From childhood to young adulthood: the importance of self-esteem during childhood for occupational achievements among young men and women

ABSTRACT

This study investigates the impact of self-esteem during childhood on men’s and women’s occupational prestige in young adulthood. By combining first-hand information from parents in the Swedish Level-of-Living surveys (LNU) 2000 and their children in the Child-LNU in 2000 and the follow-up study in LNU-2010, we are able to assess how self-esteem during adolescence is related to occupational prestige in adulthood. Multivariate analyses were used to determine whether associations between self-esteem (global and domain-specific) in childhood (aged 10–18 years) and occupational prestige in young adulthood (aged 20–28) exist and, if so, what the magnitudes of these associations are for each respective gender.

For women, there is a positive association between confidence in mathematics and prestige, even when accounting for actual math grades. Global self-esteem is positively related to later occupational prestige as well. For men, self-esteem is unrelated to occupational prestige. Only actual performance in mathematics is important for men’s occupational achievements.

These results indicate the importance of taking gender differences into account when investigating how self-esteem is related to outcomes in young adulthood. A possible implication is the importance of focusing on the development of self-esteem among children, particularly girls, in school.

KEYWORDS:

• Gender
• labour market
• self esteem
• youth work
• career

Introduction

Social background is noted to be strongly related to an individual’s destination in the social structure (e.g. Breen 2004; Jonsson et al. 2009). However, although social background accounts for a vital part of the social stratification in the labour market, a substantial part remains unexplained (Bowles, Gintis, and Osborne 2001). In fact, there are other aspects during childhood that have an impact on labour market achievements later in life. Much of the previous literature focuses on the link between cognitive skills during adolescence and adult outcomes (e.g. Murnane, Willett, and Levy 1995; Duckworth et al. 2012), but some studies also focus on the link between non-cognitive skills (such as self-esteem, sociability and communication skills) and labour market outcomes (e.g. Bowels and Gintis 2002; Harris 1940). Most prior research find strong positive associations between high self-esteem and success later in life in aspects such as work, social relations and health (Orth and Robins 2014). For instance, previous research indicates that e.g. self-esteem (Waddell 2006) and self-efficacy (Chowdry, Crawford, and Goodman 2011) during adolescence are important for labour market outcomes in adult life. Furthermore, non-cognitive skills have also been suggested as important for understanding gender inequality in the labour market (e.g. Braakmann 2009; Fortin 2008; Mueller and Plug 2006). Compared to men, women tend to have lower self-esteem (Feingold 1994; Kling et al. 1999), which may have an impact on gender differences in labour market outcomes. To increase our understanding of the unexplained part of this stratification according to both gender and social background, it is therefore important to investigate how non-cognitive skills (here measured by self-esteem, both global and domain specific) during childhood are related to labour market outcomes for young men and women.

Given that society is stratified by gender where cultured beliefs about gender affect men’s and women’s behaviours (e.g. Correll 2001; Ridgeway 1997; Wood and Eagly 2012), the importance of non-cognitive aspects may vary by gender. For instance, stereotyped assumptions about male and female characteristics are often put forward as being important for gendered occupational choices, ultimately resulting in a gender-segregated labour market both in the horizontal and vertical levels (Wood and Eagly 2012 cf. Charles and Grusky 2004). For instance, that skills in mathematic tend to be considered as ‘male’ skills have been put forward as one explanation to women’s underrepresentation in STEM-field (including science, technology, engineering and mathematics) (Correll 2001). Accordingly, there could be both gender differences in the distribution of non-cognitive skills, and/or the importance of non-cognitive skills for labour market outcomes may vary by gender. However, to our knowledge, few studies have assessed gender differences in the associations between non-cognitive skills in childhood and labour market success in adulthood.

Furthermore, longitudinal studies including rich information about both children and parents are rare in this field. In particular, there is a lack of knowledge on how aspects such as childhood self-esteem and self-perceptions are related to labour market outcomes in adulthood (cf. Gutman and Schoon 2013; Sjögren and Svaleryd 2011). The present paper attempts to partially fill this void by investigating whether associations between self-esteem in childhood and occupational prestige outcomes in adulthood exist and if so, what the magnitudes of these associations are for each respective gender.

A prerequisite for fulfilling this purpose is access to extensive longitudinal information on respondents’ living conditions both in childhood and young adulthood as well as access to information on their parents. By combining parental information from the Swedish Level of Living Survey 2000 (LNU 2000), childhood information from the Child-LNU 2000 (aged 10–18 years), and the LNU 2010 extension of young adults aged 20–28 years, we have rich longitudinal data that are well-suited for this purpose. Thus, unlike most previous studies, we are able to account for the children’s social background and their educational performance (school grades) – two factors of major importance for their social destination – and to link self-esteem in childhood to later labour market outcomes for a representative sample of Swedish boys and girls. The relationship between self-esteem and important life outcomes is debated in previous literature. Thus, it is not clear-cut whether self-esteem is a cause or a consequence of, for example, career success (e.g. Heckman, Stixrud, and Urzua 2006; Orth, Robins, and Widamen 2012). One of the strengths of the present study is that the information on self-esteem is measured before the labour market outcome.

Social stratification and non-cognitive skills

It is well-known that parents’ socioeconomic position influences their children’s educational level and occupational achievement (e.g. Breen and Jonsson 2005). However, even when assessing both social background and school performance, a large part of the social stratification is unknown. (e.g. Bowles, Gintis, and Osborne 2001). Thus, social background and educational performance is important but fail to fully explain stratification in the labour market – not least in relation to gender inequality. Previous research also highlights other aspects such as cognitive ability (e.g. Blanden, Gregg, and Macmillan 2007; Bihagen, Nermo, and Stern 2013; Carvalho 2012) and social and cultural capital (cf. Jǽger & Holm 2007) as important for understanding labour market outcomes. Furthermore, of central interest here is that recent studies indicate that non-cognitive skills are important determinants of labour market outcomes (e.g. Bihagen, Nermo, and Stern 2013; Heckman, Stixrud, and Urzua 2006; Jackson 2006; Lindqvist & Vestman 2011; Waddell 2006).

The definition of non-cognitive skills is broad and often defined as skills that are normally not captured in the notion of cognitive traits (Garcia 2014; Gutman and Schoon 2013). It is common to see cognitive and non-cognitive skills as being related to each other, where non-cognitive skills are seen as important in the development of human capital. Thus, ‘ … the formation of human capital requires, or is at least helped by, personal characteristics such as motivation, self-discipline, communication skills, energy, impulse control, perseverance, sociability, confidence, self-esteem, decisiveness, grit’ (Joshi 2014, 4). In a similar way, Gutman and Schoon (2013) refer to non-cognitive skills as a set of attitudes, behaviours, and strategies considered to underpin success in school or at work.

Self-esteem and labour market outcomes

Self-esteem is a broad concept¹ that can be defined and measured in many ways, both as a global measurement (c.f. Judge and Bono 2001; Rosenberg 1965) and as domain-specific aspects (Kling et al. 1999; Orth, Robins, and Widamen 2012). Global self-esteem refers to the overall subjective value – positive or negative – that individual’s place on themselves (Judge and Bono 2001; Orth and Robins 2014; Rosenberg 1965). Domain-specific aspects instead contain a person’s self-evaluation in many different areas such as satisfaction with appearance, social skills and academic competence (Kling et al. 1999; Rosenberg 1979). While global self-esteem could be seen as the overall value of the ‘self’, specific self-esteem could be describes as a myriad of self-related thoughts (Baumeister et al. 2003). Thus, persons can perceive themselves as talented in maths and at the same time they rate themselves low on overall self-esteem. Whether to use global or specific self-esteem (or both in combination) is well debated in previous literature, but there is no consensus in this question (see e.g. Rosenberg et al. 1995; Swann, Chang-Schneider, and Larsen McClarty 2007). Even so, studies indicate that global self-esteem and domain-specific aspects tend to be overlap but are not interchangeable (e.g. Marsh and O’Mara 2008; Rosenberg et al. 1995). For instance, prior research has shown a positive association between general self-esteem and attitudes towards appearance (Sondhaus, Krutz, and Strube 2001). Related to our focus here, Marsh and O’Mara (2008) did not find any clear relationship between global self-esteem and school performance (in line with Baumeister et al. 2003), but they found positive associations between academic self-concept (domain specific self-esteem) and educational outcomes. Marsh and O’Mara (2008) do not assert global self-esteem to be irrelevant but point out the importance of investigating academic self-concept when focusing on educational outcomes. Likewise, Rosenberg et al. (1995) show academic self-esteem to influence school performance more than global self-esteem.

Here, we focus on both global self-esteem and two domain-specific aspects: satisfaction with appearance and academic self-esteem to further our understanding of the association between self-esteem in adolescence and occupational achievement in young adulthood.

Previous studies indicate that self-esteem is important for both education (e.g. Waddell 2006) and labour market outcomes (e.g. Heckman, Stixrud, and Urzua 2006; Waddell 2006). Most previous longitudinal studies indicate that self-esteem is a predictor of, and not a consequence of, life success (Orth and Robins 2014, for a divergent view see Baumeister et al. 2003). Both Heckman, Stixrud, and Urzua (2006) and Borghans et al. (2008) found empirical evidence for a causal association between self-esteem and earnings. Likewise, Drago (2011) found a large and positive effect of self-esteem on earnings in the US. Another longitudinal study by Waddell (2006) showed that high school students with negative attitudes and poor self-esteem attained fewer years of postsecondary education, had lower earnings and were less likely to be employed at a 14-year follow-up. Additionally, Murnane et al. (2001) found self-esteem in high school to be significantly related to wages a decade later even when controlling for academic skills, based on data from the US. Likewise, Chowdry, Crawford, and Goodman (2011) longitudinal study in the UK showed that children’s beliefs in their own study-related capabilities were positively correlated to their level of education later in life. Overall, many studies indicate that children with low self-esteem are less successful in school (Mann et al. 2004).

Moreover, self-esteem also seems to be positively associated with aspirations (e.g. Gottfredsson 1981; Judge, Bono, and Locke 2000). This is naturally important for future labour market outcomes. Related to this, additional studies indicate that people with higher levels of self-esteem are more likely to apply for high-paying jobs (Heckman, Stixrud, and Urzua 2006) or more challenging jobs (Judge, Bono, and Locke 2000).

Taken together, it is reasonable to assume that both global and domain-specific self-esteem – but academic self-esteem in particular – have a significant impact on future occupational achievements.

Gender differences in self-esteem

There is a widespread idea that women score substantially lower than men on self-esteem. Several possible hypotheses have been raised to explain girls’ (real or assumed) lower self-esteem such as the following: gender roles in society where men are seen as assertive and women as communal (e.g. Wood and Eagly 2012), different treatment in school, peer interactions and women’s negative perceptions of their physical appearance and higher risk of being judged based on their physical appearance (Kling et al. 1999). Empirically, support for the notion regarding girls’ lower self-esteem compared to boys is however inconclusive. Some studies indicate that boys have higher self-esteem than girls (Brindorf et al. 2005; Feingold 1994; Kling et al. 1999), whereas other studies indicate no or small gender differences (e.g. Major et al. 1999) or that gender differences change across the life span (Robins et al. 2002). Studies on gender differences in domain-specific self-esteem, in particular satisfaction with physical appearance, do however report that girls in general have greater body dissatisfaction compared to boys (Jones and Crawford 2006; Kling et al. 1999). Women/girls also tend to be judged based on their physical appearance to a higher degree than men/boys (Jones and Crawford 2006).

Keller et al. (2015) found self-esteem to be positively correlated with obtaining high-quality jobs, but this relationship was only true for women. Keller et al. (2015) interpreted their results as being linked to gender stereotypes in society, where it was more important for women than men to be ‘liked’ by others. Thus, according to the gendered beliefs in society, men are stereotypically seen as competitive and dominant, while women are seen as nurturing and socially competent (Keller et al. 2015; Wood and Eagly 2012). In addition, as competitiveness and assertiveness often are valuable skills in prestigious occupations, they are more compatible with stereotypes about men (e.g. Keller et al. 2015; Wood and Eagly 2012 cf. Charles and Grusky 2004). Accordingly, women need self-esteem to overcome barriers to both apply for and hold higher positions in society that conform to stereotypes about men.

Gender differences in self-esteem and prestigious occupations

It is well known that the proportion of women in higher education has increased dramatically over the past decades (Bradley 2000). However, the share of women choosing to pursue so called STEM fields (science, technology, engineering and mathematics) is still far behind men (Beede et al. 2011). In general, graduation from a STEM field tends to generate occupational achievements (Beede et al. 2011 cf. Bihagen, Nermo, and Stern 2013). In particular is academic success in mathematics important for reaching higher position in the labour market (cf. ESA 2011; Rose and Betts 2001). A common explanation for gender differences in the STEM fields is different attitudes (or interest) about mathematics among boys and girls. Girls internalize gendered beliefs in society where skills in mathematics are seen as a ‘male’ qualification (Correll 2001; Hyde et al. 1990; Salikutluk and Heyne 2017). While gender differences in mathematical achievements show no clear pattern (i.e. it seems to vary between country and studies), gender difference in non-cognitive skills seem to be more stable. In general boys score higher than girls on self-evaluations in mathematical skills (Else-Quest, Hyde, and Linn 2010). Correll (2001) shows that in a comparison between women and men with equal math grades, men are more likely than women to assess themselves as competent in maths. They only do so in math and not in verbal skills. Overall, there are stereotypes regarding ‘brilliance’ (e.g. mathematics and physics) as a male characteristic that ultimately may result in women avoiding prestigious occupations (Bian, Leslie, and Cimpian 2017).

Having cognitive skills in math is most likely important for ending up in a highly prestigious occupation. Likewise, mathematic self-beliefs (e.g. how the students evaluate their own skills in mathematics) seem to be an important predictor of success in mathematics (e.g. Pitsia, Biggart, and Karakolidis 2017). Accordingly, as prior research has found self-esteem in academic ability, particularly self-esteem in math, to be important for educational outcomes (e.g. Marsh and O’Mara 2008; Pitsia, Biggart, and Karakolidis 2017; Rosenberg et al. 1995), it is reasonable to assume that both cognitive and non-cognitive skills in math are important for ending up in a highly prestigious occupation (cf. Pitsia, Biggart, and Karakolidis 2017). As prior research has shown that girls have lower self-confidence in math than boys (Correll 2001; Else-Quest, Hyde, and Linn 2010; Hyde et al. 1990), which may be one cause for women’s underrepresentation in occupations that require high skills in math.

Here, we account for both grades in math and self-esteem in math (academic self-concept), which provide us with clues about whether self-esteem in math is related to later occupational outcomes after also accounting for performance in math. In addition, if this is the case, we explore whether this varies between boys and girls.

In summary, international studies indicate that self-esteem during childhood is related to future labour market outcomes, and to our knowledge, there are no previous Swedish longitudinal studies that focus on this association and include self-reported indicators of self-esteem in youth. There is also a lack of both national and international studies that consider gender differences in these associations. It is reasonable to assume that it is especially important to study self-esteem during adolescence as these years are critical in the identity formation process (Kling et al. 1999). Furthermore, many important decisions about educational fields and occupations are made during these years. Based on previous empirical findings, we hypothesize that both global and domain-specific self-esteem is positively related to occupational prestige, especially for women.

Data and method

The LNU surveys are based on a representative sample of Swedish adults between 18 and 75 years of age and comprise both cross-sectional and panel data. The surveys have been conducted six times since 1968. The LNU 2000 included an extension referred to as the Child-LNU,² where the respondents’ children (aged 10–18 years) were also interviewed. In 2010, the children in the first wave of the Child-LNU were interviewed again as young adults, aged 20–28 years. By utilizing data from the LNU surveys in 2000 and 2010 and the extension with young adults, we have access to unique longitudinal (self-reported) information on living conditions in both childhood and young adulthood, and it is thus possible to study the extent to which these conditions are related to labour market outcomes in adulthood.

The Child-LNU 2000 consists of 1304 respondents (response rate: 85.3 per cent, of which 929 also participated in the LNU 2010 (response rate: 72 per cent).³ Of these participants, 478 individuals could be classified as holding an occupation in 2010. Respondents with missing data on any of the included variables were excluded from the analyses (i.e. listwise deletion). Thus, our final sample used in the analyses consisted of 478 individuals aged 20–28 years in 2010. As the data included individuals who lived in the same household in adolescence (i.e. siblings and step-siblings), robust standard errors (clustered at the household level) were estimated in the multivariate regressions.

Variables

Dependent variable

The Standard International Occupational Prestige Scale (SIOPS) (Treiman 1977) is used as the dependent variable. The dependent variable is a continuous variable ranging from 13 to 78. Occupational prestige captures multiple dimensions of the occupational structure and can be viewed as a metric of a structural order of occupations according to their power and control over valued recourses and rewards in society (Treiman 1977). Occupational prestige is correlated with both earnings and education. Thus, high prestige scores are used here as a proxy for successful labour market outcomes.⁴ Physicians, College, university and higher education teaching professionals, Legal professionals and Physicists, chemists and related professionals are examples of high prestigious occupations included in the sample, while Garbage collectors and related labourers, Personal care and related workers, Helpers and cleaners exemplifies occupations in the lower part of the prestige distribution in the present sample The respondents in the sample are young, and some of them are still in higher education. To increase the sample, we also include respondents in higher education who clearly indicate a professional track such as nursing, medicine and law, i.e. these respondents have been assigned the prestige value in which their education will result.

Independent variables

The above SIOPS is also used as an independent variable measuring parental occupational prestige. Here, SIOPS is used as a categorical variable and distinguishes between the following: (i) Both parents in highly prestigious occupations (prestige > 50), (ii) Father in highly prestigious occupation, (iii) Mother in highly prestigious occupation, and (iv) none of the parents in highly prestigious occupations.⁵ Thus, both fathers’ and mothers’ occupations are taken into account, which is a strength. Previous studies on intergenerational transmission of occupational status have paid more attention to the importance of fathers’ occupations than mothers’ occupations (cf. Korupp, Sanders and Ganzeboom 2002). However, studies show that mothers’ occupational belonging is also important, not least for daughters’ occupational achievements. Given that both parents in most cases are equally active in the labour market in Sweden and women have higher levels of education on average, it is reasonable to also account for the mothers’ occupational prestige when studying young adults’ occupational achievements. Parent’s prestige has, as a robustness check, also been used as a continuous variable ranging from 13 to 78, but this did not change the findings in any substantial way, see Table A1 in Appendix. The exception is that the relatively strong positive association of having both parents in high prestige occupations is less visible when using prestige as a continuous variable.

Self-esteem

The measurements of self-esteem below are based on self-reported information. There are naturally both pros and cons with using self-reported answers from children. However, as children are the main informants of their own lives, information provided by the children themselves could be considered the most valid (c.f. Nordlander 2015).

Most of the questions regarding self-esteem directed to the children are based on the self-assessment scale, ‘I think I am’ (c.f. Ouvinen-Birgerstam 1985). Global self-esteem is measured with one indicator that aims to capture the overall value the child gives oneself as a person (cf. Rosenberg 1965). The measurement is based on the item, ‘I am mostly satisfied with myself’, which has four response options: matches exactly, matches roughly, matches poorly and does not match at all. Here, the variable is used as a dummy where 1 = matches exactly.

Domain-specific self-esteem is captured by two indicators, physical appearance and confidence in math (cf. Kling et al. 1999). Physical appearance is measured by satisfied with my appearance (item ‘I am satisfied with my appearance’). The item has the same response options as the one above and is also used as a dummy where 1 = matches exactly. As a robustness check has regressions including self-esteem measured with the above scale (four items) instead of a dummy variable also been made with similar results as those presented here, see Table A1 in Appendix.

Confidence in math is based on the statement ‘I think I am good at math’, which is considered here as an indicator of having a positive self-perception of one’s own abilities in math (c.f. Gutman and Schoon 2013). As a robustness check has regressions including confidence in math as a scale (1–5) instead of a dummy variable also been made with similar results as those presented here, see Table A1 in Appendix.

We also include a dummy for high grade in math, where children with the highest grades in math (pass with special distinction, MVG) are coded as one. The grades awarded are fail, pass, pass with distinction and pass with special distinction.

All models control for age, divided into three categories: 20–22 years, 23–25 years, and 26–28 years.

Results

Table 1 below presents descriptive statistics by gender. Among boys, half of the sample participants agree with the statement ‘mostly satisfied with myself’ (global self-esteem). The proportion is somewhat smaller among girls – 43 per cent. Even if the proportion that agrees with the above statement is somewhat smaller among girls than boys, a large share of children thinks that the statements match them exactly. Thus, in general, both boys and girls have rather good self-esteem when measured by this standard. However, the satisfaction with physical appearance is somewhat larger among boys (49 per cent) than among girls (38 per cent).

Table 1. Descriptive statistics, mean/percent, (sd) in brackets.

	All:478	Boys n:235 (49.2)	Girls n:243 (50.8)	Gender diff. p-value
Both parents high prestige	22.4	23.0	21.8	0.759
Father high prestige	15.9	16.2	15.6	0.874
Mother high prestige	13.4	14.4	12.3	0.508
Both parents low prestige	48.3	48.5	48.1	0.937
Age	23.7 (2.6)	23.7 (2.6)	23.6 (2.6)	0.442
Like appearance (physical appearance)	43.7	49.4	38.3	0.015
Like myself	46.4	49.8	43.2	0.149
Confidence in math (I think I am good at math)	30.3	31.9	28.8	0.460
High grade in math	15.3	15.3	22.2	0.054
In adulthood (LNU 2010)
Prestige	44.7 (15.6)	45.5 (15.6)	43.9 (15.6)	0.113

***p < 0.001; **p < 0.01; *p < 0.05.

Approximately 22 per cent of the girls have the highest grade in math. The corresponding figure for boys is approximately 15 per cent. Despite this difference, there is no gender difference in the proportion that states that they are good at math. Thus, among both boys and girls, the share that states that they are good at math is approximately one-third. The proportion that indicates that they are good at math is slightly larger among boys (albeit not statistically significant). This mean difference indicates that girls have lower confidence in math despite the fact that they on average have higher math grades.

Regarding parental background, approximately 50 per cent have both parents in occupations with low prestige (below SIOPS 50), while approximately 20 per cent have both parents in highly prestigious occupations. Approximately 15 per cent of the children have only a father in a highly prestigious occupation, and approximately 14 per cent have only a mother in a highly prestigious occupation. Regarding the dependent variable of occupational prestige measured in young adulthood, men in general have a slightly higher prestige level, but this gender difference is insignificant.

Table 2 show a correlations matrix of all variables. Among the measurements of self-esteem, it is only the measurement of having confidence in math that is positively related to prestige. The general measurement of global self-esteem is almost uncorrelated to prestige. The indicator of satisfaction of physical appearance is weakly negatively related to prestige. Moreover, all self-evaluations are positively related to each other. The correlation between ‘like myself’ and ‘like my appearance’ is particularly strong. There is also, in line with previous research, a positive correlation between confidence in math and high grade in math (e.g. Pitsia, Biggart, and Karakolidis 2017). Parent’s occupational prestige is as expected, in accordance with theories of social stratifications, related to offspring’s prestige. Having both parents in high prestige occupations is positively correlated to own occupational prestige in young adulthood, while having both parents in low prestige occupations is negatively related to own occupational prestige. Likewise is there a positive correlation between high grade in math and social background.

Table 2. Correlations n: 478.

	Prestige	Father high prestige	Mother high prestige	Both parents low prestige	Both parents high prestige	Like myself	Confidence in math	High grade in math	Like appearance	Age
Prestige	1.00
Father high prestige	0.02	1.00
Mother high prestige	0.07	−0.17	1.00
Both parents low prestige	−0.21	−0.42	−0.38	1.00
Both parents high prestige	0.17	−0.23	−0.21	−0.52	1.00
Like myself	0.02	0.00	−0.01	0.01	0.00	1.00
Confidence in math	0.13	0.01	0.03	−0.06	0.03	0.12	1.00
High grade in math	0.29	0.07	0.05	−0.18	0.11	0.01	0.36	1.00
Like appearance	−0.06	0.02	−0.04	0.07	−0.07	0.48	0.11	0.00	1.00
Age	0.09	−0.06	−0.02	0.01	0.06	−0.25	−0.13	−0.05	−0.34	1

In the next step we take a closer look at the associations between self-esteem in childhood and occupational prestige in young adulthood, and what the magnitudes for these associations are for each respective gender, using OLS regressions. Table 3 below shows the association between conditions in childhood and occupational prestige in young adulthood in separate models for men and women. Model 1 in Table 3 indicate as expected that social background accounts for a vital part of the social stratification in the labour market. Having both parents in highly prestigious occupations during adolescence is clearly positively associated with occupational prestige in adulthood for both men and women (see Model 1). Thus, these children have on average higher prestige in young adulthood than children with parents in occupations of low prestige. For men there is also a significant positive association between having only a father (or a mother) in a highly prestigious occupation and their own occupational prestige in adulthood. The corresponding estimates for women only indicate a significantly positive association for having a mother in a highly prestigious occupation. Thus, having only a father in a highly prestigious occupation does not seem to be significantly related to women’s occupational prestige in young adulthood.

Table 3. Coefficients from OLS regression models predicting occupational prestige.

Model^a	1		2		3		4		5
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
Both parents high prestige	7.17**	9.36***	7.34**	9.28***	5.95*	7.46**	6.13*	9.41***	5.20*	7.56*
	(−2.49)	(2.57)	(−2.41)	(2.61)	(−2.41)	(2.55)	(2.43)	(2.75)	(2.37)	(2.71)
Father high prestige	2.78	5.59*	2.04	5.59*	0.30	4.92	2.29	5.57	−0.05	5.00
	(−2.92)	(3.29)	(−2.75)	(2.85)	(2.62)	(2.70)	(2.90)	(2.85)	(2.98)	(2.65)
Mother high prestige	6.96*	5.52	6.91*	5.44	6.13*	3.79	5.92*	5.50	5.34	3.87
	(−3.01)	(3.29)	(−3.05)	(3.34)	(3.09)	(3.36)	(2.92)	(3.34)	(2.98)	(3.48)
Confidence in math			8.90***	0.62	6.25**	−3.36	–	–	5.90*	−3.41
			(−2.25)	(2.19)	(2.34)	(2.36)	–	–	(2.32)	(2.10)
High grade in math					8.95***	13.58***	–	–	8.73***	13.65*
					(2.61)	(2.65)	–	–	(2.58)	(2.57)
Like myself							5.48*	−0.45	3.40	0.54
							(2.22)	(2.49)	(2.32)	(2.36)
Like appearance							−4.24	1.35	−3.76	0.94
							(2.41)	(2.52)	(2.29)	(2.37)
Constant	42.02***	42.56***	40.47***	42.42***	40.30***	41.60***	42.04***	42.34***	40.80***	41.12***
	(2.05)	(1.95)	(2.04)	(2.02)	(2.04)	(1.90)	(2.18)	(2.09)	(2.21)	(1.98)
R-squared	0.08	0.06	0.14	0.06	0.19	0.14	0.10	0.06	0.20	0.14
N	243	235	243	235	243	235	243	235	243	235

***p < 0.001; **p < 0.01; *p < 0.05.

Robust standard errors (clustered at the household level) are used, displayed in brackets.

^aAll models control for age.

In Model 2, we add a control for confidence in math. The estimates indicate a strongly significant, positive association between confidence in math and occupational prestige in young adulthood among women. Thus, having confidence in math when growing up is important for women’s future labour market outcomes. The coefficients for social background (i.e. parental occupational prestige) are more or less unaffected when introducing confidence in math in Model 2. Thus, differences in confidence in math do not seem to account for the association between social background and one’s own occupational prestige.⁶ In Model 3, we add grade in math, which also seems to be significantly and positively related to occupational prestige. Even more interesting is that the effect of confidence in math is still strong and positive. Thus, having confidence in math when growing up is important for women’s occupational prestige in adulthood, even when accounting for actual math grades.

The coefficients in Model 2 among men indicate that confidence in math, given social background, seems to be unrelated to men’s occupational prestige in young adulthood. This is rather surprising as confidence in math seems to be one of the most important factors for occupational prestige in adulthood among women.⁷ Actual math grades are, however, strongly positively related to men’s occupational attainment (Model 3). The associations between parents’ prestige and one’s own prestige decrease when accounting for grades in math for both genders, and having only a mother in a highly prestigious occupation is no longer statistically significant among men. The fact that the associations decrease between parental prestige and offspring’s prestige indicates that some of the effects between social background and occupational prestige are mediated by grades in math.

In Model 4, we add controls for global self-esteem and satisfaction with physical appearance. The presented findings indicate that being overall satisfied with oneself in childhood is positively related to occupational prestige in adulthood but only for women. In contrast to global self-esteem, satisfaction with appearance seems to be significantly and negatively related to women’s occupational prestige. This indicates that girls who are happy with their looks when growing up have lower occupational prestige as young adults than girls who are less happy with their appearance. This finding should be interpreted with some caution. Additional tests that are not presented here indicate that this effect is driven by a few girls with high occupational prestige who were really dissatisfied with their appearance in childhood. The associations between parental prestige and women’s prestige decrease some when accounting for global self-esteem, suggesting that self-esteem may mediate part of the association between social background and occupational prestige. Among men, neither global self-esteem nor satisfaction with physical appearance is significantly related to occupational prestige.

When all measurements of self-esteem and grades in math are included in the same model (Model 5), both global self-esteem and satisfaction with physical appearance become insignificant for women. This suggests that math grades may mediate the association between self-esteem and women’s later occupational prestige. The positive influence of having confidence in math, however, remains significant. Thus, for girls, confidence in math is important for future occupational attainment even after controlling for global self-esteem, social background and grades in math.

The positive association between high grades in math and occupational prestige is still strong among men in the full model as well. Thus, for men, actual performance in school seem to be very important for occupational achievement.⁸

Discussion

Prior research finds a positive association between non-cognitive skills and labour market outcomes. However, only a limited number of studies have investigated associations between non-cognitive skills in childhood and later labour market outcomes. Here, we used a sample of adolescents aged 10–18 years and follow-up data from the same individuals ten years later to investigate how non-cognitive skills in childhood measured by self-esteem are related to occupational prestige in young adulthood. The aim of the paper was to investigating whether associations between global and domain specific self-esteem in childhood and occupational prestige outcomes in adulthood exist and if so, what the magnitudes of these associations are for each respective gender.

The findings note a striking gender difference regarding these aspects. For girls, both global and domain-specific self-esteem is positively related to occupational prestige, while self-esteem seems to be unrelated to boys’ future occupational prestige. Women who have high confidence in math when growing up have higher occupational prestige as adults, even when accounting for social background and actual grades in math. Global self-esteem is positively related to later occupational prestige as well. This association was accounted for by differences in math grades, suggesting that grades may mediate the association between self-esteem and later occupational outcomes. The results indicate that not only are high grades in math of importance, but women also need to perceive themselves as being good at math for later occupational achievements. This could support the assumptions raised by Keller et al. (2015) that self-esteem is an important mechanism for women’s opportunities to achieve higher positions in society, as these positions are more in line with stereotypes about men. In particular, prior research has noted that gendered beliefs about mathematical skills, where mathematics is seen as a male skill, are important when explaining gender differences in occupational achievements (Correll 2001; Pitsia, Biggart, and Karakolidis 2017).

In line with prior research (e.g. Marsh and O’Mara 2008; Rosenberg et al. 1995) we find self-confidence in math to have more influence than global self-esteem for occupational achievement. These results support the notion that it is important to take both global and domain specific self-esteem into account.

The self-evaluation of math skills could naturally be biased by interest. Girls who are interested in mathematics may, to a larger extent, consider themselves to be good at math. Even if the results reported here cannot distinguish the impact of interest from the impact of self-confidence, it is clear that self-esteem in math is strongly related to girls’ occupational prestige even when accounting for performance in math. This relationship is absent among boys. Also global self-esteem is unrelated to boys’ prestige, while it is of importance among women. This strengthens the assumption that self-esteem is more important for girls’ later labour market outcomes. The gender differences in the impact of self-esteem could be due to gender differences in self-evaluation. It is possible that men tend to overrate their performance in math. Prior research indicates that both parents and teachers tend to overrate boys’ ability in math relative to girls’ ability (Lindberg, Hyde, and Petersen 2010). This may inflate boys’ perceptions about their own ability in math. Thus, even if boys state that they are good in math their grade in math could be quite low, maybe too low for entering educational training that leads to prestigious occupations. Related, recent research show that gender norms in the class room affect students’ performance in math. In classes with more traditional gender norms the gender gap in math is larger, to girl’s disadvantage. Thus, girls preformed worse in classes with strong masculinity norms (Salikutluk and Heyne 2017). Such findings together with those shown here indicate that gendered norms may influences both performance and self-evaluation which in the end may affect future labour market outcomes.

Even if the panel design used here has many advantages, there are also limitations. Firstly, causal interpretations should be made with caution. Omitted variables could still affect both self-esteem in childhood and occupational attainment in young adulthood. Secondly, the individuals in the studied sample are at the beginning of their careers, and some of them are still obtaining their education. Therefore, their destination in the social structure is not yet settled. A follow-up study when the respondents are older would give us more knowledge about whether the association between self-esteem and occupational prestige is long-lasting or not. In an additional analysis not presented here, we examined the same data but only included those respondents who responded to a set of personality items in adulthood. The analysis indicates that self-esteem in adulthood is significantly and positively related to women’s prestige levels, while unrelated to young men’s prestige levels (results available upon request). These results support the notion that the importance of self-esteem differs between men and women. More research is needed to understand gender differences in self-esteem. Perhaps it is even more urgent to investigate why self-esteem is differently associated with men’s and women’s later labour market outcomes. Moreover, non-cognitive skills are a broad concept and include more skills than what is under study here. Thus, we can only make statements about self-esteem and later occupational achievement and not about other non-cognitive skills. Future studies investigating relationships between other non-cognitive skills are requested as well. Finally, as some studies suggest that non-cognitive skills are malleable, especially in people’s early years (Ermisch 2008), a possible implication of the present findings is the importance of focusing on the development of self-esteem among children, both in school and in the family.

Acknowledgements

We thank the participants at the European Consortium for Sociological Research (ECSR) Conference in Berlin in September 2014 and the participants at the FORTE-funded Nordic Network ‘Jämställdhet och samspelet välfärdsstat-arbetsliv-familj. Ett nordiskt forskarnätverk’ (FORTE, dnr: 2015-01465) for their luable comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

Financial support from the Swedish Research Council for Health, Working Life and Welfare (Forte) [grant number 2013-1119] is gratefully acknowledged.

Notes

1 Self-esteem is closely related to the term self-concept which Shavelson, Hubner, and Stanton (1976) define as the individual’s self-perception. Some researchers see self-esteem as the evaluative (affective) component of self-perception while self-concept instead describe (cognitive) specific components of self-perceptions. Shavelson, Hubner, and Stanton (1976), however, argue that the term self-concept include both evaluative and descriptive elements. For example, is the statement ‘I am good in mathematics’ of both descriptive and evaluative character (see also Marsh and O’Mara 2008). Thus, both self-esteem and self-concept involve affective as well as descriptive components (Swann, Chang-Schneider, and Larsen McClarty 2007).

2 The main LNU was collected by face-to-face interviews, and the Child-LNU 2000 was collected by answering questions on a paper form via audio CD and headphones.

3 A comparison between the sample of Child-LNU in 2000 and the follow-up sample in 2010, indicate that older youths and individuals with working class background are somewhat underrepresented in the follow-up study. The attrition is also somewhat higher among children with divorced parents. For detailed information of the LNU data see Östberg, Modin, and Brolin Låftman (2014).

4 Since the SIOPS was first constructed, the labour market has undergone structural changes. Accordingly, the scale could be criticized for being out-of-date. However, several studies report prestige hierarchies to be relatively constant over time and space (e.g., Hout and DiPrete 2006). This has also been found in a recent study on prestige ratings conducted in Sweden (Ulfsdotter Eriksson 2006). Thus, the Swedish scale is highly correlated with the SIOPS but includes fewer occupations, which is why it is not used in the present study.

5 Other cut-off points have also been tested, e.g., prestige > 60. The number of respondents with only a mother in a highly prestigious occupation becomes very few (14 respondents). However, the results are essentially the same in these models, but the associations between mothers’ occupational prestige and offspring’s prestige are insignificant.

6 Additional analyses including interactions between parental prestige and confidence in math did not show any significant interactions.

7 In an additional analysis (not presented), we have pooled models for men and women including interactions between women and measurements of self-esteem. These results indicate a positive significant interaction between confidence in math and women.

8 Controlling for school grades (grade point average from the spring term in the 9th grade and final grade of compulsory school) has also been performed. The results are not substantially different from those reported here. However, the association between women’s prestige level and having a mother in a highly prestigious occupation becomes insignificant, and the association between having both parents in highly prestigious occupations and prestige decreases. This indicates that some of the associations between social background and one’s own labour market outcomes are mediated by grades. The positive effect of having confidence in math remains when controlling for grades. A supplementary test including years of education has also been conducted with similar results as those reported here.

Appendix

Table A1. Coefficients from OLS regression models predicting occupational prestige.

Model^a	1		2		3		4		5
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
Mother prestige (13–78)	0.15	0.20**	0.16	0.20**	0.17*	0.16*	0.13	0.20**	0.15	0.16*
	(0.09)	(0.07)	(0.08)	(0.07)	(0.08)	(0.08)	(0.08)	(0.07)	(0.08)	(0.08)
Father prestige (13–78)	0.14	0.24**	0.12	0.24***	0.05	0.21**	0.14	0.25	0.06	0.21**
	(0.09)	(0.08)	(0.08)	(0.08)	(0.09)	(0.07)	(0.09)	(0.08)	(0.09)	(0.07)
Confidence in math (1–5)			3.91***	0.40	2.68**	−1.35	–	–	2.79**	−1.49
			(0.91)	(1.10)	(0.95)	(1.12)	–	–	(0.98)	(1.13)
High grade in math					9.00***	11.99***	–	–	8.81***	11.86***
					(2.68)	(2.56)	–	–	(2.66)	(2.55)
Like myself (1–4)							2.68	−1.63	1.18	1.22
							(1.77)	(2.06)	(1.77)	(2.02)
Like appearance (1–4)							−2.33	2.26	−2.15	1.92
							(1.72)	(1.78)	(1.65)	(1.74)
Constant	32.25***	26.58***	19.13***	25.17***	24.91***	32.85***	32.12***	24.75***	27.82***	31.25***
	(4.24)	(4.60)	(4.82)	(2.02)	(5.03)	(5.85)	(5.77)	(6.57)	(5.86)	(7.70)
R-squared	0.08	0.11	0.14	0.06	0.18	0.17	0.09	0.11	0.19	0.18
N	243	235	243	235	243	235	243	235	243	235

***p < 0.001; **p < 0.01; *p < 0.05.

Robust standard errors (clustered at the household level) are used, displayed in brackets.

^aAll models control for age.