Jump to ContentJump to Main Navigation
Determined to Succeed?Performance versus Choice in Educational Attainment$

Michelle Jackson

Print publication date: 2013

Print ISBN-13: 9780804783026

Published to Stanford Scholarship Online: June 2013

DOI: 10.11126/stanford/9780804783026.001.0001

Show Summary Details
Page of

PRINTED FROM STANFORD SCHOLARSHIP ONLINE (www.stanford.universitypressscholarship.com). (c) Copyright Stanford University Press, 2022. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in SSO for personal use. Subscriber: null; date: 23 May 2022

Dentist, Driver, or Dropout?

Dentist, Driver, or Dropout?

Family Background and Secondary Education Choices in Denmark

Chapter:
(p.228) Chapter Eight Dentist, Driver, or Dropout?
Source:
Determined to Succeed?
Author(s):

Anders Holm

Mads Meier Jæer

Publisher:
Stanford University Press
DOI:10.11126/stanford/9780804783026.003.0008

Abstract and Keywords

This chapter, which assesses IEO in Denmark, building on the basic analyses to consider tripartite outcomes (academic tracks, vocational tracks, and exit), compares parental class and parental education inequalities at the transition made at age 16 for two cohorts born around 30 years apart. It notes that in Denmark, secondary effects make up the lion's share of observable family background inequalities in the likelihood of succeeding in the academic track in secondary education. Secondary effects arise from social-class differences in educational decision-making strategies, cultural capital, and other noncognitive resources, all factors that have previously been shown to be important in Denmark. The analysis shows that inequality in educational outcomes exists in Denmark and that secondary effects are particularly strong compared to most other countries.

Keywords:   Denmark education, school transition, family background inequalities, social-class differences

Although Denmark belongs to the Scandinavian comprehensive welfare state family, rich evidence suggests that social-class background still has a substantial impact on educational success in this national context. In terms of the quantity of inequality, social-class differences in educational attainment decreased over most of the 20th century in Denmark (e.g., Benjaminsen 2006; Thomsen 2008; for evidence on other countries, see Breen and Jonsson 2005; Breen et al. 2009). Nevertheless, class inequalities in educational attainment remain fairly strong. In terms of the quality of inequality, recent research shows that social-class differences in educational attainment in Denmark are mainly driven by social-class differences in cultural resources (Jæger and Holm 2004; Jæger 2009), but also to some extent by differences in economic and social resources (e.g., McIntosh and Munk 2007; Jæger 2007a; Jæger and Holm 2007).

In this chapter we synthesize previous research on Denmark by decomposing the total effect of family background on educational attainment into primary and secondary effects. As explained in more detail in Chapter 1 of this volume, the distinction between primary and secondary effects was popularized by Boudon (1974). Existing research suggests that both primary and secondary effects matter in Denmark. First, there is compelling evidence that family background, typically measured by social-class background and parents' educational attainment, is strongly linked to academic ability; that is, primary effects exist (e.g., Andersen et al. 2001; Jæger 2009). Second, research also documents substantial effects of family background on educational outcomes even after controlling for demonstrated academic ability; that is, secondary effects exist (e.g., Hansen 1995; McIntosh and Munk 2007; (p.229) Jæger 2009). Secondary effects have been linked to factors such as students' desire to preserve the relative social position of the family of origin (Holm and Jæger 2008) and to their beliefs about the future economic and social returns to educational decisions (Jæger 2007b).

Although previous research shows that primary and secondary effects exist, this chapter is the first to systematically identify the relative importance of primary and secondary effects of family background on educational success in Denmark. In doing so, we seek to assess the position of Denmark in comparison to other countries in terms of the relative importance of primary and secondary effects (see Erikson et al. 2005; Erikson 2007; Jackson et al. 2007). Furthermore, in addition to cross-sectional analysis we carry out cross-time analysis by comparing primary and secondary effects for two birth cohorts, born around 1954 and 1984, respectively, who made the first educational transition around 1970 and 2000, respectively. This analysis allows us to assess whether the relative importance of primary and secondary effects has changed over time.

The chapter focuses on the primary and secondary effects of family background on the transition that students make at around age 16, from elementary school to different types of secondary education. This transition is very important in the Danish educational system because at this stage students choose whether to pursue an academic or a vocational track. Furthermore, because there is very little migration between the two tracks, the decision not to enter the academic track at age 16 is in practice irreversible and has long-term consequences for students' later educational careers and socioeconomic outcomes (Jæger and Holm 2007).

The Danish educational system has some idiosyncrasies that we also address in the chapter. Unlike England or the United States, in which students decide whether to continue in an academically oriented secondary education track (A-levels in England and high school degree in the United States), the choice set in Denmark includes two qualitatively different options: academic upper secondary education and vocational secondary education. Existing research finds that students who choose the academic upper secondary education track are very different from students who choose the vocational secondary education track in terms of average academic ability, family background, and educational aspirations (e.g., Jæger and Holm 2007; Jæger 2007a, 2009). In our empirical analysis we take the more complex set of educational options in Denmark into account by (p.230) estimating primary and secondary effects in a multinomial (rather than binary) analytic framework.

The Danish Educational System

The Danish educational system combines features of a German-oriented path-dependent system and a Scandinavian equality-oriented system. There is no tracking in the elementary school system (tracking similar to that found in Germany existed previously but was abolished in two reforms, in 1958 and 1975). Students normally finish elementary school after the 9th grade at age 16. An optional 10th grade also exists. Respondents in our 1954 cohort were exposed to tracking from the 7th grade of elementary school (into an academic or a nonacademic type of lower secondary education), whereas respondents in our 1984 cohort were not exposed to any tracking. However, respondents in the 1954 cohort who were placed in the nonacademic track in lower secondary education were not formally limited with respect to their educational options after elementary school. Consequently, even when tracking existed it was much less stringent than in, for example, the German system and did not preclude any educational choices at later stages.

Educational Choices after Elementary School

The first and arguably most important educational transition that students make in the Danish educational system is the choice after elementary school. Here, students have three options: (1) leave the educational system, (2) enter upper secondary education, or (3) enter vocational secondary education.

Upper secondary education is a three-year, academically oriented track that prepares students for higher education. The most common type of upper secondary education is the Gymnasium, which focuses on general-education academic subjects (Danish, foreign languages, sciences, history, etc.). Today, around 50 percent of students enter the Gymnasium. A second type of upper secondary education, the HF (Højere Forberedelseseksamen, or higher preparatory examinations), was introduced in 1968. The HF, which grants eligibility for higher education, was designed for late-starting students who did not enter the Gymnasium immediately after elementary school but who wished to obtain upper secondary level credentials (the HF is similar to the General Educational Development certificate in the United States). Both (p.231) types of upper secondary education, the traditional Gymnasium and the HF, were available to our respondents in the two cohorts. In the 1980s two new types of upper secondary education specializing in, respectively, technical (HTX) and mercantile (HHX) subjects were introduced; in our analysis, only the 1984 cohort could enroll in these programs. The HTX and the HHX offer an academic track with a practical flavor and may therefore attract some students from lower-class backgrounds who might have the academic ability to pursue academic credentials but who may feel uncomfortable with traditional academic teaching styles in the old Gymnasium. Therefore one might expect that the academic track has become more attractive to lower-class students in the 1984 cohort compared to the 1954 cohort.

Vocational secondary education normally takes three to four years and consists of a combination of school-based training (mechanics, bricklaying, hairdressing, etc.) and on-the-job training as an apprentice with an employer. Upon completion of their training, students receive a certificate that attests to their educational qualifications, and they enter the labor market (often with the employer with whom they apprenticed).

Figure 8.1 shows transition probabilities for the 1954 and the 1984 cohorts. The figure shows that 52 percent of the respondents in the 1954 cohort left the educational system after elementary school, 26 percent completed vocational secondary education, and 22 percent completed upper secondary education. In the 1984 cohort 35 percent left the educational system after elementary school, 20 percent completed vocational secondary education, and 45 percent completed upper secondary education.1

Educational Choices after Upper Secondary Education

A second educational transition exists for students who complete upper secondary education. These students may enter one of two types of higher education: university college and university.

University college (three–four years) comprises a range of courses typically targeted at jobs in the public sector. These include, for example, schoolteacher, nurse, social worker, child care worker, and some types of engineering. Today, almost all types of courses at university colleges require applicants to have completed upper secondary education. When the 1954 cohort went through the educational system it was possible to be admitted to some courses at university colleges without first having completed upper secondary education. However, very few people used this option. (p.232)

Dentist, Driver, or Dropout?Family Background and Secondary Education Choices in Denmark

Figure 8.1. Flowchart for educational pathways

NOTE: First percentage represents the 1954 cohort, second the 1984.

University (five–six years) includes all types of education at the university level. Courses include, for example, medicine, law, economics, architecture, civil engineering, and natural sciences. Today, almost all university-level courses are built around a three-year bachelor's (BA, BSc) degree and a two-year master's (MA, MSc) degree. A doctorate normally takes three years after completing a master's degree. When the 1954 cohort went through the educational system most university degrees took five or six years to complete, and there was no formal distinction between BA and MA levels.

In Figure 8.1 we have merged university college and university into a single group labeled higher education. The figure shows that in the 1954 cohort two-thirds of those who completed upper secondary education also later completed some type of higher education. Unfortunately, data on tertiary-level education are not available for the 1984 cohort. Instead, in Figure 8.1 we plot national figures for tertiary-level completion rates for similarly aged cohorts (figures are from Undervisningsministeriet 2010, 72–77), which show that around three-quarters of those who complete upper secondary education later complete some type of higher education.

(p.233) Data and Variables

Data

We use two data sources. Our first data source is the Danish Longitudinal Survey of Youth (DLSY). The DLSY is a cohort study with a nationally representative sample of 3,151 participants who were born around 1954 (see Jæger and Holm 2007; Jæger 2007b). The DLSY respondents were first interviewed in 1968 at age 14 (when they attended the seventh grade of elementary school), and they have since been interviewed in 1970, 1971, 1973, 1976, 1992, 2001, and 2004. The response rate in the DLSY has remained high throughout its history. In the 2001 wave, 2,507, or 79.6 percent, of the original sample members were successfully interviewed. In this chapter we analyze data from the early DLSY waves (1968–1976), in which there was practically no attrition among the primary respondents.2

Our second data source is the Program for International Student Assessment Longitudinal (PISA-L) survey. The PISA-L is a follow-up survey of the original participants in the Danish PISA survey, which was carried out in 2000. The Danish PISA 2000 sample of eighth-grade elementary school students (i.e., 15-or 16-year-olds born around 1984) was reinter-viewed in 2004 when the students had reached their late teens or early 20s. The PISA-L focused on postelementary schooling decisions, occupational aspirations, and a range of other topics. Of the nationally representative sample of 3,954 participants in the 2000 PISA sample, interviews were successfully carried out with 3,084 respondents in the 2004 PISA-L survey. This yielded a response rate of 78 percent (Andersen 2005).

The DLSY and the PISA-L allow us to analyze the primary and secondary effects of social-class background on educational choices after elementary school because both surveys include information on (1) educational choice after elementary school, (2) academic ability at around age 14, and (3) family background characteristics (parental social class and parental education).

Variables

Table 8.1 shows descriptive statistics for all variables used in the analysis. The dependent variable measures type of secondary education, with the three possible outcomes being no education beyond elementary school, vocational secondary education, and upper secondary education. Our measure (p.234)

TABLE 8.1 Descriptive statistics: Percentages (categorical variables), means, and standard deviations

1954 COHORT (DLSY)

1984 COHORT (PISA-L)

Variable

%

SD

%

SD

Secondary education choice

    No education beyond elementary school

51.4

35.1

    Vocational secondary education

26.2

19.5

    Upper secondary education

22.4

45.4

Academic ability

0

1.0

0

1.0

Social class background

    Salariat

13.1

22.6

    Intermediate

32.9

50.5

    Working class

50.5

14.0

    Missing

3.5

12.9

Parental education

    High

9.0

49.2

    Medium

36.9

33.2

    Low

37.9

10.6

Missing

16.2

7.0

Gender (= female)

43.4

51.1

Number of observations

3,038

3,073

NOTE: DLSY = Danish Longitudinal Survey of Youth; PISA-L = Program for International Student Assessment Longitudinal.

of secondary education pertains to completion of the different types of secondary education rather than to enrollment, because in the 1954 cohort we have reliable information on completion but not reliable information on enrollment (but noncompletion).

The principal explanatory variables are academic ability, social-class background, and parental education. In the DLSY our measure of academic ability is the first component from a principal component analysis of three cognitive tests measuring respondents' mathematical, verbal, and spatial ability, carried out when respondents were 14 years old. This measure of academic ability accounts for 67 percent of the total variance in the three cognitive tests. In the PISA-L our measure of academic ability is the respondent's score on the PISA reading ability test carried out at age 15 (see OECD 2000). Both academic ability variables are standardized. Unfortunately, the DLSY and PISA-L do not include any information on respondents' (p.235) academic performance, for example, their grade point average (GPA) from the final examinations after ninth grade. As a consequence, we are not able to include any measure of academic performance that is observable to the respondent. However, not being able to control for academic performance such as GPA may be less of a problem in the Danish case than elsewhere for two reasons. First, unlike in other countries, student GPA from elementary school has no formal bearing on the range of available educational options after elementary school. Danish students typically apply for upper secondary or vocational secondary education in the spring of their ninth-grade year and several months before their final examinations at the end of the ninth grade. Consequently, students do not know what their GPA will be when they decide to apply for secondary education. Second, GPA is not important when secondary education schools (academic and vocational) admit students because schools are not allowed to select students based on GPA and because schools do not know applicants' GPA when they admit students (in most cases admission is based on the proximity of a student's home to the school). Consequently, because of the lack of formal requirements students have a de facto free choice in the secondary education they select (but not necessarily the particular school).3

Our measure of social class is based on the Erikson-Goldthorpe-Portocarero (EGP) class schema (see Erikson and Goldthorpe 1992). We use the dominance principle and assign the social-class position of the respondent's father or mother, whichever is higher, as our indicator of social-class background. We use three class categories: (1) salariat (EGP classes I and II), (2) intermediate (EGP classes IIIa, IVabc, and V), and (3) working class (EGP classes IIIb, VI, and VII). We also include an indicator for missing data on social-class background (for example, if neither parent was active in the labor market or if information on social class was missing). Table 8.1 shows that the distributions of social-class background differ across the two cohorts. In the 1954 cohort around half the respondents have working-class backgrounds, whereas this was the case for less than a quarter of the 1984 cohort. In the 1984 cohort only 14 percent of respondents have a working-class background, while relatively more respondents have an intermediate-class background. Consequently, our data suggest a comparatively high level of social mobility between the two generations.

Our measure of parental education is based on the highest level of education of the respondent's father or mother. We distinguish three levels of (p.236) education: (1) low (elementary school only), (2) medium (vocational or upper secondary education but no higher education), and (3) high (university college or university education). From Table 8.1 we also see marked differences across cohorts in parental education. In the 1954 cohort, less than 10 percent of the parents had high education; in the 1984 cohort, almost half the parents had high education. Our data indicate increasing levels of parental educational attainment.

Results

In this section we present results from the empirical analysis of primary and secondary effects on educational choices in Denmark. First, we provide summary evidence of primary effects in both cohorts. Then, for secondary effects we present two sets of results. Our first set of results concerns primary and secondary effects with regard to the probability of completing upper secondary education, that is, the academic track in Danish secondary education. We carry out this analysis to ensure that our results are comparable with previous studies analyzing the educational systems in the United States and Great Britain, in which students decide whether to continue in the academic track. Next, we present results that take into account the particular trichotomous choice set that Danish students face after elementary school (i.e., the choice between no education, upper secondary education, and vocational secondary education). In the binary case we decompose the total effect of social-class background and parental education on secondary education choice into primary and secondary effects using the method described in Chapter 2 of this volume. In the trichotomous case we use our own methodological approach, which is similar conceptually to the one used in the binary case but somewhat different technically.

Primary Effects

Table 8.2 shows results from linear regressions of academic ability on social-class background and parental education. These effects can be interpreted as benchmark estimates of primary effects, and because the academic ability variables are standardized, the effects of the family background variables are scaled as changes in the distribution of academic ability in fractions of a standard deviation. For both cohorts we find that both social-class background and parental education are statistically significant predictors of (p.237)

TABLE 8.2 Results from linear regressions of academic ability on social-class background

Variable

1954 cohort

1984 cohort

Social-class background

    Salariat

0.519***

0.633***

    Intermediate

0.234***

0.443***

    Working

    R2

0.05

0.05

Parental education

    High

0.599***

0.856***

    Medium

0.233***

0.590***

    Low

    R2

0.05

0.08

N

2,452

2,579

NOTE: Models do not include observations with missing values on social-class background or parental education.

(***) p 〈 0.001.

academic ability at around age 14. In the 1954 cohort, respondents whose parents were in the salariat or intermediate class or who had high or medium education display 0.5–0.6 and 0.2–0.3 of a standard deviation, respectively, higher academic ability compared to respondents whose parents were in the working class or who had low (i.e., elementary school) education. In the 1984 cohort results are very similar for social-class background, while differences in academic ability by parental education appear to be somewhat larger. Consequently, it seems that over time parental education became a stronger predictor of academic performance. Our findings resemble previous results on Denmark (e.g., Hansen 1995; Jæger 2007a, 2009; Jæger and Holm 2007).

Secondary Effects for Upper Secondary Education

Table 8.3 shows results from binary logistic regressions of the probability of completing upper secondary education. This type of education is the academic track in secondary education in Denmark, and it is the gateway to higher education. We model the probability of completing upper secondary education as a function of either social-class background or parental education and of academic ability at around age 14. The effects of social class and parental education on the probability of completing upper secondary education conditional on academic ability can be interpreted as (p.238)

TABLE 8.3 Results from logistic regressions of upper secondary education. Log-odds coefficients and average partial effects (APEs)

1954 COHORT

1984 COHORT

Variable

Coefficient

APE

Coefficient

APE

Academic ability

1.505***

0.215

0.489***

0.112

Social-class background

    Salariat

1.425***

0.239

0.526***

0.132

    Intermediate

0.815***

0.120

0.319**

0.073

    Working

    Pseudo R2

0.214

0.044

Academic ability

1.490***

0.203

0.458***

0.106

Parental education

    High

2.271***

0.392

0.786***

0.181

    Medium

0.935***

0.126

0.620***

0.140

    Low

    Pseudo R2

0.243

0.047

N

2,452

2,579

NOTE: Models do not include observations with missing values on social-class background or parental education.

(**) p 〈 0.01;

(***) p 〈 0.001.

benchmark estimates of secondary effects of family background on educational success.

Table 8.3 reports log-odds coefficients and average partial effects (APEs). The APEs express the expected change in the probability of completing upper secondary education (averaged over the whole sample) for respondents with different values on the social-class and parental education variables. In the 1954 cohort we find that, after taking into account academic ability, there are highly significant differences in the likelihood of completing upper secondary education for respondents with different social-class backgrounds and parental educational levels. The estimated log odds ratio between the salariat and the working class is 1.425, which is equivalent to an APE of 0.239. Consequently, after holding constant academic ability, respondents with salariat backgrounds are on average 24 percent more likely to complete upper secondary education compared to respondents with working-class backgrounds. This result is similar to those reported in previous research. Educational gradients are also strong in the 1954 cohort. Together, these results suggest very large secondary effects in the 1954 cohort. We also find highly significant social-class and parental (p.239) education effects in the 1984 cohort, but their substantive magnitude is much smaller. In the 1984 cohort the APE for the contrast between salariat and working-class background is 0.132 (about half that in the 1954 cohort). The gradient by parental education is also much smaller in the 1984 cohort. The difference in the strength of secondary effects might be a result of the large expansion of upper secondary education and the introduction of the practically flavored academic tracks (HTX and HHX) that occurred between the two cohorts. We discuss this possibility in more detail below.

Primary and Secondary Effects in Choosing Upper Secondary Education

In this section we decompose the total effect of social-class background and parental education on secondary education choice to calculate the relative importance of primary and secondary effects. We use the methods described in Chapter 2 of this volume (see also Erikson et al. 2005 and Jackson et al. 2007).

We begin by assessing the impact of primary effects on educational transitions. Tables 8.4a and 8.4b show synthesized transition rates into upper secondary education for respondents with different social class and parental education backgrounds. We analyze the impact of primary effects by assuming that respondents with a certain family background characteristic (e.g., working class) happen to have the academic ability distributions of respondents with a different family background characteristic (e.g., salariat). We analyze the impact of secondary effects by assuming that respondents with a certain family background characteristic happen to have the decision probabilities of respondents with a different family background characteristic. In Tables 8.4a and 8.4b the Decision columns refer to respondents' social-class or parental education group, respectively, and the Performance rows refer to academic ability distributions for those groups.

Table 8.4a shows that in the 1954 cohort the transition rate into upper secondary education was 49.9 percent for respondents of salariat background and the academic ability of the salariat. If respondents with salariat background had instead had the academic ability of those of working-class background, their transition rate would have been 35.0 percent on average, i.e., their synthesized transition rate is 14.9 percentage points lower. By contrast, respondents of working-class background have an actual transition rate of 14.3 percent and would have had a transition rate of 25.9 percent (p.240)

TABLE 8.4A Estimated factual and synthesized transition rates by social-class background

1954 COHORT

Decision

Performance

Salariat

Intermediate

Working

Salariat

49.9

39.0

25.9

Intermediate

41.6

30.4

18.3

Working

35.0

24.8

14.3

1984 COHORT

Decision

Performance

Salariat

Intermediate

Working

Salariat

52.7

46.9

39.4

Intermediate

50.5

44.7

37.4

Working

45.4

39.8

32.7

NOTE: Calculated using ldecomp in Stata (see Buis 2010).

TABLE 8.4B Estimated factual and synthesized transition rates by parental education

1954 COHORT

Decision

Performance

High

Medium

Low

High

64.0

38.0

22.1

Medium

54.3

29.0

15.7

Low

47.5

23.4

12.0

1984 COHORT

Decision

Performance

High

Medium

Low

High

50.4

46.4

32.3

Medium

46.4

42.5

28.8

Low

41.2

37.4

24.7

NOTE: Calculated using ldecomp in Stata (see Buis 2010).

had they had the ability distribution of the salariat, i.e., a positive difference of 11.6 percentage points. In general, for both cohorts the synthesized transition rates show that respondents of all class backgrounds increase their transition rate if they have the academic ability of a higher social class and decrease their transition rate if they have the academic ability of a lower (p.241) class. As shown in Table 8.4b, the picture is similar when we use parental education rather than social class as our indicator of family background.

When comparing the two cohorts we find that the gaps in transition rates for respondents with different social-class and parental education backgrounds decrease from the 1954 cohort to the 1984 cohort. In other words, it seems that primary effects are moderately more important in the 1954 cohort than in the 1984 cohort because substituting academic ability distributions between social classes has a larger effect on the synthesized transition rates in the 1954 cohort than in the 1984 cohort. This result might be attributable to the existence of a moderate level of tracking in the 1954 cohort (from the seventh grade) that did not exist in the 1984 cohort and that might have led to more substantial inequalities in academic ability before the transition into upper secondary education. However, as we show in our multinomial decomposition below, the picture is somewhat more complex because, in addition to upper secondary education, students have the option of choosing to enroll in vocational secondary education.

Tables 8.5a and 8.5b show the relative importance of secondary effects for educational success calculated using both social-class background and parental education as indicators of family background. The tables express the relative importance of secondary effects as the percentage of the total effect of, respectively, social-class background and parental education on transition rates into upper secondary education that is not attributable to academic ability. From Table 8.5a we see that in the 1954 cohort secondary effects account for between 60 and 70 percent of the total effect of social class on the likelihood of completing upper secondary education. Figures for parental education (Table 8.5b) are remarkably consistent at around 70 percent. Our results suggest that secondary effects account for most of the observed social-background inequalities in the likelihood of completing upper secondary education.

In the 1984 cohort we find that the secondary effects of social-class background are very similar to those found in the 1954 cohort. In the 1984 cohort secondary effects account for about 60 percent of the total class differences in transition rates when comparing the salariat and intermediate class and 65 percent when comparing the salariat and the working class. The secondary effects of social class are slightly lower when comparing the intermediate and working classes (60.2 percent in the 1984 cohort compared to 70.6 percent in the 1954 cohort). When using parental education (p.242)

TABLE 8.5A Relative importance of secondary effects, social-class background (%)

1954 COHORT

Class

Salariat

Intermediate

Working

Salariat

59.1 (0.128)

65.4 (0.065)

Intermediate

59.1 (0.128)

70.6 (0.091)

Working

65.4 (0.065)

70.6 (0.091)

1984 COHORT

Class

Salariat

Intermediate

Working

Salariat

60.0 (0.143)

65.1 (0.134)

Intermediate

60.0 (0.143)

60.2 (0.318)

Working

65.1 (0.134)

60.2 (0.318)

NOTE: Estimates show the relative importance of secondary effects, expressed as the percentage of the total social-class effect on transition rates into upper secondary education that is not attributable to academic ability. Standard errors in parentheses (calculated using the R package DECIDE).

TABLE 8.5B Relative importance of secondary effects, parental education (%)

1954 COHORT

Education

High

Medium

Low

High

72.6 (0.097)

73.7 (0.063)

Medium

72.6 (0.097)

73.5 (0.076)

Low

73.7 (0.063)

73.5 (0.076)

1984 COHORT

Class

High

Medium

Low

High

49.6 (0.140)

67.0 (0.085)

Medium

49.6 (0.140)

73.8 (0.096)

Low

67.0 (0.085)

73.8 (0.096)

NOTE: Estimates show the relative importance of secondary effects, expressed as the percentage of the total parental education effect on transition rates into upper secondary education that is not attributable to academic ability. Standard errors in parentheses (calculated using the R package DECIDE).

as the measure of family background we find that in the 1984 cohort secondary effects account for between 50 percent (difference between high and medium educated parents) and 74 percent (difference between medium and low educated parents) of the total educational gradient in the likelihood of completing upper secondary education. Consequently, it seems that across the two generations born 30 years apart secondary effects have become relatively less important in explaining differences in educational outcomes.

(p.243) A Multinomial Decomposition of Primary and Secondary Effects

Unlike in England or the United States, secondary education in Denmark includes two very different tracks: the academic upper secondary track and the hands-on vocational track. Thus, upon completion of elementary school at around age 16, Danish students must decide not only whether to continue in the educational system but also which of the two tracks they wish to pursue. In this section we take this particular institutional feature of Danish secondary education into account by presenting results from a multinomial decomposition of primary and secondary effects of social-class background and parental education on the choice of secondary education. Existing research shows that students' choices of track differ significantly in terms of their family background characteristics and academic ability (e.g., Jæger and Holm 2007; Jæger 2009). Consequently, it is likely that the relative importance of primary and secondary effects will differ across the two educational choices.

Because the choice set is now trichotomous (no education, vocational secondary education, or upper secondary education) rather than binary (no education or upper secondary education), the decomposition method used in the previous sections no longer applies. Instead, we employ a technically different but substantively similar method introduced by Karlson, Holm, and Breen (forthcoming), which is adapted for multinomial outcome variables. The method yields results very similar to those of the methods employed in the previous sections when applied to binary models (Karlson and Holm 2010). The idea behind the method in Karlson, Holm, and Breen is simple. First, we run a linear regression of academic ability on the social-class and the parental education variables (one regression for each family background variable). Second, we use the residuals from these regression models (which by definition are uncorrelated with the social-class and parental education variables) instead of the original academic ability variable in a multinomial logistic regression of secondary education choice. We use the new residualized academic ability variable instead of the original ability variable to correct for differences in the scale of the residual variances in the multinomial logistic regression model. This procedure allows us to directly compare log-odds coefficients across nested models. We then proceed by estimating two multinomial logistic models: one model that includes the residualized academic ability variable and one model that includes the (p.244) original academic ability variable. The first model measures the total effect of social class and parental education on educational choice, because in this model social class and parental education is uncorrelated with academic ability (because all shared variance between the social class and parental education and academic ability variables has already been purged in the linear regression). The second model, which uses the original academic ability variable, measures the conditional effect of social class and parental education on educational choice (i.e., the effect of social class and parental education conditional on its indirect effect running through academic ability). The two models are statistically equivalent in the sense that they have the same fit to the data. We interpret the percentage-wise change in the log-odds coefficients of the social class and parental education variables from the first model (total social class and parental education effect) to the second model (conditional social-class and parental education effect) as a measure of the relative importance of secondary effects. Unlike in the previous analysis, we include gender as an explanatory variable (a dummy variable for women) because there are significant gender differences in the choice of (especially vocational) secondary education in Denmark.

Tables 8.6a and 8.6b summarize results from the multinomial decompositions in the 1954 and the 1984 cohorts. Columns M1 and M2 show results from the two multinomial logistic regression models using, respectively, the residualized and the original academic ability variable. The RSE (relative secondary effect) column shows our measure of secondary effects, that is, the percentage reduction of the log-odds coefficient for the social-class and parental education variable when comparing M1 and M2.

We begin by analyzing primary and secondary effects in the likelihood of completing upper secondary education versus no education. In the 1954 cohort we find highly significant effects of both academic ability and social-class background and parental education when comparing the likelihood of choosing upper secondary education over no education. Our RSE for social-class background is 63.2 when comparing salariat versus working class and 72.6 when comparing intermediate versus working class. These results indicate that secondary effects account for between 63 and 73 percent of social-class differences in the likelihood of choosing the academic upper secondary track over no education. Results are very similar for different levels of parental education. Consequently, we obtain essentially the same results for the decomposition analysis of primary and secondary (p.245)

TABLE 8.6A Results from multinomial logistic regression models of secondary education choice on parental social class. Reference category is low education (no education beyond elementary school)

1954 COHORT

1984 COHORT

M1

M2

RSE

M1

M2

RSE

VOCATIONAL SECONDARY EDUCATION VS. LOW EDUCATION

Academic ability

−0.042

−0.042

−0.650***

−0.650***

Social-class

    background

    Salariat

−0.361**

−0.334**

92.5

−0.750***

−0.338

45.1

    Intermediate

0.302**

0.314***

104.0

−0.226

0.062

27.4

    Working

Gender (= female)

−0.326***

−0.326***

−1.008***

−1.008***

UPPER SECONDARY EDUCATION VS. LOW EDUCATION

Academic ability

1.226***

1.226***

0.239***

0.239***

Social-class

    background

    Salariat

2.099***

1.326***

63.2

0.607***

0.455***

74.9

    Intermediate

1.275***

0.926***

72.6

0.461***

0.355***

77.0

    Working

Gender (= female)

−0.054

−0.054

−0.107

−0.107

Pseudo R2

0.119

0.119

0.067

0.067

NOTE: M1 = model with residualized academic ability; M2 = model with observed academic ability; RSE = relative secondary effect (social-class coefficient in M2 as percentage of coefficient in M1).

(**) p 〈 0.01;

(***) p 〈 0.001.

effects as those reported in Tables 8.5a and 8.5b using a different technique. Together, these results again show that in the 1954 cohort secondary effects account for most of the inequalities in the likelihood of completing upper secondary education.

Results for the 1984 cohort are remarkably similar to those for the 1954 cohort. Tables 8.6a and 8.6b show that academic ability and social class and parental education all matter, with those of high academic ability and more privileged class backgrounds or better-educated parents being more likely to choose upper secondary education compared to those of low ability and less privileged backgrounds. Furthermore, the relative sizes of secondary effects, as expressed by the RSEs, resemble those found in the 1954 cohort. We find that in the 1984 cohort secondary effects account for over 70 percent of social-class differences in the likelihood of choosing upper secondary education over no education (the RSEs are 74.9 and 77.0 for parents' social class and 78.9 and 86.5 for parents' education). In sum, our (p.246)

TABLE 8.6B Results from multinomial logistic regression models of secondary education choice on parental education. Reference category is low education (no education beyond elementary school)

1954 COHORT

1984 COHORT

M1

M2

RSE

M1

M2

RSE

VOCATIONAL SECONDARY EDUCATION VS. LOW EDUCATION

Academic ability

−0.031

−0.031

−0.627***

−0.627***

Parental education

    High

−0.726**

−0.703**

96.8

−0.779***

−0.242

31.1

    Medium

−0.047

−0.038

80.9

−0.046

0.261

−567.4

    Low

Gender (= female)

−0.310**

−0.310**

−1.037***

−1.037***

UPPER SECONDARY EDUCATION VS. LOW EDUCATION

Academic ability

1.221***

1.221***

0.224***

0.224***

Parental education

    High

2.962***

2.071***

69.9

0.910***

0.718***

78.9

    Medium

1.268***

0.923***

73.5

0.813***

0.703***

86.5

    Low

Gender (= female)

−0.014

−0.014

−0.091

−0.091

Pseudo R2

0.134

0.134

0.070

0.070

NOTE: M1 = model with residualized academic ability; M2 = model with observed academic ability; RSE = relative secondary effect (social-class coefficient in M2 as percentage of coefficient in M1).

(**) p 〈 0.01;

(***) p 〈 0.001.

results suggest that despite an overall decline in the socioeconomic gradient in educational attainment over time in Denmark (Benjaminsen 2006) the relative balance between primary and secondary effects has changed very little. This is a noteworthy result.

Cross-temporal results are somewhat different when we compare the relative importance of primary and secondary effects on the likelihood of completing vocational secondary education compared with no education. In the 1954 cohort we find no significant effect of academic ability on the likelihood of choosing vocational secondary education over no education. This result suggests that in this cohort those who choose vocational secondary education do not have better academic skills than those who choose not to pursue any type of secondary education (previous research finds similar results, see, e.g., Jæger and Holm 2007; Jæger 2009). However, as regards secondary effects, we do find highly significant social-class effects after accounting for academic ability and less strong parental education effects. Our results thus suggest that respondents whose families belong to (p.247) the salariat or whose parents are highly educated are less likely to choose vocational secondary education over no education compared to respondents with working-class backgrounds or respondents whose parents are poorly educated. When comparing the social-class and parental education effects across M1 and M2 in the 1954 cohort, we find that their effects on the likelihood of choosing vocational education over no education change very little when we adjust for primary effects. This result suggests that for the 1954 cohort the difference in the likelihood of choosing vocational secondary education over no education is almost entirely driven by secondary effects. This finding might be attributable to vocational occupations generally having low social status in the 1954 cohort (being associated with working-class jobs), especially among socioeconomically advantaged families. Consequently, it may be preferable for children from advantaged backgrounds who know that they are not going to enter upper secondary education (maybe because of poor academic ability or lack of motivation) not to pursue any type of secondary education rather than to obtain a vocational education because the latter alternative may be socially stigmatizing. Other mechanisms, for example, advantaged families exploiting social or informational resources rather than (or as a supplement to) investing in academic skills to promote their children's long-term socioeconomic success (e.g., Jæer and Holm 2004, 2007)might also explain why secondary effects dominate.

In the 1984 cohort the results are somewhat different. In this cohort we find a significant negative effect of academic ability on the probability of choosing vocational secondary education over no education. Although counterintuitive, this finding has been reported in previous research (e.g., Jæger 2009). One explanation may be that students are increasingly being selected into vocational secondary education (being distinctly hands-on rather than academic) on the basis of poor academic skills. This trend would concentrate academic low achievers in vocational secondary education. Regarding secondary effects, we find only marginally significant effects of social-class background and parental education after controlling for academic ability. RSEs are also rather low and should be interpreted with caution, because in most cases the class and parental education effects are not statistically significant. Our analysis suggests that primary effects (negative selection on ability) are much more important in explaining family background differences in the choice of vocational secondary education (p.248) versus no education in the 1984 cohort than in the 1954 cohort. A possible explanation for this might be that when the 1954 cohort completed elementary school the vocational education system was much less institutionalized and recognized (in the sense of providing students with formally recognized and marketable skills) than it is today. As a consequence, gaining an apprenticeship position and subsequent success in the labor market (and eventually acquiring high social status) depended to a larger extent on family social connections and access to information, that is, factors that were unrelated to a child's actual academic ability but which are arguably captured by the social-class and parental education variables. By contrast, when the 1984 cohort left elementary school the vocational secondary education system had become much more institutionalized and educational institutions themselves helped students find apprenticeship positions. Possibly, this change in the organization of vocational secondary education has reduced the usefulness of parents' informal resources (secondary effects), and in combination with increasing negative selection on academic ability, these trends may have increased the relative importance of primary effects.

Our results for the multinomial decomposition can be summarized as follows. The first conclusion is that in both the 1954 and the 1984 cohorts we find that family background differences in the likelihood of completing upper secondary education are mainly driven by secondary effects. Our results show that in both cohorts secondary effects account for approximately 70 percent of the association between social-class background and educational outcome and the association between parental education and educational outcome. Given the considerable expansion of upper secondary education between the 1954 cohort and the 1984 cohort, it is remarkable that the relative balance of primary and secondary effects has remained so stable over time (around one quarter of the 1954 and similarly aged cohorts completed upper secondary education, while this was the case for around half of the 1984 and similarly aged cohorts). Indeed, in the context of the expansion of upper secondary education one would expect primary effects to have become less important over time, since academic ability is arguably less discriminating in terms of securing access to upper secondary education today than it used to be. However, upper secondary education remains the only gateway to higher education in Denmark, which means that family factors other than those that go into the production of children's academic ability (intergenerational status preservation, other noneconomic returns to (p.249) education, etc., see Jæger 2007a; Jæger and Holm 2007) arguably continue to be important for educational decision making. This situation would explain the persisting importance of secondary effects.

Our second main conclusion is that primary effects have become relatively more important vis-à-vis secondary effects over time with respect to explaining family background differences in the likelihood of choosing vocational secondary education relative to no education. A possible explanation of this finding might be that the vocational education system has become much more institutionalized over the period that we study and parents' noneconomic resources (social connections, information) are of lesser importance today compared to the past.

Conclusion

We decompose the total effect of family background on choice of secondary education in Denmark into primary and secondary effects. We compare the relative magnitude of primary and secondary effects across two cohorts: an older cohort that made the transition from elementary to secondary education around 1970 and a younger cohort that made this transition around 2000. We also extend the traditional approach to modeling primary and secondary effects to accommodate the institutional structure of the Danish secondary education system, which offers two (rather than one) tracks: the academic upper secondary track and the hands-on vocational track.

Our empirical analysis suggests that in Denmark secondary effects make up the lion's share of observable family background inequalities in the likelihood of succeeding in the academic track in secondary education. Secondary effects arguably arise from social-class differences in educational decision-making strategies, cultural capital, and other noncognitive resources, all factors that have previously been shown to be important in Denmark. Interestingly, we find that the relative importance of secondary effects on the likelihood of completing upper secondary education has changed very little over the period that we study (1970–2000) despite the rapid expansion of upper secondary education and lower overall inequality in educational opportunity. We argue that the role of upper secondary education as the exclusive gateway to higher education, alongside its consistently academic profile (a profile that remained largely unchanged over (p.250) the period that we study), is likely to be the principal explanation of this finding. The contrast between those who do not complete any type of secondary education and those who complete vocational secondary education shows radically different results. We find that in the 1954 cohort academic ability does not matter and that almost all observable differences in the likelihood of completing vocational secondary education by social-class background and parental education are driven by secondary effects. We speculate that parents' social resources (and in particular their ability to help their child find an apprenticeship position or a job) and a fragmented vocational educational system at the time were key ingredients in generating strong secondary effects. By contrast, in the 1984 cohort we find that primary effects now dominate and secondary effects are of little importance. We argue that this change over time may be due to a rapid institutionalization and professionalization of vocational secondary education in Denmark, which has reduced the usefulness of parents' social resources and has led to an increasingly negative selection of students into vocational secondary education based on poor academic skills.

Our analysis shows that inequality in educational outcomes exists in Denmark and furthermore that secondary effects, at least with respect to the likelihood of completing upper secondary education, are particularly strong compared to most other countries (with the exception of Italy). Consequently, in Denmark, social-class background and parental education have large direct effects on educational outcomes after accounting for their impact on academic ability. The reason for the dominance of secondary effects might be attributable to, first, no tracking in elementary school; second, enrollment in upper secondary education at the end of elementary school not formally being linked to academic ability (unlike, for example, in Germany); and third, upper secondary education being very academically oriented. These conditions allow academically weak students to gain easy access to upper secondary education (thereby reducing the impact of primary effects on enrollment), but once students get in, they do not have an easy time because of the strong academic orientation of upper secondary education (thereby increasing the importance of secondary effects on completion, such as those resulting from differences in parental cultural and social resources). In this sense the Danish hybrid between a German-oriented path-dependent and a Scandinavian equality-oriented educational system gives rise to strong secondary effects.

(p.251) Notes

References

Bibliography references:

Andersen, Annemarie Møller, Niels Egelund, Torben Pilegaard Jensen, Michael Krone, Lena Lindenskov, and Jan Mejding. 2001. Forventninger og færdigheder: danske unge i en international sammenhæng. Copenhagen: AKF, DPU og SFI-Survey.

Andersen, Dines. 2005. 4 år efter grundskolen. 19-årige om valg og veje i ungdomsuddannelserne. Copenhagen: Institute of Local Government Studies.

Benjaminsen, Lars. 2006. Chanceulighed i det 20. århundrede. Copenhagen: PhD diss., Sociologisk Institut, Københavns Universitet.

Boudon, Raymond. 1974. Education, Opportunity, and Social Inequality: Changing Prospects in Western Society. New York: Wiley.

Breen, Richard, and Jan O. Jonsson. 2005. “Inequality of Opportunity in Comparative Perspective: Recent Research on Educational Attainment and Social Mobility.” Annual Review of Sociology 31:223–43.

Breen, Richard, Ruud Luijkx, Walter Müller, and Reinhard Pollak. 2009. “Nonpersistent Inequality in Educational Attainment: Evidence from Eight European Countries.” American Journal of Sociology 114:1475–521.

Buis, Maarten. 2010 “Direct and Indirect Effects in a Logit Model.” Stata Journal 10:11–29.

Erikson, Robert. 2007. “Social Selection in Stockholm Schools: Primary and Secondary Effects on the Transition to Upper Secondary Education.” In From Origin to Destination: Trends and Mechanisms in Social Stratification Research, edited by Stefani Scherer, Reinhard Pollak, Gunnar Otte, and Markus Gangl, 35–76. Frankfurt, Germany: Campus Verlag.

(p.252) Erikson, Robert, and John H. Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies. London: Clarendon Press.

Erikson, Robert, John H. Goldthorpe, Michelle Jackson, Meir Yaish, and David R. Cox. 2005. “On Class Differentials in Educational Attainment.” Proceedings of the National Academy of Sciences 102:30–33.

Hansen, Erik Jørgen. 1995. En generation blev voksen. Copenhagen: Socialforskningsinstituttet.

Holm, Anders, and Mads Meier Jæger. 2008. “Does Relative Risk Aversion Explain Educational Inequality? A Dynamic Choice Approach.” Research in Social Stratification and Mobility 26:199–220.

Jæger, Mads Meier. 2007a. “Educational Mobility across Three Generations: The Changing Impact of Social Class, Economic, Cultural, and Social Capital.” European Societies 9:527–50.

———. 2007b. “Economic and Social Returns to Educational Choices: Extending the Utility Function.” Rationality and Society 19:451–83.

———. 2009. “Equal Access but Unequal Outcomes: Cultural Capital and Educational Choice in a Meritocratic Society.” Social Forces 87:1943–71.

Jæger, Mads Meier, and Anders Holm. 2004. “Penge, (ud)dannelse, forbindelser eller brains? En test af fire forældreressourcers betydning for unges uddannelsesvalg i Danmark.” Dansk Sociologi 15:67–83.

———. 2007. “Does Parents' Economic, Cultural, and Social Capital Explain the Social Class Effect on Educational Attainment in the Scandinavian Mobility Regime?”Social Science Research 36:719–44.

Karlson, Kristian B., and Anders Holm. 2010. “Decomposing Primary and Secondary Effects: A New Decomposition Method.” Research in Social Stratification and Mobility 29:221–37.

Karlson, Kristian B., Anders Holm, and Richard Breen. Forthcoming. “Comparing Regression Coefficients between Same-Sample Nested Models Using Logit and Probit: A New Method.” Sociological Methodology.

McIntosh, James, and Martin D. Munk. 2007. “Scholastic Ability vs. Family Background in Educational Success: Evidence From Danish Sample Survey Data.” Journal of Population Economics 20:101–20.

OECD. 2000. PISA Technical Report. Paris: OECD.

Thomsen, Jens Peter. 2008. Social differentiering og kulturel praksis på danske universiteter. Roskilde, Denmark: Roskilde Universitetsforlag.

Undervisningsministeriet [Ministry of Education]. 2010. Tal, der taler 2009. Uddannelsesnøgletal 2009. Copenhagen: Ministry of Education.

Notes:

(1.) Educational noncompletion is somewhat overestimated in the 1984 cohort because respondents were interviewed in their late teens or early 20s. Consequently, some respondents who reported at age 19 not having completed any education beyond elementary school would eventually complete some type of secondary education (see Jæger 2009).

(2.) For more information on the DLSY, see http://www.sfi.dk/dlsy. The missing data that we observe on the family background variables arise from information on parental social class and parental education being collected from parents themselves in a separate postal survey in 1968 (the primary DLSY respondents also made retrospective reports of their parents' education and occupation in later surveys, and we have used this information whenever information from parents themselves was not available).

(3.) Jæger (2009) reports that in 1996, of all students who applied for upper secondary education, 97.2 percent were admitted.