norm who also perform well. Several of the groups in the lower half of the hierarchy were still performing worse than those in occupations requiring higher education or more in 2023. We also find a larger wage growth dispersion span in 2023 when comparing the reference occupational group of advanced education with both higher education and administration and services. As these are all large occupational groups in the labour market, this indicates that occupational hierarchy not only influences wage growth dispersion, where those higher up in the hierarchy benefit more, the effects of occupational hierarchy on wage growth dispersion seem to be increasing over time.
Wage in 2018 showed that those with the lowest wages also had the highest risk of low-wage increases, which is another potential source of accentuated wage inequality. In 2023 there was a noticeably higher probability of good wage improvements among those with the highest wages and a lower probability among those with lower wages compared with the medium-wage group, although it should also be mentioned that those with the lowest wages experiences relatively improved wage growth over time.
Overall, a pattern of wage growth emerges where those with high wages and high positions had better opportunities for good wage growth than those further down the occupational hierarchy and wage scale. The results indicate a disintegration of the governing function of the IA system, with wage growth following lines of privileged positions among the occupational groups and not only the industrial norm.
The results show that individual's structural position of occupational hierarchy and wage level have unique effects on wage growth, while the effects of control factors related to individual characteristics are rather limited. One noteworthy result is that men have a higher probability of better wage growth than women. The continued lower probability for women to experience wage growth illustrates the tension within the IA to address wage inequality, despite amendments to fight such disparities (Müller et al., 2018). Moreover, the mixed results for the lower-wage groups and the poor results for occupational groups at the lower end of the hierarchy do not offer clear-cut results of decreasing wage gaps. Our results paint a different picture compared to the generally stable results of wage ratios found by Karimi et al. (2024), which could indicate that new patterns emerge when taking individual structural position into account rather than aggregate wage group only.
How can the results be interpreted regarding the growth model theory and what do they imply for the future of the balanced growth model? Erixon and Pontusson (2022) found signs of a shift in the direction of a finance and consumption-led growth model due to strongly increased private debt, partly made possible by cheap money during the low-inflation years. The high-inflation period has made this debt more costly, which has put a greater stress on employees to raise their wages to maintain their standard of living. The Swedish wage-setting system function according to the principle of competitive corporatism (Bengtsson and Ryner, 2018; Thelen, 2001) with tight cross-industry coordination, which is in line with our empirical results, but also local wage setting flexibility. It seems likely that this local flexibility is utilised by occupational groups to make use of their individual privileged position to achieve greater wage increases under circumstances when wages play a more crucial role, such as a high-inflation period, while at the same time not causing growing wage dispersion among industries. There is previous research indicating that despite greater local flexibility in the individual and differentiated wage setting, the strong wage norm of the IA gravitates local settlements close to the norm (Ulfsdotter Eriksson et al., 2021). Our results, however, show that the outcome of settlements in real wage growth vary strongly in relation to occupational groups position and wage level, which is an indication that the wage norm of the IA is wavering concerning equality expectations and shifting towards a model of growing wage increase dispersion driven by private debt (cf. Erixon and Pontusson, 2022).
Our results indicate an imbalance in the balanced model, with a hollowing out of the wage equality component. The model arguably retains its efficacy regarding growth but causes growing inequalities along unintended structures. As our study shows, the inequality of outcomes follows a clear hierarchical gradient in the occupational structure. Thus, the present bargaining model mainly favours more privileged categories in the labour market, which skews away from the norm of the wage setting system. This is also in line with Baccaro and Howell (2017), who argued that Sweden has gradually begun to become neoliberal through what we see as a process of institutional drift (Streeck and Thelen, 2005). The institutional purpose of maintaining competitiveness through industry-based wage restraints is maintained, but the institutional goal of a wage norm is undermined through wage drift at the occupational and wage group levels.
The results point towards a challenge of maintaining the balanced growth model which benefits both the export and domestic sectors of the economy. Growing inequality is likely to undermine the support for the wage-setting model, especially if larger number of employees loose even more economically due to rising inflation despite wage restraint. Unions in particular risk losing credibility if they continue to praise the IA as both efficient and just economic policy, while their members see a system that favours the rich and privileged. A more nuanced and open debate between the social partners, in particular union and employer association representatives from both the export and domestic sectors as well as key state actors such as the SNMO, about the benefits and drawbacks of the IA could clarify what kind of economic equality can be pursued within the current framework as well as alternative wage models through which to pursue wage equality. It could also clarify to what extent the social partners are willing to deviate from the goal of wage equality to retain or achieve wage growth. Such a debate would help in our understanding of the stability and support of the balanced growth model and contribute to the growth model theory in general.
Addressing some of the limitations of the study, our analysis relies to a large extent on the change observed between two specific years. Future research based on more data points over time would strengthen the evidence of structural change. While Sweden is considered an ideal-typical example of balanced growth it is not unique and comparative research including more countries could further clarify the mechanisms and development of this growth model. More detailed analysis of the role and mechanisms involving occupational power resources would also contribute to the research field. Mixed or qualitative methods could potentially explain whether local agreements and negotiations is the main arena where occupational wage increase differences occur.
Supplemental Material
sj-docx-1-jir-10.1177_00221856251388523 - Supplemental material for Who benefits from the Industrial Agreement? Uncovering the trends and structures of wage inequality at play in the Swedish wage-setting model
Supplemental material, sj-docx-1-jir-10.1177_00221856251388523 for Who benefits from the Industrial Agreement? Uncovering the trends and structures of wage inequality at play in the Swedish wage-setting model by Patrik Vulkan, Johan Alfonsson and Tomas Berglund in Journal of Industrial Relations
Footnotes
ORCID iDs
Ethical considerations
Consent to participate
Funding
Declaration of conflicting interests
Data availability
Supplemental material
Biographical notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
