Abstract
Since the 2007–2008 financial crisis, it has been suggested that the middle class is undergoing proletarianisation, with stagnant wages and reduced work autonomy making their conditions more like those of the working class. However, there is no consensus on whether this shift has occurred or on how to define the two classes. This article proposes a class concept rooted in value domination, which compels individuals to the market for survival. While both classes depend on the market, the middle class is defined by their greater individual market power and a labour power with higher value, resulting in higher wages and better working conditions. Using this definition, the proletarianisation thesis is examined in Sweden. Findings indicate a widening wage gap, with middle class wages diverging from those of the working class, and no significant reduction in middle class autonomy relative to the working class. These results suggest that proletarianisation has not occurred, and class distinctions have, in fact, intensified.
Keywords
Introduction
In recent years, an increasing number of voices have argued that since at least the financial crisis of 2007–2008, the middle class has been the losers in today’s labour market. It has been suggested that the middle class is experiencing a proletarianisation, meaning their conditions have become more similar to those of the working class (Broström 2023; Ehrenreich & Ehrenreich 2013; Moody 2017; Vaughan-Whitehead et al. 2016). However, there is no consensus on what class actually is, or whether this development has truly occurred.
Some scholars have argued that the middle class has become as dispensable and replaceable as assembly line workers or those in warehouses and foundries (Ehrenreich & Ehrenreich 2013). Following this it has been suggested that the middle class has become underpaid, overworked, and that their income levels have decreased (Grabka et al. 2016; Moody 2017; Pressman 2007; Vaughan-Whitehead et al. 2016). It has also been argued that the middle class has lost its autonomy at work and has become as replaceable as the working class, and thus proletarianised (Broström 2023; Ehrenreich and Ehrenreich 2013; Moody 2017: 32; Pressman 2007; Vaughan-Whitehead et al. 2016).
However, a recent study by Moawad and Oesch (2024) argues that, using an occupational-based class definition, in terms of household income, there is no empirical evidence for middle class proletarianisation in the United States, United Kingdom, Germany, France, Spain and Poland; in fact, the working class was worse off. The results held true for Austria, Denmark, Ireland, the Netherlands and Switzerland as well. In another multinational study, Dallinger (2013), who defined the middle class as the middle three quintiles of the income distribution, showed that while the gap between middle and top disposable incomes grew significantly in the 19 countries studied, there was no evidence of a decline in income for the middle class. Thus, the results seem to point in different directions, highlighting the need for further research.
Regarding the decrease in autonomy and control for the middle class, the argument put forward is not based on empirical evidence but rather suggests that a shift towards lean production, New Public Management, and just-in-time production has undermined the middle class’s autonomy at work, making this class more similar to the working class (Broström 2023; Ehrenreich & Ehrenreich 2013; Moody 2017).
One issue with the existing research on the decline of the middle class is the lack of a clear theoretical definition of the difference between the working class and the middle class, and also the use of different operationalisations of class in different studies. In the broader empirical class literature, the tendency has often (but not always, see, for example, Goldthorpe 2000; Wright 1985 and ‘Previous definitions of the middle class and working class’ below) been to focus on finding ways of measuring class rather than trying to define what it actually is (see, for example, Erikson & Goldthorpe 1993; Evans 2000; Marshall et al. 2005 [1989]; Oesch 2006). There has been a strive not to ‘ontologically overloading’ the concept (Oesch 2006: 24–25) and class schemas has been seen as a work instrument that should be evaluated for its ability to capture empirical regularities (Erikson & Goldthorpe 1993: 46). This empiricist tendency in class studies seems to aim for a class measurement that captures the highest correlation with various outcome variables. However, if a class definition is not rooted in an ontological understanding of its nature, it becomes impossible to assess how accurately a class schema reflects its ontological reality. This article will therefore adopt an ontologically rooted definition of class, grounded in the value domination within the capitalist mode of production, and argue that, as individuals are forced to survive in the market under this mode of production, value and market power is the key to understanding class.
The focus will be on studying the Swedish case, as it offers a compelling opportunity to explore class development in terms of wages and autonomy. Sweden, as a representative of the Nordic labour market model, has strong unions, has seen real wage increases over the past 20 years that exceed the EU-15 and OECD (Organisation for Economic Co-operation and Development) averages, and operates a quasi-centralised wage-setting model praised for delivering substantial real wage increases across all occupations (Kjellberg 2023; OECD 2024; Swedish National Mediation Office 2024). Does this hold true, or has the middle class been worse off?
Against this background, the development of wages and autonomy at work among the middle and working classes in Sweden will be explored. Given the lack of consensus on class definitions and income trends, and the scarcity of empirical research on the changing autonomy of the middle and working classes at work, these issues require further attention.
The aim of this article is, first, to establish a definition of class grounded in the ontological principles of capitalism. Second, using this definition, the article aims to investigate whether the middle class and working class in Sweden have become more similar in terms of wages and autonomy at work.
The article will be structured as follows: first, a background of the Swedish case will be provided, along with a review of previous research on wage development and autonomy among the middle class and working class in Sweden. Following this, the definition of class used in previous studies will be outlined, and it will be argued that these approaches contain several issues. Then a class concept based on value domination in capitalism will be developed and operationalised, before doing the same for the outcome variables. After this, the empirical results will be presented, concluding with a discussion of what might explain the findings.
Background
Sweden has been regarded as an ideal type of the Nordic labour market model, characterised by a social democratic universal welfare state and strong labour market partners, which, through negotiations, set the rules of the game. The common view has been that the state plays a minor role in regulating the labour market, leaving it to the partners to regulate through collective agreements (Alfonsson et al. 2023; Kjellberg 2023). As a result, Sweden has no minimum wages.
Since the 1990s, Sweden has followed the international trend of austerity and neoliberalism (Altermark & Plesner 2025; Skyrman et al. 2023). The public sector has been downsized, and taxes have been lowered – on labour but most significantly on capital. This has resulted in Sweden experiencing the highest increase in inequality in the OECD. The Gini coefficient (including capital income) has risen from 0.23 in 1990 to 0.32 in 2021 (Swedish Fiscal Policy Council 2024). In addition, a greater proportion of people are now employed in the private sector, increasing from 62% to 70% (Alfonsson et al. 2022). Union density is high in Sweden but has decreased, from a peak of 85% in the early 1990s to 68% today. The most significant drop has been among the working class, where membership has fallen from 86% to 58% (Kjellberg 2025).
Looking at wage developments over the long period 1970–2020, real wage growth in Sweden has been below the average of most EU countries (Theodoropoulou 2019). However, this long trend masks differences over time related to different wage-setting models, which have had varying success in raising real wages. From the 1950s to the early 1980s, Sweden had a centralised wage-setting model that aimed to reduce wage inequality. The trade union slogan ‘equal pay for equal work’ was central to the solidaristic wage-setting model. Although wage growth stagnated at the end of this model, it produced high real wage increases, averaging 2.3% per year (Swedish National Mediation Office 2024).
However, the model collapsed in the early 1980s when employers and unions in the manufacturing industry abandoned the agreement in favour of a decentralised model. This shift opened the door to a more market-oriented wage-setting system, where conditions in each sector determined wage increases. On average, the decentralised period was characterised by low real wage increases and the average annual real wage increase was 0.93% (Swedish National Mediation Office 2024).
In the mid-1990s, the government pressured the social partners to develop a wage-setting model aligned with the European wage increase norm. This was seen as a way to enhance the competitiveness of Sweden’s export industry and curb inflation. In response to the government’s threat of intervention, the social partners agreed to a new wage-setting model called the Industrial Agreement. This quasi-centralised system allows the export industry to negotiate wages first, with other sectors then following suit, using the export industry’s agreements as a benchmark. While there is room for individual variations at the workplace level, the overall wage increases gravitate around the industry benchmark. Although this model is considered a success, average wage growth for the period 1998–2023, has been just 1.3%. The period from 1998 to 2009 saw higher wage increases, averaging 2.4%, while from 2010 to 2023, this figure dropped to just 0.4%. While Sweden’s real wage growth since 2000 has been modest by historical standards, it has performed better than the EU-15 and OECD average (1.55% real wage increase per year compared with 0.7% for the OECD average and 0.55% for EU-15) (OECD Employment Outlook 2024; Swedish National Mediation Office 2024). Given the wage-setting model, where all sectors follow the wage norm, and strong unions, one would expect the middle class and working class to experience similar wage increases.
However, in the previous research it has been argued that the middle class is the losers in Sweden. Based on an income-based class definition, where the middle class is seen as those living in household with an income between 60% and 200% of the median disposable income, Anxo (2016) has argued that the middle class in Sweden has decreased and that part of the middle class has seen a downwards mobility and fallen below the 60% threshold, and another, smaller part of the middle class has moved upwards and gained incomes above the 200% threshold. Thus, it is argued that the Swedish middle-class incomes have seen a polarisation where the incomes have decreased for one part of it, and for a smaller part there has been an increase. Using a slightly different definition of the middle class, including those ranging from 75% to 125% of the median income, Pressman (2007) claims that Sweden saw the biggest middle class decrease, and thus proletarianisation, among the 11 surveyed countries between the years 1980–2000.
Broström (2023), defining the middle class after how much control an occupation has in work, has argued that parts of the middle class have lost control at work and thus seen a proletarianisation. She claims that ‘some part of the middleclass has gained real wage increase’ and that other parts of the middle class has seen wage cuts or at least no wage increases (Broström 2023: 141, 184). Further she argues that there are middle class professions that have lost so much autonomy that they have become working class (Broström 2023: 144). Like many other countries, Sweden has seen the rise of New Public Management, and lean-production models have been implemented in many sectors (Alfonsson 2020: 128, 141). But has this led to the claimed decline in the autonomy of the middle class to the extent that it is now on a par with the working class?
In summary, while some previous research suggests that parts of the middle class have experienced declining wages and reduced autonomy at work, and thus have undergone proletarianisation, a key issue with these studies is their inconsistent definitions of class, with each study using different measures. This highlights the importance of developing a coherent approach to class measurement.
Previous definitions of the middle class and working class
The definition of class is a contested area, with no clear-cut agreed definition. Even within the body of research focusing on the supposed proletarianisation of the middle class, a variety of class definitions are employed. Two major approaches can be distinguished in the literature: one defines class based on income, while the other focuses on occupations and the skills and/or autonomy linked to them.
The first approach, in which class is viewed as an expression of income, might be easier to study due to the availability of data; however, when it comes to distinguishing between different class groups, the definition becomes less clear. The first issue is whether to consider wages alone or all forms of income. Another question is whether to measure income at the individual or household level. And there is no definitive argument on where to draw the boundaries between class groups. In the literature that uses an income-based definition, the most common approach to defining the middle class is by median household income, although the demarcation line varies. For instance, some studies define the middle class as those between the third and eighth income deciles (Vaughan-Whitehead et al. 2016), others as those in the middle three income quintiles (Dallinger 2013), or within 60%–200% of the median income (Anxo 2016; Vaughan-Whitehead et al. 2016) or 7%–150% of the median (Grabka & Frick 2008; Pressman 2007). The working class is typically defined as those below these thresholds, while the upper class is positioned above them. Where these boundaries are drawn can significantly affect findings on class development over time (Atkinson and Brandolini 2011, 2013; Foster & Wolfson 2010). This income-based definition could be described as a content-based definition, where a specific characteristic – income in this case – is used to define class, representing a non-relational understanding of class (Hörnqvist 2016).
As Moawad and Oesch (2024) have emphasised, in the income-based class definition nearly everyone with a wage income is defined as middle class, effectively excluding the working class from the analysis. The 60% threshold aligns with the EU’s relative poverty line and thus labelling everyone in the working class as poor. While it is true that an income gap often exists between the working class and the middle class, class is not equal to income. Income reflects a deeper structure in which some employees are granted better wages than others. To understand what class is, the underlying structural factors that create income differences between groups must be examined.
The other approach offers a better understanding for class as the outcome of a relation on the labour market. In this framework, exemplified by scholars such as Erikson and Goldthorpe (1993), Wright (1985, 2000), and Oesch (2006, 2013), class is primarily seen as a hierarchical classification of occupations. The hierarchical scale is based on factors such as the skills required for a specific occupation, the authority individuals have over the work process and others, and the autonomy they have in their own work.
In Goldthorpe’s approach, the key distinction within the group of employees is based on their relationship with their employer. Class here is defined by social positions that are differentiated by the employment relationship they hold (Goldthorpe 2000). Central to the definition is the division between employees with what he terms a ‘labour contract’ and those with a ‘service relationship’. Those with a labour contract provide work under the supervision of employers or managers and are compensated for the time spent working (Erikson & Goldthorpe 1993: 41–42; Goldthorpe 1997: 42). This type of contract typically applies to jobs that are easily supervised and require minimal skill (Goldthorpe 2000). By contrast, the service relationship prevails for employees whose work requires specialised skills, is more difficult to monitor and whose departure would result in a loss of value to the employer (Goldthorpe 2000). This relationship is characterised by employees offering a service to the organisation and receiving not only financial compensation but also rewards for their loyalty. This includes better salaries, career pathways, and greater autonomy at work (Goldthorpe 1997: 42). From this division, the analysis has historically focused on blue-collar industrial workers as the working class and white-collar service sector workers as the middle class.
Goldthorpe’s class definition comes from an inductive approach and the theoretical argument is made post festum. The definition is based on empirical regularities, and he developed the theoretical argument for why different occupations are regulated by different employment relations at a later stage (see Goldthorpe 2000). His theoretical argument is based on rational choice/action theory, with a focus on the subjective rationality of employers to use different contracts for different types of occupations to meet their efficiency needs. Class is then a consequence of the aggregate rational actions of employers. But he does not develop why the different contracts should be seen as the basis for a class division. Rather, this seems to be an empirical fact, as shown by the correlation between the employment contract and different empirical outcomes (see Goldthorpe 2000). Goldthorpe’s approach is characterised by a methodological individualism which, as will be shown, differs substantially from the Marxist perspective which grounds class in the structural logic of the mode of production.
An empirical critique of Goldthorpe has been developed by Oesch (2006: 27ff), who argues that the division between labour and service contracts mirrors the labour market of the golden age of Fordism in the 1960s and 1970s. When the labour market shifted, with industry declining and the service sector expanding, this distinction became obsolete. Today, many service jobs share similar working conditions with manual labour, making the hierarchical distinctions between them less relevant (Oesch 2006: 49).
Building on Goldthorpe and Erikson’s approach, Oesch developed a class model that incorporates both vertical and horizontal dimensions. The aim is to move away from a strictly hierarchical class definition and better capture the complexities of class in the post-Fordist era. In the vertical dimension, the focus is on occupations and the skills they require, with those demanding higher skills placed higher on the class scale. The more important these skills are for employers and productivity, the greater efforts employers will make to retain these employees, leading to better working conditions (Oesch 2006: 64–65).
The horizontal axis accounts for differences in work logics. His argument is that people in similar labour market situations and work contexts share common experiences that can be understood as class positions (Moawad & Oesch 2024; Oesch 2006). Those who share similar work conditions are likely to have similarities in income, working conditions, and training requirements, and thus, will face similar challenges and focus on similar political issues in society (Moawad & Oesch 2024; Oesch & Rennwald 2018).
Oesch’s justification for incorporating work logics into his class definitions is grounded purely in an empirical approach. He, like Goldthorpe and Erikson, strives to avoid overloading the class schemas with theoretical concepts, instead valuing them for the empirical insights they offer (Oesch 2006: 20). The goal is to avoid ‘ontologically overloading’ the class concept (Oesch 2006: 24–25).
However, by sidelining the ontological foundations and focusing exclusively on empirical outcomes, this approach risks drifting towards a content-based definition of class, similar to income-based definitions. By focusing on empirically observable features rather than the ontologically deep structure that contains the generative mechanism that shapes empirical outcomes, class becomes more about providing an effective tool for capturing variation in outcomes rather than addressing the underlying structures responsible for these empirical realities. For a deeper understanding of class, we need not only to recognise that skills influence wages and working conditions, but also to understand why this is the case.
Although Oesch demonstrates that work logics can influence empirical outcomes, we must understand why these logics should be considered part of a class definition. To argue for or against their inclusion, an ontological reference point is needed. Without such a foundation, class could be defined by almost anything – from income to work logics – making it impossible to argue why one class schema is superior to another, aside from the strength of their empirical correlation measurements.
Therefore, to clearly define class, we need to move beyond empirical expressions and ground the definition in an ontological framework. This requires us to step back and understand the fundamental role of class in the capitalist mode of production.
A neo-Marxist perspective on the division of the middle class and working class
From a Marxist perspective classes are social relations of domination and exploitation (Wright 2009) that exist within the capitalist mode of production, which is, following Marx (1976 [1867]), characterised by the dominance of value production. Value is defined by the equalisation of qualitatively different types of labour into a uniform abstract form – human labour as such (p. 131ff; see also Arthur 2004; Postone 1993: 124ff). For the value domination to be possible, the worker must be, as Marx [1976 [1867]: 271–272) put it, ‘free in the double sense’ – free to sell their labour power to whomever they choose, and ‘free’ from any other way of making a living besides selling their labour power. As a result of this double freedom, workers have no means of survival other than to sell their labour power on the market. Workers are therefore subsumed to market forces. For capitalists it is their social position as owners of the means of production that is at stake. If they fail in competition with other capitalists, they may go out of business, lose their capital, and thus need to become sellers of labour to survive. For capitalists to continue existing, they must create surplus value and use this surplus to renew and expand their production. Only by doing so can they survive competition with other capitalists and remain capitalists. Surplus value comes from the exploitation of workers, whose labour creates more value than is needed to reproduce their labour power. Thus, while both classes are subsumed to the value logic, workers are exploited, and capitalists are the exploiters.
The distinction between workers and capitalists is relatively straightforward, following the division between employers and employees. However, differences within the group of employees and how to divide them into distinct class groups are not as clear-cut. In the Marxist tradition, there has been much debate about how to differentiate between groups of wage earners (Butler & Savage 2013; Moody 2017; Toscano & Woodcock 2015; Wright 1980, 1985, 2005). Wright’s perspective is particularly influential in this regard. He argues that the division between the middle class and the working class should be understood through two key factors: authority over the work process and expertise (Wright 2000, 2005). The first factor refers to managers, who are delegated responsibility by owners to ensure that other employees produce as efficiently as possible. These individuals, although employed, work in the interests of the employers and dominate other employees. Because of their central role, they can earn a higher wage – which Wright (2000: 16–17) calls loyalty rent – than they otherwise would. Managers comprise only a small part of the middle class, about 10% according to Wright’s (2000: 45) figures. The larger part of the middle class occupies this position due to their skills or expertise. Their working conditions and wages can rise above the average because of the scarcity of their skills or systematic barriers to increasing the supply of such skills, for example by the need for diplomas and specialised education. This leads to what Wright calls a ‘skill rent’, which allows those with higher skills to secure better wages and working conditions. According to Wright (2000), skills and expertise are ‘assets embodied in the labour power of people, which enhance their power in the labour markets and labour processes’ (p. 19).
Although Wright’s approach clarifies how capitalist market relations shape differences in occupational outcomes based on market power, and although his starting point is a Marxist approach, he has left Marx’s theory of value aside (see, for example, Wright 1999, 2010: 101). This suggests that class is merely an outcome of market power relations and market-based stratification. He thus separates class from the underlying structure of value production in capitalism. However, if we define capitalism as a mode of production in which value dominates social relations in general and relations of production and exchange in particular, then value cannot be ignored in understanding class. From this perspective, value is a social relation rooted in class differences based on differential relations of ownership and exchange of commodities. The capitalist class owns the means of production, where commodities – and thus value – are created. Being ‘double free’, workers must sell the only commodity they possess: their labour power. This marks a qualitative difference between the two groups. However, the differences within the category of employees are not qualitative but quantitative, based not only on market power but also on the amount of value embodied in the labour power being sold. My argument is that employees’ wages, working conditions, and other outcomes are influenced by both the value of their labour power and their market power.
Following Marx (1976 [1867]: 325ff), the value embodied in labour power is determined by the socially necessary labour time required to produce and reproduce that specific labour power. The more socially necessary working time is required to produce and reproduce a specific labour power, the more potential value it has. This means that labour power that requires extensive education or training holds more value than that which does not (this does not mean they are more valuable to society or regarded as such by others).
Given that the labour market works essentially like any other market, the price of labour power is affected by supply and demand. Both the middle class and the working class are dependent on the market, but what distinguishes them is that the middle class, in addition to their labour power having more value, also has a better market position than the working class. The middle class, by possessing labour power with high value that is in demand, can secure better wages and working conditions than the working class. They achieve this through their individual capital, which is transformed in the market into individual power resources. The working class, on the other hand, has a lower degree of individual power resources and must instead rely on collective power resources to improve their position. Union power, political power, and institutional power are central here (see, for example, Korpi 1978). The separation between the middle class and the working class is, therefore, a question of degree of value and of market power, where the middle class has a higher degree of market power, while the working class has less.
From this definition, in which the middle and working classes are distinguished by quantitative differences in the amount of value and market power, empirical indicators such as high income, authority, and control over one’s work are more often associated with the middle class. However, it is not these characteristics that define the middle class, but rather the higher value of their labour power and their privileged position in the market. Class expressions may, within capitalism, vary over time, but the fundamental definition of class, based on value and labour market power, remains constant. According to this definition, proletarianisation is a process in which the value or market power of the middle class is reduced to the same level as that of the working class.
To sum up, class is defined as relations within the capitalist mode of production, which is dominated by value production. The difference between the middle class and the working class results from differing access to individual power resources, where the former possesses a labour power with higher value that is in greater demand in the labour market, granting them better working conditions than the latter.
Operationalising class from a Marxist perspective
From this definition of class, the question that begs an answer is how to capture class empirically. Following Marx, there is no direct translation from the ontological real world to the empirical (if it would be, science would be obsolete, as Marx (1981 [1894]: 956) has stated). Instead, we need to find the best way to capture the real object – in this case, the value of specific labour powers and labour market power tied to specific labour market positions. This essentially comes down to what can be empirically measured, as well as the availability of data.
It was argued that the division between the middle class and working class follows the value of labour power and labour market power, where those with labour power that has high value and is in high demand and low supply secure better positions. Supply and demand can fluctuate, meaning labour power that at one point has high value could move from a favourable labour market position to a less favourable one if demand falls or supply increases beyond demand.
To capture both the value of labour power and the demand for it, the most effective approach is to examine the skill levels required for specific occupations. Although not a perfect measure, the skill level required for an occupation reflects the average necessary social time needed to produce the labour power and thus the value of labour. Occupations with higher skill requirements encounter a more closed market, which gives them more market power and therefore greater potential for improvement in their position. Thus, occupational skill requirements could be used to capture both sides of the class definition.
Why then not use education to capture class? This would be easier to measure. The reason for this is that values only exist when they are realised on the market. Following Marx, when a commodity is produced, it holds no direct value, it holds potential value until it is sold. If it is not sold the potential value is not realised and thus is lost (see, for example, Marx 1973 [1857–1858]: 91ff; and which has been developed by Sohn-Rethel 1978). A sandwich in a café that is prepared in the morning and waited to be sold holds a potential value. If it is sold the value is realised. If it at the end of the day did not sell and has to be thrown away the potential value is lost. The same goes for skills and education. A person can have a double PhD but if the labour power cannot be sold on the market, the potential value is not realised and thus lost. Thus, if we want to identify those in more privileged positions, we must look for individuals who have successfully sold their labour power, which can be assessed by examining the skill and education levels required for occupations.
Measuring class using education and skill level linked to occupations may initially seem like a straightforward method, but this approach faces three challenges that need to be addressed. The first problem is where to draw the line between different occupations. What skill level marks the tipping point at which an occupation stops being classified as working class and instead becomes classified as middle class? The second problem is differences in skills needed in different spatial contexts. A medical doctor will need more or less the same education no matter what country they live in, but there are some occupations that can differ. Third, the occupational structure changes and an occupation that one day needed special skill might the other day not need it. For example, this is what Marx (1976 [1867]: chap 14) and Braverman (1998) emphasise when analysing the rise of industrial (monopoly) capitalism. Former craftsman occupations were deskilled through standardisation, mechanisation and automation.
The first problem is difficult to completely overcome. It is an inherent challenge in the effort to bridge the gap between the empirical world and the real world, as the measurements used will never fully capture the real phenomenon. A straightforward class measure that can be used to capture the definition of class as an expression of value and power in the labour market is a simplified and somewhat adjusted version of Oesch’s full class schema, 1 in which the horizontal axis related to work logic is excluded. In Oesch’s vertical class definition, there, first, is a division between large employers and employees/small employers. Second, among the group of employees and small employers, he makes a division between what he calls small business owners, a higher-grade service class, lower-grade service class, skilled workers and non-skilled workers. Following the theoretical argument in the previous section, small business owners are part of the employer group and are therefore excluded from the employee class groups in my analysis. The division among employees follows the division into four different skill levels, referred to here as the upper-middle class, the middle class, the skilled working class and the unskilled working class. The two highest levels are considered the middle class, while the two lower levels are classified as the working class. The upper-middle class includes managers and professionals, representing occupations with the highest skill levels (e.g. managers, engineers, medical doctors, dentists, pharmacists and secondary school teachers). The middle class-group includes those occupations that hold the second highest skill level, this corresponds to occupations that would need some sort of university education (for example, nurses, primary school teachers, librarians). Supervisors and associate manager occupations that do not need a university degree are also included here. The skilled working class is defined as occupations that need upper-secondary education, or some sort of specialised skills, for example hairdressers, painters, electricians. The unskilled working class includes occupations that do not require any education or specific skills, for example garbage collectors, cleaners and industrial assembly workers (see, for example, Moawad & Oesch 2024).
The second issue requires an understanding of the specific class structure of the country being studied. In Sweden, there are some differences compared with Oesch’s class schema regarding certain occupations. For example, police officers in Sweden require a university degree and are thus placed in the middle class-group, whereas in other countries they are not. To address this, Oesch’s definition is not applied directly. Oesch uses the international occupation coding ISCO-08 (and earlier ISCO-88) developed by the ILO. However, the ISCO classifications are explicitly defined as a framework that, due to variations in occupational structures across countries, must be adapted to the specific context. In Sweden, SSYK-12 is the country-specific version of ISCO-08. Because of this, while Oesch’s definition is used as a foundation for classifying occupations based on skill and educational requirements, the classification has been adapted to better capture the Swedish context.
The third issue relates to changes in occupational structures and the skills required for various jobs over time. For instance, 30 years ago, police officers in Sweden did not need a university degree, and many new occupations have emerged in the IT sector over the past 20 years. This means that occupational definitions must be continually updated, as older classifications may become obsolete. To address this, four different occupational measures are used, each covering distinct time periods and capturing the specific occupational structure of its respective era.
The occupational measure considered the most robust and most aligned with the definition is SSYK-12, the Swedish adaptation of ISCO-08, for which there are data from 2015 to 2022. ISCO-08 is used for the period 2011–2022, following Oesch’s coding of ISCO-08 with some modifications to better fit the Swedish context (for example, the police officers). For the period 2005–2015, SSYK-96 is used, the Swedish adaptation of ISCO-88. As with SSYK-12, Oesch’s classification has been used but slightly adapted to better fit the Swedish context. SSYK-96 is deemed as the least precise measurement. This is because certain differences cannot be reconciled within the data. For example, supervisors are classified differently: in SSYK-12, they are placed in their own category, while in SSYK-96, they are more often (but not always) grouped with non-supervisors in the same occupational category. As such, they do not measure exactly the same thing. The measures are based on the most precise level, the so-called 4-digit level, which means that between 350 and 430 occupational groups are included in each measure, which are then subdivided according to the classification described.
To make longitudinal measurements more robust, and to ensure that observed changes are not simply the result of variable redefinitions, a fourth classification is included: the Socio-Economic Index (SEI). This is based on the Swedish occupation classification SNI82, which is somewhat outdated, but its classification of occupations aligns well with the definition used here. 2 The advantage of SEI is that data are available for the period between 1997 and 2022, for which there are data for wages (2005–2022) and autonomy at work (1997–2021). The different measurements have a Pearson correlation with SSYK-12 of between 0.89 and 0.95.
Operationalisation of the outcome variables
In order to examine whether the gap between the working class and the middle class has narrowed, as suggested by the literature on the proletarianisation of the middle class, it is necessary to operationalise not only class but also income and autonomy at work. The idea here is that if the gap between the middle class and the working class has narrowed in terms of these outcome variables, this indicates that the value and/or market power of the middle class has declined, and thus proletarianised.
Starting with income, previous studies have often used household disposable income to analyse class and income (Dallinger 2013; Foster & Wolfson 2010; Moawad & Oesch 2024; Pressman 2007). However, disposable income includes various factors that are not directly related to labour market power or value of the labour power, such as tax levels, social benefits and capital income. These elements can be influenced by political decisions that can benefit the working class, the middle class or the capitalists to varying degrees, but since the focus here is on changing class dynamics in the labour market, the emphasis must be on individuali wages rather than disposable household income.
To investigate changes in wages among class groups, two sources of data are used: labour force survey data combined with register data. The latter provides actual wages for each individual, while the former gives information about occupation and whether they are employees or employers. Data are available for the years 2005–2022, which captures both the period before the financial crisis of 2008, when wages developed quite well, as well as the COVID-19 pandemic and the start of the inflation crisis. About 150,000 answers are included in the data each year. The same individual can be present up to 4 times a year so the number of answers will be higher than the individuals included. Weights are applied to ensure that the results are representative of the entire population. Everyone who is employed and earning a wage are included. By using actual wages rather than applying regulations (such as a minimum number of hours worked), the aim is to capture the true working conditions of each group. For instance, the working class is more likely to be in temporary employment compared with the middle class (Alfonsson 2022), which can affect their wages, but this is a crucial aspect of being part of the working class, and reflects their actual earnings. However, to minimise the impact of extreme values, median wages will be used instead of the mean. In studying wage increase among classes all four class schemas (SSYK-96, SSYK-12, ISCO-08 and SEI) are used.
To explore autonomy at work and potential changes of this, data from the Swedish Work Environment Survey are used and the focus is on three answers to questions that capture different aspects of autonomy at work. These answers are: Can mostly/never partly decide when different tasks should be done; Is mostly/never involved in deciding the organisation of my own work; and Can at most decide my work pace for half of the working time. These answers to the questions capture control over when, how and at what pace tasks are completed, thus collectively measuring workplace autonomy. Depending on the year, between 3300 and 10,500 people, a representative sample of the working population, answer the survey every 2 years, starting in 1997 and going to 2021.
Since data on class are limited in the Work Environment Survey, autonomy will be examined in relation to three class schemas: SSYK-96, SSYK-12 and SEI (ISCO-08 is not available). The SSYK-96 and SSYK-12 classifications, which at their highest four-digit level of detail consist of 355 and 429 occupations respectively, but due to the limited number of responses for many occupations, data are used at the two-digit level, where there are 27 and 46 groups of occupations respectively. This sacrifices some precision, a limitation that the SEI does not face. To ensure sufficient representation within each class group and to prevent a small number of occupations from disproportionately influencing the results in years with fewer responses, the classes have been grouped into the broader categories working class and middle class. This reduces precision but provides more robust results. Another limitation is that in the SSYK-96 and SSYK-12 data in the work environment survey, it is not possible to exclude employers from the analysis, whereas this is possible with the SEI data. Nevertheless, using three different measurements allows for comparison, and if all three show similar patterns, this will be interpreted as a robust result, despite the limitations.
Results
Having defined class and operationalised this definition and the outcome variables, the second purpose will now be explored. Is there empirical evidence that the middle and working classes have become more similar? As argued in the introduction, there are two main arguments for the deterioration of the middle class: declining autonomy and wage stagnation. Is there any evidence to support this notion of middle class proletarianisation?
Wage development in the middle and working classes
Regarding wages, the question is whether wage growth has been worse for the middle class than for the working class, which would indicate that proletarianisation has occurred. Figure 1 shows the development of real wages for the upper middle class, the middle class, the skilled working class and the unskilled working class, using the four different class measures as operationalised and described above. An average of all four classes measured is also shown (dashed line).
Real wage increase in four different classes.
Unsurprisingly, the salary levels in Swedish kronor reflect the class hierarchy, with the upper middle class earning the most and unskilled workers the least. All groups have seen real wage increases over the period, but they also experienced significant losses during the inflation crisis of 2022. The different class measurements align fairly well with each other. Following the trend lines, it is clear that the higher up the class hierarchy, the steeper the increase in wages has been.
To better understand the increase in each class group, the yearly average real wage increases are examined in Table 1. Four different periods are analysed, which are not theoretically motivated but instead align with the data periods of the various class measurements.
Real wage increase among the four different classes across four periods.
The figures are based on median yearly wages, adjusted for inflation using 2023 prices. The time periods are chosen to follow the data available for each measurement and where there are overlapping time periods for different measurements. This is done to prevent data breaks from causing spurious changes.
Over the total period between 2005 and 2022, where SEI is the only available measurement, the middle-class groups have experienced much better wage development than the working-class groups, the middle-class have seen more than double the wage increase than the unskilled working class. Looking at the average development for all four class measures, there are only 3 years (2007, 2008 and 2020) when the unskilled working class experienced higher wage increases than the upper middle class. However, there are some differences between the different periods.
In the period between 2005 and 2015, SEI and SSYK-96 offer slightly different pictures. Both show the middle class as having the best wage growth, but SSYK-96 shows the worst development in the upper-middle class, while SEI shows the worst in the skilled working class. The main difference between SSYK-96 and SEI is found in the earlier period, from 2005 to 2011, where SSYK-96 shows a higher wage increase for both working-class groups compared with SEI (figures not presented). During the 2011–2015 period, where there are three measurements (SSYK-96, SEI and ISCO-08), all show that the working-class had the lowest wage increase, while the middle class saw the best wage growth (figures not presented). Given that SEI aligns with the more robust measurements (SSYK-12 and ISCO-08) in other periods, this suggests SEI is a more reliable indicator, and that SSYK-96 may be less accurate. Nevertheless, the period saw good wage development across all groups.
For the period between 2011 and 2022, using both SEI and ISCO-08, the unskilled working class had the lowest wage increase, followed by the skilled working class. The middle class saw the largest increase – about double that of the skilled working class and eight times that of the unskilled working class.
In the overlapping period between 2015 and 2022, with data from three measurements, we see that the unskilled working class experienced a real wage decrease of between –2.9% and –4.85%, the skilled working class saw a small decrease to zero growth, the upper middle class saw an increase of between 1.5% and 2%, while the middle class had the largest increase, around 4.3%. The working-class outcome from the inflation crisis was significantly worse than that of the middle class. For example, in 2022, the two middle-class groups had lost 5 years of real wage growth, the skilled working class lost 6–8 years (depending on the measurement), and the unskilled working class lost 9–10 years. Overall, there is no support for the idea that middle-class wages have become more similar to those of the working class. On the contrary, the wage gap has widened.
Class and autonomy at work
The other argument in the proletarianisation thesis is that, in Sweden and elsewhere, the middle class has lost autonomy at work, making its working conditions more similar to those of the working class (Broström 2023; Ehrenreich & Ehrenreich 2013; Moody 2017). To investigate this, the official Swedish work environment survey is used and the focus is on the three questions mentioned earlier. Since the argument suggests that the middle class autonomy has become similar to those of the working class, the relative difference between the two is of interest. Therefore, Figures 2 to 4 will show the relative difference in percentage points between the classes. A higher percentage point indicates that a greater share of individuals within a class schema have reported low control at work. This means that the numbers presented reflect how much larger proportion of the working class experiences low control compared with the middle class.
Regarding the first question about the ability to control when different work tasks should be carried out, the trends for the different class measures follow each other closely, as shown in Figure 2.
Percentage point difference between the middle class and working class in response to the question, ‘Can mostly not/never partially decide when different tasks should be carried out’.
In 1997, both SEI and SSYK-96 started at the same level, decreased slightly, rose in 2001, and then dropped again in 2003. During the final years of the first period, the measurements continue to mirror each other. Over the course of the measured period, there is no significant reduction in the gap between the middle class and the working class. On average, the working class scores around 20–25 percentage points higher than the middle class throughout the entire period. No significant change in trend is observed, although the last year of measurement shows the largest difference between the groups.
Regarding the question of the ability to decide on the organisation of their own work, as can be seen in Figure 3, the two class measurements during the first period align quite well and generally move in the same direction.
Percentage point difference between the middle class and working class in response to the question, ‘Are mostly not/never involved in deciding the organisation of their own work’.
No specific upwards or downwards trend is evident, with the greatest variation occurring in the final 2 years of measurement. Throughout the entire period, the working class scores on average about 21–29 percentage points higher than the middle class on this question.
The third question, concerning control over the pace of work, is examined in Figure 4. The trend differs from the others in two ways: the gap between the working class and the middle class is not as large, and the measurements vary significantly across different years and between the different class measurements.
Percentage point difference between the middle class and working class in response to the question, ‘Can decide the work pace for at most half of the working time’.
We can observe that the trends for SEI and SSYK-96 seem to somewhat align, although the latter scores lower in terms of the difference between the classes. In the final period using SSYK-12, there is a dramatic increase from a small to a larger difference. Due to this variability, it is difficult to determine whether there is an upwards, downwards, or stable trend over the years. However, the results suggest that the middle class generally has a slightly higher degree of control over their work pace compared with the working class. Beyond that, given the large variations, the challenges in comparing different metrics, and the relatively small differences in some years, it is difficult to identify a clear trend in Figure 4.
To sum up the results, we can see that while there is no substantial difference in control over work pace, the middle class has significantly greater autonomy and control over when and how their work is done compared with the working class. Overall, no trend suggests that the middle class has lost more autonomy than the working class. Therefore, the class difference in this regard remains large and intact.
Discussion
Given the results, we can conclude that, opposite to what is stated in previous research regarding the proletarianisation of the middle class, it is the situation for the working class in Sweden that has deteriorated over the last few decades. In terms of wage, the gap started big and ended even bigger. The working class experienced the worst wage performance, with low to negative real wage growth between 2015 and 2022, a trend not seen in the middle class. The working class also lost many more years of real wage growth during the inflation crises of 2022. The autonomy gap is overall very large and has not been reduced. As for proletarianisation, it is the working class that has fallen deeper into its proletarian condition, while the middle class has moved somewhat in the opposite direction. The gap between the middle class and the working class has widened, not reduced. Why is this the case? In a country like Sweden, with a social democratic tradition, and strong unions, the aim has been to reduce or at least maintain the gap between these groups. The wage-setting model has been praised for delivering good wage increases for all, but as has been shown here, it has not.
To understand this, we need to return to the ontological definition of class. It was previously argued that class must be understood within the capitalist context value domination and of market power. The middle class is defined as a group with labour power that holds more value and that is in demand in the market, while the working class has labour power with lower value, which is less in demand. The middle class possesses individual power resources, which they can leverage to secure higher wages and better working conditions. The middle class can use their individual power resources to avoid precarity and secure better wages, while the working class cannot. In contrast, the working class is instead dependent on collective power resources. As a result, the working class is more vulnerable to institutional changes than the middle class.
During the studied period, union power among the working class has been reduced. 86% of the working class was unionised in the early 1990s, which had dropped to 58% by 2023. The decrease was much smaller in the middle class group, from 82% to 73% (Kjellberg 2025). At the same time, several other changes have affected the institutional power resources that could have protected the working class. For example, it has become harder to receive unemployment benefits, and benefit levels have been reduced, putting pressure on the unemployed to accept work with poorer conditions. The Employment Protection Act, which governs matters such as the use of temporary contracts and the order of layoffs during redundancies, has been liberalised, increasing employers power (Alfonsson 2024).
Given the lower union power and lower individual labour market power, the reduction in employment protection and unemployment benefits, could have had a more significant impact on the working class. The wage-setting model has not been able to compensate for this loss of power, which has led to worsened conditions for the working class. The middle class, with its higher individual market power, has managed to avoid a similar decline in working conditions, even in the face of reduced institutional power. This highlights the importance of focusing on institutional power to reduce inequality between classes.
Given that the working class has experienced worse economic development than the middle class and is more reliant on collective power resources – resources that have been diminished – they have faced a lost promise of prosperity. This divergence could also make it more difficult to foster class solidarity between the middle class and the working class, as their economic trajectories and the challenges they face have differed.
In Sweden, as well as in other countries, it has often been social democratic governments that have implemented austerity measures, weakening the institutional power of the working class and leaving them adrift (Alfonsson 2024; Altermark & Plesner 2025; Skyrman et al. 2023). When social democratic parties abandoned the working class, the working class, in turn, abandoned them (see, for example, Vestin 2019). Based on the analysis in this article and the class definition applied, one way to improve the situation for the working class would be to strengthen their institutional power. If political reforms were introduced to increase the power of the working class, their wages and working conditions would improve. The focus on the middle class and its potential proletarianisation has shifted attention in the wrong direction, pushing the development of the working class into the political shadows.
Using a class definition based on the value of labour power and on labour market power within the capitalist mode of production allows us to understand both the causes of increased proletarianisation within the working class and how to reverse this trend.
Footnotes
Acknowledgements
I would like to thank the two anonymous reviewers for their valuable feedback and Jesper Prytz for his comments on the manuscript.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Forte, the Swedish Research Council for Health, Working Life and Welfare, dnr 2020-00359.
