Equity in higher education and graduate labour market outcomes in Australia

ABSTRACT

The rate of higher education participation in Australia has increased over the past decade for individuals from disadvantaged backgrounds. This study contributes to the knowledge on the outcomes of disadvantaged individuals who complete higher education by looking at the labour market outcomes of university graduates from equity groups. The number of Indigenous graduates and graduates with disabilities was found to be very low, suggesting that more needs to be done to improve higher education completion for these two groups. The labour market outcomes for other equity groups are mixed, with those from low socio-economic status backgrounds and regional and remote Australia performing well in the labour market, while graduates from non-English-speaking backgrounds and female graduates in science, technology, engineering and mathematics fields experience substantial disadvantage in the labour market. The findings suggest that selection processes prior to the graduates’ entry into the labour market are important.

KEYWORDS:

• Higher education
• equity
• disadvantage
• labour market
• employment
• earnings

Introduction

Encouraging participation in higher education by individuals from disadvantaged backgrounds has been on the higher education policy agenda of many countries and was a key recommendation of the Bradley, Noonan, Nugent and Scales (2008) review of Australian Higher Education. There has been progress in this regard in Australia, as noted by Koshy (2014).

A significant development in the Australian higher education sector lies in the uncapping of Commonwealth-funded university student places under the student demand-driven system of 2012. Under the demand-driven system, higher education student enrolments have been increasing, which has led to doubts about maintaining academic standards and calls for university students places to be capped, or for a minimum Australian Tertiary Admission Rank (ATAR) for university admission to be imposed. Yet, as Norton (2013) points out, imposing minimum ATARs would impact negatively on low socio-economic status (SES) students most. At the same time, a study by Pitman, Koshy and Phillimore (2015) showed that Australia’s higher education expansion has not led to any decline in educational quality and standards. The findings from these two studies thus favour higher education policies that maintain access for underprivileged individuals.

Previous studies on outcomes of Australian university students from equity groups have been limited in terms of the scope of the outcomes analysed, concentrating mainly on university academic outcomes. For example, Mills, Heyworth, Rosenwax, Carr and Rosenberg (2009) assessed first-year students’ university academic outcomes from one single university, Lim (2015) and Chesters and Watson (2013) examine university course completion rates, while Habel (2012) considers equity pathways to higher education. Lim (2015) examined the probability of completing university degree courses for various disadvantaged groups using data from the Longitudinal Study of Australian Youth and found that students with low SES are less likely to complete their course compared to students with high SES, as are students from regional locations. Students from an Asian language background are more likely to complete their university course, compared to those from ‘other’ language backgrounds. There are relatively few studies looking at the employment outcomes of university graduates from disadvantaged backgrounds, with similar gaps in the international literature. Further information on employment outcomes for students from equity groups would be beneficial in informing higher education policy. In particular, it would inform policies to help disadvantaged groups at particular stages of their academic life.

This study widens the evidence base in that it assesses a range of employment-related outcomes of disadvantaged students and, further, utilises data from multiple universities within one Australian state. Outcomes assessed include the probability of employment, qualification-job match, job quality and earnings. This assessment of the graduates’ labour market performance contributes by examining key outcomes which are primary motivating factors behind policies to expand higher education access and equity. In addition, individuals in the key equity groups tend to belong to groups who face labour market disadvantage. The analysis focuses on students from low SES backgrounds, regional or remote areas, non-English-speaking backgrounds (NESB) and females who graduated in science, technology, engineering and mathematics (STEM) fields of study. Two other recognised equity groups, Indigenous students and students with a disability, could not be included in the analysis due to small sample sizes.

Literature review

Studies of the outcomes of Australian equity groups in higher education have typically focussed on the academic outcomes, such as access, participation, retention and academic performance at university. These studies tend to find that students from low socio-economic status perform well academically compared to their peers, after controlling for prior academic achievement (Win & Miller, 2005). The policy implications of these and other studies encourage the participation of equity groups in higher education (Coates & Krause, 2005).

As Edwards and Coates (2011) point out, however, it is important to monitor the outcomes of graduates from disadvantaged backgrounds beyond graduation. Nevertheless, studies which examine the post-graduation labour market outcomes of equity groups in Australia are relatively scarce. Coates and Edwards (2009) examined the short-term (1 year) and long-term (5 years) labour market outcomes for Australian graduates in the key equity groups mentioned earlier. A main finding by Coates and Edwards (2009) was that graduates in equity groups have similar employment outcomes compared to the general university population. Edwards and Coates (2011) examined the employment rates of graduates and found negligible differences in the short-term employment rates between graduates from advantaged or disadvantaged backgrounds. At 3 and 5 years after graduation, however, graduates from disadvantaged backgrounds were slightly less likely to be working full-time and slightly more likely to be working part-time. Graduates from disadvantaged backgrounds were also less likely to be working in managerial or professional occupations, although this was argued to be attributable to differences in the fields studied at university. The annual salaries of graduates from disadvantaged backgrounds were found to be comparable to their peers from more advantaged backgrounds.

Edwards and Coates (2011) found that the number of Indigenous people who participated in higher education was less than 1 per cent of their sample. In contrast, 2011 Australian census estimates show Indigenous people to comprise 2.5 per cent of the population. Hence, the participation of Indigenous Australians in higher education was recommended to be a priority. Nevertheless, Edwards and Coates (2011) did find that Indigenous graduates fared better than non-Indigenous graduates in terms of being employed, and in terms of being more positive about the overall benefits of their degree to their employment.

Studies that look at the labour market outcomes of NESB graduates in Australia have generally done so together with a focus on migrant outcomes. However, it has also been established that the language background of migrants plays a part in determining labour market outcomes. For example, Kler (2006) and Green, Kler and Leeves (2007) find that migrants from English-speaking countries of origin fare better than migrants from Asian countries in terms of labour market outcomes such as labour market mismatch and earnings. These findings are congruent with other studies of the general labour market in Australia, where migrants have been found to experience disadvantage in employment outcomes, particularly migrants from NESBs.

There is a richer literature that looks at the gender wage gap for university graduates. This literature, however, does not tend to focus on graduate labour market differences by gender within STEM fields. In Australia, as in other countries, women are substantially under-represented in STEM courses, particularly in physics, engineering and technology (Sikora, 2015). Edwards and Coates (2011) examined gender differences for Australian university graduates and noted that there are gender differences in the fields studied at university and that the choice of fields studied would have substantial influence on the future occupations and earnings of the graduates. Female graduates were found to be more highly represented in the fields of health and education, while male graduates were more likely to have studied information technology and engineering. They found a gender wage gap in favour of males (of AUD$2000 in annual salary) exists for graduates 1 year after graduation, after controlling for factors such as field of study, occupation and industry of employment, work status and age. This gap widened at 3 and 5 years after graduation, with the graduate gender wage gap at 5 years around AUD$7800.

Finally, graduates originating from regional and remote areas were found to have lower rates of labour force participation and higher unemployment rates compared to graduates from metropolitan areas 1 year after graduation (Edwards & Coates, 2011). Differences in the labour force participation and unemployment rates by geographic area of origin dissipated in the longer term, however, leading to similar employment outcomes by the 5-year mark. Nonetheless, a reverse trend was observed for earnings outcomes. Specifically, the earnings for graduates in full-time work were identical 1 year after graduation, regardless of their area of origin. At 5 years after graduation, however, graduates from metropolitan areas were found to have a modest earnings advantage, compared to the graduates from regional and remote areas.

Data

Data description

The data for this study were drawn from multiple sources. Confidentialised unit record data were obtained from four anonymous universities within one Australian state. These universities were varied in institution type and included institutions from three Australian university groupings: (1) the research intensive Group of Eight universities, (2) the Australian Technological Network and (3) the Innovative Research Universities. Student records for domestic students who completed a bachelor’s degree (pass or honours), and who were admitted to their university study on the basis of the completion of Year 12, were obtained from these universities. These records were then linked to the Australian Graduate Survey, a survey of short-term labour market outcomes of graduates from Australian universities. The Australian Graduate Survey is an annual national census of graduates who have completed a university qualification from an Australian university and had been conducted since 1972 (Graduate Careers Australia, 2015a). The present study uses the Graduate Destination Survey component of the Australian Graduate Survey, which contains information on the graduates’ employment and further study characteristics at 4 months post-completion of their higher education. The Australian Graduate Survey is conducted in April and October every year, which reflects the typical mid- and end-year university degree completion cycles. The study sample comprises graduates who completed their bachelor’s degrees between 2010 and 2014.¹

The Australian Graduate Survey is administered by the universities individually, although the survey instrument, code of practice and coding instructions are developed by Graduate Careers Australia, a not-for-profit organisation with expertise in graduate employment issues in Australia (Graduate Careers Australia, 2015b).² The individual institutional response rates for the sample used in the study ranged from 54 to 65 per cent. The survey questionnaire, contact instruments, code of practice and other relevant documentation can be accessed through the Graduate Careers Australia website.

As the Australian Graduate Survey was administered by the individual universities, it was possible for the data linkage to university student records to be done by each university. Broadly speaking, the linked data set consists of three sorts of variables: (1) personal and demographic characteristics, (2) university study characteristics and (3) labour market characteristics. The personal characteristics include gender, age, English-speaking background (ESB), disability status, Indigenous background, residential postcode and country of birth. The university characteristics include course weighted average mark, Australian Tertiary Admission Rank, level of study, major field of study and further study status. The labour market characteristics include employment status, job-seeking status, industry of employment, earnings, type of employment contract, sector of employment, employer size and self-reported importance of the university degree type to employment. The sample consists of 10,718 graduates.

Definition of equity groups

The equity groups are defined for the purposes of this study as follows. First, the residential postcodes of the graduates at the time they applied to enrol at university are linked to the Index of Relative Socioeconomic Advantage and Disadvantage within the Socio-Economic Indexes for Areas (SEIFA) Indices produced by the Australian Bureau of Statistics (Australian Bureau of Statistics 2011a). Graduates who were in the lowest quartile of the Index of Relative Socioeconomic Advantage and Disadvantage are defined as belonging to the low socio-economic equity group. It must be acknowledged that such area-based measures of SES have important limitations, and Dockery, Koshy and Seymour (2016) demonstrate the potential for misclassification when SES is measured on a postcode basis using the SEIFA as opposed to one based on family and household-level characteristics (see also Lim, Gemici, Rice, & Karmel, 2011). However, no alternative measures of student SES background are available in the data, and the results relating to the postcode-based measure have important policy implications as it aligns with the government’s method for setting targets and monitoring low SES participation in higher education.

The second equity group of graduates are those who originate from regional or remote regions in Australia. As with low SES, this measure was based on their original location prior to starting their degree course at university and defined according to the Australian Standard Geographical Classification Remoteness Structure (Australian Bureau of Statistics, 2011b). Graduates who resided in Inner Regional, Outer Regional, Remote and Very Remote areas, as opposed to major capital cities, were defined to be in this second equity group.

The third, fourth and fifth equity groups consist of those from NESBs those who had a physical or mental disability and those who identified as being Indigenous. These are all based on self-report in the university records. Finally, the last equity group of women in non-traditional study areas refers to female graduates in the STEM fields of study, namely natural and physical sciences, information technology and engineering, as defined within the Australian Standard Classification of Education (Australian Bureau of Statistics, 2001).

Descriptive statistics

Table 1 presents selected descriptive statistics for the sample. Up to 27 per cent of the graduates came from a low SES background. Fourteen per cent of the graduates were from a regional or remote area of origin, while 2 per cent of the graduates were from a NESB and who have been in Australia for less than 10 years. Female graduates who graduated from a STEM field of study represent 9 per cent of total graduates. The proportion of graduates who have a disability and who identified as being Indigenous were 1.6 per cent and 0.3 per cent, respectively. For these two latter groups, the proportions appear to be strikingly low. However, these were consistent with the higher education statistics from the Department of Education and Training (2015). Furthermore, low proportions of Indigenous graduates, at less than 1 per cent, had been previously reported in the Review of Indigenous Higher Education (DIISRTE 2013) and Edwards and Coates (2011). Hence, for these two groups of graduates, the low proportions and numbers are a reflection that more needs to be done to encourage access, participation and completion of higher education. Due to the low numbers in these two equity groups, statistical analyses were not feasible and hence further statistical analysis in this paper will only be conducted for the remaining equity groups.

Table 1. Descriptive statistics.

Variable	Mean	Std. dev.
Demographics
Female	0.616
Age, expressed in years	23.155	3.375
Born in Australia	0.795
Equity group
Low SES (SEIFA)	0.266
Regional or remote area of origin	0.142
NESB	0.020
Indigenous	0.003
Disability	0.016
Females in STEM	0.090
Academic characteristics
University 1	0.428
University 2	0.155
University 3	0.141
University 4	0.276
University course weighted average mark	69.517	7.348
STEM field of study	0.229
Further study after graduation	0.280
Worked in final year of study	0.846
Employment characteristics
Employed full-time	0.450
Employed part-time	0.377
Employed in the public sector	0.250
Professional or semi-professional occupation	0.516
Quality job	0.350
Job match	0.510
Permanent job (>1 year)	0.414
Large firm (>99 employees)	0.543
Medium firm (20–99 employees)	0.167
Small firm (<20 employees)	0.232
Likely to have changed employers since graduation	0.297
Annual salary, in Australian dollars	39,231	32,606
Observations	10,718

Standard deviations are presented for continuous variables.

SES: Socio-economic status; SEIFA: Socio-Economic Indexes for Areas; STEM: science, technology, engineering and mathematics; NESB: non-English-speaking background.

Academic characteristics of the graduate sample are also presented in Table 1. The average course weighted average mark for the graduate sample was around 70 percentage points. The course weighted average mark was also looked at for the equity groups of the low SES graduates, graduates from regional and remote areas, graduates from NESB and female graduates in STEM fields. The mean course weighted average marks across these equity groups were very similar in value, at around 68–70 percentage points (not shown in table). Hence, the academic performance of equity groups was very similar to their peers.

Methodology

Probit models

The employment outcomes of the graduates are examined using binary probit models.³ In these models, several measures of job outcomes are examined, with a regressor included for each equity group. The probit model can be expressed in the following manner. Let Emp be a binary variable that takes on the value of 1 denoting an employment outcome (full-time or part-time), 0 otherwise. The probability of attaining the employment outcome is conditional on a vector Z of explanatory variables, so that . The expected value conditional on Z is then

(1)

where denotes the slope coefficients associating the explanatory variables with the unobserved latent-dependent variable. The unobserved probability of attaining the employment outcome is assumed to be normally distributed, N(0,1).

The second probit model examines a more refined measure of employment, in particular, the probability for employed graduates of being employed in a job where the graduate’s university degree is self-reported to be ‘matched’ to his or her job. Specifically, the graduates were asked whether their university qualification was useful to their job, with possible responses to this question being (1) formal requirement, (2) very important, (3) somewhat important, (4) not important and (5) don’t know. For this model, graduates with responses of ‘don’t know’ were excluded from the sample.⁴ Hence, the outcome variable for this second probit model is defined as follows. A graduate is considered ‘matched’ (value = 1) if the response was that the university degree was a formal requirement or very important to their job, 0 otherwise.

The third probit model builds on the outcome variable used in the second probit model described above and examines an alternate measure of employment termed ‘job quality’. The graduates (who have found employment) are defined as being in a job of good quality if they satisfy the two requirements of (1) they are ‘matched’ to their job as described above and (2) if they report to be not looking for another job.⁵ Hence, for this third probit model, the outcome variable ‘job quality’ takes on the value of 1 if the graduate reports that their university degree was a formal requirement or very important to their job and they were not looking for another job, 0 otherwise.

Earnings models

The determinants of earnings of the graduates were examined using an ordinary least squares (OLS) model of earnings based on the human capital specification developed by Mincer (1974). This model can be written as

(2)

where ln Y denotes the annual salary of the graduates in Australian dollars, expressed in logarithmic format. X is a vector of variables hypothesised to impact on graduate earnings and includes dichotomous variables representing membership of the various equity groups.

Furthermore, the Blinder–Oaxaca decomposition method (Blinder, 1973; Oaxaca, 1973) was used to look at the earnings differentials for certain equity subgroups. The main aim for using the Blinder–Oaxaca decomposition method was to separate the earnings differential between subgroups into different components. The Blinder–Oaxaca decomposition can be expressed as

(3)

where subscripts A and D denote the advantaged and disadvantaged groups, respectively, represents the earnings difference between the groups and V is a matrix of regressors.

The first component of the earnings differential (E) has often been referred to as the ‘endowment’ effect, as it represents the earnings differences arising from differences in human capital characteristics between groups. The ‘endowment’ effect has also been called the ‘explained’ portion of the earnings gap. The second component of the earnings differential (C) is termed the ‘coefficient’ effect and represents the portion of the earnings gap arising from between-group differences in how human capital was translated into earnings. This ‘coefficient’ effect represents the ‘unexplained’ portion of the earnings gap and is often taken as a proxy for the portion of the earnings gap that can be attributed to discrimination. A third component of the earnings differential (CE) is the ‘interaction’ effect and stands for the earnings gap arising from joint differences in the ‘endowment’ and ‘coefficients’ effects. This ‘interaction’ effect can be apportioned to either the ‘explained’ or ‘unexplained’ component of the earnings gap, depending on the approach chosen. In the present study, the approach by Cotton (1988) will be used, where the weighted average of the ‘interaction’ effect will be taken and apportioned to the ‘explained’ and ‘unexplained’ portions of the earnings gap, with the weight to be used being given by the employment share of the groups examined. This avoids arbitrarily nominating one of the groups as the ‘reference’ group against which the other group is compared.

Results

Low SES graduates

Graduates from low SES backgrounds have comparable employment outcomes to the graduates from more privileged backgrounds. This is consistent across the three models of job outcomes in the probit models presented in Table 2. Specifically, the estimated effects of the graduate being from a low SES background on the probability of being employed, in a matched job or in a job of good quality are all statistically indifferent from the graduates from more advantageous SES backgrounds. Furthermore, the results from the OLS model of earnings in Table 3 revealed that there is no earnings disadvantage for low SES graduates.

Table 2. Results from the probit models of employment outcomes.

	Employment		Job match		Job quality
Variables	Coefficient	Average marginal effect	Coefficient	Average marginal effect	Coefficient	Average marginal effect
Female	0.071*	0.015*	0.053	0.020	0.046	0.017
(Male = 0)	(0.040)	(0.008)	(0.034)	(0.013)	(0.035)	(0.013)
STEM field of study	−0.142***	−0.030***	0.186***	0.069***	0.217***	0.080***
(Non-STEM = 0)	(0.051)	(0.011)	(0.048)	(0.018)	(0.048)	(0.018)
Weighted average mark	0.004*	0.001*	0.020***	0.008***	0.021***	0.008***
	(0.002)	(0.000)	(0.002)	(0.001)	(0.002)	(0.001)
Work in last year of degree	1.308***	0.274***	−0.362***	−0.133***	−0.316***	−0.117***
(No work = 0)	(0.037)	(0.007)	(0.051)	(0.019)	(0.050)	(0.018)
Born in Australia	0.100**	0.021**	−0.062*	−0.023*	−0.033	−0.012
(Not Australia = 0)	(0.040)	(0.008)	(0.037)	(0.014)	(0.037)	(0.014)
Further study	−0.560***	−0.117***	−0.775***	−0.286***	−0.699***	−0.258***
(No further study = 0)	(0.034)	(0.007)	(0.033)	(0.011)	(0.035)	(0.012)
Lowest quartile of SES	0.052	0.011	0.026	0.010	−0.025	−0.009
(Not lowest SES = 0)	(0.039)	(0.008)	(0.034)	(0.012)	(0.034)	(0.013)
Regional or remote area	−0.063	−0.013	0.150***	0.055***	0.136***	0.050***
(Not in regional or remote = 0)	(0.048)	(0.010)	(0.043)	(0.016)	(0.044)	(0.016)
NESB	−0.096	−0.020	−0.027	−0.010	−0.192*	−0.071*
(English-speaking background = 0)	(0.105)	(0.022)	(0.111)	(0.041)	(0.115)	(0.042)
Female × STEM field of study	−0.118	−0.025	−0.518***	−0.191***	−0.541***	−0.200***
(Male = 0, non-STEM = 0)	(0.074)	(0.015)	(0.072)	(0.026)	(0.073)	(0.027)
Constant	−0.236		−0.867***		−1.209***
	(0.157)		(0.147)		(0.149)
Observations	10,422		8678		8678
Log likelihood	−3949		−5589		−5453
Pseudo R-squared	0.175		0.070		0.060

Standard errors in parentheses. ***, ** and * denote significance at1, 5 and 10 per cent levels, respectively.

SES: Socio-economic status; STEM: science, technology, engineering and mathematics.

Table 3. Results from the OLS model of earnings, graduate sample and by gender.

Variables	Full sample	Males	Females
Female	−0.070***
(Male = 0)	(0.022)
STEM field of study	0.086***	0.072**	−0.072*
(Non-STEM = 0)	(0.031)	(0.032)	(0.041)
Weighted average mark	−0.002*	−0.004***	−0.001
	(0.001)	(0.002)	(0.002)
Work in last year of degree	0.058	0.056	0.062
(No work = 0)	(0.037)	(0.052)	(0.052)
Born in Australia	0.040	0.001	0.060*
(Not Australia = 0)	(0.026)	(0.034)	(0.036)
Further study	−0.259***	−0.327***	−0.220***
(No further study = 0)	(0.027)	(0.039)	(0.035)
Full-time employment	0.781***	0.790***	0.775***
(Part-time = 0)	(0.027)	(0.042)	(0.035)
Employed in public sector	−0.015	−0.059	0.007
(Private sector = 0)	(0.024)	(0.040)	(0.031)
Professional or semi-professional occupation	0.187***	0.229***	0.159***
(Not professional or semi-professional = 0)	(0.023)	(0.037)	(0.029)
Medium employer	0.108***	0.103**	0.109***
(Small employer = 0)	(0.030)	(0.050)	(0.037)
Large employer	0.184***	0.164***	0.196***
(Small employer = 0)	(0.023)	(0.034)	(0.030)
Permanent contract	0.448***	0.429***	0.455***
(Casual or short-term contract = 0)	(0.029)	(0.045)	(0.038)
Fixed-term contract	0.417***	0.348***	0.450***
(Casual or short-term contract = 0)	(0.035)	(0.055)	(0.045)
Lowest quartile of SES	0.002	−0.001	0.005
(Not lowest SES = 0)	(0.022)	(0.033)	(0.029)
Regional or remote area	0.059**	0.082*	0.049
(Not in regional or remote = 0)	(0.030)	(0.044)	(0.039)
NESB	−0.262***	−0.297*	−0.235*
(English-speaking background = 0)	(0.099)	(0.172)	(0.121)
Female × STEM field of study	−0.156***
(Male = 0 or non-STEM = 0)	(0.051)
Constant	9.447***	9.679***	9.202***
	(0.103)	(0.130)	(0.157)
Observations	7829	2951	4878
Adjusted R-squared	0.411	0.446	0.385

Standard errors in parentheses. ***, ** and * denote significance at 1, 5 and 10 per cent levels, respectively.

SES: Socio-economic status; STEM: science, technology, engineering and mathematics; NESB: non-English-speaking background.

Graduates from regional and remote areas

Lim (2015) found that the lowest likelihood of university completion was for students from regional areas. This selection bias appeared to translate into favourable labour market outcomes for those who did manage to complete their university studies. The model of employment probability indicated that graduates from regional and remote areas have comparable success in finding employment, as graduates from metropolitan areas. In addition, graduates from the regional and remote areas experienced increased probabilities in securing a job matched to their qualifications or of good quality. The job outcomes for regional and remote graduates are hence superior to those from metropolitan areas. Graduates from regional and remote areas were also found to have higher earnings compared to their peers from metropolitan areas, although the positive earnings effect was only observed for male graduates in this equity group.

Graduates from NESBs

The probability of finding a job did not differ statistically for NESB graduates, compared to ESB students. However, for graduates who were employed, NESB graduates were less likely to be in a good quality job, statistically significant at the 10 per cent level. NESB graduates were found to earn substantially less than ESB graduates, by around 26 per cent. This earnings disadvantage was substantial for NESB graduates of both sexes, with male (female) NESB graduates earning 30 per cent (24 per cent) less than their respective ESB counterparts. The Blinder–Oaxaca decomposition of the earnings difference by language status showed that 40 per cent of the NESB earnings gap was attributable to differences in human capital endowments between NESB and ESB graduates (Table 4). The remaining 60 per cent of the earnings gap was unexplained.

Table 4. Estimates from the Blinder–Oaxaca decomposition.

Item	NESB (i)	Gender (ii)	Female*STEM (iii)	Female in STEM (iv)
Predicted earnings for privileged group	10.227***	10.330***	10.252***	10.198***
	(0.012)	(0.019)	(0.012)	(0.016)
Predicted earnings for unprivileged group	9.801***	10.153***	9.865***	9.865***
	(0.128)	(0.015)	(0.048)	(0.048)
Raw wage gap	0.426***	0.176***	0.387***	0.333***
	(0.128)	(0.024)	(0.050)	(0.051)
● Due to endowments	0.258**	0.047***	0.282***	0.253***
	(0.102)	(0.018)	(0.038)	(0.065)
● Due to coefficients	0.251**	0.080***	0.211***	0.065
	(0.106)	(0.020)	(0.046)	(0.041)
● Due to interaction	−0.083	0.049***	−0.106***	0.016
	(0.072)	(0.013)	(0.033)	(0.026
Unexplained gap	0.249	0.110	0.202	0.067
Explained gap	0.176	0.066	0.185	0.266
Percentage of unexplained	59	63	52	20
Percentage of explained	41	37	48	80
Observations	7829	7829	7829	4878

Standard errors in parentheses. ***, ** and * denote significance at 1, 5 and 10 per cent levels, respectively. Privileged (unprivileged) groups are (i) ESB (NESB), (ii) males (females), (iii) males or non-STEM major (females in STEM major) and (iv) females in non-STEM major (females in STEM major).

Female graduates from STEM fields of study

Overall, female and male graduates had similar labour market outcomes in terms of their propensity to be employed, the importance of skills for their jobs and job quality, though women had around 7 per cent lower earnings. Taking males and females together, graduates from STEM fields of study actually displayed inferior outcomes in terms of employment probability, although STEM graduates experienced more positive outcomes in terms of job matching and job quality. STEM graduates did have higher earnings but this can be attributed purely to a STEM earnings premium for males of around 7 per cent. Although STEM graduates fared worse than non-STEM graduates in terms of the propensity to be employed, this applied similarly to male and female graduates. That is, females with STEM qualifications were not significantly less likely than males with STEM qualifications to be in employment. However, the disadvantage faced by women is apparent in terms of job matches. Female graduates from STEM fields are markedly less likely than their male counterparts to report that they are in a job for which their STEM qualification was a prerequisite or very important.

The sample for the final model reported in Table 4 is restricted to females only. Relative to women who graduated in other fields, women who graduated in STEM fields earned around 16 per cent lower salaries. Most of this latter earnings ‘penalty’ seems to be attributable to differences in observable characteristics of females who undertake STEM studies and those who do not. For the wider sample (model iii), however, differences in returns to characteristics do make a substantial contribution, although the lower earnings for female STEM graduates are still driven by differences in observable characteristics.

Discussion

The results from the probit models of job outcomes and the earnings analyses indicate that there are mixed outcomes for the graduates from various equity groups. Specifically, graduates from regional or remote areas appear to have slightly superior labour market outcomes compared those from metropolitan areas, while graduates from low SES backgrounds have comparable outcomes to their higher SES peers.⁶ Nonetheless, it needs to be borne in mind that a selection process has taken place. One point of selection occurs at entry into university, as students with disadvantaged backgrounds face greater obstacles to participation in higher education, such as financial barriers and greater opportunity costs, and hence have reduced propensities to participate in higher education (Le & Miller, 2005). Existing student selection practices by the universities were also found by Pitman (2016) to be inadequate in fostering access for equity groups. A further process of selection takes place during university study, in terms of the impact of SES on attrition and completion rates in university, as found by Lim (2015). This suggests that the graduates in the sample consisted of individuals who have accessed and completed higher education amidst substantial obstacles and hence are likely to possess positive but unobservable characteristics which have translated well into the labour market.

Students with an Asian language background, and who would hence fall within the NESB group here, were found by Lim (2015) to have the highest likelihoods of completing university. Hence, the inferior labour market outcomes experienced by NESB graduates are suggestive of marked disadvantages experienced by NESB graduates upon entry to the labour market. Bear in mind that the NESB graduates in the sample are Australian residents, and the majority were either born in Australia or have been in Australia for over 10 years. As the decomposition analysis indicates, the earnings differential by language status is not a result of human capital endowments and can be attributed to other sources.

The labour market outcomes for female graduates in STEM fields are highly suggestive of gender-based barriers or discrimination against females in the rewards to STEM qualifications. First, female graduates with STEM qualifications appear to be less likely than males to find a job using those skills. Second, there is the contrasting return, in terms of earnings, to having completed studies in a STEM field for males (positive) and females (negative). This may well be because jobs requiring STEM qualifications – and the higher paying jobs in particular – are male dominated and this creates barriers to entry and progression in those occupations for females.

These findings are disconcerting for efforts to increase female participation in STEM courses. The evidence from this sample is that there are well-founded reasons for females to steer clear of those non-traditional fields. They not only fare worse than females who enter non-STEM fields, they also do markedly worse than their male classmates. Given the obvious barriers to enrolling in such fields in the first place, it is particularly worrisome that those women who successfully graduate then also appear to face discrimination in the workforce. Policies to increase the number of young women choosing to enter STEM courses need to also address gender inequality in labour market opportunities. The stark under-representation of females in STEM fields of study not only contributes to this segregation and the limitation of women’s career options but also impedes the efficiency of the labour market by vastly limiting the supply of capable persons into fields identified as strategically important for innovation and Australia’s economic potential. A better understanding of labour market outcomes for those women who do venture into these male dominated areas is clearly important for the formation of policy to encourage young women into STEM-related career paths.

Conclusion

This study examined the labour market outcomes of Australian graduates from disadvantaged backgrounds, utilising a data set that linked university student records from four universities, to the Australian Graduate Survey. The study of labour market outcomes of graduates from equity groups in Australia is important and evaluates a key outcome of equity policies aimed at closing inequality for disadvantaged groups. There are some limitations of this study which should be addressed in further research. First, the study looked at data from four Australian universities located in the same state. While analysing labour market outcomes for graduates from the same state is advantageous due to the presence of a more homogenous labour market environment, future studies could attempt to obtain and use data from more universities to provide findings more representative of the Australian graduate labour market. Second, it is probable that individuals from disadvantaged backgrounds who successfully complete university do so in the face of substantial barriers and obstacles. Hence, the findings in this study possibly relate to those who have greater motivation and possess greater amounts or quality of human capital endowments. Third, the labour market outcomes considered in this paper occur in the short term after graduation. The advantage of this approach is that the labour market outcomes of these graduates are considered at a career stage where there are less confounding factors that could have influenced outcomes, such as divergence in career developmental pathways. This is particularly important as equity group members typically face labour market disadvantage as well (e.g. females and migrants). Nonetheless, the short-term nature of the labour market outcomes considered in this paper needs to be noted in the interpretation of the findings.

This study has some important implications for future policies towards higher education for disadvantaged groups. First, greater support in the form of policies encouraging access to and completion of higher education is required. Indeed, prior research has indicated that providing educational opportunities to the disadvantaged allows them to catch up and sustain their academic achievement (Lamb, Jackson, Walstab, & Huo, 2015).

Second, the number of graduates from the equity groups of Indigenous Australians and individuals with disabilities is well below their representation in the wider population. This highlights the need for further action to boost participation and, more importantly, completion for individuals in these two groups.

Third, the labour market outcomes for graduates from equity groups were mixed. Graduates from low SES backgrounds and regional and remote areas experienced comparable labour market outcomes to their peers, given that they had entered higher education institutions. This is likely to be attributable to the selection process mentioned above, where graduates from these two areas of disadvantage and who possess unobservable positive human capital characteristics are able to successfully complete their university study despite their disadvantage. Graduates from these two groups hence also experience success in the labour market. However, graduates in the equity groups of NESB and female graduates in STEM fields were found to have adverse labour market outcomes in comparison to their peers. The positive labour market outcomes experienced by graduates from low SES backgrounds and from regional or remote areas indicate that the increased higher education participation rates by these groups have borne fruit and contributed to sustained success in the labour market. From this perspective, participation in higher education for these groups could be further encouraged.

Fourth, substantial labour market disadvantage for graduates from NESB and female graduates in STEM fields was found. For graduates from NESB, a substantial earnings penalty was uncovered, and which could not be explained by differences in their human capital endowments. This is suggestive of discrimination against this group. Research by Booth, Leigh and Varganova (2012) found that NESB graduates were less likely to be invited for interviews by employers. Hence, policies should be introduced to encourage employers to use equal opportunity employment practices that are monitored on a regular basis by a government agency. Female graduates in STEM faced barriers to securing (good) employment outcomes as well as earnings disadvantage. In particular, it was found that female graduates with STEM qualifications did not find employment in STEM-related fields. Hence, it is imperative that policies be introduced to encourage and stimulate female employment in STEM fields. Even when female graduates in STEM fields find employment in STEM areas they have an earnings penalty that suggests some form of employer or workplace discrimination.

To conclude, further research that identifies causes of these disadvantages, as well as policies to improve the labour market outcomes of these groups, would be welcome. This paper marks one step towards exposing the extent and nature of disadvantages faced by certain equity groups after leaving university education. It offers both endorsement and reservations for policy to promote greater equity in higher education in Australia. For some equity groups, expanded participation will have reduced labour market disadvantage, while for other equity groups, such policies need to be complemented with measures to address barriers faced after graduation.

Acknowledgements

The authors acknowledge funding from the National Centre for Student Equity in Higher Education based at Curtin University. The authors are also grateful for the provision of data and data support from the data offices of four universities who participated in this study. The views expressed in this paper are those of the authors, and should not be attributed to the funding body, the data providers or the authors’ institutions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Pooled data from 2010 to 2014 were used in order to overcome the issue of small numbers of observations in some of the equity groups, such as NESB and students with disabilities.

2. From 2015, this survey has been renamed the Graduate Outcomes Survey and will be administered by the Social Research Centre based at the Australian National University.

3. In reporting the results of these binary probit models, both the probit regression coefficients and average marginal effects are reported.

4. One hundred and thirty-nine graduates or 3.85 per cent of the sample responded ‘don’t know’ to this question. After excluding these responses and excluding graduates not in employment, there were 8678 observations in the sample.

5. Several other probit models were estimated using different measures of ‘job quality’ in addition to those reported in this paper. These included (1) obtaining a full-time job, (2) obtaining a professional or semi-professional occupation, (3) the field of study matched the requirements of the job, (4) the skills of the graduate matched the requirements of the job and (5) occupational status as measured by the AUSE106 index (McMillan, Beavis, & Jones, 2009). In the interests of conciseness, the results from these models are not reported, although it can be noted that the conclusions to be drawn as to the performance of equity groups in the labour market are similar to those reported below. Full results from these analyses are available on request.

6. Alternatively, it is possible that students from the equity groups perform better in the labour market in the short term if they are less likely to be engaged in further study. As such, the probit models of job outcomes were estimated on subsamples of graduates in further study, and not in further study. The results from these models were qualitatively similar to those reported in Table 2, with the only difference being that graduates from regional or remote areas were less likely to be in employment if they were in further study, statistically significant at the 1 per cent level.