2.1. Theoretical framework

Trends in job quality in the Republic of Korea can be considered in the light of competing theories. For one group of theories, the expectation is that all dimensions of job quality will either improve or decline simultaneously. For another group, the expectation is that some dimensions will rise while others fall. A third group focuses on polarization, with some social groups gaining or declining relative to others in respect of some or all job quality dimensions.

In the first group, a widely held optimistic perspective is that all dimensions of job quality will improve as a country becomes more affluent. This perspective assumes that working conditions are normal goods (in the economic sense) and that the trade-off between wages and other dimensions is reasonably stable. In macroeconomic terms, this implies that employers offer higher wages and better working conditions across all dimensions in times of positive economic growth (as in the Republic of Korea), and lower wages and poorer conditions in times of economic recession. This economic argument is in line with the sociological proposition dating from the “industrialism” perspective of the 1960s and renewed in the “post-industrialism” perspective, which emphasized the knowledge economy. The latter held that firms and countries would increasingly compete on the basis of innovation and knowledge, and that the complexity, skill requirements and level of autonomy afforded in jobs would gradually rise accordingly (Gallie 2007; Knox, Warhurst and Pocock 2011). From this perspective, then, the Republic of Korea’s rapidly growing economy would be expected to translate over time into improved job quality without the need for social regulation and, accordingly, to engender better health and well-being.

That expectation is qualified, however, in that regulation may be necessary to generalize job quality improvements among all workers, especially where the drivers of improvement are inhibited by constraints on labour mobility and labour market competition. Moreover, a contrasting, power-resources perspective suggests that job quality could be adversely affected in the absence of constraining regulation. If, over time, rents are redistributed to capital from labour (Howell and Kalleberg 2019; Henley 2022), job quality could diminish in all dimensions. The declining power of labour is said to be manifested in falling union density, less collective bargaining and a reduced voice for labour representatives in national politics, which together limit the regulatory control of labour markets and are currently all common features in several Western countries. Until the disruptions of the COVID-19 pandemic, the spread of domestic and global outsourcing, plus expanded competition through trade, may have further tipped the balance of power away from workers across US and European workplaces (Howell and Kalleberg 2019).

In the Republic of Korea, trade union density has remained low for most of the twenty-first century, at only around one in ten workers, although the unions gained some ground among younger workers after 2017, when the government regularized non-standard workers in the public sector. Moreover, unions are prohibited from having direct links to political parties and have been able to exert relatively little pressure on policymaking. Workplace regulation has been limited, focusing primarily on working hours, minimum wages and employment equality. Regulation of working hours was first introduced in 1953. It was tightened in 2023 to allow a maximum of 40 hours a week, with an additional 12 hours of overtime work permitted by agreement, plus 16 hours of holiday (weekend) work, bringing the maximum working hours per week to 68. In 2018, the maximum work week was further restricted to 52 hours, comprising 40 regular hours and 12 overtime hours permitted by agreement, which now also include holiday (weekend) hours. Minimum wages, first introduced in 1988, are uprated annually based on the economic situation. Workplace sexual harassment was proscribed by the 1988 Equal Employment Opportunity Act.

If such regulations are in practice significant drivers of change in job quality, they are likely to have a varying impact on the different dimensions of job quality and therefore bear out the second group of theories, according to which job quality dimensions can change in opposite directions. Workers’ changing attitudes may also lead to differentiation (Gallie, Felstead and Green 2012). For example, as more and more women enter the labour force, and as gender norms and the associated preferences change among new generations of workers, workers’ preferred trade-offs between dimensions may become different, with greater importance being attached to working time. If employers respond, even partially, to workers’ preferences in order to remain competitive in the labour market, working time may improve even while other working conditions deteriorate. Alternatively, differentiation could stem from technological or organizational trends. For example, the theory of effort-biased technical change implies increasing work intensification, but this can be accompanied by rising wages for those accepting to work more intensively (Green et al. 2022).

The third group of ideas anticipates the polarization of working conditions across the workforce (Kalleberg 2011). The theory of task-biased technical change predicts rises in job quality for those with the education and skills needed to perform the proportionately increasing non-routine tasks that remain after digitalization (Autor, Katz and Kearney 2006). The digitalization of work also brings new risks in terms of job quality for some groups of workers, especially those who are marginalized and do not have the opportunity to participate significantly in work organization and design (Berg et al. 2023). The rise of platform work is reigniting and intensifying workforce segmentation into a secure, formal primary sector versus an increasingly insecure secondary sector (Hassel and Sieker 2022).

In this article, we focus on polarizations associated with two above-mentioned striking characteristics of the Republic of Korea: education and gender. The job quality gap between better- and less-well-educated workers would likely be driven by the relative demand for and supply of better-educated workers, and the OECD (2017, 67) documents the existence of a job quality premium for intrinsic working conditions in Europe that complements the conventional university wage premium. In the Republic of Korea, the proportion of the university-educated population in the labour force has increased dramatically, particularly for the young female population (see figure 1). This rapidly growing relative supply of university-educated workers leads to the expectation of a positive but declining job quality premium and increasing differentiation among tertiary-educated workers (Green and Zhu 2010).

Figure 1

Tertiary-educated population (percentages)

Source: OECD tertiary education database.

Reflecting sharply defined gender roles in traditional Korean society, women have long experienced substantial occupational and industrial segregation, lower earnings than men even when doing similar jobs, and fewer opportunities for promotion (Patterson and Benuyenah 2021). Given the implied high levels of gender disparity, a gender perspective might anticipate that the wage gap will be reinforced by similar differentials in favour of men with respect to all or most dimensions of job quality. Stier and Yaish (2014), for example, find that most men do better than women in respect of five out of six aspects of job quality across 27 countries,³ although Green (2021) finds a more mixed picture in Europe. To the extent that traditional views of gender roles and attitudes have continued to decline as the economy has grown, the introduction of anti-discriminatory legislation (beginning with the 1988 Equal Employment Opportunity Act) promotes the participation of women in the labour force to levels comparable to those in many Western countries (see figure 2). Gender job quality gaps could therefore be expected to narrow over time, as long as there is adequate regulatory compliance.

Figure 2

Labour force participation rate (percentages)

Source: ILOSTAT Labour Force Statistics database.

Lastly, given that job quality varies significantly across industries (Eurofound and ILO 2019), the aggregate changes anticipated within each of the above perspectives may derive from changes not only within each industry but within the industrial structure. The Republic of Korea was already a developed country at the start of the period under consideration, and the pace of industrialization has slowed since 2000 (Lee 2016). Despite the decelerated changes in industrial composition, however, it remains possible that aggregate job quality trends stem in part from a structural industrial change.

2.2. Existing evidence on job quality trends in the Republic of Korea

Eurofound and the ILO (2019) have compared the levels of job quality dimensions in the Republic of Korea, Europe, the United States and several other countries. In addition to the long working hours previously noted, the working hours of an especially high proportion of Korean workers are fixed by the employer. Overall, both the Working Time Quality and the Skills and Discretion indices are found to be distinctly worse in the Republic of Korea than in either Europe or the United States. For example, Korean workers have relatively little latitude in terms of decision-making. By contrast, Work Intensity is reported to be lower in the Republic of Korea, while for other dimensions (e.g. Prospects and the Physical Environment), the comparison is mixed (Lee 2018).

Studies of job quality trends in the Republic of Korea, and of how the country might fit into any of the above perspectives, are, however, piecemeal and relatively scarce. Annual working hours have been decreasing quite rapidly for more than three decades, consistent with the affluence perspective. However, the few studies that have looked at changes in other job quality variables are less encouraging. Lee (2018), for example, finds that Work Intensity indicators worsened between 2010 and 2017, while those for both job security and short-term flexibility during working hours declined. Increased exposure to vibrations, low temperatures, chemicals and infectious materials were also observed between 2006 and 2010 (Kim et al. 2015). Further, a recent study reports a significant decline in the share of well-resourced employees with good working conditions from 2010 to 2015 (Murtin et al. 2024). That said, less than seven years may be too short a period in which to assess secular changes. Moreover, all job quality dimensions must be comprehensively covered over a period of at least a decade, preferably more, in order to obtain an adequate understanding of whether there has been secular progress in job quality in the Republic of Korea as it continues to converge to the level of material affluence of Western countries.

2.3. Wage premium for university graduates

Studies of the wage premium associated with higher levels of education in the Republic of Korea show a substantial fall from very high levels in the 1980s through to the mid-1990s, followed by a slow rise between 2000 and 2011 (Park 2014). Furthermore, there is increasing heterogeneity in graduates’ earnings depending on the type of university education (two- or four-year) and the university’s prestige (Yi and Kim 2016; S. Lee 2022; Lee and Vignoles 2022; Han, Bae and Sohn 2012; Jung and Lee 2016; Lee, Jeong and Hong 2018). Taken together, this prior evidence suggests that the mean gap in earnings between graduates of two- and four-year university programmes may have shrunk or become even zero for graduates from many universities, bar those from a few top four-year university programmes. To our knowledge, however, no studies examine the changing premium for Korean graduates in respect of working conditions other than wages.

2.4. Gender discrimination

While the gender pay gap remains large by international standards, it has gradually improved since the introduction of anti-discrimination legislation. Whereas women earned only 40 per cent of what men earned during the 1980s, by 2021 the ratio had improved to 69 per cent.⁴ Notwithstanding the extensive literature on this gap (e.g. Monk-Turner and Turner 2004; K. Lee 2022), no studies have been conducted to date of how male–female job quality gaps, other than for wages, have been changing in the Republic of Korea.

To document secular trends in job quality for Korean workers, and to better understand whether these trends conform to any of the competing theoretical perspectives outlined above, we investigated trends and changing job quality gaps by gender and level of education over the period 2006–20. Using the Korean Working Conditions Survey, we constructed seven job quality indices, using the same protocols described in detail in Eurofound (2012), and addressed the following questions:

1. How do the seven job quality indices change over time between 2006 and 2020?

2. Do university graduates benefit from a “job quality premium” compared with those without a degree? How does that premium change over time?

3. Does gender disparity exist in wider aspects of the labour market, besides earnings? If so, has any change in this disparity reinforced or counterbalanced improvements in the gender wage balance?

3.1. Dataset

We used the Korean Working Conditions Survey, conducted by the Occupational Safety and Health Research Institute, which has benchmarked and adopted questions similar to those of the European Working Conditions Survey. We took into account the first six repeated cross-sectional surveys (2006, 2010, 2011, 2014, 2017 and 2020), which target employed people aged 15 and over. Trained interviewers visited each household and conducted the questionnaire using computer-assisted personal interviewing; online survey methods were also partially used in 2020, during the COVID-19 pandemic. The target sample size was 10,000 individuals in 2006 and 2010, and 50,000 in the subsequent surveys. Clustered random sampling methods were used, and survey weights are included in all our analyses, yielding nationally representative estimates.

3.2. Job quality indicators

Along with most job quality scholars, we used an objective definition of job quality that considers the job attributes that help workers meet their needs through their work. Those attributes are distinct from, though expected to be related to, subjective feelings such as job satisfaction. They can be grouped in multiple ways covering both extrinsic aspects of jobs, which are often detailed in labour contracts, and intrinsic aspects related to tasks, relationships and the physical setting (Muñoz de Bustillo et al. 2011; Eurofound 2012; Felstead et al. 2019). While the detailed list of job quality dimensions and associated indicators varies substantially between studies (Stefana et al. 2021), most lists are variations, subsets or intersections of the seven dimensions adopted by Eurofound and the European Parliament (Eurofound 2012; Berg et al. 2023): Earnings, Prospects, Skills and Discretion, Social Environment, Physical Environment, Work Intensity and Working Time Quality. Each dimension, in turn, is measured using composite indicators. Because of the generality of this approach, and because of the close similarity between the European and the Korean Working Conditions Surveys, we used the Eurofound (2012) approach here. We constructed the same composite indicators for each of the seven job quality dimensions, or indices, using the Korean Working Conditions Survey (see table 1 for a brief description).

“Earnings” is net monthly earnings (in Korean won at 2010 prices), since this most directly captures how earnings, whether from wages or self-employment, serve to cover living costs. In common with other surveys, the Earnings index had a number of missing values; these missing values were imputed with predicted earnings from regression analysis using the respondents’ gender, age, age squared, level of education and industry.

The Prospects index includes one negative component (job insecurity, that is, the probability of job loss) and one positive component (the probability of career progression).

The Skills and Discretion index was constructed based on the skills used for the job and the worker’s degree of discretion or autonomy at work, as measured by the replies to the following questions: Are you able to choose or change the order in which you perform your tasks, your methods of work, and your speed or rate of work? Are you able to apply your own ideas in your work? Can you influence decisions that are important for your work?

The Social Environment index has two components: social support and the absence of abuse. Social support was measured by asking the following question: Do your managers (colleagues) help and support you? The absence of abuse was measured by asking questions such as: During the course of your work, have you been subjected to verbal abuse, unwanted sexual attention, threats or humiliating behaviour? Some questions (for instance, those about support from colleagues and managers) were not asked of self-employed respondents, leading to many missing values for the Social Environment index. We therefore assigned the value of 0 (zero) to self-employed workers who did not have employees.

The Physical Environment index also has two subdimensions: the absence of chemical/physical hazards and ergonomic risks. Chemical or physical hazards were investigated by measuring exposure to severe vibration from tools and machinery, noise, high or low temperatures, smoke or vapours, and chemical products or substances. Ergonomic risks were examined by measuring the extent to which jobs involved tiring or painful positions, carrying or moving heavy loads, and repetitive movements.

The Work Intensity index encompasses the pace of work, emotional demands and performance pressure from various sources, including colleagues and managers.

Lastly, the Working Time Quality index includes working hours and how frequently employees work non-standard hours, such as night work. Apart from Earnings, all job quality dimensions were normalized to a range of 0 to 100.

Although the dimensions and indices refer to objective qualities, we relied on workers’ reports of their job characteristics, as opposed to reports by employers or third parties, in the surveys. This allows the researcher to hear directly from the worker doing the job, who is best informed about it. However, when examining trends or making cross-cultural comparisons, some of the reports may need to be interpreted with care, owing to the possibility that any reporting biases – for example, social esteem bias – may change over time or between cultures.

3.3. Empirical approach

To address our research questions, we started with a descriptive analysis of job quality means in each survey year. We then regressed each job quality index against the survey year to estimate the average trend growth, expressed as:

$J{Q}_i=\alpha +\beta \mathit{\text{yea}}{r}_i+{\epsilon }_i$      (1)

We then shifted the focus to the job quality premium for graduates, the gender job quality gap, and the associated trends. To that end, we ran a conventional Mincer-type model comprising sex (0 = female; 1 = male), level of education (0 = high school or below; 1 = university educated) and age squared. We also included the interaction terms of gender and level of education with year to estimate the trend in job quality gaps over time, expressed as:

$\begin{array}{c}J{Q}_i={\alpha }^{\prime }+{\beta }_1\mathit{\text{yea}}{r}_i+{\beta }_2\mathit{\text{ed}}{u}_i+{\beta }_3\mathit{\text{ed}}{u}_i\cdot \mathit{\text{yea}}{r}_i+{\beta }_4\mathit{\text{mal}}{e}_i+\\ {\beta }_5\mathit{\text{mal}}{e}_i\cdot \mathit{\text{yea}}{r}_i+{\beta }_6\mathit{\text{ag}}{e}_i+{\beta }_7\mathit{\text{ag}}{e}_i^2+{\epsilon }_i^{\prime }\end{array}$      (2)

In equations (1) and (2), JQ_i indicates one of the seven job quality indices, year_i indicates the year in which the survey was conducted, and ɛ_i and ${{\epsilon }^{\prime }}_i$ are normally distributed error terms. In the case of Earnings, following standard conventions, the variable entered is the log of monthly earnings. For examining trends, the main parameters of interest are the regression coefficients β in equation (1) and both β₃ and β₅ in equation (2).

4.1. Job quality trends

Table 2 shows the mean values by year and the estimated trend (β) for each job quality dimension. Three of the seven job quality indices (Prospects, Skills and Discretion, and Physical Environment) trend downwards between 2006 and 2020; the remaining four (Earnings, Social Environment, Work Intensity and Working Time Quality) trend upwards over the same period. As Work Intensity is a negative indicator, this means that four of the seven job quality indices worsen between 2006 and 2020. These changes are statistically significant at the 5 per cent level. Among the deteriorating dimensions, Prospects worsens most rapidly, falling at the rate of a 26 per cent standard deviation decrease per decade. By comparison, the ten-year rates of decline for the other dimensions are 23 per cent for Work Intensity, 15 per cent for Skills and Discretion, and just 7 per cent for Physical Environment.

A closer look at the decline in Prospects reveals that a downward trend in job security (β = –1.43) outweighs a smaller improvement in career progression (β = 0.39).⁵ The Skills and Discretion index moves downward largely owing to declining levels of workplace autonomy. While ergonomic hazards decrease, the magnitude of worsening physical and chemical hazards outweighs the magnitude of the improved level of ergonomic hazards, leading Physical Environment to decline, although only by a small amount.

In contrast, Working Time Quality improves swiftly, equivalent to a 76 per cent standard deviation increase over ten years. All components of this index improve, with long working hours, night work and weekend work in particular showing a downward trend. For example, in 2006, 45 per cent of individuals reported working long hours, defined as more than 48 hours per week, whereas in 2020, only 20 per cent did so. Similarly, the proportion of individuals engaged in night work is 23 per cent for 2006 but only 8 per cent for 2020.

Earnings also rise at a ten-year rate of 14 per cent of a standard deviation. Lastly, the Social Environment index exhibits only a very modest improvement, with a ten-year rate of 6 per cent of a standard deviation. This is mainly due to improved support from managers and colleagues; no significant changes are observed in terms of the absence of abuse at work. Socially driven bias may be a concern when it comes to interpreting this index, because the public discourse on sexual harassment at work, stimulated by the #MeToo movement’s prominence in the Republic of Korea as of 2018, might well have heightened awareness and prompted survey respondents to recall and report incidents. To assess the movement’s potential impact on the Social Environment index, trends were compared before and after the inclusion of the sexual harassment variable. The results show that the modest upward movement in the Social Environment index remains significant.

To assess how far these changes in job quality are associated with the changes in industrial composition, we ran separate regression models that included industry dummies. The differences in the regression coefficients (β) before and after controlling for the industrial composition are all small, implying that the trend changes in each job quality index are not accounted for by industrial structural change (see table 2).

4.2. Trends in the university premium and the gender gap

To address our second and third research questions, we shifted the focus to the job quality premium for university-educated workers (compared with high school and below) and gender job quality gaps. Tables 3A and 3B show the differences in the seven job quality dimensions between men and women and between graduates and non-graduates.

First, men experience better job quality than women in only three out of the seven dimensions: Earnings, Prospects, and Skills and Discretion. The gender gap for Earnings shows the largest gap in favour of men, with a mean log earnings gap of 0.46 (equivalent to 64 per cent of the standard deviation), followed by Skills and Discretion (20 per cent), and Prospects (8 per cent). By contrast, women experience better Social and Physical Environments, lower Work Intensity and higher Working Time Quality. The gender gap for the Physical Environment index shows the largest pooled gap in favour of women, equivalent to 28 per cent of the standard deviation, followed by Working Time Quality (21 per cent), Work Intensity (9 per cent) and Social Environment (4 per cent). Two job quality indices show an upward trend in the gender gap: Prospects (5 per cent of the standard deviation increase for a decade) and Skills and Discretion (4 per cent). Thus, while the gender pay gap declines, the gender gap in Prospects and Skills and Discretion, which favours men from the start, grows even larger over the period.

Second, in all years there is a positive job quality premium for graduates compared with non-graduates in six of the seven job quality dimensions. Graduates have the largest premium in Earnings, with 0.53 higher log monthly earnings on average than non-graduates, equivalent to 74 per cent of the standard deviation of the index. Among non-pecuniary indices, the Skills and Discretion index shows the largest premium for graduates – over the whole period, the pooled gap is 5.78, equivalent to 57 per cent of the standard deviation of the index – followed by Physical Environment (53 per cent), Prospects (41 per cent), Working Time Quality (35 per cent) and Social Environment (21 per cent). The exception is Work Intensity, for which university-educated workers do not have consistently better results over time than those without a degree. From the point of view of the job quality premium for graduates, four of the seven job quality dimensions (i.e. Earnings, Prospects, Social Environment and Physical Environment) trend downwards. Only in respect of Working Time Quality does the graduate premium show a small upward trend, amounting to a 3 per cent of standard deviation increase per decade.