Schooling inequality, higher education and the labour market: Evidence from a graduate tracer study in the Eastern Cape, South Africa

ABSTRACT

An emerging body of research has shown that there are large inequalities in access to higher education in South Africa. There remains a gap, however, in identifying how factors such as schooling background, academic performance, race and gender are linked with key higher education outcomes. In particular, the significance of these factors for first-choice degree attainment at university and in the subsequent transition to the labour market is of interest. This article addresses these questions by presenting a descriptive and multivariate analysis of data collected through a tracer study which interviewed graduates from two Eastern Cape universities. The results suggest that schooling background, race and gender are associated with study choices and unemployment. These findings have important implications both for equity and for the efficiency of higher education institutions. The article concludes with a discussion of potential policy responses and the implications for equity in higher education.

KEYWORDS:

• Higher education
• unemployment
• skills

1. Introduction

Inequality in primary and secondary schooling outcomes is a persistent feature of the education system in post-apartheid South Africa.¹ In turn, differences in schooling quality carry over into the post-school sector where choices for further education and training are often limited by schooling background and socio-economic status. In particular, research has suggested that South Africa's relatively low participation rate in higher education (Cloete, 2004) and the large number of young people who are not in employment, education or training (NEETs) are products of low schooling quality and a post-school training sector which does not meet the needs of the majority of school leavers (Cosser & du Toit, 2002; Pillay, 2004; Cloete, 2009). By one estimate, at least 10% of 18 to 24 year olds in South Africa are classified as NEETs but would qualify for at least some type of post-schooling education or training (Sheppard & Cloete, 2009).

In addition to the low levels of participation and completion in higher education, there are at least two further areas of concern for higher education in South Africa. The first is the substantial gap between programme or degree preference and enrolment, on the one hand, and completion, on the other (Cosser, 2009). Even if the government is successful in achieving its stated goal of shifting the balance of enrolments away from Humanities subjects towards Science, Engineering and Technology (SET) and Business and Commerce (Department of Education, 2001; Cloete, 2004; Pillay, 2004), its achievement would be blunted by the large number of students who do not complete their first-choice qualifications, particularly in cases where first choices are in desired fields of study such as SET.

The second area of concern is the heterogeneity in employment outcomes among those who have completed a university degree, for both equity and efficiency reasons. Equity considerations require identification of ways of achieving more equal outcomes for all graduates regardless of education history, demographics or socio-economic background. Efficiency considerations require that human resources are deployed optimally within the labour market by matching graduates with jobs that meet their skill profiles.

This study attempts to link schooling, demographic, socio-economic and academic factors to first-choice degree completion and labour market outcomes. More specifically, this study investigates those factors that are most directly associated with whether the degrees that university graduates obtain reflect their first qualification choices, and also examines the effects of those factors and degree types on labour market outcomes.

The remainder of the article is structured as follows. The next section reviews the literature on the transition from schooling to higher education in South Africa with a particular focus on programme choice, followed by the recent literature on graduate employment and unemployment, and the main findings of the graduate tracer studies which have been conducted in South Africa to date. Section 3 describes the graduate tracer study design and the analysis upon which the empirical section of the article is based. In Section 4, the results of the descriptive statistics and a multivariate analysis are presented in two parts. Finally, Section 5 discusses the results and reflects on the implications for higher education in South Africa.

2. Review

2.1. Programme choice and the transition from school to university

One fairly widely held conclusion from the literature on the transition to higher education in South Africa is that access to and success in tertiary education are still very closely associated with socio-economic status (Branson et al., 2009a). However, success in higher education has been relatively narrowly defined and is often restricted to enrolment and graduation rates, employment and earnings. One clear exception is research (Cosser & du Toit, 2002; Cosser et al., 2004; Cosser, 2009, 2010) on the links between the course preferences of secondary students and degree outcomes at university. On the whole, this work has concluded that there is evidence of a wide disparity between learner preferences or ambitions and actual higher education enrolments (Cosser, 2010).

In addition to identifying a gap between intentions and enrolments with respect to degree choice, this work has suggested that the mismatch between intentions and outcomes differs both by race and field of study. For example, Cosser et al. (2004) found that, among school leavers who intended to pursue a degree in SET, white students were more likely to enrol in such a programme than students from other race categories. Concomitantly, the preference gap for black Africans who intend to study an SET subject has been particularly and persistently larger (Cosser, 2009).

Two further contributions from this literature are, first, that poor academic performance during the last year of schooling (matric) is the main reason for the gap between programme intentions and enrolment (Cosser et al., 2004). Second, however, is the finding that programme preferences also change considerably after enrolment; that is, during the course of university study (Cosser et al., 2004). This particular finding requires more attention and is one of the areas of focus for the present study.

2.2. Post-apartheid trends in graduate employment

The literature on graduate outcomes in both developed (Teichler, 2002, 2007) and developing countries (Mugabushaka et al., 2003, 2007; Al-Samarrai & Bennell, 2007) has been concerned largely with whether graduates find jobs and how university curricula align with the demands of the labour market. Similarly, the focus of the South African literature has been on whether or not graduate unemployment is a significant and growing problem. On the one hand, several studies (Bhorat, 2004; DPRU, 2006; Pauw et al., 2006b; Kraak, 2010) have suggested that graduate unemployment is increasing in South Africa. The general view according to this literature is that there is still a mismatch between the skills demanded by employers/firms and the training provided by universities (Bhorat, 2004; Bhorat & Oosthuizen, 2005; Kraak, 2005; DPRU, 2006). In particular, there has been some suggestion (see also Mugabushaka et al., 2007) that Humanities and Arts graduates are less likely to find employment than those from fields such as Engineering and the Medical Sciences (du Toit & Roodt, 2008).

On the other hand, more recent research (van der Berg & van Broekhuizen, 2012; Altbeker & Storme, 2013; van Broekhuizen, 2013) has suggested that the problem of graduate unemployment in South Africa has been exaggerated since the unemployment rate for people with university degrees has consistently been below 6% (according to the broad definition of unemployment). Graduate unemployment, like unemployment generally, was highest in about 2001, at which point about 8.4% of university graduates were unemployed, as were just over 18% of diploma holders (van der Berg & van Broekhuizen, 2012). Economic expansion between 2002 and 2007 reduced these rates of unemployment greatly and they remain very low in comparison with the rate of unemployment for people who have only a school education (Branson et al., 2009a; Altbeker & Storme, 2013). Much of the ‘problem’ of graduate unemployment therefore seems associated with students who have attended Technical Vocational Education and Training colleges (van der Berg & van Broekhuizen, 2012; van Broekhuizen, 2013).

Not all graduates, however, experience the labour market on equal terms. A number of studies on graduate employment in the post-apartheid period (Moleke, 2005a; Pauw et al., 2006b; Branson et al., 2009b; Bhorat et al., 2010; Letseka et al., 2010) suggest that race, gender and type of institution (i.e. historically white universities vs. historically black universities [HBUs]) are still significant determinants of labour market outcomes. Although some of the disadvantage in labour market outcomes is related to the field of study, there is evidence to suggest that black African graduates, and particularly those from HBUs, are significantly less likely to find employment immediately after graduation, even after controlling for field of study (Moleke, 2005a). The reasons for the poorer employment prospects for graduates of HBUs are not clear, but firm-level research has suggested that some employers may still perceive HBUs as having a lower quality of graduates (DPRU, 2006; Pauw et al., 2006a). An alternative explanation for graduate unemployment (Kraak, 2010) which has gained some traction is that graduates with general degrees, and particularly those from the HBUs, enter the labour market without any substantive social networks among private-sector firms and enterprises.

2.3. Graduate destination studies in South Africa

Much of the analysis on South African university graduates and their labour market characteristics has been based on data from national Labour Force Surveys. However, there have also been a handful of dedicated graduate tracer (or destination) studies conducted in South Africa. The first national graduate study (Moleke, 2005a, 2005b) to focus on university graduates in South Africa traced 2672 students who graduated between 1990 and 1998. The study (Moleke, 2005a) found the rate of unemployment among university graduates to be generally low (about 94% of graduates found employment within a year of graduation) and, where unemployment did occur, it was only for short periods.

Despite these low levels of unemployment, the study showed that black African graduates, women, those with degrees in the Humanities and graduates from HBUs were all significantly more likely to report having experienced a period of unemployment (Moleke, 2005a). Among the employed, employment sectors also seem to differ by race and the findings suggest that the public sector is often the first employer for black African and coloured graduates, while white and Indian/Asian graduates obtain their first job in the private sector (Moleke, 2005a).

In 2005, the Human Sciences Research Council (HSRC) (see Cosser & Letseka, 2010) extended this earlier work with a graduate tracer study which investigated how the field of study is chosen and what determines the success of university outcomes and transition to the labour market. Once again, one of the key findings was that black African graduates and those who obtained a degree in the Humanities, in particular, were more likely to be unemployed (Bhorat et al., 2010; Moleke, 2010). The findings lead to the unfortunate conclusion that race is still one of the strongest indicators of both graduation and employment, even after controlling for type of institution and field of study (Bhorat et al., 2010).

Finally, the most recent South African graduate tracer study was conducted by the Cape Higher Education Consortium (CHEC, 2013). The CHEC study aimed to trace all 2010 graduates from the four public universities in the Western Cape. The respondents were contacted in 2012 to identify employment and unemployment transitions and outcomes in the two years following graduation. Overall the study found that 84% of the interviewed cohort was employed at the time of the interview. Similar to the two earlier studies, two of the key findings were that the burden of unemployment was highest among black African graduates and that the institutional differences were significant (CHEC, 2013).

3. Methods

3.1. Research design

The data analysed in this article come from a graduate tracer study which interviewed successful graduates from the two traditional universities² in the Eastern Cape Province of South Africa, namely Rhodes University and the University of Fort Hare (UFH). Respondents were drawn from a stratified³ random sample⁴ of all graduates who completed a three-year or four-year bachelor's degree in either 2010 or 2011. Aggregate information on graduates was accessed through the Higher Education Management Information System and unit record information on each graduating group was obtained from the respective university administrations. Responses from a total of 469 graduates from Rhodes and 742 graduates from Fort Hare were successfully captured (n = 1211). Given the difficulties in obtaining reliable contact details for graduates, the survey was administered both telephonically and through an online survey platform. On the whole, the response rates (39% of the Fort Hare sample and 47% of the Rhodes sample) were appreciably higher than for past South African tracer studies. Once the data collection was complete, statistical weights were estimated in order to correct for non-response.

3.2. Analysis

The closest empirical antecedent to the present study is the analysis by Bhorat et al. (2010), which estimated the probability of unemployment from the HSRC's graduate tracer study in 2005. The analysis for the present study has two distinct parts. The first consists of a descriptive analysis of degree choice and the transition to the labour market among graduates from the two universities. The second part then identifies, through a multivariate analysis (logit estimation), how schooling quality,⁵ academic achievements, field of study as well as race, gender and socio-economic status are associated with the probabilities of completing a first-choice degree and of finding employment.

3.3. Limitations

There are several important limitations to the study's design. First, the study team did not have full access to student records. As a result, the information captured in the survey is the result of retrospective, self-reported evaluations and cannot be verified by administrative records. Second, and related to the first, the survey respondents were also asked to act as secondary sources of information on such household characteristics as parental employment, income and education and tertiary education among siblings. Selection bias, which is commonly associated with tracer studies, could apply because the contactability of respondents may be correlated with a number of the outcomes being investigated (e.g. employment status).

3.4. Graduate sample characteristics

Since Rhodes University is classified as a historically white university and Fort Hare is an HBU, it is not surprising that the racial composition of graduates from the two universities is very different. Most graduates from Rhodes (57% of the 2010 and 2011 cohorts) are still ‘white’ while only 35% are ‘black African’. The vast majority (93%) of graduates from Fort Hare, however, are classified as black African and less than 5% are white. While there has been some transformation, particularly at Rhodes, over the past 20 years, it is clear that the racial compositions of the two universities still reflect their historical positions in South Africa's higher education system.⁶

There is very little evidence to support the claim that race is strongly associated with field of study (Table 1). In particular, the suggestion (e.g. Moleke, 2005b) that black African students, and particularly those who study at HBUs, are significantly more likely to enrol in programmes (such as Humanities) which have a lower likelihood of employment does not appear to apply to Rhodes and Fort Hare graduates. Moreover, black African graduates from Fort Hare are not significantly more likely to have completed a degree in the Humanities than similar graduates from Rhodes.

Table 1. Field of study, by university and population group.

	Black African	Coloured	Indian/Asian	White	Total
Rhodes University
SET	21.72 (3.70)	1.92 (1.96)	40.66 (10.24)	19.16 (2.21)	20.60 (1.91)
Business / Commerce	34.98 (4.41)	19.23 (11.91)	23.08 (7.67)	21.08 (2.37)	26.05 (2.18)
Education	1.69 (0.97)	3.85 (2.82)	0.00 (0.00)	0.44 (0.22)	0.95 (0.37)
Humanities	41.61 (4.19)	75.00 (12.06)	36.26 (10.10)	59.32 (2.86)	52.40 (2.38)
Total	100.00	100.00	100.00	100.00	100.00
University of Fort Hare
SET	18.54 (1.47)	8.20 (5.87)	0.00 (0.00)	0.00 (0.00)	17.43 (1.38)
Business / Commerce	24.45 (1.58)	26.23 (12.73)	46.15 (29.15)	39.06 (8.18)	25.23 (1.54)
Education	9.27 (1.40)	11.48 (7.80)	0.00 (0.00)	25.00 (8.08)	9.97 (1.36)
Humanities	47.73 (1.94)	54.10 (13.76)	53.85 (29.15)	35.94 (9.40)	47.37 (1.88)
Total	100.00	100.00	100.00	100.00	100.00

Notes: The data are weighted. Standard errors are in parentheses.

In light of the large differences in schooling quality in South Africa, one important finding is that the schooling histories of the two sets of graduates are very different (Table 2). About half of the Rhodes cohort attended public elite schools (compared with about one-third of Fort Hare graduates). These are often described as former Model C schools and, while classified as public institutions, the tuition fees are often high (typically prohibitively so for low-income households), the learner to teacher ratios are low and the schools are relatively well resourced in terms of infrastructure. There is also a considerable ‘elite’ element to the Rhodes graduate group. About 30% attended private schools with very high tuition fees. Over half (53%) of Fort Hare graduates, on the other hand, attended low-cost public schools. These schools are generally associated with lower academic achievements, high learner to teacher ratios and relatively poor infrastructure.

Table 2. Type of school attended.

	Rhodes University	University of Fort Hare	Total
Public elite	49.88 (2.41)	33.68 (1.92)	40.35 (1.52)
Public low cost	14.77 (1.79)	52.66 (2.02)	37.05 (1.50)
Private elite	29.60 (2.17)	3.72 (0.73)	14.38 (1.06)
Private low cost	5.09 (1.06)	9.17 (1.17)	7.49 (0.82)
Home schooling	0.17 (0.17)	0.16 (0.16)	0.17 (0.12)
Farm school	0.49 (0.35)	0.61 (0.30)	0.56 (0.23)
Total	100.00	100.00	100.00

Notes: The data are weighted. Standard errors are in parentheses.

4. Findings

4.1. Degree preferences and completion

Despite the far lower levels of schooling quality among the Fort Hare graduates, degree preferences between the two groups were similar. For example, the same percentage (30%) of graduates from both universities reported that, during their final year of school, they planned to study a discipline within the field of SET (Table 3). At the same time, the Fort Hare cohort exhibited a slightly higher preference (relative to that of Rhodes graduates) for Commerce and a slightly lower intention to study Humanities. In terms of realising these intentions, about 47% of the Rhodes graduates and 41% of the Fort Hare graduates went on to complete a degree in their first-choice subject. In other words, Rhodes graduates were only slightly (and not significantly) more successful in completing the degree which they intended to study while still in school.

Table 3. Intended field of study while still in matric (first choice).

	Rhodes University	University of Fort Hare	Total
SET	30.02 (2.21)	29.67 (1.77)	29.81 (1.38)
Commerce	24.84 (2.18)	29.18 (1.73)	27.49 (1.35)
Education	0.76 (0.39)	3.78 (0.79)	2.60 (0.51)
Humanities	44.38 (2.39)	37.37 (1.87)	40.10 (1.47)
% who went on to complete the intended degree at university	47.41 (2.37)	40.99 (1.84)	43.46 (1.46)

Notes: The data are weighted. Standard errors are in parentheses.

However, these figures mask large differences between fields of study (Table 4). At Rhodes, for example, about 60% of graduates who intended to study an SET discipline successfully completed an SET degree (but not necessarily in the same discipline or subject as was originally intended). Among Fort Hare graduates, however, fewer than half (48%) of those who intended to obtain an SET degree did so. Across the four CESMs, Rhodes graduates were significantly more likely than Fort Hare graduates to complete a degree in the field in which they originally intended to enrol. Among all graduates who changed their study category (between matric and university graduation) the largest percentage switched to Humanities. For example, among those who intended to study SET and Education, 26% graduated in a Humanities discipline instead⁷ (data not shown).

Table 4. Graduation in intended field of study, by first-choice field of study in matric.

	Rhodes University	University of Fort Hare	Total
SET	59.75 (4.34)	47.90 (3.56)	52.55 (2.78)
Business / Commerce	81.28 (3.71)	66.96 (3.45)	72.00 (2.67)
Education	–	65.90 (9.70)	59.94 (9.45)
Humanities	92.27 (1.77)	84.80 (2.25)	88.02 (1.50)

Notes: The data are weighted. Standard errors are in parentheses.

The main reasons for changing from the initial intended course of study also differ between the two groups (Table 5). The main reason that UFH graduates changed their intended course of study (32%) was that their marks were not good enough to gain entry or to continue to completion. The Fort Hare group was also more likely to change degrees due to future employment considerations (7% believed there to be a lack of jobs in their initial study area) or because of the lack of scholarship funding for their first-choice subjects (14%). Among the Rhodes graduate group the main reason was a loss of interest (48%).

Table 5. Reasons for not completing intended course of study.

	Rhodes University	University of Fort Hare	Total
Lack of jobs in South Africa	2.79 (1.13)	7.36 (1.43)	5.74 (1.01)
No scholarship	5.79 (1.74)	14.24 (2.07)	11.25 (1.49)
Marks not good enough	23.29 (3.13)	31.64 (2.55)	28.68 (2.00)
No places available	9.81 (2.25)	24.32 (2.41)	19.19 (1.78)
Started but could not continue	13.37 (2.42)	5.42 (1.32)	8.23 (1.21)
Lost interest	48.11 (3.62)	20.47 (2.14)	30.25 (1.97)

Notes: The data are weighted. Standard errors are in parentheses. Cells are not mutually exclusive and therefore columns do not total to 100%.

Examination of the reasons for not completing the intended course by CESM category (data not shown) suggests several interesting findings. For example, among those who did not complete an intended SET course at Fort Hare, nearly one-quarter indicated that their marks were too low to continue and 23% reported that they could not find a place in the relevant programme. Somewhat surprisingly, an even higher percentage (38%) of Rhodes graduates who did not complete an intended SET degree reported that their marks were too low. However, 32% also indicated that they simply lost interest in the subject (compared with 22% from Fort Hare).

4.2. Transition to the labour market

Turning now to the transition from university to the labour market, Figure 1 identifies the unemployment rates of graduates. The most striking finding is the difference in unemployment rates between Rhodes and Fort Hare graduates. The unemployment rate of 7% among Rhodes graduates corresponds with the national average for university graduates (see Pauw et al., 2008; van der Berg & van Broekhuizen, 2012) while the unemployment rate (20%) among Fort Hare graduates is almost three times higher.

Figure 1. Broad unemployment rates (as of 1 March 2014), by field of study. Notes: The data are weighted.

Contrary to some of the expectations described in the literature, the descriptive statistics do not provide any evidence that the risk of unemployment for Humanities graduates is significantly higher than for other fields of study. While Humanities graduates from Rhodes are more than twice as likely to be unemployed than SET and Business graduates, the difference is not statistically significant (i.e. confidence intervals overlap at the 95% level of significance). At Fort Hare, SET graduates actually have a higher rate of unemployment but, again, the difference is not significant. The significantly lower risk of unemployment among Education graduates (about 9% among the Fort Hare sample) is probably the result of the practical application of a teaching degree and relatively easy absorption into the teaching profession. One key area where there are significant differences between graduates is in employment sectors (data not shown). The vast majority (73%) of Rhodes graduates are employed in the private sector while 67% of Fort Hare graduates work for the government (see also Moleke, 2005a).

Finally, among the graduates who are employed (but excluding the small number who are self-employed), there are some important differences in job search strategies which may explain some of the differences in labour market outcomes (Table 6). The single most common way that Rhodes graduates found their current job was through personal contacts or networks (30%). Moreover, if the categories of ‘relatives’, ‘social media’ and ‘personal contacts’ are combined, then just under half of Rhodes graduates found their current employment through a social network. Fort Hare graduates, on the other hand, relied more on newspaper advertisements (36%) than on any other specific search strategy.

Table 6. Means of finding employment (among employees, i.e. not the self-employed).

	SET	Commerce	Education	Humanities	Total
Rhodes University
Employment agency	13.16 (4.06)	12.73 (3.58)	0.00 (0.00)	7.25 (1.87)	9.84 (1.61)
Relatives	6.04 (2.66)	5.15 (2.12)	0.00 (0.00)	7.75 (1.92)	6.59 (1.27)
Linked to bursary	1.55 (1.54)	4.43 (2.27)	0.00 (0.00)	2.10 (1.04)	2.63 (0.89)
Social media	4.96 (2.46)	15.08 (3.97)	21.05 (18.05)	11.82 (2.32)	11.56 (1.74)
Personal contacts	21.02 (4.56)	27.07 (4.56)	27.63 (18.35)	34.12 (3.41)	29.56 (2.38)
Newspaper	8.76 (3.91)	9.78 (3.44)	60.53 (18.99)	11.08 (2.24)	10.86 (1.71)
Campus recruitment	3.77 (2.17)	8.69 (2.90)	6.58 (6.66)	5.23 (1.61)	5.95 (1.24)
University of Fort Hare
Employment agency	0.00 (0.00)	4.50 (1.57)	0.00 (0.00)	1.62 (0.80)	1.92 (0.56)
Department of Labour	10.03 (3.18)	6.16 (1.89)	11.90 (5.62)	8.82 (1.80)	8.68 (1.28)
Relatives	1.21 (1.20)	3.08 (1.36)	1.82 (1.81)	3.42 (1.22)	2.79 (0.73)
Linked to bursary	12.11 (3.44)	4.92 (1.70)	21.56 (6.59)	15.48 (2.32)	12.91 (1.52)
Social media	5.84 (2.54)	7.84 (2.09)	0.00 (0.00)	4.36 (1.29)	4.99 (0.91)
Personal contacts	8.39 (2.86)	13.84 (2.66)	7.84 (3.82)	11.78 (2.15)	11.32 (1.37)
Newspaper	36.62 (5.04)	40.69 (3.87)	20.55 (6.08)	37.50 (3.13)	36.23 (2.10)
Campus recruitment	8.82 (2.99)	4.16 (1.55)	3.30 (3.23)	4.75 (1.34)	5.08 (0.96)

Notes: The data are weighted. Standard errors are in parentheses. Cells are not mutually exclusive and therefore columns do not total to 100%.

4.3 Estimations

In order to explore the descriptive findings on first-choice degree completion and graduate unemployment in greater detail, two sets of logit regressions were estimated. Each set of estimations is presented first as a pooled sample and, given the vast differences between the two universities, both institutionally and in terms of graduate characteristics, and then the results are presented separately by institution. In addition to the variables listed in the first column of Tables 7 and 8, both sets of regressions also control for Senior Certificate Examination levels and symbols, academic achievement at university, parental levels of education or parental employment status.⁸

In the first set (Table 7), the correlates of graduating with a first-choice degree are presented. The results from the pooled sample (column I) indicate that those graduates who intended to pursue an SET or Commerce degree were significantly less likely, relative to those who planned to complete a Humanities degree, to have completed their first-choice degree. Black African graduates and those from low-income schools are also less likely to have completed a first-choice degree, even after controlling for schooling achievements and other background factors. The interaction effects (column II) suggest further that black African graduates from Fort Hare and black African female graduates (on the whole) are significantly less likely to have completed their first-choice degree.

Table 7. The correlates (logit estimations) of completing a first-choice university degree.

	Pooled		Rhodes		University of Fort Hare
	(Column I)	(Column II)	(Column III)	(Column IV)	(Column V)	(Column VI)
UFH	0.007 (0.213)	0.928** (0.389)	---	---	---	---
Female	−0.162 (0.133)	0.191 (0.241)	0.0176 (0.227)	0.104 (0.274)	−0.263 (0.171)	0.805 (0.726)
Black African	−0.340* (0.196)	0.230 (0.287)	−0.141 (0.243)	−0.035 (0.392)	−1.074*** (0.366)	−0.379 (0.515)
Matric
Mathematics or science higher grade	−0.191 (0.184)	−0.205 (0.183)	−0.418 (0.258)	−0.445* (0.259)	−0.100 (0.281)	−0.142 (0.287)
English higher grade	−0.072 (0.150)	−0.056 (0.151)	0.282 (0.325)	0.287 (0.330)	−0.217 (0.174)	−0.222 (0.174)
Mathematics upper-class pass	−0.003 (0.197)	−0.044 (0.198)	−0.475* (0.247)	−0.481* (0.247)	0.786** (0.330)	0.758** (0.334)
Science upper-class pass	0.101 (0.254)	0.130 (0.254)	0.407 (0.308)	0.445 (0.313)	−0.302 (0.473)	−0.255 (0.487)
Low-quintile school	−0.492*** (0.143)	−0.220 (0.301)	−0.462* (0.279)	0.021 (0.421)	−0.452*** (0.172)	−0.248 (0.255)
Ambitions
SET_matric	−1.296*** (0.165)	−1.092*** (0.205)	−1.196*** (0.262)	−1.057*** (0.279)	−1.375*** (0.224)	−1.154*** (0.320)
Comm_matric	−0.560*** (0.159)	−0.315 (0.221)	−0.753*** (0.289)	−0.510 (0.319)	−0.528*** (0.197)	−0.198 (0.330)
Edu_matric	0.343 (0.443)	−0.750 (0.678)	−2.438** (1.035)	−3.178*** (1.183)	0.497 (0.485)	−0.132 (0.762)
Interactions
Black×UFH		−1.142*** (0.423)
Low quintile ×SET_matric		−0.481 (0.333)		−0.772 (0.710)		−0.384 (0.438)
Low quintile×comm_matric		−0.547* (0.313)		−1.074 (0.669)		−0.521 (0.407)
Low quintile×edu_matric		1.329 (0.892)		---		0.728 (0.957)
Black×female		−0.492* (0.287)		−0.164 (0.478)		−1.109 (0.748)
_cons	5.546*** (1.394)	5.006*** (1.408)	6.277 (4.371)	1.787 (5.808)	5.886*** (1.671)	5.145*** (1.736)
(Unweighted) N	1156	1156	445	444	711	711

Notes: The data are weighted. Standard errors are in parentheses. The regressions also include controls for: age, parental level of education and mode of data collection.

*p < 0.10.

**p < 0.05.

***p < 0.01.

Among the Rhodes sample (column III), those from low-quintile schools are also less likely to have completed an intended degree, as are graduates who initially intended to study SET, Commerce or Education. Interestingly, there are no racial differences in degree completion among the Rhodes sample and none of the interaction terms are significantly associated with graduating with a first-choice degree. The results from the Fort Hare sample (column V) identify black African graduates and those from low-quintile schools as being significantly less likely to have successfully completed an intended degree. While UFH graduates who reported initial intentions to undertake SET or Commerce degrees are also less likely to have completed those degrees, none of the interaction effects are significant.

The correlates of unemployment among graduates are presented in Table 8. In the pooled sample (column I), the estimates are generally in line with the existing literature and suggest that there is a significantly higher risk of unemployment for Fort Hare graduates, black African graduates, women and those who completed a degree in the Humanities. However, the expected interaction (column II) between Humanities and race is not significant, which suggests that black African graduates who completed a degree in the Humanities are not at a specific risk of unemployment. Black African graduates from Fort Hare are also not at a higher risk of unemployment over and above the risks identified in the main effects regression (column I). Black African women and particularly those from low-quintile schools, on the other hand, face a significantly higher risk of unemployment.

Table 8. The correlates (logit) of unemployment among Rhodes and Fort Hare graduates.

	Pooled		Rhodes		University of Fort Hare
	(Column I)	(Column II)	(Column III)	(Column IV)	(Column V)	(Column VI)
UFH	0.735** (0.342)	0.497 (0.809)
Female	0.319* (0.188)	−1.239* (0.713)	0.102 (0.462)	−1.429 (1.377)	0.350* (0.206)	−0.148 (0.399)
Black African	1.226*** (0.358)	0.862 (0.559)	1.576*** (0.383)	0.164 (1.381)	---	---
Matric
Mathematics or science higher grade	0.116 (0.268)	0.152 (0.282)	0.494 (0.452)	0.485 (0.535)	−0.048 (0.348)	−0.021 (0.350)
English higher grade	−0.100 (0.192)	−0.098 (0.196)	−0.857* (0.506)	−0.873 (0.560)	−0.008 (0.207)	−0.014 (0.209)
Low-quintile school	0.276 (0.193)	0.735 (0.846)	−0.346 (0.677)	0.665 (1.409)	0.424* (0.217)	0.976* (0.471)
Field of study
SET	0.256 (0.265)	0.952** (0.452)	0.0368 (0.944)	−0.042 (1.063)	0.408 (0.282)	1.435*** (0.507)
Humanities	0.425* (0.225)	1.180* (0.670)	0.995 (0.732)	1.270 (0.870)	0.345 (0.239)	1.166** (0.596)
Education	−0.513 (0.519)	0.940 (0.732)	---	---	−0.627 (0.528)	1.094 (0.754)
Interactions
Black×UFH		−0.088 (0.809)
Low quintile ×Black		−0.854 (0.880)		−0.445 (1.492)
Low quintile ×UFH		1.059 (0.723)
Black×female		1.305* (0.692)		2.489** (1.076)
Low quintile ×Humanities		−1.186** (0.517)		−1.507 (1.619)		−1.116** (0.564)
Low quintile ×SET		−1.130** (0.577)		0.209 (1.804)		−1.585*** (0.628)
Low quintile ×Education		−2.442** (1.040)		---	---	−2.762*** (1.078)
Black×Humanities		0.033 (0.694)		0.140 (1.487)
Female× Humanities		−0.007 (0.383)		0.127 (1.335)		−0.024 (0.419)
Female×low-quintile school		0.667* (0.403)		0.360 (1.387)		0.822* (0.446)
_cons	−1.299 (1.574)	−1.175 (1.646)	−9.262** (4.386)	−9.538** (4.885)	1.122 (1.672)	0.719 (1.731)
(Unweighted) N	1104	1104	429	429	667	667

Notes: The data are weighted. Standard errors are in parentheses. The regressions also include controls for: age, parental employment, Senior Certificate Examination marks (self-reported), average university mark (self-reported) and mode of data collection.

*p < 0.10.

**p < 0.05.

***p < 0.01.

Among Rhodes graduates (where unemployment levels are very low), there are very few significant correlates of unemployment (column III). Importantly, however, the risk of unemployment is significantly higher for black African graduates and, in particular, black African women (column IV) even after controlling for field of study, schooling quality, socio-economic factors and the employment status of their parents. The disappointing conclusion from these estimations is therefore that race and gender (and not achievements) appear to be consistent predictors of success in the labour market.

The more interesting question is what explains the relatively higher risk of unemployment among the Fort Hare sample where race is unlikely to explain differences within this group.⁹ The results show that, in the main effects regression (column V), the two significant correlates of unemployment are gender and low-income schooling. Moreover, in the final regression (column VI), one of the crucial factors associated with unemployment among Fort Hare graduates is the interaction (0.822) between these two variables. In other words, being both female and coming from a low-income school carries an extra risk of unemployment over and above the risks identified in the previous column. The interactions also show an association between field of study (particularly SET and Education) and a lower risk of unemployment (relative to Commerce graduates) among Fort Hare graduates with a poor schooling background.

5. Discussion

The results of this study have contributed to the post-schooling literature in South Africa in two main ways. First, the findings extend the existing work on degree or programme choice by suggesting that the completion of a first-choice degree is further conditioned by ‘pre-higher education’ factors such as school quality, race, gender and intended field of study. To the extent that the self-reported information on school achievements and past study intentions are accurate, this means that at least part of the preference gap identified by Cosser (2009) is explained by poor schooling backgrounds and, by extension, a lack of adequate preparation for university study.

The implication is that schooling quality and low socio-economic status do not only have the expected impacts on higher education access or performance, but they are also clearly linked with study choices and employment. This suggests that the poor quality of support interventions (e.g. career guidance) in the run-up to Grade 9 – which is the year of subject choice in South African schools – in under-resourced schools could have long-term negative effects, as can poor support for learners as they approach matric (Stumpf et al., 2012). This is an important issue of equity if learners from poorer schools who have demonstrated an academic ability (i.e. they qualified for and graduated from a university) are not able to follow their intended course of study.

Of course, this is also an issue of the efficiency of the education system in South Africa in converting potential human resources into the types of high level and scarce skills which are required in the labour market. Given the reality of poor schooling quality in South Africa, one suggestion (see Stumpf et al., 2012) has been to adopt a number of Australian initiatives to strengthen learner decision-making on further training and study. This suggestion has been made with particular reference to the Technical Vocational Education and Training sector, but it seems that, based on the findings presented here, this type of intervention could apply equally to universities, and HBUs in particular.

Second, the findings suggest that, while graduate unemployment is far higher among Fort Hare graduates, at least some of this disadvantage is actually carried over from the type of schooling obtained by graduates. The implication is therefore that interventions aimed at improving the employment prospects of graduates from HBUs should be targeted at university students from the low-quintile schools or, as already suggested, pupils from those schools who might qualify to attend university. As identified in much of the literature on NEETs in South Africa, this is the group which is currently falling through the cracks of the post-schooling system in South Africa.

In line with many of the debates in the international literature (Teichler, 2002, 2007; Schomburg & Teichler, 2006; Nunez & Livanos, 2010), much of the focus in the South African literature has been on the relevance of university curricula (and Humanities in particular) to skills needed in the job market. Critically, the findings from this study do not provide much support for a switch away from Humanities, particularly if, as Cloete (2004:74) suggests, the future of higher education depends on the creation of ‘self-programmable labour’. The results of both the descriptive statistics and the multivariate analysis, for example, do not support the conventional wisdom that black African students (including those who are enrolled in HBUs) are more likely to enrol in subjects with poorer prospects for immediate employment (Moleke, 2005b; Pauw et al., 2008). Moreover, the evidence from the Eastern Cape universities does not suggest that Humanities graduates are significantly more likely to be unemployed after controlling for other factors.

Given the aforementioned, the remaining crucial question is whether the far higher rate of unemployment among Fort Hare graduates is really the result of the oversupply of certain skills (Woolard et al., 2003; Pauw et al., 2008) or whether factors such as poor matching (Altman & Marock, 2011), poor signalling – including the effects of perceptions and preferences of prospective employers about graduates from HBUs such as Fort Hare – or a lack of appropriate social networks in the labour market (Kraak, 2010) apply. While not definitive, the evidence presented in this article tends to support the latter three factors.

Poor access to professional social networks, particularly for first-generation students, may explain not only the high rates of university drop-out (see Letseka & Maile, 2008), but also degree choices and the unsuccessful transition to the labour market (Kraak, 2010; Altman & Marock, 2011). In support of this conclusion, the descriptive findings on job search strategies have suggested that different search strategies are successful in the private and public sectors, and that Rhodes graduates appear to be more successful in leveraging their networks to obtain private-sector employment (Al-Samarrai & Reilly [2008] report similar findings from a Tanzanian graduate study). One policy implication is therefore that proactive matching interventions (see Altman, 2007; Altman & Marock, 2011) may be required to link university graduates with appropriate labour market opportunities.

6. Conclusion

University graduates, and especially those from poorly resourced schools and low-income communities, are an important human resource precisely because they have demonstrated a tangible ability to succeed. Perhaps the main contribution of the graduate tracer study presented in this article is the conclusion that policy should focus most closely on university students from poorly resourced schools and as early as possible in their university studies, in addition to interventions in those schools themselves, because poor academic performance appears to be a major factor in students’ abandoning their first-choice fields of study. Also, the results of the tracer study provide support for the literature which shows that, while the problem of graduate unemployment in South Africa is relatively small, it is highly concentrated in historically disadvantaged universities and particularly among graduates from poor secondary schools. Rather than confronting study choices, per se, to address graduate unemployment, policy should focus on improving the match between these graduates and the labour market, not only by addressing the supply-side issues explored in this study, but also by considering evidence in the literature about the shaping of labour demand by employer preferences and employment practices.

Acknowledgements

The authors would like to thank Rod Bally and Kevin Whitfield from the University of Fort Hare for their assistance in coordinating the research. Thanks must also go to the CHEC research team and to Haroon Bhorat for providing insight into the study methodology. Finally, the authors are indebted to Ulandi du Plessis and her field team, all of the respondents who gave generously of their time and the helpful comments of two anonymous reviewers.

The ideas, opinions, conclusions or recommendations expressed in this article are strictly those of the authors and do not represent those of the Department of Higher Education and Training.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The research was conducted under the Labour Market Intelligence Partnership, a research consortium led by the HSRC, South Africa, and funded by the Department of Higher Education and Training.

Notes

¹South Africa has one of the world's highest measures of intra-class correlation, which is a standardised indicator of variation in academic achievements (Branson & Zuze, 2012).

²Traditional universities offer theoretically-oriented degrees and are less vocational compared with comprehensive universities and universities of technology (see de Villiers et al., 2013).

³The sample was stratified by field of study – as categorised by the South African Classification of Educational Subject Matter (CESM) manual. The categories included: Science Engineering and Technology (SET), Business and Commerce, Education, and Humanities.

⁴Random (stratified) samples of 50% of Rhodes graduates and 70% of UFH graduates were drawn. The higher percentage sampled from UFH was due to the lower response rate among graduates from that institution. The end result is that about 25% of all graduates from each university participated in the study (the total population of graduates was 4927).

⁵Schooling quality is defined here in terms of school resources. Public schools are divided into quintiles according to their poverty ranking. In the regressions, the school poverty variable is constructed as a dummy variable where quintile 5 schools and elite privates schools are the reference category (0) and quintiles 1–4 are coded as ‘1’.

⁶There are a very small number of graduates from the coloured and Indian/Asian population groups from the two universities.

⁷Similarly, 20% of those who intended to study a Commerce degree but graduated in a different CESM switched to a degree in Humanities.

⁸In order to test for a possible bias associated with the method of data collection, both sets of estimations include a dummy control for whether the survey was completed online.

⁹The number of Indian/Asian and white graduates from Fort Hare who are unemployed is so low that they were excluded from the final two estimations in the table.