Beyond Graduation: Socio-economic Background and Post-university Outcomes of Australian Graduates

Abstract

Research consistently shows that higher-education participation has positive impacts on individual outcomes. However, few studies explicitly consider differences in these impacts by socio-economic background (SEB), and those which do fail to examine graduate trajectories over the long run, non-labor outcomes and relative returns. We address these knowledge gaps by investigating the short- and long-term socio-economic trajectories of Australian university graduates from advantaged and disadvantaged backgrounds across multiple domains. We use high-quality longitudinal data from two sources: the Australian Longitudinal Census Dataset and the Household, Income and Labour Dynamics in Australia Survey. Low-SEB graduates experienced short-term post-graduation disadvantage in employment and occupational status, but not wages. They also experienced lower job and financial security up to 5 years post-graduation. Despite this, low-SEB graduates benefited more from higher education in relative terms—that is, university education improves the situation of low-SEB individuals to a greater extent than it does for high-SEB individuals.

Full text

Vol.:(0123456789)
Research in Higher Education
https://doi.org/10.1007/s11162-019-09578-4
1 3
Beyond Graduation: Socio‑economic Background
andPost‑university Outcomes ofAustralian Graduates
WojtekTomaszewski1,2 · FranciscoPerales1,2· NingXiang1,2· MatthiasKubler1,2
Received: 6 December 2018
© Springer Nature B.V. 2019
Abstract
Research consistently shows that higher-education participation has positive impacts on
individual outcomes. However, few studies explicitly consider differences in these impacts
by socio-economic background (SEB), and those which do fail to examine graduate tra-
jectories over the long run, non-labor outcomes and relative returns. We address these
knowledge gaps by investigating the short- and long-term socio-economic trajectories of
Australian university graduates from advantaged and disadvantaged backgrounds across
multiple domains. We use high-quality longitudinal data from two sources: the Australian
Longitudinal Census Dataset and the Household, Income and Labour Dynamics in Aus-
tralia Survey. Low-SEB graduates experienced short-term post-graduation disadvantage in
employment and occupational status, but not wages. They also experienced lower job and
financial security up to 5years post-graduation. Despite this, low-SEB graduates benefited
more from higher education in relative terms—that is, university education improves the
situation of low-SEB individuals to a greater extent than it does for high-SEB individuals.
Keywords Higher education· Post-graduate outcomes· Longitudinal trajectories· Panel
data· Australia
Background
The benefits of attaining university-level educational qualifications are well documented.
Individuals who complete university education generally enjoy better labor-market pros-
pects. For example, across OECD countries 7% of university-educated adults aged
25–34 year-olds are unemployed, compared to 9% for those with upper-secondary and
post-secondary qualifications, and 17% of those with lower credentials (OECD 2017). In
Australia, the focal country in this study, employment rates are substantially higher for
individuals holding postgraduate (82%) and bachelor (80%) degrees than for individuals
* Wojtek Tomaszewski
w.tomaszewski@uq.edu.au
1 Institute forSocial Science Research, The University ofQueensland, Long Pocket Precinct, 80
Meiers Road, Indooroopilly, QLD4068, Australia
2 Australian Research Council Centre ofExcellence forChildren andFamilies Over theLife Course,
The University ofQueensland, Indooroopilly, Australia

Research in Higher Education
1 3
without post-school qualifications (54%) (ABS 2017a). University graduates are also more
likely to receive higher wages and work in more prestigious occupations, internationally
(Card 1999; Desjardins and Lee 2016; Heckman etal. 2016), and in Australia (Cassells
etal. 2012; Daly etal. 2015). The positive outcomes of university education are not con-
fined to the labor market, with research documenting positive influences in other domains,
including mental health (Heckman etal. 2017), general health (Cutler and Lleras-Muney
2008), and subjective wellbeing (Oreopoulos and Salvanes 2011).
Because of this, sociologists have long been interested in the social patterning in access
to and completion of higher education, as well as in how the benefits of higher-education
participation differ across social strata. Of key importance has been the role of socio-
economic background (SEB), as its associations with education are pivotal to the study of
social mobility and equality of opportunity. Over two decades ago, Hout (1984) reported
no association between social origins and occupational status among US higher-education
graduates, a finding which some interpreted as a sign of the meritocratic function of uni-
versity (Breen and Jonsson 2007). Yet more recent studies paint a more complex picture,
suggesting differing returns to university participation by SEB, with such returns depend-
ing also on factors such as qualification level or field-of-study (Torche 2011). This paper
contributes to the literature on the returns to higher education for low- and high-SEB grad-
uates in several ways. First, it expands the focus from employment outcomes to broader
measures of health and wellbeing—hence providing a more comprehensive picture of the
benefits of higher-education participation. Second, it examines how post-graduation trajec-
tories in outcomes evolve over time using longitudinal data and methods—thereby offering
a better window into the short- and long-term outcomes of low- and high-SEB graduates.
Third, it focuses on both absolute and relative returns to higher education for low- and
high-SEB graduates.
Conceptual Framework
In this section, we discuss key perspectives from multiple disciplines, including sociology,
economics, and industrial relations, theorizing the relationship between higher-education
participation and individual outcomes. While these approaches generally aim to explain
the ‘higher education participation → personal outcomes’ association, we build on them to
derive testable hypotheses about the ‘higher education → differences in personal outcomes
by SEB’ nexus. We first discuss perspectives suggesting that higher-education attainment
should result in more similar outcomes between low- and high-SEB individuals (human
capital, signaling, and rational action theories). Collectively, we refer to the mechanisms
proposed by these theories as ‘levelling forces’. Second, we discuss perspectives postu-
lating that higher-education attainment should result in more disparate outcomes between
low- and high-SEB individuals (social and cultural capital, effectively maintained inequal-
ity, and life-course theories). We refer to the mechanisms suggested by these theories as
‘stratifying forces’. In doing so, we explicitly recognize that multiple—often competing—
mechanisms might operate at the same time, contributing to either narrowing or widen-
ing differences in personal outcomes between low- and high-SEB graduates. That is, we
acknowledge that—far from being mutually exclusive—‘levelling’ and ‘stratifying’ forces
operate concurrently. Importantly, while we use these frameworks to develop hypotheses
about overall differences in post-graduation outcomes by SEB, the aim of this study is not

Research in Higher Education
1 3
to provide specific tests of each theory (e.g., by including variables approximating social
networks, productivity or socio-cultural capital in the models).
Levelling Forces
Several perspectives lead to the expectation that low-SEB graduates will benefit from
degree attainment to a similar extent as high-SEB graduates. Under human capital the-
ory (Becker 1964), university participation is a key mechanism whereby people learn new
knowledge and skills that increase their labor-market productivity. Accordingly, stud-
ies have documented causal effects of higher-education participation and attainment on a
range of outcomes, with the effects of university education being driven by learnt cognitive
and non-cognitive skills (Heckman etal. 2016). If university education raises productiv-
ity to a similar level for low- and high-SEB individuals, we should expect no differences
in the returns to their university qualifications. Signaling theory (Spence 1973) also pre-
dicts that university-degree attainment will be associated with better outcomes, but differs
from human capital theory in the proposed mechanisms. Because employers are unable
to directly assess the productivity of job applicants, they use their educational credentials
(e.g., a university diploma) as a ‘signal’ of productivity. From this perspective, employers
should not differentiate between low- and high-SEB applicants in their hiring practices, so
long as they have attained commensurate levels of education. Arguments based on rational
action theory (Goldthorpe 1996) lead to a similar set of expectations: because the rela-
tive costs of attending university are higher for low- than high-SEB individuals, low-SEB
individuals weigh the potential costs and benefits of higher-education participation more
carefully than their high-SEB peers (Flaster 2016). Only those low-SEB individuals that
have the highest success chances (e.g., through demonstrated excellent academic aptitude)
choose to pursue higher education. These positively-selected low-SEB individuals are
likely to accrue cognitive and non-cognitive skills from university participation at similar
rates as their high-SEB peers.
Stratifying Forces
Unlike the theories discussed thus far, several other perspectives postulate greater returns to
higher-education participation amongst high-SEB than low-SEB individuals. Social capital
theory (Coleman 1988) draws attention to the importance of access to information chan-
nels for individuals to navigate social structures. Low-SEB graduates have less developed
social networks and their networks usually comprise other relatively under-resourced low-
SEB individuals (Lin 1999). This hinders their ability to access information about high-end
jobs, or navigate selection processes (Coleman 1988; Lin 1999). Similarly, cultural capital
theory (Bourdieu 1984) posits that employers are biased towards hiring individuals similar
to them (homo-social reproduction), restricting low-SEB graduates’ ability to access high-
status, high-paying occupations. The theory of effectively maintained inequality (Lucas
2001) posits that, as higher-education participation becomes more common, high-SEB par-
ents increase their investments in their children so that they can differentiate themselves
from other university graduates. This includes subsidizing higher-status tertiary-education
options, including more prestigious disciplines (e.g., medical, engineering) and institutions
(e.g., Australian Go8 institutions). Finally, according to the life-course approach (Elder
et al. 2003), inter-relationships between life domains are important in structuring indi-
vidual outcomes. As such, low-SEB graduates may be more likely to experience negative

Research in Higher Education
1 3
life events in domains other than employment or education (e.g., health problems, family
breakdown, financial difficulties) than their high-SEB peers (Umberson etal. 2014). Expo-
sure to these stressors may restrict the ability of low-SEB graduates to pursue, focus on and
develop their work careers, so as to benefit from their educational attainment to the same
extend as their high-SEB peers.
Theoretical Expectations andExisting Empirical Evidence
Socio‑Economic Stratication ofGraduate Outcomes
As outlined before, some theories postulate similar outcomes for low- and high-SEB grad-
uates, while others hypothesise inferior outcomes for low- than high-SEB graduates. As
noted earlier, we recognize that ‘levelling’ and ‘stratifying’ forces will operate concur-
rently. For examples, university degrees may provide all graduates with the same skills and
signals to employers but, at the same time, high-SEB graduates may graduate from better
universities, enjoy higher levels of socio-cultural capital, and experience fewer challenges
in other life domains. Therefore, despite ‘levelling forces’ potentially generating similar
skills amongst low- and high-SEB graduates that send comparable signals to employers, it
is likely that high-SEB graduates can draw on their social or cultural capital to obtain com-
petitive advantages or avoid other life stressors. Therefore, all in all, we hypothesize that:
Hypothesis 1 Low-SEB graduates will achieve worse post-graduation outcomes than
high-SEB graduates.
Several studies from OECD countries have generated empirical evidence consistent
with this proposition. For Example, Hansen (2001) documented that high-SEB individuals
in Norway received higher economic returns to university participation than low-SEB indi-
viduals, net of qualification level and field of study. Similarly, Triventi (2013) found that
European graduates in Norway, Italy and Spain whose parents had also university qualifi-
cations were more likely to have attained a high-status occupation 5years post-graduation
than similar graduates whose parents did not hold university qualifications. However, no
such pattern was observed amongst German graduates. The limited Australian evidence
available is nevertheless mixed. Richardson etal. (2016) found that low-SEB graduates
were less likely than high-SEB graduates to be employed 6months post-graduation, while
Li and colleagues (2017) found no significant employment differences. In the longer run,
Edwards and Coates (2011) found that, 5years post-graduation, low- and high-SEB grad-
uates had similar rates of employment and employment in a high-status occupation and
median annual salaries.
The present study provides more robust Australian evidence encompassing labor and
non-labor personal outcomes. This is an important contribution, as it recognizes that uni-
versity education can be a driver for positive personal outcomes beyond the realm of work.
Critically, it also expands the international evidence base by postulating and testing (i) dif-
ferences in the longitudinal trajectories of low- and high-SEB graduates, and (ii) the rela-
tive—rather than absolute—returns to university participation for these two groups. The
next sections present theoretical arguments that enable us to postulate testable hypotheses
about these.

Research in Higher Education
1 3
Change Over Time intheRelationship Between Socio‑Economic Background
andGraduate Outcomes
As argued before, social and cultural capital are ‘stratifying forces’ leading to better
graduate outcomes for high-SEB than low-SEB individuals. Yet previous studies sug-
gest that social and cultural capital play a more prominent role immediately after gradu-
ation: high levels of social capital may enable high-SEB graduates to deploy their social
networks to obtain (better) first jobs earlier than their low-SEB peers (Coleman 1988;
Jackson etal. 2005; Lin 1999). In contrast, human capital—which was characterized
as a ‘levelling force’—may play a more important role over the long run (Jacob etal.
2015). If low- and high-SEB graduates possess similar skills, their demonstrated job
performance should serve as a more direct signal to employers than their social ori-
gins as time elapses. Altogether, these arguments suggest that less favorable initial out-
comes for low-SEB graduates (as proposed in Hypothesis 1) should fade over time, as
these graduates socialize into their work environments, learn skills on-the-job, and pro-
vide employers with opportunities to directly assess their performance. Therefore, we
hypothesize that:
Hypothesis 2 Any differences in the post-graduation outcomes of low- and high-SEB
graduates will fade over time.
To our knowledge, only one empirical study has compared to some extent the post-grad-
uation longitudinal trajectories in outcomes of low- and high-SEB individuals. Jacob etal.
(2015) examined the effect of parental education on university graduates’ occupational out-
comes at labor-market entry and 5years post-graduation in Germany and the UK. Their
findings are consistent with our second hypothesis: high-SEB individuals had a compara-
tive advantage over low-SEB graduates in entering high-status occupations, but this effect
was stronger at labor market entry than 5years after graduation.
Relative Returns toUniversity Education bySocio‑Economic Background
The reviewed theories have been predominantly applied to investigate absolute differences
in outcomes between low- and high-SEB university graduates. However, a separate and
equally important question is whether or not low-SEB graduates gain more or less from a
university degree in relative terms (i.e., compared to themselves prior to graduation). Even
if high-SEB graduates have better labor-market outcomes than low-SEB graduates, the ben-
efits accrued with graduation may be greater for low- than high-SEB graduates in relative
terms. Low-SEB individuals and their families experience less favorable objective circum-
stances than high-SEB graduates (e.g., financial situation, living standards). As a result,
access to high-paying jobs within the graduate job market will often translate into signifi-
cant improvements in income and financial prosperity for low-SEB individuals (Brand and
Xie 2010). This may not be true for high-SEB graduates, for whom the same employment
outcomes may not represent commensurate changes in objective circumstances. For exam-
ple, a young medicine graduate working as a doctor who comes from a family of blue-col-
lar workers will experience more substantial relative improvement in their circumstances
than an otherwise similar medicine graduate working as a doctor who comes from a family
of doctors. Based on these considerations, we formulate a final hypothesis:

Research in Higher Education
1 3
Hypothesis 3 Positive before/after graduation differences in outcomes will be larger
amongst low-SEB than high-SEB graduates.
Our review of the available literature identified only one previous study examining the
relative rather than absolute returns to higher education by SEB. Drawing on US longitu-
dinal data, Brand and Xie’s (2010) found that low-SEB graduates benefit more from higher
education than high-SEB graduates in terms of their earnings.
Data
We use data from two authoritative sources: the Australian Census Longitudinal Dataset
(ACLD) and the Household, Income and Labour Dynamics in Australia (HILDA) Survey.
The Australian Census Longitudinal Dataset
The Australian Census of Population and Housing (the Census) is undertaken by the Aus-
tralian Bureau of Statistics (ABS) every 5years, collecting information from the complete
Australian population (ABS 2017b). To evaluate the short-term labor-market outcomes
of recent university graduates we analyze data from the ACLD, a longitudinal extension
of the Census (ABS 2018a). The ACLD 2011–2016 panel is a linked dataset that com-
bines information from two consecutive censuses (2011 and 2016) for a 5.7% random
sample of the Australian population. Of the 1,221,057 records selected from the 2011
Census, 76% were linked to 2016 records. The majority of these records (72.7%) were
linked using deterministic matching based on personal and demographic characteristics,
with the remainder being linked by probabilistic matching (ABS 2018b). This resulted in
927,520 linked records. We focus on a sample of young people aged 15–17 in 2011 and
20–22 in 2016 (n = 48,399). This allows capturing socio-economic background informa-
tion when cohort members attended secondary education in 2011, as well as early employ-
ment post-university destinations in 2016. We then restricted the sample to those young
people who completed a Bachelor degree between 2011 and 2016 (n = 3040). The final
analytic sample varies depending on the outcome variable of interest, ranging from 3023
individuals (employment) to 1207 individuals (weekly income for individuals in full-time
employment). The age of the selected cohort of young people (15–17years in 2011) means
that cohort members are observed at ages 20–22years in 2016. Hence, the outcomes for
most of these young people are observed up to 2years post-graduation (OECD 2017). The
advantages of ACLD are its reliability, robustness and large sample size to study small
subpopulations. Its disadvantages include the limited scope of the information collected
(which restricts our analysis to labor-market outcomes) and the relatively short-term time-
frame post-graduation (up to 2years).
The Household, Income andLabour Dynamics inAustralia (HILDA) Survey
The Household, Income and Labour Dynamics in Australia (HILDA) Survey is an annual
household panel survey covering the 2001–2016 period that contains rich information
from a sample of individuals aged 15 and older (Watson and Wooden 2012). The initial
HILDA Survey sample is largely representative of the Australian population in 2001. The

Research in Higher Education
1 3
HILDA Survey data are collected using a complex, multi-stage sampling strategy at the
household level, and a mixture of self-complete questionnaires and computer-assisted face-
to-face interviews. Sample sizes range between 12,226 and 17,400 individuals across the
16 HILDA Survey waves utilized. Pooling all HILDA Survey waves we obtained a sam-
ple of 12,074 observations from 1105 individuals who were observed at least twice and
obtained a Bachelor degree during the life of the panel. This sample is used to examine
the differences before/after attaining a degree on health and wellbeing outcomes. It will be
referred to as the before/after sample. To examine trends in outcomes post-graduation, we
exclude those observations prior to individuals obtaining their degrees (7076 observations
dropped). This yields a subsample of 4998 observations from 935 individuals. This will be
referred to as the trajectory sample. Of note, we do not exclude individuals with informa-
tion in some but not all of the outcome variables. Hence, the final analytic numbers will
depend on the outcome under consideration. The HILDA Survey offers distinct analytic
advantages: it collects rich information on non-labor outcomes (e.g., health and wellbeing)
and its panel structure allows examining how post-graduation outcomes evolve for up to
15years. A disadvantage of the HILDA Survey is its comparatively small sample size for
the target population, as ‘only’ 1105 individuals are observed to graduate from university.
Measures
Socio‑Economic Background
We use information on parental occupation to operationalize SEB. In ACLD, we extract
information about the occupational status of parents co-residing with our sample of young
people in 2011. Young people in households in which at least one parent worked in a mana-
gerial/professional occupation were considered to be ‘high-SEB’, and the remaining young
people as ‘low-SEB’. In the HILDA Survey, paternal and maternal occupation information
was captured using respondent-reported retrospective data pertaining to when the respond-
ent was 14years of age. Individual in households in which at least one parent worked in a
managerial/professional occupation qualified as ‘high-SEB’, and the remaining individuals
as ‘low-SEB’. In both datasets, managerial/professional occupations are those in codes 1
and 2 of the Australian and New Zealand Standard Classification of Occupations 2006 at
the major-group level (ABS 2006).
Outcome Variables
Three labor-market outcome variables are used in the ACLD analyses. Employment status
is captured through a binary indicator taking the value 1 if the individual was employed,
and the value 0 otherwise (including unemployment and not in the labor force). Work in
a managerial/professional occupation is denoted by a binary variable taking the value 1 if
the individual worked in a managerial/professional occupation (defined as for the parents
above), and the value 0 if the individual worked in another occupation—non-employed
individuals are assigned missing values. Finally, high income is captured through a binary
variable taking the value 1 if the individual’s gross individual weekly income was over
AU$1250 per week, and the value 0 otherwise. Of note, income information in the Cen-
sus is banded (e.g., AU$1000–AU$1249 per week, AU$1250–AU$1499 per week;
AU$1500–AU$1749 per week, etc.). As such, we cannot select a specific percentile of the
income distribution (e.g., the usual 20th or 25th percentile) when defining our high-income

Research in Higher Education
1 3
variable. The AU$1250 threshold identifies a small—but not too small—proportion of top
income earners (about 17% of Bachelor degree holders in full-time employment). Using
the immediately preceding or immediately posterior thresholds of AU$1000 and AU$1500
would have resulted in too many (46%) or too few (5%) individuals in the top-earning
group.
In the HILDA Survey analyses, we focus on four outcome variables pertaining to labor-
market circumstances, health and wellbeing. Hourly wages are generated by dividing
current weekly gross wages and salary from all jobs by weekly hours usually worked in
all jobs. The resulting figure is adjusted to 2016 prices using the Consumer Price Index.
To correct for a right-skewed distribution, in regression models we use the natural log
of hourly wages. Job-security satisfaction is determined from a question asking partici-
pants about their satisfaction with job security on a scale from 0 (totally dissatisfied) to
10 (totally satisfied). Mental health is captured using the mental health subscale of the
SF-36, a 5-item additive scale with transformed scores ranging from 0 to 100 (Ware and
Sherbourne 1992). Financial prosperity is based on a question asking participants to rate
their “prosperity given current needs and financial responsibilities” using the follow-
ing response options: 1 = Prosperous, 2 = Very comfortable, 3 = Reasonably comfortable,
4 = Just getting along, 5 = Poor and 6 = Very poor. In regression models, we treat this as a
continuous-level variable.
Control Variables
In multivariate models we control for a parsimonious set of potential confounds measured
in 2011. In ACLD analyses these include gender (male; female), residence in a regional
or remote area (based on the Remoteness Area classification of the Australian Statistical
Geography Standard, ABS 2018c), and area-level socio-economic disadvantage (based
on the lowest quintile of the Index of Education and Occupation of the Socio-Economic
Indexes for Areas, ABS 2018d). In the HILDA Survey, controls include time-varying vari-
ables capturing respondents’ age (in years), gender (male; female), attainment of a post-
graduate qualification (attained; not attained) and partnership status (partnered; not part-
nered). When modelling health and wellbeing outcomes in the HILDA Survey, we also
control for employment status (employed; not employed). Tables1 and 2 present descrip-
tive statistics for all analytic variablesin ACLD and HILDA respectively.
Analytic Approach
ACLD Analyses
Analyses of ACLD rely on standard, cross-sectional logistic regression models of the fol-
lowing form:
where EO is a given employment outcome measured in 2016, SEB is a binary indicator
for low SEB, C is a vector of control variables, the βs represent coefficients or vectors of
coefficients to be estimated, and e is the usual random error in regression. The key model
coefficient is β1l, which gives the estimated difference in employment outcomes between
high-SEB and low-SEB individuals. To facilitate the interpretation of results, we present
(1)
ln (
p(EO)
1−
p
(
EO
)
)
=SEB𝛽1+C𝛽2+
e

Research in Higher Education
1 3
Table 1 Descriptive statistics of ACLD data. ACLD 2011–2016, unweighted data extracted using TableBuilder
% Range Obs. Population
Outcomes People aged 15–17years in 2011
Employed 80.1 0–1 3023 with a Bachelor degree in 2016
Works in professional/managerial occ. 48.1 0–1 2429 with a Bachelor degree and in employment in 2016
Weekly income of $1250 or more 16.7 0–1 1207 with Bachelor degree and in full-time employment in 2016
Key predictor
High SEB 51.4 0–1 3023 People aged 15–17years in 2011 and with Bachelor degree in 2016
Controls
Lowest SEIFA quintile 9.2 0–1 3023 People aged 15–17years in 2011 and with Bachelor degree in 2016
Regional/remote 17.8 0–1 3023 People aged 15–17years in 2011 and with Bachelor degree in 2016
Female 61.6 0–1 3023 People aged 15–17years in 2011 and with Bachelor degree in 2016

Research in Higher Education
1 3
average marginal effects (AMEs) (for details see Greene 2012, Chapter17; for applications
see Manly etal. 2019; Tieben 2019).
HILDA Survey Analyses: Growth‑Curve Modelling
Two sets of analyses are executed using the HILDA Survey: one examining long-term
post-graduation trajectories in outcomes and one examining changes in outcomes before
and after individuals obtain a university degree. To track the post-graduation trajectories
of low- and high-SEB graduates, we fit growth-curve models (Singer and Willett 2003:
Chapter8). Growth-curve models are statistical techniques that expand multilevel, random-
intercept models to “allow for the estimation of inter-individual variability in intra-indi-
vidual patterns of change over time” (Curran etal. 2010, p. 2). These models are useful to
determine the evolution of an outcome with time elapsed since a given event. In our case,
the event is graduation from an undergraduate university degree, and the outcome are dif-
ferent variables capturing health, subjective wellbeing and labor-market circumstances. We
fit linear growth-curve models, as the outcomes of interest in this part of the analysis are—
or can be treated as—continuous:
where i and t denote individual and time; HW is an outcome variable capturing a given
dimension of health and subjective wellbeing, YSG is a time-varying continuous variable
capturing the number of years since graduation (ranging from 1 to 15), SEB is a time-
constant binary indicator of low-SEB; C is a vector of time-changing control variables,
the βs represent coefficients or vectors of coefficients to be estimated, e is the usual ran-
dom error in regression, and u is an individual-specific random intercept capturing unob-
served effects. The interaction effect between YSG and SEB (i.e., β3) is the parameter of
(2)
HWit =YSGit 𝛽1i+SEBi𝛽2+(YSGit ∗+SEBi)𝛽3+Cit𝛽4+ui+eit
Table 2 Descriptive statistics of HILDA survey data
HILDA Survey (2001–2016)
Trajectory sample Before/after sample
Mean/% SD Range Obs. Mean/% SD Range Obs.
Degree attainment
Observed degree attainment 51% 0–1 12,074
Years after degree attainment 4.74 3.28 1–14 4998
Key predictor
Low SEB 38% 0–1 4998 37% 0–1 12,074
Outcomes
Mental health 73.27 15.83 4–100 4543 73.32 15.71 4–100 11,056
Financial prosperity 4.03 0.79 1–6 4534 4.01 0.80 1–6
Log of hourly wages 3.47 0.41 − 0.73 to 5.74 3883
Job-security satisfaction 7.96 2.02 0–10 4488
Controls
Age (in years) 30.41 8.23 18–74 4998 25.67 8.66 15–74 12,074
Male 41% 0–1 4998 40% 0–1 12,074
Postgraduate degree attained 17% 0–1 4998 7% 0–1 12,074
Partnered 55% 0–1 4998 34% 0–1 12,074

Research in Higher Education
1 3
key interest, as it gives the differences in post-graduation trends in outcomes between low-
and high-SEB graduates. In some specifications we used a polynomial specification for the
YSG variable (and its interaction with low-SEB) to capture quadratic trends since gradua-
tion. We do this when its addition significantly improves the model fit.
HILDA Survey Analyses: Fixed‑Eect Panel Regression Models
Our second set of HILDA analyses compares the outcomes of individuals before and after
attaining an undergraduate university degree. Using the HILDA Survey, we can ascertain
when an individual graduates by comparing his/her highest educational qualification at a
given wave (time t) and the previous wave (time t − 1). Based on this comparison, we first
derive a dummy variable capturing the time at which the highest educational qualification
recorded in the data moves from any qualification lower than a degree at time t − 1 into
‘undergraduate degree’ at time t. We then create an additional dummy variable (G) that
distinguishes all observations prior to graduation (value 0) and all observations subsequent
to graduation (value 1). This variable is then interacted with the dummy variable captur-
ing the low-SEB for use in fixed-effect panel regression models. These models compare
the health and subjective wellbeing of the same individuals before and after they obtain
their degree.1 In practice, the fixed-effect model is estimated by regressing deviations in
person-specific means in the outcome variable on deviations in person-specific means in
the explanatory variables (Allison 2009; Perales 2019). We fit linear fixed-effect models,
as the outcomes of interest in this part of the analysis are continuous. An initial version of
our model can be formally represented as:
where all notation is as for Eq.(2) above. Because fixed-effect models are estimated using
within-individual change over time, they cannot accommodate time-constant predictors.
However, they can accommodate interactions between time-constant and time-varying pre-
dictors (Allison 2009; Perales 2019). Our key interest is in one such interaction, namely
that between low-SEB (time constant) and attainment of a degree (time varying). Hence,
the models we actually fit are as follows:
where GL and GH represent graduating from a degree by low- and high-SEB individuals,
respectively. A comparison of the estimated β coefficients on these two terms via Wald
tests provides the requisite evidence of whether or not degree attainment impacts the out-
comes of low- and high-SEB individuals to the same extent.
(3)
HWit
−HW
i
=
(
G
it
−̄
G
i)
𝛽1+
(
C
it
−̄
C
i)
𝛽2+
(
e
it
−̄e
i)
(4)
HWit −HWi=

DLit −DLi

𝛽1+

DHit −DHi

𝛽2+

Cit −̄
Ci

𝛽3+

eit −̄ei

1 These fixed-effect models are not to be confused with difference-in-difference models (Donald and Lang
2007). Difference-in-difference models compare the pre/post outcomes of a group of individuals exposed
to a ‘treatment’ (in our case degree attainment) and a control group of individuals not exposed to the same
‘treatment’. Difference-in-difference models require additional assumptions. This includes the parallel trend
assumption—namely that, in the absence of the treatment, differences in outcomes between the treatment
and control groups would be constant over time.

Research in Higher Education
1 3
Results
ACLD: Comparison ofOutcomes After Degree Attainment (Hypothesis 1)
Consistent with Hypothesis 1, results of the ACLD analyses (Table 3) yield bivariate evi-
dence of poorer outcomes for low-than high-SEB graduates concerning employment (mean
high-SEB: 82%, mean low-SEB: 78%), and employment in a managerial/professional
occupation (mean high-SEB: 52%, mean low-SEB: 44%). t-tests indicate that these dif-
ferences are statistically significant. However, the proportion of high-income earners is
not significantly different by SEB (mean high-SEB: 16%, mean low-SEB: 17%, p: 0.73).
Results from multivariate logistic regression (Table 4) largely confirm the descriptive
results: high-SEB graduates enjoy better outcomes concerning employment (AME = 0.038,
p < 0.01) and work in managerial/professional occupations (AME = 0.079, p < 0.001), but
not income (AME = – 0.015, p > 0.05).
HILDA Survey: Trends Over Time After Degree Attainment (Hypothesis 2)
Results from the first set of HILDA analyses, compare post-graduation trends in outcomes
between low- and high-SEB graduates using growth-curve models (Table 5). Due to the
Table 3 Descriptive analyses of
ACLD data
ACLD 2011–2016, unweighted data extracted using TableBuilder
^ Two-sample t tests with unequal variances
Employed Worked as manager
or professional
Weekly
income
≥ $1250
Low-SEB 78.2% 44.2% 17.0%
High-SEB 81.9% 51.7% 16.3%
t test (p-value)^0.012 < 0.001 0.729
n (individuals) 3023 2429 1207
Table 4 Results from logistic regression models of ACLD data (average marginal effects)
ACLD 2011–2016, unweighted data extracted using TableBuilder
*p < 0.05; **p < 0.01; ***p < 0.001
a In 2016; population aged 15–17 in 2011 with a Bachelor degree in 2016
b Population aged 15–17 in 2011 with a Bachelor degree and in employment in 2016
EmployedaEmployed as manager or
professionalb
Weekly income
≥ $1250b
High-SEB 0.037* 0.038** 0.075*** 0.079*** − 0.007 − 0.015
Controls
Lowest SEIFA quintile − 0.024 0.022 − 0.061
Regional/remote area 0.043* 0.070** 0.035
Female 0.085*** − 0.002 − 0.053*
n (individuals) 3023 2429 1207
Pseudo R20.015 0.007 0.009

Research in Higher Education
1 3
complexity of these analyses and the number of parameters that need to be interpreted jointly,
the results of these models are easier to grasp by visually inspecting the marginal effects in
Fig. 1. Overall, hourly wages and financial prosperity increase with time since graduation,
while mental health and job-security satisfaction remain stable. Concerning differences in out-
comes by SEB (Hypothesis 1), the picture is mixed. The hourly wages and mental health of
low-SEB graduates (red lines) appear to be on par with those of high-SEB graduates (blue
Table 5 Results from growth-curve models using HILDA Survey data (coefficients)
HILDA Survey (2001–2016). Before/after sample
# p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001
Log hourly wage Job-security sat. Mental health Financial prosperity
Key explanatory variables
Low-SEB − 0.00 − 0.30** − 0.59 − 0.26***
Years after degree 0.05*** 0.07* − 0.07 − 0.02
Years after degree, squared − 0.00** − 0.01* 0.00#
Low-SEB * years after degree 0.01 0.08 0.06**
Low-SEB * years after
degree, squared
− 0.00 − 0.00**
Controls
Age 0.01*** − 0.02* − 0.05 − 0.01***
Male 0.05** − 0.05 1.21 0.02
Postgrad − 0.01 − 0.08 − 0.85 0.04
Partnered 0.05** 0.22** 1.55** 0.02
Constant 2.97*** 8.38*** 74.03*** 4.46***
n (observations) 3883 4488 4543 4534
n (individuals) 875 902 899 898
Fig. 1 Marginal effects from growth-curve models. HILDA Survey (2001–2016). Based on results from
growth-curve models presented in Table 5. Covariates held at their means and random effects at zero.
Whiskers denote 90% confidence intervals

Research in Higher Education
1 3
lines). Differences between the two groups are not statistically significant, as can be inferred
from overlapping 90% confidence intervals. However, consistent with Hypothesis 1, job-secu-
rity satisfaction and financial prosperity are comparatively worse amongst low-SEB graduates
in the first 4years post-graduation. Furthermore, consistent with Hypothesis 2, low- and high-
SEB trajectories for these outcomes converge over time. That is, there is a ‘catch up’ effect for
low-SEB graduates resulting in outcomes comparable to those of high-SEB backgrounds.
HILDA Survey: Within‑Individual Changes inOutcomes Before andAfter Degree
Attainment (Hypothesis 3)
Results from fixed-effect models comparing the relative returns to a university degree for low-
and high-SEB individuals are presented in Table6. Attaining a degree significantly improves
the mental health of low-SEB (β = 1.14; p < 0.05) but not high-SEB (β = 0.78; p > 0.05)
individuals. Yet, in Wald tests, differences in these estimates are not statistically significant
(p = 0.49). Low-SEB individuals also report significant improvements in perceived financial
prosperity after attaining an undergraduate degree (β = 0.09; p < 0.001), which again is not the
case for high-SEB individuals (β = 0.02; p > 0.1). The difference in the magnitude of these
effects is statistically significant in a Wald test (p < 0.05). Altogether, these results suggest that
obtaining a university degree is associated with significant gains in mental health and financial
prosperity, but these gains are restricted to low-SEB individuals. Therefore, these results pro-
vide support for Hypothesis 3; that is, low-SEB graduate appear to benefit more from univer-
sity degree in relative terms.
Table 6 Results from fixed-effect
panel regression models using
HILDA Survey data (model
coefficients)
HILDA Survey (2001–2016). Trajectory sample
# p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001
Mental health Financial prosperity
Key explanatory variables
High-SEB 0.78 0.04
Low-SEB 1.14* 0.09***
Controls
Age − 0.10* − 0.01**
Postgrade 0.12 − 0.02
Partnered 1.36*** 0.02
Employed − 0.04 0.09***
Constant 74.89*** 4.10***
βLow-SEB = βHigh-SEB (p-value
of Wald test)
0.49 < 0.05
n (observations) 11,056 11,029
n (individuals) 1101 1101

Research in Higher Education
1 3
Discussion andConclusion
In this paper, we have leveraged longitudinal data from two high-quality, longitudinal,
nationally representative Australian datasets—the ACLD and the HILDA Survey—to com-
pare the absolute and relative returns to university degrees of low- and high-SEB gradu-
ates, and how these evolve with time since graduation. In doing so, we contributed to the
literature on the returns to higher education, as well as the literature on social stratification.
Key study contributions included the modelling of a broad set of outcomes that go beyond
labor-market indicators, considering long-run trends in post-graduation trajectories, and
undertaking explicit comparisons of the absolute and relative returns to higher education.
‘Stratifying Forces’ Prevail, but‘Levelling Forces’ alsoMatter (Hypothesis 1)
Our first hypothesis was formulated based on conceptual premises from social and cultural
capital theory, the theory of effectively maintained inequality, and life-course theory, all of
which highlighted the role of “stratifying forces” post-graduation. Specifically, we hypoth-
esised that low-SEB graduates would achieve worse post-graduation outcomes than high-
SEB graduates. Consistent with this theoretical prediction, we found that low-SEB gradu-
ates received lower returns to higher-education qualifications than high-SEB graduates for
employment and managerial/professional work (ACLD) and job-security satisfaction and
financial prosperity (HILDA). These results echo those from previous studies in Norway
(Hansen 2001), Italy and Spain (Triventi 2013), as well as previous Australian evidence
(Richardson etal. 2016). However, some of our results were consistent with the predictions
of human capital, signaling and rational action theories, which pointed to higher education
as a “levelling force” and expected low-SEB graduates to exhibit outcomes comparable
to those of their high-SEB counterparts. This applied to the likelihood of having a high
weekly income (ACLD) and hourly wages and mental health (HILDA Survey). Similar
patterns of effects have been reported in earlier international (Hout 1984) and Australian
(Li etal. 2017) studies. Altogether, our findings for different outcomes lent some support
to different perspectives. This heterogeneity in associations underscores the importance of
considering multiple outcome variables when examining differences in the returns to edu-
cation by social origin. Further, they suggest that some of the “leveling” and “stratifying”
mechanisms discussed before may apply more prominently for some of the outcomes. For
instance, high-SEB graduates having better chances of employment and managerial/profes-
sional work may be due to their superior social networks and cultural capital. Meanwhile,
the comparatively lower levels of financial prosperity reported by low-SEB graduates
might result from more complicated life courses and greater associated financial responsi-
bilities—such as paying off education loans or supporting dependents. Yet other observed
associations may be driven by different processes. For example, the similar income and
earnings of low- and high-SEB graduates may emerge due to the high regulation of gradu-
ate-job salaries in the Australian labor market.
Dierences inOutcomes Fade Over Time (Hypothesis 2)
One of the key contributions of this study was the consideration of longitudinal trajecto-
ries in post-graduation outcomes. Based on the “stratifying” and “levelling” frameworks,
we expected that any differences in the post-graduation outcomes of low- and high-SEB
graduates would fade over time. Consistent with this hypothesis (Hypothesis 2), for those

Research in Higher Education
1 3
outcomes in which an initial penalty associated with having a disadvantaged background
was observed, this disappeared over time—fading at about 4 years after graduation. This
‘catch up’ effect by low-SEB graduates was observed for job-security satisfaction and
financial prosperity. This pattern of results indicates that the relevance of different “strati-
fying” and “levelling” forces may shift over graduates’ post-university life courses. Spe-
cifically, the ‘closing gaps’ scenario observed in our data is consistent with the proposition
that social capital may play a greater role at labor market entry, while human capital may
play a greater role thereafter (Lin 1999; Jacob etal. 2015). The latter could be due to an
erosion in any initial differences in productivity by social origins through work experience
(Heckman etal. 2016), or the superior social networks of high-SEB graduates being more
important in opening job opportunities immediately after graduation than later on (Jacob
etal. 2015). Overall, the longitudinal associations in our analyses resemble those found
in previous international (Jacob et al. 2015) and Australian (Edwards and Coates 2011)
research.
Relative Returns are Greater forLess Advantaged Graduates (Hypothesis 3)
Our final hypothesis, Hypothesis 3, posited that before/after graduation differences in
outcomes will be larger amongst low-SEB than high-SEB graduates. In other words, we
expected that the relative returns to degree attainment would be greater amongst low-SEB
than high-SEB, due to relatively more substantial changes to their circumstances brought
about by university participation. Consistent with this, our analyses yielded evidence that
a significant within-individual before-after graduation improvement was observed for low-
SEB graduates but not for high-SEB graduates. This applied to both mental health and
perceived financial prosperity—although the difference was only statistically significant for
the latter. The pattern is consistent with arguments that similar outcomes post-university
(e.g., income or wages) reflect more pronounced relative benefits (e.g., greater perceived
financial prosperity) for low- than high-SEB graduates because of the poorer financial con-
ditions that low-SEB graduates experienced pre-graduation (e.g., lower financial support
from family) (Brand and Xie 2010).
Limitations andFurther Research
Despite the importance of our findings, some study limitations must be acknowledged.
First, our analyses do not account for self-selection into university participation/comple-
tion and—as explained previously—this selection is likely to be more pronounced amongst
low-SEB individuals (Goldthorpe 1996). Low-SES individuals are less likely than high-
SES individuals to access higher education in the first place, and more likely to drop out of
a higher-education program after gaining access. Hence, the subsample of low-SES indi-
viduals observed after attaining a degree may not be representative of all low-SES indi-
viduals, but may instead comprise a subset of highly capable low-SES individuals. An
implication of this potential source of sample selection is that our estimates should not be
readily taken as evidencing causal relationships, as the positive selection of low-SES indi-
viduals into the sample may have attenuated the observed differences in post-graduation
outcomes by socio-economic status. Taken together with our findings, this potential selec-
tivity can also be seen as suggesting that even the smartest and most determined low-SEB
students captured in our sample fail to achieve post-graduation labour-market outcomes
comparable to those of their—less positively selected—high-SEB counterparts. Future

Research in Higher Education
1 3
studies could model these selection processes explicitly using fit-for-purpose estimation
approaches. Second, despite drawing on large, nationally representative datasets, we had
relatively small sample sizes in our target group of university graduates. As such, we were
unable to incorporate further granularity into the analyses—e.g., stratifying the models by
gender, or comparing undergraduate versus postgraduate degrees. Future research leverag-
ing larger datasets (e.g., administrative data) could circumvent this issue. Third, our data
lacked robust proxies to test the specific mechanisms proposed by the theories discussed
in our conceptual framework (e.g., social networks, productivity, or socio-cultural capital),
which prevented us from investigating their individual contributions to overall differences
in the returns to university education between low- and high-SEB individuals. Finally, our
analyses do not consider the possibility that attendance to university, without completion,
may exert some influence on individuals’ subsequent health and employment outcomes
(see Toutkoushian etal. 2013). Theorizing and testing this premise should be the focus of
further research.
Concluding Remarks
Our findings carry important implications for policy and practice. Overall, they suggest
that in the contemporary Australian context social origin continues to play a role in shap-
ing up the labor-market and personal outcomes of university graduates. This is manifested
by the lower chances that low-SEB graduates have—at least initially—to find employment
and access managerial/professional occupations, and by their poorer job-security satisfac-
tion and perceived financial prosperity. Other study findings, however, could be read with
more optimism: low-SEB graduates eventually ‘catch up’ with their high-SEB peers in
some of the longitudinal outcomes considered, and benefit comparatively more from their
university degrees in relative terms. We also found some support for the meritocratic or
levelling function of higher education—including comparable income, wages, and mental
health amongst low- and high-SEB graduates.
All in all, our findings contribute to those from a broader body of work in Australia
(Harvey etal. 2016) and internationally (e.g., European Union 2014) that demonstrates
that low-SEB individuals are less likely to choose to attend higher education, enact choices
to attend higher education, and complete their higher-education courses. These processes
represent significant barriers to equality of opportunity, and the mechanisms that produce
them need to be identified and addressed. Adding to this pool of evidence, our findings
suggest that addressing educational inequalities by SEB requires additional attention to
post-graduation outcomes, to complement the current emphasis on access and completion.
Policies should explicitly consider the need to ensure that all graduates make a success-
ful transition from education to employment and enjoy equal chances to succeed post-
graduation—regardless of their social origins. This will require coordinated education and
labor-market policies. Universities have also an important role to play here, and should pro-
vide not only high-quality curricula, but also training on employability skills and adequate
career guidance. Strengthening the latter could help reduce the length of time it takes for
low-SEB graduates to ‘catch up’ with their high-SEB peers.
Funding This study was funded by the National Centre for Student Equity in Higher Education (NCSEHE)
at Curtin University (Grant reference Number RES-51444/CTR-11202).

Research in Higher Education
1 3
Compliance with Ethical Standards
Conict of interest The authors declare that they have no conflict of interest.
References
ABS. (2006). Australian and New Zealand Standard Classification of Occupations (ANZSCO). Canberra:
Australian Bureau of Statistics.
ABS. (2017a). Education and work, Australia, May 2017. Canberra: Australian Bureau of Statistics.
ABS. (2017b). Census of population and housing: Understanding the Census and Census data, Australia,
2016 (2900.0). Canberra: Australian Bureau of Statistics.
ABS. (2018a). Microdata: Australian Census longitudinal dataset, 2011-2016 (2080.0). Canberra: Austral-
ian Bureau of Statistics.
ABS. (2018b). Information paper: Australian Census longitudinal dataset, methodology and quality assess-
ment, 2011-2016 (2080.5). Canberra: Australian Bureau of Statistics.
ABS. (2018c). Australian statistical geography standard (ASGS): Volume 5—Remoteness structure, July
2016 (1270.0.55.005). Canberra: Australian Bureau of Statistics.
ABS. (2018d). Census of population and housing: Socio-economic indexes for areas (SEIFA), Australia,
2016 (2033.0.55.001). Canberra: Australian Bureau of Statistics.
Allison, P. (2009). Fixed-effect regression models. London: Sage.
Becker, G. S. (1964). Human capital: a theoretical and empirical analysis, with special reference to educa-
tion (1st ed.). Chicago: University of Chicago Press for the National Bureau of Economic Research.
Bourdieu, P. (1984). Distinction: A social critique of the judgement of taste. London: Routledge.
Brand, J. E., & Xie, Y. (2010). Who benefits most from college? Evidence for negative selection in hetero-
geneous economic returns to higher education. American Sociological Review, 75(2), 273–302.
Breen, R., & Jonsson, J. O. (2007). Explaining change in social fluidity: Educational equalization and edu-
cational expansion in twentieth-Century Sweden. American Journal of Sociology, 112(6), 1775–1810.
Card, D. (1999). The causal effect of education on earnings. In O. Ashenfelter & D. Card (Eds.), Handbook
of labor economics (Vol. 3A, pp. 1801–1863). Amsterdam: Elsevier.
Cassells, R., Duncan, A., Abello, A., D’Souza, G., & Nepal, B. (2012). Smart Australians: Education and
Innovation in Australia. Melbourne: AMP.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94,
95–120.
Curran, P. J., Obeidat, K., & Losardo, D. (2010). Twelve frequently asked questions about growth curve
modeling. Journal of Cognition and Development, 11(2), 121–136.
Cutler, D. M., & Lleras-Muney, A. (2008). Education and health: Evaluating theories and evidence. In S.
H. James, R. F. Schoeni, G. A. Kaplan, & H. Pollack (Eds.), Making Americans Healthier: Social and
Economic Policy as Health Policy. New York: Russell Sage Foundation.
Daly, A., Lewis, P., Corliss, M., & Heaslip, T. (2015). The private rate of return to a university degree in
Australia. Australian Journal of Education, 59(1), 97–112.
Desjardins, R., & Lee, J. (2016). Earnings and employment benefits of adult higher education in compara-
tive perspective: Evidence based on the OECD Survey of Adult Skills (PIAAC). Los Angeles: UCLA.
Donald, S. G., & Lang, K. (2007). Inference with difference-in-differences and other panel data. The Review
of Economics and Statistics, 89(2), 221–233. https ://doi.org/10.1162/rest.89.2.221.
Edwards, D., & Coates, H. (2011). Monitoring the pathways and outcomes of people from disadvantaged
backgrounds and graduate groups. Higher Education Research & Development, 30(2), 151–163.
Elder, G. H. J., Johnson, M. K., & Crosnoe, R. (2003). The emergence and development of life course
theory. In J. T. Mortimer & M. J. Shanahan (Eds.), Handbook of the life course (pp. 3–22). New York:
Kluwer.
Flaster, A. (2016). Kids, college, and capital: Parental financial support and college choice. Research in
Higher Education, 59(8), 979–1020.
Goldthorpe, J. H. (1996). Class analysis and the reorientation of class theory: The case of persisting differ-
entials in educational attainment. British Journal of Sociology, 47(3), 481–505.
Greene, W. H. (2012). Econometric analysis. Boston: Pearson.
Hansen, M. N. (2001). Education and economic rewards: Variations by social-class origin and income
measures. European Sociological Review, 17(3), 209–231.
Harvey, A., Burnheim, C., & Brett, M. (Eds.). (2016). Student equity in Australian higher education:
Twenty-five years of a fair chance for all. Singapore: Springer.

Research in Higher Education
1 3
Heckman, J. J., Humphries, J. E., & Veramendi, G. (2016). Returns to education: The causal effects of edu-
cation on earnings, health and smoking. Cambridge: National Bureau of Economic Research.
Heckman, J. J., Humphries, J. E., & Veramendi, G. (2017). The non-market benefits of education and abil-
ity. Bonn: IZA Institute of Labor Economics.
Hout, M. (1984). Status, autonomy, and training in occupational mobility. American Journal of Sociology,
89(6), 1379–1409.
Jackson, M., Goldthorpe, J. H., & Mills, C. (2005). Education, employers and class mobility. Research in
Social Stratification and Mobility, 23, 3–33.
Jacob, M., Klein, M., & Iannelli, C. (2015). The impact of social origin on graduates’ early occupational
destinations—An Anglo-German comparison. European Sociological Review, 31(4), 460–476.
Li, I. W., Mahuteau, S., Dockery, A. M., & Junankar, P. N. (2017). Equity in higher education and graduate
labour market outcomes in Australia. Journal of Higher Education Policy and Management, 39(6),
625–641.
Lin, N. (1999). Social networks and status attainment. Annual Review of Sociology, 25(1), 467–487.
Lucas, S. R. (2001). Effectively maintained inequality: Education transitions, track mobility, and social
background effects. American Journal of Sociology, 106(6), 1642–1690.
Manly, C. A., Wells, R. S., & Kommers, S. (2019). Who are rural students? How definitions of rurality
affect research on college completion. Research in Higher Education. https ://doi.org/10.1007/s1116
2-019-09556 -w.
OECD. (2017). Education at a glance 2017: OECD indicators. Paris: Centre for Educational Research and
Innovation, OECD.
Oreopoulos, P., & Salvanes, K. G. (2011). Priceless: The nonpecuniary benefits of schooling. Journal of
Economic Perspectives, 25(1), 159–184.
Perales, F. (2019). Modeling the consequences of the transition to parenthood: Applications of panel regres-
sion methods. Journal of Social and Personal Relationships. https ://doi.org/10.1177/02654 07519
84752 8.
Richardson, S., Bennett, D., & Roberts, L. (2016). Investigating the relationship between equity and gradu-
ate outcomes in Australia. Perth: The National Centre for Student Equity in Higher Education at Curtin
University.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling change and event occur-
rence. New York: Oxford University Press.
Spence, M. (1973). Job market signaling. The Quarterly Journal of Economics, 87(3), 355–374.
Tieben, N. (2019). Non-completion, transfer, and dropout of traditional and non-traditional students in Ger-
many. Research in Higher Education. https ://doi.org/10.1007/s1116 2-019-09553 -z.
Torche, F. (2011). Is a college degree still the great equalizer? Intergenerational mobility across levels of
schooling in the United States. American Journal of Sociology, 117(3), 763–807.
Toutkoushian, R. K., Shafiq, M. N., & Trivette, M. J. (2013). Accounting for risk of non-completion in pri-
vate and social rates of return to higher education. Journal of Education Finance, 39(1), 73–95.
Triventi, M. (2013). The role of higher education stratification in the reproduction of social inequality in the
labor market. Research in Social Stratification and Mobility, 32, 45–63.
Umberson, D., Williams, K., Thomas, P. A., Liu, H., & Thomeer, M. B. (2014). Race, gender, and chains
of disadvantage: Childhood adversity, social relationships, and health. Journal of Health and Social
Behavior, 55(1), 20–38.
Union, European. (2014). Education and training monitor, 2014. Brussels: European Commission.
Ware, J. E., & Sherbourne, C. D. (1992). The MOS 36-item short-form health survey (SF-36): I. Conceptual
framework and item selection. Medical Care, 30(6), 473–483.
Watson, N., & Wooden, M. P. (2012). The HILDA survey: A case study in the design and development of a
successful household panel survey. Longitudinal and Life Course Studies, 3(3), 369–381.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.