How large is the missing middle and what would it cost to fund?

ABSTRACT

The concept of the missing middle has gained currency within South African post-school education and training (PSET) discourse in recent years. The term has been defined as representing those too wealthy to benefit from National Student Financial Aid Scheme (NSFAS) funding, but who struggle to afford higher education. The missing middle currently refers to those students from households with incomes between R350 000 and R600 000. Whilst some 6% of South African households have incomes between R350 000 and R600 000, differing participation rates and average household income levels, amongst students from different race groups result in an estimated size of the missing middle in 2019 of 343 000 students out of a total PSET population of 1.4 m. The cost of funding full bursaries for all these students at 2019 prices is estimated at R19.2 bn. If a sliding scale of support is introduced, the total cost falls to R11.4 bn.

KEYWORDS:

• Post-school education and training
• funding
• missing-middle

1. Introduction

The National Student Financial Aid Scheme (NSFAS) was established (NSFAS Act 56 of 1999, as amended) to ‘provide financial aid to eligible students who meet the criteria for admission to a further education and training programme or to a higher education programme’. Amongst its identified functions are:

• To allocate funds for loans and bursaries to eligible students;

• To develop criteria and conditions for the granting of loans and bursaries to eligible students in consultation with the Minister (of Higher Education and Training); and

• To advise the Minister on matters relating to student financial aid.

A set of new principles for the allocation of Department of Higher Education and Training (DHET) funding, was introduced for new entrants to the post-school education and training (PSET) system in South Africa for the academic year 2018, and thereafter.

The new scheme changed the landscape of financial support for poor and working-class students wishing to access PSET in two main ways. Firstly, the former loan and bursary scheme was changed to a pure bursary scheme with no repayable loan component. Secondly, it redefined the financial definition of poor and working-class students. While previously this was defined as those from households with incomes of less than R122 000 per annum, poor and working-class students are now defined as those from households with incomes of less than R350 000 per annum.

Whilst the boundary moved upwards so that an additional 24% of households were incorporated into the poor and working-class category,¹ the new scheme does not address the implications of a hard distinction between poor and working-class and middle class. For example, those who move from a household income of R349 000 to R351 000, face a marginal tax rate that is substantial and penal.²

The issue of those that just fall on the wrong side of the income level threshold defining poor and working-class, has been a parallel theme of the #feesmustfall movement. Within the higher education funding context, those households that fall between the upper ‘poor and working-class’ boundary (now set at R350 000) and R600 000 have been designated as the ‘missing middle’ (DHET, 2017).

The concept of a ‘missing middle’, however, has a longer history, originally stemming from economic and political analysis of development. In comparison with countries of similar average incomes, population sizes, development challenges and resources, South Africa has a larger than average gap between those who are wealthy or just comfortably off, on the one hand, and the working poor, on the other hand. In comparison to other middle-income countries,³ South Africa’s rich are extremely rich, the poor very poor and there is a very large hole in the middle: the missing middle (Levy et al., 2014).

Estimates of household income nationwide produced by Xpert Decision Systems (XDS) analytics using South African Credit and Risk Reporting Association (SACRRA) data as at 18 January 2019⁴ indicate that 62.6% of households have pre-tax income less than R122 000 per annum, 89.7% of households have pre-tax income less than R350 000 per annum and 95.8% below R600 000.

Under the former description of poor and working-class (household incomes up to R122 000) the missing middle thus represented some 33.2% (95.8%–62.6%) of South African households. Under the new dispensation this has fallen to 6.1% (95.8%–89.7%). Whilst household size and participation rates differ across household income strata, it is assumed, to the first level of approximation, that the higher participation rates and smaller household units in the higher income strata cancel each other out so that, on current (2019) total undergraduate recruitment and TVET student registration numbers of approximately 1 400 000,⁵ this equates to around 85 000 current registered students that would fall in the household income bracket of R350 000 to R600 000. The 4.2% of students from households with incomes in excess of R600 000 would number some 60 000 students.

However, the household income distribution in South Africa is extremely skewed to the right. Bulmer (1979) asserts that any distribution with an absolute skewness value greater than 1 is ‘highly skewed’. The household income distribution in South Africa (calculated by the authors from the published household income data) has a Pearson’s moment coefficient of skewness of +3.42, indicating an extreme level of skewness. Consequently, standard Normal distribution values will not produce accurate predictions of the size of the missing middle. Earlier research output from the NSFAS Research and Policy Unit (2019a) is also used towards establishing the size of the missing middle.

The greater inclusivity of the 2018 financial eligibility rules for NSFAS bursary funding has led to the plight of missing middle students taking a more central role in campus politics and policy review. Despite that, there is a dearth of studies attempting to estimate and forecast the cost implications of different higher education funding policy decisions in South Africa (Wangenge-Ouma & Cloete, 2008; Gurgand et al., 2011; Jacobs et al., 2019). The contribution of this paper is to provide a clear and evidence-based estimation of the current size of the missing middle and the cost of including them within the NSFAS bursary system.

2. Methodology and results

The household income data from Xpert Decision Systems (XDS) is generated from the South African population at large. However, from analysis of 2018 NSFAS funded student data, it is clear that the racial balance amongst NSFAS recipients, university students, TVET college students and the wider South African population is very different. These are tabulated in Table 1.

Table 1. The racial mix of South African society, TVET students, university students and NSFAS recipients.

Grouping	African	Coloured	Indian/Asian	White
South Africa	81.0%	9.0%	2.0%	8.0%
TVET students	92.1%	6.5%	0.3%	1.1%
University students	72.6%	6.4%	5.2%	15.8%
TVET NSFAS recipients	94.9%	4.8%	0.1%	0.2%
University NSFAS recipients	95.2%	3.1%	0.7%	1.1%

Source: NSFAS Vital Statistics (2019a).

The average household income levels of the different racial groups in South Africa are also very different and are tabulated in Table 2.

Table 2. Average annual household income by population group of the household head as at 2014/15 and updated to 2019^b.

Race of household head	Average annual household income as at 2014/15^a	Average annual household income from column 2 updated to 2019 population averages
African	R92 983	R107 847
Coloured	R172 765	R200 382
Indian/Asian	R271 621	R315 040
White	R444 446	R515 492

^aSource: Statistics South Africa (2017), 2014/2015, Living Conditions of Households in South Africa, Table 3.3.

^bConsistent with 2019 population averages.

The racial proportions from the South African population from Table 1 applied to the average household incomes in Table 2 lead to a national average household income of R131 853. These data are from 2014/15. The income distribution of XDS that is used in this study is from 2019 with a national mean value of R152 930. Assuming the same inflationary growth in average household incomes results in assumed annual household income averages by race as recorded in column 3 of Table 2.

The different racial mixes for the total population and sub-samples recorded in Table 1 and average household incomes recorded in column 3 of Table 2 mean that estimating the size of the missing middle is rather more complex than simply applying the 6.1% of the population who have incomes in the R350 000 to R600 000 range to the numbers in PSET at any point in time.

The estimation process starts with the planned numbers of registrations in TVET colleges and undergraduate university programmes for 2019 (570 000 and 870 000, respectively) (NSFAS, 2019b). The racial percentages of students in PSET from Table 1 are then applied to generate a racial profile of 2019 PSET students, recorded in Table 3.

Table 3. Racial make-up of 2019 PSET students (forecast).

	African	Coloured	Indian/Asian	White	Total
TVET students	524 970	37 050	1710	6270	570 000
Undergraduate university Students	631 620	55 680	45 240	137 460	870 000
Total PSET students	1 156 590	92 730	46 950	143 730	1 440 000

In NSFAS (2019b) it is reported that the percentage of university students funded by NSFAS under the new funding regime is 55% of undergraduate numbers and 53% of total registrations for TVET students. Applying these percentages to the total TVET and undergraduate figures from Table 3 leads to forecast NSFAS supported students at both TVET and undergraduate level. Applying the relevant racial mix of NSFAS recipients from Table 1 to these NSFAS supported numbers and subtracting that value from the figures in Table 3 generates the number of students currently not in receipt of NSFAS funding, by race and educational institution type, as reported in Table 4.

Table 4. Estimated number of students currently in PSET and not in receipt of NSFAS funding by race and educational institution type.

	African	Coloured	Indian/Asian	White	Total
TVET students	238 277	22 549	1408	5666	267 900
Undergraduate university students	176 088	40 847	41 891	132 197	391 022
Total PSET students	414 365	63 396	43 298	137 862	658 922

Amongst the numbers in Table 4 are students who do not meet NSFAS funding eligibility rules, including rules relating to aspects other than household income. These fall into three main categories: students taking courses that are not approved/eligible for NSFAS funding; TVET students who are resitting courses; university students who have not academically progressed. Amongst TVET students, some 20% of registrations are resitting courses and 12% are taking courses that are not funded by NSFAS (NSFAS, 2019b). For university students, 7% of NSFAS applicants are rejected on academic grounds (a combination of inadequate academic progress, having taken more than the prescribed time to graduate or taking an unfunded course). It is assumed that a similar rate will occur amongst non-NSFAS funded students. These non-eligible students are removed from the non-NSFAS funded students from Table 4 to leave a potential pool of academically eligible non-NSFAS funded students reported in Table 5.

Table 5. Estimate of academically eligible non-NSFAS funded students by race and educational institution type.

	African	Coloured	Indian/Asian	White	Total
TVET students	70 287	10 693	861	3659	85 500
Undergraduate university students	143 693	36 972	38 937	122 900	342 503
Total PSET students	213 980	47 665	39 798	126 560	428 003

From the mean household income information, reported in column 3 of Table 2, we see that much of the skewness in the household income distribution flows from the considerable difference in average household income between the racial groups. It is assumed that within the different racial groups, the extent of skewness is much reduced. This allows the use of standard Normal tables to predict the number of students from households with an income between R350 000 and R600 000 and those from households with an income above R600 000 within each of the racial groups. The individual distributions cannot be identified from the income data we use as income frequencies by race are not available.

The full household income distribution is used to generate a standard deviation of the data set. This amounts to some R17 500. This figure is used in combination with the average incomes by racial group reported in Table 3 to forecast the percentage of each group that has household income between R350 000 and R600 000 and those above R600 000. These are reported in Table 6.

Table 6. Probability of households having an income below R350 000 and R600 000 and, thereby, the percentage that lie in the missing middle range, by race.

	African	Coloured	Indian/Asian	White
P (household income ≤ 350)	87.5%	76.2%	56.6%	22.6%
P (household income ≤ 600)	99.1%	97.2%	91.3%	65.7%
P (being financially eligible part of missing middle)	92.8%	88.2%	80.0%	55.7%

Thus, of the Africans in PSET who are not currently funded by NSFAS and are academically eligible (as reported in Table 5), 92.8% ((99.1–87.5)/(100–87.5%)) are predicted to form part of the missing middle whilst 55.7% ((65.7–22.6)/(100–22.6)) of academically eligible Whites in PSET, who are not currently funded by NSFAS, form part of the missing middle.

Applying these percentages to the numbers in Table 5 leaves us with our final estimate of the size of the current missing middle, which is recorded in Table 7, along with the percentage racial composition of the cohort.

Table 7. Estimate of the size and composition of the missing middle (mm).

	African	Coloured	Indian/Asian	White	Total
TVET	65 226	9431	689	2038	77 384
Undergraduate university	133 347	32 609	31 150	68 456	265 562
PSET	198 573	42 041	31 838	70 494	342 946
TVET %	84.3%	12.2%	0.9%	2.6%	100.0%
Undergraduate university %	50.2%	12.3%	11.7%	25.8%	100.0%
PSET %	57.9%	12.3%	9.3%	20.6%	100.0%

The data in Table 7 indicate that:

• The size of the missing middle on 2019 recruitment levels is around 340 000 students;

• Of these, 77 000 are at TVET colleges and 265 000 at universities;

• The racial mix of the missing middle is very different between TVET colleges and universities

• At TVET colleges, the overwhelming majority of these students will be African;

• At universities, whilst Africans remain the largest grouping, they constitute only 50% of the missing middle with 12% being Coloured, 12% Indian/Asian and 26% White;

• In comparison with the percentages for NSFAS funded students reported in Table 1, the missing middle students at university are much less African and much more Coloured, Indian/Asian and White. This is expected given average household incomes and is particularly marked amongst the group of undergraduate university students.

Multiplying the number of students by the average cost of a NSFAS funded student at TVET college and university (NSFAS, 2019b) produces a total cost of funding the missing middle of R19.2 bn. Of this, R1.7 bn would apply to the TVET sector and R17.5 bn to the university sector.

3. Discussion

3.1. Funding cliff edge

Funding the missing middle fully would still leave a ‘cliff edge’ type funding in that households with an income of R559 999 would receive full bursary support and those with a household income of R600 001 would receive nothing (please refer to footnote 1 which applies here too). The massive marginal tax rate for those households on the upper boundary of financial bursary eligibility remains.

A solution to this, in the same vein as the previous ‘expected family contribution’ operated by NSFAS, would be to introduce funding support for the missing middle on a sliding scale. To eliminate the marginal tax rate issue fully, the bursary offered could fall in value by 1% for each R2500 increase in household income, over the current R350 000 limit. For a TVET student this amounts in 2019 terms, to a fall in bursary of R219 per R2500 increase in income and for an average university student, R658. Consequently, the suggestion is that a student from a household with income of R350 000 would receive a 100% bursary, a student from a household with income of R475 000 would receive a 50% bursary and a student from a household income with income of R600 000 would receive a 0% bursary.

Due to the skewed nature of the household income distribution, with more households having incomes closer to the lower R350 000 limit than the R600 000 upper limit, the cost of bursaries of such a system would be more than the simple 50% of funding the whole of the missing middle with full bursaries. That is, more than R9.6 bn. To estimate the true cost, the distribution data and the bursary costs reported by NSFAS (NSFAS, 2019b) are applied to the numbers in Table 7. The cost of implementing the proposed sliding scale would be just over R11.4 bn, R1.1 bn for TVET students and R10.3 bn for university students.

The proposed sliding scale results in a 40% reduction in total cost and eliminates the considerable and regressive tax rates at the maximum of the bursary range. It also has a less marked impact on TVET funding as compared to university funding.

3.2. Caveats

Several assumptions have been made which have been documented above. One other assumption that is critical to the validity of this estimate, which has yet to be discussed, is the percentage of students currently funded by NSFAS. The values used are those current in 2018. As identified in NSFAS (2019b), the participation rate of African and Coloured students is much lower than that for White and Indian/Asian students. If the participation rates across races were to move towards equality then there would be an additional 100 000 eligible NSFAS students attending PSET institutions, largely at universities and costing an additional R6 bn in NSFAS bursaries. These increases would occur under current legislation with the missing middle still excluded from bursary funding. This potential cost needs to be considered when reviewing the affordability of the extension of NSFAS bursary support to the missing middle.

4. Summary and conclusions

Since the extension of bursary support to students from households with a family income below R350 000 there has been growing focus on the plight of the missing middle (household incomes from R350 000 to R600 000). Claims such as ‘the missing middle constitutes 40% of the youth’ (Sunday Times Live, 2019) are made but not substantiated.

Using data from NSFAS on students funded in 2018 and household income information, it is estimated in this paper that the current size of the missing middle attending public institutions is 343 000, comprising 266 000 at university and 77 000 at TVET colleges. To fund these fully would cost, at 2019 rates, R19.2 bn, R1.7 bn for TVET students and R17.5 bn for university students. This represents some 50% of the current NSFAS budget.

If a sliding scale of funding the missing middle were introduced, then the total cost could fall to R11.4 bn. Such a scale would eliminate the current considerable, regressive, marginal tax rate suffered by those at the very bottom end of the missing middle if they pursue post-school education or training.

A related factor in assessing PSET funding priorities and the availability of funding to support the missing middle, is the low rates of participation amongst the lowest income households. If these participation rates increase to the level of the higher income households, then there will be an additional 100 000 NSFAS fundable students in the system, largely at universities, which would cost an additional R6 bn. This will be an automatic consequence of improved basic education and participation rates amongst the currently marginalised.

These costs, R6 bn to fund improved participation by African and Coloured students, R19.2 bn to fully fund the current missing middle and R11.4 bn to fund a sliding scale support for the missing middle, need to be considered in the context of the medium term expenditure framework approved budget for NSFAS bursaries of R31 bn in 2019/20 and R37 bn in 2021/2022.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Calculated from data provided by Xpert Decision Systems (XDS) analytics using South African Credit and Risk Reporting Association (SACRRA) data.

2 Whilst there is a level of flexibility afforded by NSFAS on the R350 000 income limit, the resultant funded/unfunded decision still leaves the same substantial tax rate at whatever limit is set. The move to fully fund students following the December 2017 announcement, meant that the historical concept and calculation of an ‘expected family contribution’ fell away.

3 Levy et al.’s (2014) investigation compares the average monthly earnings for the largest income earner per household, by ventiles, for South Africa, Mexico and Turkey.

4 Personal communication.

5 The Department of Higher Education and Training (DHET) sets recruitment targets for universities and TVET Colleges annually. Refer to NSFAS Research Report 5: Predicting university and TVET demand for NSFAS Bursaries.