Workplace Skill Investments – An Early Career Glass Ceiling? Job Complexity and Wages Among Young Professionals in Sweden

Abstract

Despite higher educational investments, women fall behind men on most indicators of labour market success. This study investigates whether workplace skill investments set men and women off on different tracks in which the human capital acquired through higher education is either devalued or further developed. A survey sample of Swedish men and women who recently graduated from five educational programmes, leading to occupations with different gender composition, is analysed (N ≈ 2300). Results show that, a few years after graduation, men are more likely than women to acquire complex jobs and that this difference contributes to early career gender gaps in wages and employee bargaining power. The findings do not support the notion that child-related work interruptions provide a main mechanism for sorting women into less complex jobs.

Keywords

Employee bargaining power gender wage gap initial on-the-job training job complexity work interruptions workplace skill investment

Across the world, women take the educational road to gender equality. In the 2000s, the gender gap in education has been reversed in the member countries of the Organization for Economic Cooperation and Development (OECD). In 2000, adult men had higher tertiary attainment rates than adult women whereas in 2012, the situation was inverted (OECD, 2014: 31).

Clearly, however, this road is long, winding and not sufficiently outlined on maps provided by human capital theory. Despite higher educational investments, a gender difference to women’s disadvantage can be observed on most indicators of labour market success. Unexplained gender wage gaps persist throughout the OECD (e.g. Magnusson, 2010; Perales, 2013) and these gaps are particularly large among highly educated employees and in skilled occupations (e.g. Boye et al., 2017; Magnusson, 2010). Thus, even as women surpass men in terms of credentials, they fall behind in rewards.

Faced with these puzzles, several researchers have pointed to the role of workplace skill development. Because workplace training is an employer investment requiring long-term employment relations to be profitable, such investments may discriminate against women who can be expected to interrupt their work to care for children (e.g. Polachek, 1981). Also, workplace training is discussed as a mechanism driving labour market segregation (e.g. Estévez-Abe, 2006; cf. Dale, 1987). Thus, workplace skill investments could set men and women off on different tracks in which the human capital acquired through higher education is either devalued or further developed and over time, such processes can contribute to ever larger wage gaps.

The study presented here examines these propositions by analysing if men are more likely than women to acquire complex jobs, requiring substantial on-the-job training; if gendered access to complex jobs produces early career differences in wages and employee bargaining power and, finally, if such differences are explained by female work interruptions. The analysis is based on a sample of Swedish men and women who recently graduated from five higher educational programmes, leading to occupations with different gender composition (N ≈ 2300).

Previous research and contribution

While educational attainments are based on individual choices (however constrained), workplace training represents an employer investment. Theoretically, such investments are characterized by their long investment horizon. Indeed, researchers from vastly different perspectives (e.g. Becker, 1964/1993; Osterman, 1984) argue that workplace learning and training is key to understanding employers’ interest in long-term employment contracts.

Based on such premises, researchers present workplace skill investments as an important factor explaining labour market gender inequalities. The basic argument is that because such investments presuppose long-term employment relations, employers will hesitate to invest in women who are presumed to interrupt their careers to care for children and families (Polachek, 1981; Polavieja, 2008). In a similar vein, it is argued that women will avoid or be excluded from occupations requiring substantial on-the-job training (Dale, 1987; Estévez-Abe, 2006; Polachek, 1981). Finally, because on-the-job training is considered to be more important in high-skilled jobs (OECD, 2003), gender differences in the access to training could be part of the ‘glass ceiling’-effect encountered by high-achieving women. In short, workplace skill investments could devalue human capital acquired in the educational system, contribute to horizontal and vertical stratification and produce a gender wage gap that is particularly large among the highly educated.

These accounts are both plausible and problematic. In general, the focus on workplace skill investments offers a promising avenue for advancing our understanding of present-day labour market stratification, including gender stratification. Tåhlin and colleagues (Le Grand and Tåhlin, 2013; Tåhlin, 2007a, 2011) have demonstrated that job complexity – or the skill requirements of a job’s work tasks – represents the main dimension of the vertical differentiation in work content. The concept of complexity incorporates both the educational requirements of the job and the initial job training requirements¹ and empirically, both these factors correlate more strongly with wages than does individual education (Tåhlin, 2011). Presumably then, men and women with similar educational attainments will face different wage trajectories if the skill requirements of jobs differ systematically by gender.

At the same time, the prevailing accounts connecting gender and workplace skill investments rely on assumptions – notably, about female work interruptions – that appear outdated and oversimplified. Although women still carry the main responsibility for children and household and are more likely than men to take parental leaves and work part-time, both attitudes and practices are changing, particularly among the highly educated and in contexts promoting gender equality, such as Scandinavia (e.g. Edlund and Öun, 2016). Furthermore, while labour market gender segregation appears persistent, several occupations undergo rapid change, particularly because women are entering prestigious, previously male-dominated occupations requiring higher education (e.g. England, 2010). More generally, it seems problematic to call upon traditional behaviours to explain gender gaps that persist or widen as women behave less traditionally, for example by acquiring university degrees. In sum, there is a need to empirically examine the proposed mechanisms.

This study focuses on gender differences in the access to complex jobs. As noted, the definition of complex jobs rests on two pillars, the educational requirements of the job and the workplace training required to perform the job tasks. The educational requirements inherent in the job are painstakingly controlled for in this study as men and women are matched in terms of educational programme and occupation. Therefore, the empirical measure of job complexity can be narrowed down to the requirements for initial training. We argue that employers’ workplace training investments could set men and women off on different tracks and that, over time, initial sorting on job complexity could produce substantial wage gaps through reinforcing feedback processes.

Feedback processes of job complexity

Although individuals are likely to be sorted to complex jobs by ability, such jobs – requiring independent thinking and autonomous judgement – also provide good learning environments, further enhancing worker skills. Conversely, the value of education will be undermined in routine jobs where skills cannot be fully used or further developed (Tåhlin, 2011). For example, studies have demonstrated reciprocal effects between complex jobs and intellectual flexibility (e.g. Kohn and Schooler, 1978). Thus access to jobs with high initial training requirements will not only have a direct effect on wages but create a causal feedback loop as learning improves workers’ productivity and, thereby, their wages and promotion opportunities.

Second, workplace training will strengthen the bargaining power of the employee. Since on-the-job training is an investment, employers are assumed to have a strong incentive to keep trained employees in the company. However, the notion that workplace training causes employers and employees to ‘get bonded together’ (Becker, 1964/1993: 20) in a long-term relationship based on mutual dependence is increasingly contested. While human capital theory claims that skills conveyed by on-the-job training are firm-specific, binding employees to the company providing them, empirical research suggests that most training provides skills that are transferable to various firms (Leuven, 2005; OECD, 2004). Moreover, Tåhlin (2007a) shows that employer–employee dependence relations are commonly asymmetric and that both wages and mobility can be related to the relative dependence of the parties – specifically, the extent to which the employee is difficult for the employer to replace and highly employable by others, or vice versa. If complex work tasks enhance worker productivity in ways that are useful also to other employers, individuals who acquire jobs with high initial training requirements have better opportunities for both ‘exit’ and ‘voice’. Thus, a second feedback loop is created, further strengthening the long-term effect of initial on-the-job training.

A third feedback loop may appear in relation to within-family specialization. If workplace training sorts men into demanding jobs with higher wages, it could spur a traditional division of paid and unpaid work even between spouses with similar education (cf. Dale, 1987). Over time, differences in skill investments will increase, either because specialization maximizes household utility (Becker, 1991) or because the relative income of spouses affects their bargaining power in the household (Lundberg and Pollak, 1996). If this is the case, training would be related to female work interruptions and part-time work but the causal processes would be more complex than commonly assumed.

In sum, there is good reason to explore gender differences in initial training requirements of jobs. The wage impact of initial training requirements has been established in research from different countries (e.g. Duncan and Hoffman, 1979; Gronau, 1988; Tåhlin, 2011). Also, this measure has been shown to explain part of the gender wage gap (e.g. Gronau, 1988; Grönlund and Magnusson, 2013; Tåhlin, 2007b). However, gender differences in initial training requirements have not been thoroughly explored or related to education, occupation and work interruptions (but see Grönlund, 2012). Furthermore, gender differences in perceived replaceability, reflecting the bargaining power of employees, have not been explored or related to workplace training.

This article explores the potential importance of the mechanisms discussed above by examining gender differences in access to complex jobs and the extent to which such differences relate to both wages and perceptions of replaceability, reflecting the bargaining power of the employee. By using a specific sample, this study puts the theoretical assumptions to a strong test. The study focuses on younger individuals brought up in Sweden. Here, welfare state policies promoting gender equality and a dual-earner/dual-carer family model have been in place since the early 1970s (e.g. Korpi, 2000) and since then, levels of female employment have been high by international standards. Attitudes supporting a gender equal sharing of paid and unpaid work are stronger here than in many other countries (Edlund and Öun, 2016) and although parental leaves are still gendered, Swedish fathers on average take more than three months of leave, a significant amount by international standards (e.g. Duvander and Viklund, 2014). With the choice of the Swedish context and a sample comprising men and women working in the same occupation, we intended to minimize gender differences. The idea is that if the predicted gender patterns appear in this sample, then the account of gendered skill investments provided by human capital theory may be regarded as a valid and important mechanism for gender inequality even in modern, dual-earner societies.

Aim and hypotheses

The aim of the article is to examine if, a few years after graduation, women are less likely than men to have complex jobs requiring substantial on-the-job training; if differences in job complexity translate into differences in wages and replaceability and whether gender patterns can be attributed to female work interruptions.

The analysis is based on the following hypotheses:

H1. Women are less likely than men to have complex jobs, requiring more than one year of workplace training. The difference is explained by women’s work interruptions.

H2. Because of their lower access to complex jobs, women have lower wages than men in the same occupation.

H3. Because they less often have complex jobs, women believe themselves to be easier for the employer to replace than do men in the same occupation.

Data and method

The questionnaire was distributed in 2013 to Swedish men and women that graduated in 2007–2010 from five higher educational programmes: Degree of Master of Science in Engineering (hereafter: civil engineers); Degree of Master of Laws (lawyers); Degree of Master of Science in Psychology (psychologists); Degree of Bachelor of Science in Social Work (social workers); and the Police Programme (police officers). The first four programmes are all university-based but differ in length (3.5–5 years). The Police Programme is a post-secondary education and comprises 1.5 years of studies plus six months of trainee service.

The motivation for choosing these programmes was to expose the impact of gender, by keeping confounding factors such as occupation in check. The programmes are similar in the sense that they all lead to a specific professional title, which means that individuals have already made an occupational choice. Meanwhile, the programmes differ in their gender-mix. According to the Swedish occupational register, women constituted 20 per cent of the civil engineers, 26 per cent of the police officers, 50 per cent of the lawyers, 72 per cent of the psychologists and 84 per cent of the social workers in the labour force in 2010. Finally, the sample was stratified such that 500 men and 500 women were sampled from each programme.

The sample was drawn from the National Register of Higher Education and the Swedish Register of Education. Sampling, distribution and coding were administered by Statistics Sweden. The response rate was 55 per cent. This study uses a subsample comprising employed individuals working >15 hours a week in the occupation they were trained for (N ≈ 2300).

Variables and analytical strategy

The analysis is divided into three parts, based on hypotheses 1–3.

The first part (H1) concerns gender differences in complex jobs, defined as having a job that requires at least one year of initial on-the-job training.² The respondents were asked how long it would take for someone with the right education and qualifications to learn to do the respondent’s job reasonably well. Response categories were: (1) <3 months; (2) 3 months–1 year; (3) 1–3 years; (4) 3–5 years; and (5) 5< years. As the study focuses on skill and career opportunities, the main interest concerns the jobs requiring extensive skill development. Therefore, the measure was dichotomized to indicate an initial on-the-job training period of one year or more (categories 3–5). This indicator, capturing the initial training requirements inherent in the job, has been used in several studies and validated against both US and Swedish wage data (e.g. Duncan and Hoffman, 1979; Gronau, 1988; Tåhlin, 2007b). Tåhlin’s (2011) measure of job complexity also includes the educational requirements of the job; however, in this sample these are already accounted for as men and women are matched on education. Continuous workplace training, a potential third dimension, is not included here. Tåhlin found measurement properties for the indicator of continuous training to be less clear than for initial training. The focus on initial training also seems theoretically motivated as the initial training period should be easier for employers and employees to appreciate than the long-term prospects. Moreover, the initial workplace training may be crucial for determining long-term wage and career trajectories.

To analyse the likelihood of having a complex job, a linear probability model (LPM) is estimated. In an LPM, a binary dependent variable is regressed on the independent variables using linear regression. This method ensures comparability between models (Mood, 2010), which is important for analysing the influence of factors proposed to explain gender differences. A drawback of LPM as compared to logistic regression, the traditional method of choice with binary dependent variables, is the risk of heteroscedastic and non-normal residuals. Heteroscedasticity-robust standard errors are used to correct for this. It has been shown that LPM effect estimates are unbiased and consistent estimates of average effects and hence LMP is appropriate when the focus lies on effects at a certain point of the distribution, such as the average, and not the functional form of the relationships per se (Mood, 2010). As a sensitivity test, the models were estimated using logistic regression with average marginal effects and these results corresponded to the LPM results.

The analysis focuses first on the impact of gender, then variables are entered stepwise to capture the mechanisms proposed to link gender and workplace training. The main independent variable is parenthood which is used as a proxy of work interruptions related to child-rearing. These could include both full-time parental leaves, periods of part-time work, and absence due to care for sick children. This indicator measures the gendered nature of care responsibilities, and includes three dichotomous measures. No children in household indicates the absence of children below age 18 in the respondent’s household. The other two measures are distinguished by the timing of children in the household. All children born before graduation indicates that all children were born before the year that the respondent graduated from the educational programme in question. Child born after graduation indicates that at least one child was born after the year of graduation. Distinguishing between children born before and after graduation is a way to capture any child-related labour market absences after graduation. With a child born after graduation the employee is more likely to have taken parental leave or worked part-time after entering the current occupation and presumably, such leaves are a stronger signal of low commitment to making a career in the occupation than leaves completed before graduation. Also, children born after graduation will be younger, demanding more intense care. Next, work experience and seniority are entered. Work experience indicates the number of years that the respondent has been in gainful employment. The measure includes periods of labour market absence during which the respondent was employed (e.g. parental leaves), but excludes extra work carried out parallel to studies. Seniority measures the number of years the respondent has worked with the current employer. Finally, usual weekly work hours are controlled for.

To complement this analysis, additional regressions are carried out to study the link between length of parental leave and workplace training (see Online Appendix). This analysis comprises parents only (N = 994) and because parental leave length is highly correlated with gender, it is run also for fathers and mothers separately. Parental leave is measured in four categories: 0–2 months; 3–6 months; 7–11 months; and 1 year or more, and only captures the parental leave taken with the youngest child.

The second part analyses whether gender differences in the access to complex jobs produce an early career wage gap (H2). Here, ordinary least squares (OLS) regressions are estimated and the natural logarithm of monthly wages (before taxes and transfers) is used as the dependent variable. Using a logged variable means that the coefficients (b) can be recalculated to show the percentage difference in the dependent variable for a one unit difference in the independent variable using the following formula: 100*(exp(b)–1). As in the first part, the independent variables are entered stepwise. Job complexity is entered in a separate model after controlling for parenthood, work experience and seniority, presuming that the impact of female work interruptions would run through job complexity. The final model controls for work hours.

The third part applies LPM to analyse determinants of low replaceability, considered here as a proxy of the employee’s bargaining power. Specifically, the analysis explores whether men perceive themselves to be more difficult for the employer to replace than do women, and whether such a gender difference can be explained by job complexity and female work interruptions (H3). To capture replaceability, the respondents were asked how difficult they thought it would be for their employer to replace them if they left the job. Response categories were: (1) very difficult; (2) fairly difficult; (3) not that difficult; (4) fairly easy; and (5) very easy. Because complex jobs involving substantial training would make the employee difficult to replace, the measure was constructed to capture a situation where the employee is fairly or very difficult to replace (categories 1 and 2). Previously, some studies have used this measure to capture an asymmetric dependence relation signifying employee power vis-a-vis the employer (Edlund and Grönlund, 2008; Tåhlin, 2007a). As shown by Tåhlin (2007a), it represents a situation where demand for labour exceeds supply, pushing wages up, and thus it can be regarded as a proxy for bargaining power. In these studies, replaceability was paired with a measure of employability (i.e. the prospects for finding new employment). However, as the analysis showed no gender difference in employability this measure does not add any additional information here. Also, because the indicator of employability can be more transient and sensitive to current conditions on the labour market it should be less relevant for capturing the implications of training investments. Thus, the focus is on the measure of replaceability. The analysis follows the same steps as the wage analysis.

Due to the stratification of the sample, all analyses include controls for occupation (civil engineer, lawyer, psychologist, social worker and police officer) as well as graduation year (2007–2010, not shown in the tables). The sampling implies that women and men that have made gender-atypical occupational choices have been oversampled to minimize the influence of factors other than gender. The stratification also implies that the variable occupation reflects central characteristics of the occupation (e.g. skill requirements) rather than its gender composition. However, to give the reader a more complete understanding of the results and relate to the discussion about workplace skill investments and segregation, the ‘unadjusted’ gender differences from weighted bivariate regressions, in which the share of men and women in each occupation corresponds to actual share in the population (i.e. individuals graduating from these programmes 2007–2010), are presented in the Online Appendix. Also, other additional analyses and sensitivity tests have been carried out and are reported below (full analyses available from the corresponding author).

Results

Table 1 presents descriptive statistics. As shown, more men than women had complex jobs requiring an initial workplace training period of at least one year. Men had higher monthly wages than did women and were also more likely than women to say that they would be fairly or very difficult for the employer to replace. Although the men and women in this study were at the beginning of their career, there was a gender difference in work experience to the advantage of men. An additional analysis (not displayed) showed that the gender difference in experience is explained by the difference in age; men were on average 34 years old and women one year younger. There was no gender difference in seniority. The share of mothers and fathers was the same and an equal share of mothers and fathers had had children before and after graduation, respectively. The stratified sample implies that men and women are equally distributed across the five occupations and there was no gender difference in graduation year. There is, however, a slight overrepresentation of women in the sample.

Table 1.

Descriptive statistics.

	All		Men		Women		Significance test of the gender difference
	per cent/mean	N	per cent/mean	N	per cent/mean	N	Significance test of the gender difference
Man	45.2	1082
Woman	54.8	1312
Complex job	26.8	638	32.7	352	22.0	286	***
Monthly wage (SEK)	31232.7		32078.7		30534.0		***
Hard to replace	47.6	1132	54.9	589	41.6	543	***
Work hours/week	41.7		41.8		41.5
Work experience (years)	7.7		8.2		7.3		***
Seniority (years)	3.2		3.1		3.2
Child born after graduation	35.3	839	36.8	394	34.1	445
All children born before grad.	9.6	229	9.0	96	10.2	133
No children in household	55.1	1308	54.3	581	55.7	727
Age	33.6		34.2		33.0		***
Civil engineer	19.8	474	18.9	204	20.6	270
Lawyer	18.4	440	18.5	200	18.3	240
Psychologist	22.4	537	23.1	250	21.9	287
Social worker	20.5	491	20.8	225	20.3	266
Police officer	18.9	452	18.8	203	19.0	249
Graduation year:
2007	25.3	605	24.8	268	25.7	337
2008	22.3	534	22.8	247	21.9	287
2009	25.9	619	25.1	272	26.5	347
2010	26.6	636	27.3	295	26.0	341

Note: ***p < 0.001.

The gender difference in job complexity

Table 2 displays the results from an LPM that explored the likelihood of having a complex job, requiring at least one year of on-the-job training. Model 1 confirms that women were less likely to be in such jobs than were men in the same occupation. A main theoretical argument for explaining such a gender difference is that women are more likely to make work interruptions to care for children. Model 2 controls for the presence and timing of children in the household. As shown, parents who had all their children before graduation were more likely to be in complex jobs than were those who had a child after graduation. Employees without children occupied a middle position but differences between this category and the other categories are not statistically significant. This group was younger than the two groups of parents and it is possible that employers abstain from investing in skill development of young employees without children due to the likelihood that they will become parents in the near future. Crucially, however, controlling for parenthood does not reduce the gender difference in skill development. Since parenthood is regarded as a proxy for child-related labour market absence, this lack of effect indicates that although motherhood often implies a greater interference with labour market work than does fatherhood, this difference does not explain women’s lower access to complex jobs. Interactions between gender and parenthood, included in Model 3, further show that the impact of parenthood does not vary between women and men. An additional analysis explored the effect of parental leaves on a subsample of parents (see Online Appendix Table A2). The results show that parental leave was unrelated to job complexity among both fathers and mothers, with the exception that fathers who took a very short or no leave (0–2 months) worked in less complex jobs than fathers who took a longer leave. This finding is in line with studies showing that high-skilled men (with higher education) take longer leaves (e.g. Duvander and Viklund, 2014). Consequently, controlling for parental leave did not reduce the difference in job complexity between fathers and mothers (the small reduction in the gender coefficient between Models 1 and 2 was not statistically significant, Wald test, p = 0.404). In sum, parents of small children were less often in complex jobs than those with older children, but parenthood did not impact more negatively on women’s than on men’s access to complex jobs and the greater labour market absence that follows from motherhood, as compared to fatherhood, did not explain why women were less often employed in complex jobs.

Table 2.

The gender difference in the likelihood of having a complex job (initial on-the-job training of 1+ years). Linear probability model.^a,b

	Model 1	Model 2	Model 3	Model 4	Model 5
Woman	−0.1100***	−0.1115***	−0.1145***	−0.1130***	−0.1089***
	(0.0182)	(0.0182)	(0.0295)	(0.0183)	(0.0182)
Child born after graduation (Ref.)
All children born before grad.		0.0812*	0.0395	0.0738*	0.0697
All children born before grad.		(0.0338)	(0.0526)	(0.0360)	(0.0359)
No child in household		0.0298	0.0341	0.0391	0.0223
		(0.0199)	(0.0305)	(0.0200)	(0.0201)
Woman* All childr. born before grad.			0.0733
Woman* All childr. born before grad.			(0.0680)
Woman* No child in household			−0.0076
Woman* No child in household			(0.0387)
Work experience				−0.0004	−0.0005
				(0.0020)	(0.0020)
Seniority				0.0243***	0.0237***
				(0.0057)	(0.0054)
Work hours					0.0093***
					(0.0018)
Civil engineer (Ref.)
Lawyer	0.0429	0.0415	0.0405	0.0607*	0.0402
	(0.0299)	(0.0299)	(0.0299)	(0.0298)	(0.0297)
Psychologist	−0.0942***	−0.0932***	−0.0933***	−0.0696*	−0.0410
	(0.0258)	(0.0267)	(0.0267)	(0.0274)	(0.0277)
Social worker	−0.0338	−0.0402	−0.0415	−0.0322	−0.0122
	(0.0277)	(0.0283)	(0.0283)	(0.0293)	(0.0294)
Police officer	0.1700***	0.1704***	0.1698***	0.1600***	0.1897***
	(0.0315)	(0.0318)	(0.0318)	(0.0331)	(0.0333)
Constant	0.3559***	0.3382***	0.3402***	0.2330***	−0.1565
	(0.0286)	(0.0309)	(0.0340)	(0.0410)	(0.0835)
N	2285	2285	2285	2285	2285
R²	0.0595	0.0622	0.0629	0.0749	0.0860

Notes:

Robust standard errors in parentheses.

All models control for graduation year. Significance levels ***p < 0.001, **p < 0.01, *p < 0.05.

As noted, women in this sample on average had less work experience than men (due to their younger age). However, as seen in Table 2, Model 4, this difference does not explain why women were less often in complex jobs, and neither does seniority (although longer seniority is associated with a higher likelihood of having a complex job). Model 5 shows that work hours are positively related to job complexity and, also, that work hours explain the association between parenthood and job complexity. In other words, parents with children born before graduation were more likely to have complex jobs than parents with younger children and non-parents and this difference was related to their propensity to work longer hours. Again, however, further analysis (not displayed) including interaction terms showed that this applied to men and women alike. Also, the gender difference in job complexity remained stable even as work hours were accounted for. The slight reduction of the gender difference between Models 4 and 5 was not found to be statistically significant (Wald test, p = 0.066).

In sum, H1 is supported regarding the gender gap in the access to complex jobs but not in terms of the mechanisms explaining it. Human capital differences between men and women are few, particularly in this sample in which men and women have graduated from the same educational programmes during the same time period. The only significant gender difference was found in work experience and this difference did not explain why women had jobs with lower initial training requirements. Parenthood – considered here as a proxy reflecting gender differences in child-related work interruptions and adaptions – did not explain gender differences in job complexity and neither did the length of parental leave with the youngest child, which was added in additional analyses. Taken together, these results suggest that women’s workload in the family and longer child-related labour market absences are not per se the cause of the gender difference in the access to complex jobs.

Job complexity and wages

The next step focused on the gender difference in wages and analysed the extent to which job complexity and female work interruptions contributed to an early career wage gap between men and women. Table 3, displaying OLS regressions on logged monthly wages, shows that even in this select sample, there was a gender wage gap amounting to about 4 per cent (Model 1). This gap remained stable after controlling for the presence and timing of children in Model 2, but was slightly reduced in Model 3 (a Wald test showed a statistically significant reduction of the coefficient, p = 0.029). Additional analyses showed that this reduction was entirely due to the gender difference in work experience. Controlling for age, the indicator of work experience did not reduce the gender wage gap. Hence, the impact of work experience does not appear to reflect women’s child-related work interruptions, but simply their lower age. Model 4 shows that having a complex job is positively related to wages. Further, the model shows that the gender wage gap is reduced by 10 per cent after controlling for job complexity, a reduction that was found to be statistically significant (Wald test, p = 0.000). Controlling for work hours in Model 5 does not further reduce the gender wage gap. The small reduction in the gender coefficient between Models 4 and 5 was not statistically significant (Wald test, p = 0.195).

Table 3.

The gender difference in wage (log wage). OLS regression.^a,b

	Model 1	Model 2	Model 3	Model 4	Model 5
Woman	−0.0433***	−0.0438***	−0.0412***	−0.0368***	−0.0337***
	(0.0068)	(0.0068)	(0.0068)	(0.0068)	(0.0064)
Child born after graduation (Ref.)
All children born before grad.		0.0211	0.0013	−0.0016	−0.0054
All children born before grad.		(0.0123)	(0.0131)	(0.0130)	(0.0122)
No child in household		0.0144	0.0177*	0.0163*	−0.0032
		(0.0076)	(0.0076)	(0.0076)	(0.0072)
Work experience			0.0027***	0.0027***	0.0025***
			(0.0007)	(0.0007)	(0.0007)
Seniority			0.0056***	0.0047**	0.0042**
			(0.0017)	(0.0017)	(0.0016)
Complex job				0.0385***	0.0240**
				(0.0078)	(0.0074)
Work hours					0.0110***
					(0.0006)
Civil engineer (Ref.)
Lawyer	−0.0419***	−0.0421***	−0.0387***	−0.0409***	−0.0635***
	(0.0109)	(0.0109)	(0.0109)	(0.0108)	(0.0102)
Psychologist	−0.1085***	−0.1058***	−0.1069***	−0.1042***	−0.0707***
	(0.0104)	(0.0107)	(0.0109)	(0.0108)	(0.0103)
Social worker	−0.2948***	−0.2947***	−0.3033***	−0.3020***	−0.2784***
	(0.0106)	(0.0109)	(0.0112)	(0.0111)	(0.0105)
Police officer	−0.4448***	−0.4433***	−0.4558***	−0.4620***	−0.4239***
	(0.0110)	(0.0111)	(0.0113)	(0.0113)	(0.0108)
Constant	10.5761***	10.5675***	10.5232***	10.5140***	10.0555***
	(0.0103)	(0.0112)	(0.0139)	(0.0139)	(0.0290)
N	2263	2263	2263	2263	2263
R²	0.5410	0.5420	0.5481	0.5530	0.6077

Notes:

Standard errors in parentheses.

See note b, Table 2.

In sum, H2 is supported in that there was a significant gender wage gap appearing early in the career. Moreover, part of this wage gap was explained by workplace skill investment, as men were more often awarded with complex jobs requiring substantial on-the-job training. However, this wage gap was not explained by female work interruptions, either directly or indirectly, since the difference in job complexity was not related to the indicators used to capture such interruptions. Finally, training requirements explain only a small part of the wage gap. A statistically significant gender wage gap amounting to 3.3 per cent remained through to the final model.

Job complexity and perceptions of replaceability

Finally, the notion that workplace training renders employees more valuable to the employer and strengthens their bargaining power was examined. Table 4 displays the results from LPM estimations of the likelihood for employees to report that they would be fairly or very difficult for the employer to replace – a situation presumed to reflect an asymmetric dependence relation favouring the employee. As shown in Model 1, women less often than men perceived themselves to be difficult to replace. This gender difference is not explained by parenthood (Model 2), neither by experience or seniority (Model 3), although a longer seniority increases the likelihood of being difficult to replace. Model 4 shows that part of the gender difference in replaceability is explained by job complexity. The reduction of the gender difference between models 3 and 4 was found to be statistically significant (Wald test, p = 0.000). Accounting for work hours, however, does not further reduce the gender difference in replaceability.

Table 4.

The gender difference in perceived low replaceability. Linear probability model.^a

	Model 1	Model 2	Model 3	Model 4	Model 5
Woman	−0.1403***	−0.1405***	−0.1402***	−0.1249***	−0.1246***
	(0.0202)	(0.0202)	(0.0203)	(0.0204)	(0.0204)
Child born after graduation (Ref.)
All children born before grad.		0.0076	−0.0021	−0.0124	−0.0128
All children born before grad.		(0.0374)	(0.0394)	(0.0390)	(0.0390)
No child in household		0.0053	0.0101	0.0052	0.0032
		(0.0227)	(0.0227)	(0.0225)	(0.0227)
Work experience			0.0008	0.0008	0.0008
			(0.0022)	(0.0022)	(0.0022)
Seniority			0.0115*	0.0082	0.0082
			(0.0051)	(0.0050)	(0.0051)
Complex job				0.1356***	0.1342***
				(0.0229)	(0.0230)
Work hours					0.0011
					(0.0020)
Civil engineer (Ref.)
Lawyer	−0.2039***	−0.2039***	−0.1952***	−0.2032***	−0.2056***
	(0.0329)	(0.0329)	(0.0332)	(0.0328)	(0.0331)
Psychologist	−0.0400	−0.0390	−0.0300	−0.0200	−0.0167
	(0.0312)	(0.0321)	(0.0332)	(0.0329)	(0.0335)
Social worker	−0.1727***	−0.1726***	−0.1725***	−0.1682***	−0.1659***
	(0.0320)	(0.0328)	(0.0341)	(0.0337)	(0.0340)
Police officer	−0.3647***	−0.3641***	−0.3726***	−0.2942***	−0.3905***
	(0.0311)	(0.0314)	(0.0322)	(0.0322)	(0.0330)
Constant	0.7471***	0.7439***	0.6873***	0.6553***	0.6092***
	(0.0303)	(0.0331)	(0.0422)	(0.0420)	(0.0956)
N	2274	2274	2274	2274	2274
R²	0.0873	0.0873	0.0898	0.1030	0.1031

Note: ^aSee notes a and b, Table 2.

These findings support H3 regarding the gender difference in replaceability and the impact of job complexity. Hence, due to the placement in jobs involving more initial training, men are likely to acquire a stronger bargaining position than women with similar human capital endowments. Presumably, these differences could have additional feedback effects on gender differences in careers and wages. Again, however, these processes seem unrelated to female work interruptions.

In sum, even in this select sample, significant gender gaps appear in job complexity, wages and perceptions of replaceability. Further, gender differences in wages and replaceability are related to the differential access to complex jobs. However, the findings do not support the notion that child-related work interruptions provide a main mechanism for sorting women into jobs with lower training requirements.

As mentioned above, female work interruptions may also affect workplace skill investments indirectly, through occupational choice. In this stratified sample, men and women are studied under conditions of maximum similarity and the impact of occupation is factored out. However, to better understand the mediating effect of occupational choice, the findings from the stratified sample can be compared to those found in weighted bivariate regressions, in which the share of men and women in each occupation corresponds to actual share in the population (Table A1, Online Appendix). In weighted regressions, the unadjusted gender wage gap amounted to almost 10 per cent, which is more than twice the size of the gap reported above. Meanwhile, however, gender differences in job complexity and replaceability were only slightly higher than those found in the stratified sample. Thus, occupational choice is not a main factor driving the gender difference in job complexity in this sample.

Further analyses considering interactions between gender and occupation (not displayed) strengthen this impression. Indirectly, theoretical accounts linking workplace skill investments to women’s occupational choices suggest that non-traditional choices could have other effects. Presumably, then, a woman who chooses to work in a male-dominated occupation would signal a strong career commitment and therefore employers would have less reason to discriminate against her when investing in employee training. However, the analyses did not support this notion. Gender wage gaps were found to be smaller in the male-dominated occupations than in the female-dominated occupations but the largest gap was found in the gender-integrated occupation of lawyers (cf. Magnusson, 2013; Purcell and Elias, 2004). More to the point, the gender difference in access to complex jobs did not vary significantly by occupation.³

Taken together, these findings suggest that while gender differences in access to workplace training are a potentially important factor for understanding present-day gender inequalities, such differences cannot be easily explained by female work interruptions.

Discussion

The thrust of this article was to explore the notion that gendered workplace skill investments may explain why inequalities in wages and careers persist at a time when women invest heavily in higher education.

A main finding is that even in this select sample, comprising men and women with higher education working in the same occupations, men were more often in complex jobs, requiring at least one year of initial on-the-job training. Also, already a few years after graduation, men had higher wages and a better bargaining position vis-a-vis the employer. At least partly, these gender gaps are explained by men’s better access to initial on-the-job training.

Comparing the above findings to recent studies on the gender wage gap, it seems reasonable to assume that the gender wage gap found here will widen as the careers of the studied men and women progress (c.f. Purcell and Elias, 2004). For example, Boye et al. (2017) report a gross average gender wage gap of 17 per cent in skilled occupations in Sweden for the year 2010. The analysis presented here suggests that workplace skill investments can provide a mechanism by which women and men are sorted into different tracks where training and learning over time will contribute to increasing inequalities even between men and women with similar human capital endowments. Thus, a glass ceiling may be present already early in the career, although its effects may not become apparent until later.

Although the division of work within the household may play a part in these processes (Boye et al., 2017; Magnusson, 2010), so can workplace training and this study points to the importance of further examining the mechanisms behind gender differences in the access to complex jobs. In this study, these differences cannot be linked to female work interruptions. Clearly, measures of observed behaviour may not tell the whole story since work interruptions can be anticipated by the employer and discounted already at the time of hiring. However, also claims about employer expectations must be further scrutinized. It can be argued that if employers and employees can make reasonable predictions about the future – as is implied in the human capital investment perspective – anticipated and actual behaviour should converge. Of course, due to information asymmetry, employers may consider all women at risk and avoid hiring them in jobs requiring substantial training. However, simply assuming that gender differences in labour market outcomes reflect deliberations on female work interruptions is increasingly problematic and this article was designed to put such assumptions to a strong test. It is far from obvious why women in this sample – comprising a young cohort of highly educated men and women working in the same occupations and raised in the Swedish context – would be expected to be less career oriented than men. The argument about employer anticipation is empirically contradicted by a large Swedish field experiment which found no evidence for systematic recruitment discrimination of either mothers or women in general (Bygren et al., 2017) and in the present article, factors that would signal a higher/lower career commitment are not found to affect the gender gap in job complexity. First, choosing a male-dominated occupation does not increase women’s likelihood (relative to men) of having a complex job. Second, although parents with children born before graduation are more likely to be in complex jobs than parents with younger children and non-parents, this is equally true for mothers and fathers and the presence and timing of children does not explain any of the gender gap in training.

The conclusions from this study are limited by the fact that only five occupations were included (but see Grönlund, 2012). Nevertheless, our study suggests that the search for mechanisms explaining gendered workplace skill investments (and ensuing gaps in wages and careers) should move beyond the perspective of gendered human capital investments to focus on how these investments are developed or devalued in the labour market. In particular, we believe more attention should be paid to the organizational context.

Stainback et al. (2010) present a theoretical framework for an organizational approach to stratification involving three main mechanisms, all of which can be relevant for understanding gender differences in job complexity. The first mechanism is the inertial tendencies of organizational structure, logic and practice. Here, cognitive biases relating to stereotypical gender beliefs affect both the perception of competencies and the division of labour (Ridgeway and Correll, 2004). As a case in point, Purcell and Elias (2004) found that gender segregation at the workplace contributed to the gender pay gap among undergraduates at an early stage in their career. The second mechanism is the relative power of actors within workplaces. Here, several researchers have pointed to social closure processes, by which male employees monopolize skill and authority, and exclude women from opportunities for on-the-job training (Tomaskovic-Devey and Skaggs, 2002; cf. Cockburn, 1991). Finally, institutional and competitive pressures from the organizational environments constitute a third mechanism which can produce both stability and change. For example, modern organizational models, emphasizing flexibility and the integration of functions, are discussed as a challenge to gender segregation and stereotypical gender typing of tasks, although outcomes are uncertain (Abrahamsson, 2014). Apart from these broad mechanisms, the study of job complexity also requires a focus on the work content per se. Here, Chan and Anteby (2016) have observed that task segregation might produce inequality even within the same job. For example, the fact that women are more often assigned with tasks with low promotability (Babcock et al., 2017), doing the ‘housework’ of organizations (Heijstra et al., 2017) may hamper their access to more complex tasks. In future research, these theoretical perspectives should be applied not only in qualitative case studies but also in quantitative studies using both individual and workplace level data to study skill development and wages.

All in all, it is notable that women and men with equal educational attainments are sorted into jobs with different requirements for initial training. If this sorting entails radically different wage and career trajectories, workplace skill investments could represent a ‘glass ceiling’ which women encounter immediately after graduation.

Footnotes

Funding

For financial support we gratefully acknowledge the Swedish Research Council for Health, Working Life and Welfare (FORTE) grant number 2011-0816.

Supplementary material

Supplementary material is available for this article online.

Notes

Katarina Boye is Associate Professor of Sociology at the Swedish Institute for Social Research, Stockholm University, Sweden. She studies the intersection between family and labour market and focuses particularly on division of work between men and women and its labour market consequences.

Anne Grönlund is Professor of Sociology, employed at the Department of Social Work, Umeå University, Sweden. Her research concerns a range of labour market issues as well as work–family reconciliation, with a focus on gender and gender segregation. Important themes include flexibility and autonomy at work, work–family conflict, skill development and wages.

References

Abrahamsson

(2014) Gender and the modern organization, ten years after. Nordic Journal of Working Life Studies 4(4): 109–36.

Babcock

Recalde

Vesterlund

Weingart

(2017) Gender differences in accepting and receiving requests for tasks with low promotability. American Economic Review 107(3): 714–47.

Becker

(1964/1993) Human Capital, 3rd Edition. Chicago, IL: University of Chicago Press.

Becker

(1991) A Treatise on the Family. Cambridge, MA; London: Harvard University Press.

Boye

Halldén

Magnusson

(2017) Stagnation only on the surface? The implications of skill and family responsibilities for the gender wage gap in Sweden, 1974–2010. British Journal of Sociology. Epub ahead of print 31 March 2017. DOI: 10.1111/1468-4446.12252.

Bygren

Erlandsson

Gähler

(2017) Do employers prefer fathers? Evidence from a field experiment testing the gender by parenthood interaction effect on callbacks to job applications. European Sociological Review 33(3): 337–48.

Chan

Anteby

(2016) Task segregation as a mechanism for within-job inequality: women and men of the transportation security administration. Administrative Science Quarterly 61(2): 184–216.

Cockburn

(1991) In the Way of Women. Men’s Resistance to Sex Equality in Organizations. London: MacMillan.

Dale

(1987) Occupational inequality, gender and life-cycle. Work, Employment and Society 1(3): 326–51.

10.

Duncan

Hoffman

(1979) On-the-job training and earnings differences by race and sex. The Review of Economics and Statistics 61(4): 594–603.

11.

Duvander

Viklund

(2014) Kvinnors och mäns föräldraledighet. In: boye

Nermo

(eds) Lönsamt arbete – familjeansvarets fördelning och konsekvenser, SOU 2014:28. Stockholm: Fritzes, 23–61.

12.

Edlund

Grönlund

(2008) Protection of mutual interests? Employment protection and skill formation in different labour market regimes. European Journal of Industrial Relations 14(3): 245–64.

13.

Edlund

Öun

(2016) Who should work and who should care? Attitudes towards the desirable division of labour between mothers and fathers in five European countries. Acta Sociologica 59(2): 151–69.

14.

England

(2010) The gender revolution: uneven and stalled. Gender & Society 24(2): 149–66.

15.

Estévez-Abe

(2006) Gendering the varieties of capitalism. A study of occupational segregation by sex in advanced industrial societies. World Politics 59(1): 142–75.

16.

Gronau

(1988) Sex-related wage differentials and women’s interrupted careers – the chicken or the egg. Journal of Labor Economics 6(3): 277–301.

17.

Grönlund

(2012) On-the-job training – a mechanism for segregation? Examining the relationship between gender, occupation and on-the-job investments. European Sociological Review 28(3): 408–20.

18.

Grönlund

Magnusson

(2013) Devaluation, crowding or skill specificity? Exploring the mechanisms behind the lower wages in female occupations. Social Science Research 42(4): 1006–17.

19.

Heijstra

Steinthorsdóttir

Einarsdóttir

(2017) Academic career making and the double-edged role of academic housework. Gender and Education 29(6): 764–80.

20.

Kohn

Schooler

(1978) The reciprocal effects of substantive complexity of work and intellectual flexibility: a longitudinal assessment. American Journal of Sociology 84(1): 24–52.

21.

Korpi

(2000) Faces of inequality: gender, class and patterns of inequalities in different types of welfare states. Social Politics 7(2): 127–91.

22.

Le Grand

Tåhlin

(2013) Class, occupation, wages and skills: the iron law of labour market inequality. Comparative Social Research 30: 3–46.

23.

Leuven

(2005) The economics of private sector training: a survey of the literature. Journal of Economic Surveys 19(1): 91–111.

24.

Lundberg

Pollak

(1996) Bargaining and distribution in marriage. The Journal of Economic Perspectives 10(4): 139–58.

25.

Magnusson

(2010) Why is there a gender wage gap according to occupational prestige? An analysis of the gender wage gap by occupational prestige and family obligations in Sweden. Acta Sociologica 53: 99–117.

26.

Magnusson

(2013) More women, lower pay? Occupational sex composition, wages and wage growth. Acta Sociologica 56(3): 227–45.

27.

Mood

(2010) Logistic regression: why we cannot do what we think we can do, and what we can do about it. European Sociological Review 26(1): 67–82.

28.

OECD (2003) Employment Outlook. Paris: OECD.

29.

OECD (2004) Employment Outlook. Paris: OECD.

30.

OECD (2014) Education at a Glance 2014: OECD Indicators. OECD Publishing. Available at: https://dx-doi-org.web.bisu.edu.cn/10.1787/eag-2014-en.

31.

Osterman

(ed.) (1984) Internal Labor Markets. Cambridge, MA; London: MIT Press.

32.

Perales

(2013) Occupational sex-segregation, specialized human capital and wages: evidence from Britain. Work, Employment and Society 27(4): 600–20.

33.

Polachek

(1981) Occupational self-selection: a human capital approach to sex differences in occupational structure. The Review of Economics and Statistics 63(1): 60–9.

34.

Polavieja

(2008) The effect of occupational sex-composition on earnings: job specialization, sex-role attitudes and the division of domestic labour in Spain. European Sociological Review 24(2): 199–213.

35.

Purcell

Elias

(2004) Researching graduate careers seven years on. A research project jointly funded by the Economic and Social Research Council and the Higher Education Careers Services Unit, Research paper no. 4, Higher Education and Gendered Career Development.

36.

Ridgeway

Correll

(2004) Unpacking the gender system: a theoretical perspective on gender beliefs and social relations. Gender & Society 18(4): 510–31.

37.

Stainback

Tomaskovic-Devey

Skaggs

(2010) Organizational approaches to inequality: inertia, relative power, and environments. Annual Review of Sociology 36: 225–47.

38.

Tåhlin

(2007a) Class clues. European Sociological Review 23(5): 557–72.

39.

Tåhlin

(2007b) Skills and wages in European labour markets. In: Gallie

(ed.) Employment Regimes and the Quality of Work. Oxford: Oxford University Press, 35–76.

40.

Tåhlin

(2011) Vertical differentiation of work tasks: conceptual and measurement issues. Empirical Research in Vocational Education and Training 3(1): 55–70.

41.

Tomaskovic-Devey

Skaggs

(2002) Sex segregation, labor process organization and gender earnings inequality. American Journal of Sociology 108(1): 102–28.

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.

0.00 MB

0.09 MB