Revisiting graduate unemployment in South Africa

71point4 > Blog archive > 2019 > February > 13 > Revisiting graduate unemployment in South Africa
71point4 > Blog archive > 2019 > February > 13 > Revisiting graduate unemployment in South Africa

Revisiting graduate unemployment in South Africa

Posted by: Chris Garbers
Category: Education

Is there a graduate unemployment problem in South Africa? We look at available data to answer that question.

Our conclusion: it certainly looks like it, and it looks like a problem that impacts on young black women disproportionately. But available data is just not good enough for us to be absolutely certain. If we care enough, we should be looking at better data that lies in university administrative systems and SARS.

Graduate unemployment appears to be on the rise

In 2013 the Centre for Development and Enterprise (CDE) published a study on graduate unemployment in South Africa. That study, based on data from the 2011 Quarterly Labour Force Survey (QLFS), found that rates of unemployment for working-age graduates were low (just under 5%) and not a serious concern (see Altmeter & Storme, 2013)1.

But a more nuanced analysis of the data points to worrying trends. As indicated by Figure 1, more recent data shows that graduate unemployment has increased and has grown more rapidly than in other education cohorts (e.g. individuals with only a diploma, matric, or less than matric)2. That said, graduates still achieve a much more favourable labour market outcome than their less-educated compatriots: the average graduate unemployment rate in 2018 was 7%, compared with 12% for diplomates, 28% for those with matric, and 31% for those with less than matric.

Figure 1

Narrow unemployment rate by level of education (working age)

 

Source: QLFS 2008Q1 – 2018Q3. The numbers indicate the average unemployment rate across all surveys within a year for each segment.

 

To some extent, this increase in graduate unemployment can be attributed to South Africa’s weak growth performance over the last decade. Indeed, between 2002 and 2007 when the average annual real GDP growth rate was 4.4%, graduate unemployment declined from 6% to 3%. In contrast, the near doubling of the graduate unemployment rate between 2008 and 2018 occurred in an environment where GDP growth averaged just 1.8% per annum3. Over that same time period, the supply of university graduates expanded at around 5% per year; QLFS data indicates that the total number of graduates in South Africa increased from 1.1 million in 2008 to 1.7 million in 20184. According to the Higher Education Management Information System5(HEMIS), in 2017 South African universities delivered in the region of 190 000 new graduates per year6, up from 118 000 in 2008. At the same time, according to the QLFS the formal sector appears to absorb an average of 41 000 additional graduates per year7.

Figure 2

Graduates per university: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through http://www.heda.co.za.

 

Figure 3

Gender composition of graduates per university: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through http://www.heda.co.za.

 

Figure 4

Racial composition of graduates per university: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through http://www.heda.co.za.

 

Data limitations inhibit meaningful analysis of graduate unemployment trends

Business cycle effects aside, this aggregate perspective of the graduate labour market can be somewhat misleading. While the term “graduates” in common parlance is typically associated with young, recent graduates, in fact, a “graduate” as defined by the 2013 analysis includes all individuals with a university degree, irrespective of age8 and work experience. In line with that definition, a 48-year old with a Business Science degree and 25 years work experience, and a 29-year old very recent PhD would both be classified as “graduates”.

In part, the focus on all graduates, both young and old, is a function of the limitations of the principal data source underpinning this analysis, namely the QLFS. As shown in the table below, there are relatively few survey respondents who are graduates, fewer who are young and fewer still who are unemployed. Refining this further by race or gender, not to mention field of study becomes impossible. In 2008Q1 for instance, no unemployed Indian/Asian female graduates between the ages of 18 and 34 were even sampled. A detailed analysis on graduate unemployment based on labour force survey data – especially when calculated for different graduate-segments – should therefore be interpreted with caution.

average number of graduates sampled per year

To overcome this sample size limitation, we have therefore pooled the data across quarters to generate a larger sample. For instance, in 2017, there were a total of 6 480 graduates who were surveyed. Of these, 2 225 were under the age of 35 and 391 were unemployed. Our analysis therefore reports annual statistics, derived as the average of all surveys administered within a given year9.

Using this pooled data, and noting its limitations, our analysis unsurprisingly shows that there are very different rates of unemployment for young graduates and older graduates: as per Figure 5, in 2018, the unemployment rate of graduates under the age of 35 was 12%, compared to 4% for those aged 35 and older10. Viewed from a different angle, of those graduates who are unemployed, 60% are under the age of 35.

Apart from age, several papers note that graduate unemployment also has distinctive racial characteristics11. Figure 6 shows that the unemployment rate for white young graduates is 3% while the unemployment rate for young (i.e. under 35) black graduates is 18%, a noticeable increase from 13% in 2008.

But within this, as illustrated in Figure 7, gender plays a critical role. In fact, the very high rates of unemployment among black female graduates in particular has been a driving factor behind the increasing trend in young graduate unemployment between 2008 and 2018. According to the QLFS, unemployment rates in that cohort doubled from 11% to 22% over that time period.

Figure 5

Narrow unemployment rate for graduates by age group

 

Source: QLFS 2008Q1 – 2008Q4, 2013Q1 – 2013Q4, and 2018Q1 – 2018Q3. Notes: The numbers indicate the average unemployment rate across all surveys within a year for each segment. Estimates are weighted.

 

Figure 6

Narrow unemployment rate for young African/black and white graduates

 

Source: QLFS 2008Q1 – 2008Q4, 2013Q1 – 2013Q4, and 2018Q1 – 2018Q3. Notes: The numbers indicate the average unemployment rate across all surveys within a year for each segment. Estimates are weighted. Young Indian/Asian/coloured graduates are excluded due to small sample sizes.

 

Figure 7

Narrow unemployment rate for young African/black and white graduates by gender

 

Source: QLFS 2008Q1 – 2018Q3. Notes: The numbers indicate the average unemployment rate across all surveys within a year for each segment. Estimates are weighted. Young Indian/Asian/coloured graduates are excluded due to small sample sizes. 95% confidence intervals (not shown here) overlap between 2008 and 2018 for young black males, white females, and white males. In line with this, we cannot be certain that unemployment rates in 2018 for these three segments were, statistically speaking, different to what they were in 2008. The notable exception is young black female graduates; the data indicates a statistically significant (95%) deterioration in unemployment rates for this cohort.

 

Of course, we cannot be certain that what we are seeing in the survey data is, statistically speaking, real. As is clear from the chart, the data is unstable, reflecting small sample sizes and inherent limitations with the survey instrument.

Nevertheless, figure 8 shows that in 2017, 45% of all new graduates produced by universities in South Africa were black women – and black women are now the single most dominant race / gender segment who obtain degrees in any year, by far. Putting this together with the weak capacity of the economy to absorb graduates, it is plausible that the unemployment rate of young black female graduates in particular has increased more rapidly than other segments.

Figure 8

Race and gender profile of university graduates: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through http://www.heda.co.za.

 

Figure 9

Black female graduates per university: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through https://www.heda.co.za.

 

Figure 10

Black female graduates per field of study: 2017

 

Source: HEMIS data as provided by IDSC and the University of Johannesburg through http://www.heda.co.za.

 

But wait, there is (and should be) more

No doubt those who, prior to reading this paper, were inclined to believe there is a graduate unemployment problem in South Africa will feel vindicated by these findings; on the face of it they were right, particularly when it comes to young, black, female graduates. We should note, however, that those who continue to insist we do not have a graduate unemployment problem will point to the very real limitations of the underlying data – sample sizes are small, and we have pooled data as a workaround. Further, they would argue that even if unemployment rates are higher than we would like, this is not a problem of graduate unemployment per se, but one of economic growth. By implication then, the problem (and the solution) does not lie with institutions that produce graduates, with the entities that might employ graduates or even with graduates themselves.

Thus, it would seem that even after analysis of available data, the debate about graduate unemployment in South Africa remains no more data driven than it was in the past. That is a great pity, because this debate actually matters for the economy, not to mention the promise a degree holds for young people and their families who sacrifice a great deal for a university education.

We can do better. There is abundant data that lies in various administrative systems. Data contained in the Higher Education Management Information System (HEMIS) includes unique identifiers for each graduate as well as the institution, field of study and time to completion for those who are no longer studying12. At the same time, SARS data contains the selfsame unique identifier (the ID number) together with employer and income data13. Combining these two data sets (adhering to sound practise with regard to identity masking and data protection) provides fertile ground for rich and very helpful research; formally employed graduates will appear in the tax data enabling an analysis of (among others) the time taken for graduates to find a formal job, the types of jobs that graduates find, and the level of remuneration that graduates can expect. It could help identify which institutions produce easily employed graduates, and possibly more critically, which fields of study are most likely to be rewarded with formal employment. Not only would this help policymakers and institutions to identify areas of strength and weakness, but it would also enable students to better assess whether an opportunity for further study is an investment worth pursuing. Clearly, given the stakes and the very helpful outcomes of such research, this learning endeavor should be a priority in higher education.

Featured image source: 

AFP/GianLuigi Guercia (https://africacheck.org/factsheets/factsheet-many-south-african-students-graduate/)

References

Altbeker, Anthony, and Evelien Storme. 2013. Graduate unemployment: A much exaggerated problem. Edited by Ann Bernstein. April. Johannesburg: Th Centre for Development; Enterprise. https://www.cde.org.za/graduate-unemployment-in-south-africa-a-much-exaggerated-problem/

Altman, Miriam. 2007. Youth labour market challenges in South Africa. Pretoria: Human Sciences Research Council; Human Sciences Research Council.

Kerr, Andrew. 2016. “Job flows, worker flows, and churning in South Africa.” WIDER Working Paper. United Nations University World Institute for Development Economics Research. https://www.wider.unu.edu/sites/default/files/wp2016-37.pdf.

Lam, David Allen, Murray Leibbrandt, and Cecil Mlatsheni. 2007. “Dynamics of Labor Market Entry and Youth Unemployment in South Africa : Evidence from the Cape Area Panel Study.” IPC Working Paper Series. University of Michigan International Policy Center.

Pauw, K, M Oosthuizen, and C van der Westhuizen. 2008. “Graduate unemployment in the face of skills shortages: A labour market paradox.” South African Journal of Economics 76 (1): 45–57.

Van Broekhuizen, Hendrik. 2016. “Graduate unemployment and Higher Education Institutions in South Africa.” Stellenbosch Economic Working Papers. Stellenbosch University. https://www.ekon.sun.ac.za/wpapers/2016/wp082016.

Van Der Berg, Servaas, and Hendrik Van Broekhuizen. 2012. “Graduate unemployment in South Africa: A much exaggerated problem.” Stellenbosch Economic Working Papers. Stellenbosch University. https://www.ekon.sun.ac.za/wpapers/2012/wp222012.

 

  1. The principle data source for this analysis is the QLFS. That survey is conducted every quarter (January – March, April – June, July – September, and October – December) of each year. Each survey comprises a sample of around 37 000 working age individuals (aged between 15 and 64 years). The survey includes detailed demographic data as well as labour market indicators such as employment, occupation, sector and industry of employment, and conditions of employment. With regard to education (question 1.7 in the survey), in line with the analysis prepared by the Centre for Development and Enterprise (2013) and Statistics South Africa’s (Stats SA) reporting protocol, respondents who cite their highest level of education as “Bachelors Degree”, “Bachelors Degree and Post Graduate Diploma”, “Honours Degree”, and “Higher Degree (Masters/PhD)” are grouped together as graduates.
  2. The compounded annual growth rate (CAGR) of graduate unemployment was 6.8% between 2008 and 2018. Over the same period, the CAGR of unemployment for diplomates, matriculants, and those with less than matric was 4.8%, 1.9%, and 1.9% respectively
  3. Economic growth data for South Africa was obtained from the World Bank at https://data.worldbank.org
  4. The stock of graduates is derived using population weights provided in the QLFS. QLFS data indicates an average growth rate of 4.5% in the graduate labour force between 2008 and 2018. HEMIS data for the institutions plotted in figure 2 over the 2008 – 2017 period puts this number at 5.5%. We report the average of these two figures.
  5. The higher education management information system (HEMIS) database contains information on all graduates from higher education institutions in South Africa. This dataset is the national repository for unit-record information on all students who have enrolled in and subsequently graduated from public higher education institutions in South Africa (Van Broekhuizen, 2016). This dataset contains information on graduates’ field of study, institution of study, as well as demographic information (e.g. race and gender).
  6. This figure excludes foreign graduates.
  7. In total, between 2008 and 2018, an average of 123 863 formal jobs were created each year. For graduates an average of 41 139 formal jobs were added each year. We note the data is not stable. For instance, in 2014 the QLFS indicates a total of 15 394 new graduate jobs. In 2015 it records 122 553.
  8. As noted, the analysis is restricted to those of working age, that is, aged 15-64.
  9. This is similar to the pooling approach used by Van Der Berg and Van Broekhuizen (2012:18). Yearly labour force statistics are calculated by taking the average across the four surveys conducted within each year. Since Stats SA is yet to release the data for QLFS Q4, the 2018 statistics are based on Q1, Q2, and Q3 data.
  10. Defining the youth as individuals aged between 18 and 34 is consistent with South Africa’s Youth Employment Service (YES), a business-led collaboration with government and labour that aims to create work opportunities for the youth. See https://www.yes4youth.co.za for more details.
  11. Several prior studies have flagged these age, gender, and racial traits of graduate unemployment, and identify a variety of factors that explain them. These studies rely, in the main, on survey data, and therefore suffer the same weaknesses as the analysis presented here. For instance, Altman (2007) and Pauw, Oosthuizen, and Van Der Westhuizen (2008) point to the importance of networks and prior work experience that equip graduates with job searching and workplace relations skills and enhance their employment prospects. Lam, Leibbrandt, and Mlatsheni (2007) note that whites are more likely than other population groups to have worked before leaving school. Van Broekhuizen (2016) finds that graduates from historically advantaged institutions have lower unemployment rates than graduates from historically disadvantaged universities and notes that historically disadvantaged universities still account for a large share of black graduations. See Van Der Berg and Van Broekhuizen (2012) and Van Broekhuizen (2016) for a review of previous studies on graduate unemployment and their associated shortcomings.
  12. For example, Van Broekhuizen (2016) uses HEMIS data to test whether the institutional legacy of universities bears an influence on graduate unemployment.
  13. Kerr (2016) uses the SARS dataset to offer insights on labour market churn in South Africa.
Author: Chris Garbers

Leave a Reply

2 Comments

  • Inger

    Thorough and worthwhile effort. Congratulations on style and presentation!

  • Rachel Garbers

    Very informative, well-written and to the point. I enjoyed reading this! It provides excellent perspective given the limitations in the data. I hope we can improve the quality of data in the future.