102
PSP Discussion Paper Series
19711
December 1996
Poverty and Inequality in the Distribution
of Public Education Spending in South Africa
Florencia Castro-Leal
December 1996
Povertv and Social Policy Deparment
Human Capital Development
The World Bank



P>SP Discussion Papers reflect work in progress. Thev are intended to make lessons emerging from the
current work program available to operational staff quickly and easily, as well as to stimulate discussion
and comment. Thev also serve as the building blocks for subsequent policy and best practice papers. The
i views expressed here are those of the authors and should not be attributed to the World Bank or its Board
of Executive Directors or the countries thev represent.



Southern Africa Department                            Poverty and Social Policy Department
WorldBank                                                WorldBank
Poverty and Inequality in the Distribution of Public Education Spending
in South Africa
Florencia Castro-Leal *
December, 1996
* The author gratefully acknowledges the support receivedfrom Ann Duncan in producing this paper.
The author also wishes to thank Stephan Klasen, Jan Leno, Gurushri Swamy and Carolyn Winterfor
very helpful comments on an earlier draft Yisgedu Amde and Kalpana Mehra provided excellent
computational analysis of household survey data. Many thanks to Precy Lizarondofor providing quick
and efficient wordprocessing assistance



I
I



Table of Contents
ABSTRACT ...........v
ACKNOWLEDGMENTS                       ...vi
EXECUTiVE SUMARY                       ...vii
.       INTRODUCTION .......................         ,.I
IL      SOCIOECONOMIC AND DEMOGRAPHIC GROUPS   .............................2
Hi.    SCHOOL ENROLLMENT ...................................... .   .       4
Over-aged children attending primary school are common among
the poor             ,    ,                            .....................4
Disparities in enrollments are wide for secondary but widest for
tertiary education ...4
Girls are as likely as boys to attend primary school and more likely than
boys to attend secondary school               ..5
The most important reason for being out-of-school is that school expenses
are too high ,,,......,.....8
Primary school-age Africans find school expenses too high and
no school locally ..... ,...                                    _., 9
More than one in every ten secondary school-age Africans and Coloureds are
out-of-school                     , ... 10
One in every three secondary school-age girls who is out-of-school is because
she became pregnant.                                      .   I11
IV.    PUBLIC SPENDING ON EDUCATION ............................... ............... . 12
Increases in public education spending outpaced population growth in the
early 1990s ........2.... .,,..                       . ... l2
Public primary education spending increased but public technical education
spending decreased ...,,,,,,,,..,    ......                 12
Disparities are substantial in public education spending per student,.,,,,,...., 13
V.      THE DISTRIBUTION OF PUBLIC EDUCATION SPENDING                 .    15
Public education spending in South Africa is not pro-poor  ..,,,,,.,,,,, 16
Race-specific subsidies reveal severe inequities in the South African
education system .                                                19
Public education spending per capita is lowest for Africans .      20
iii



Targeting the Eastern Cape and the Northern Transvaal can improve the
poverty focus of public education spending ................................................... 21
The poor are unable to afford the direct costs of education .............................. 22
Reduced education expenses can increase enrollments among the poor: ..............
Uniforms, transport and meals for primary and secondary and
exemption of fees for tertiary ................ .................................. 24
VL      CROSS-COUNTRY COMPARISONS OF EDUCATION INCEI)ENCE
ANALYSIS .25
South Africa has one of the worst distributions of public education spending... 27
VII. POLICY IMPLICATIONS .28
Reduce inequities in education by geographically targeting public spending
to the poor .28
Reduce education expenses on uniforms, transport and meals for poor
primary and secondary students .28
Higher education scholarship program and credit market development . 29
REFERENCES .........                                                       30
APPENDIX A ......... 31
iY



ABSTRACT
South Africa's public education resources have increased significantly. Education
spending has increased to 7.3 percent of the Gross Domestic Product (GDP) in 1993/94
compared with 5.8 percent in 1987/88. The annual growth rate of education expenditures
in real terms shows an absolute expansion in per capita public funding of education. This
paper explores the distribution of all public education resources across socioeconomic and
demographic groups in South Africa. The methodology is known as Benefit Incidence
Analysis and it measures how well public services are targeted to certain groups of the
population.
The analysis suggests that the South African education system is hampered by
disparities across income groups, regions and races. The shares of public education
resources benefiting the ultra-poor and the poor are substantially lower than their shares of
school-age population. Inequality in the distribution of public education resources widens
by educational level from primary to tertiary for all population groups. The analysis of
household direct costs suggests that reduced costs of uniforms; transport; meals for poor
primary and secondary school-age children, and scholarships in tertiary education
exempting the poor from the payment of school fees can increase enrollments among the
poor. These programs can be accompanied by cost recovery mechanisms that do not
discriminate against the poor to free public education funds to finance the additional
expense. Regionally, important gains in allocating public resources to the poor and the
ultra-poor can be achieved by targeting the Eastern Cape and the Northern Transvaal.
V'






ACKNOWLEDGMENTS
This study is a background paper for the Report on Poverty and Inequality in
South Africa, currently in progress, being prepared at the request of the Tender
Conmmittee by a consortium consisting of Data Research Afiica (DRA), Centre for African
Research and Transformation (CART), Centre for Reconstruction and Development
(CRD) and Community Agency for Social Inquiry (CASE), working in cooperation with
the World Bank and the United Nations Development Programme.
The author gratefully acknowledges the support received from Ann Duncan in
producing this paper. The author also wishes to thank Stephan Klasen, Jan Leno,
Gurushri Swamy and Carolyn Winter for very helpful comments on an earlier draft.
Yisgedu Amde and Kalpana Mehra provided excellent computational analysis of
household survey data. Many thanks to Precy Lizarondo for providing quick and efficient
word processing assistance.
The views expressed in this study are those of the author and should not be
attributed in any manner to the World Bank, to members of its Board of Executive
Directors, or to the countries they represent.
Vi



I
I
I



EXECUTIVE SUMMARY
South Africa's public education resources have increased substantially in recent
years. The expansion in public education spending outpaced population growth in the
early 1990s and the allocation of state expenditures to public education grew by more than
the rate of growth in the Gross Domestic Product (GDP). However, the education system
in South Africa is hampered by disparities across income groups, regions and races.
The shares of public education resources benefiting the ultra-poor and the poor are
substantially lower than their shares of school-age population (see Table A). In 1993,
poor households received 40 percent of al public education spending for 60 percent of
the South African school-age population, while the share going to richest households was
23 percent for only 8 percent of the school-age population. Ultra-poor households
received 21 percent of public education resources for 34 percent of the school-age
population.
Table A
Incidence ofpublic education spending by household group, 1993
_ercentage share
Educational level         Education spendin bene/ijng.
(School-age population)    Ultra-poor  Poor        Richest
(Poorest 20   (Poorest 409)   (Richest 20%)
l _____________________  percent)
Primary                   27            48             19
(Children 6 to 12)       (36)          (62)           (8)
Secondary                 18            36            25
(Children 13 to 17)      (35)          (61)           (9)
Tertiary                  11            24            32
(Population 18 to 22)    (29)          (54)           (9)
All education             21            40             23
fPopulation 6 to 22)     (34)          (60)            (8)
Source: DBSA (1993) and SALDRU (1993)
Inequality in the distribution of education resources widens by educational level
from primary to tertiary across different income groups and also across different races.
Public education spending per capita is lowest for Africans at every educational level (see
Figure A). However, not all African students belong to the poor and the ultra-poor.
vii



Thus, important gains in allocating public education resources to the poor and the ultra-
poor can be achieved by targeting the Eastem Cape and the Northern Transvaal. These
two provinces have by far the highest poverty rates and concentrate more than 50 percent
of the former Homeland population.
Figure A
Annual education -spending per capita, by level of schooling and race, 1993
1": 
990
.U se DBA(99)90dSLRU(93
Scoo ex90  se ar*hSotipratresndsorgntecndanet sho
80
Sorc:                                               Sec(99)anoSLDU 193
in South Africa. Most South African families have to spend a significant amount of
resources in school fees, uniforms, transport, meals and books and stationary in order to
send their children to school. Poor families pay more than 40 percent of per capita non-
food household expenditures per each child sent to primary school compared to only 6
percent for non-poor families. Education costs increase to half of per capita non-food
household expenditures for poor famnilies sending children to secondary school, in contrast
to 10 percent for the non-poor. Tertiary education is prohibitive for the poor. Each poor
student enrolled at the tertiary level needs more than two-and-a-half times the per capita
amount spent on non-food items compared to less than one-third for non-poor students.
This analysis suggests that reduced costs of uniforms, transport and meals for poor
primary and secondary school-age children and scholarships in tertiary education
exempting the poor from the payment of school fees can increase enrollments among the
poor. These programs can be accompanied by cost recovery mechanisms that do not
discriminate against the poor to free public education funds to finance the additional
expense.
viii



I. INTRODUCTION
South Africa's public education resources have increased significantly. Education
spending has increased to 7.3 percent of the Gross Domestic Product (GDP) in 1993/94
compared with 5.8 percent in 1987/88.1  The annual growth rate of education
expenditures in real terms is more than 5 percent in contrast to a 2 percent growth rate for
the population in the 6-18 age group.2 In other words, there has been an absolute
expansion in real per capita public funding of education. This paper explores the
distribution of all education resources across socioeconomic and demographic groups.
There are two main factors that influence prevailing inequities in the amount of
public education resources received by the population. First, there is the allocation of
public resources to individuals in different socioeconormic groups, regions, races and by
gender as a result of their consumption of public services. Second, there is the allocation
of government spending within the education sector. The distribution of public education
spending can be examined across different socioeconomic groups, regions, races and by
gender through the allocation of per unit public subsidies according to individual
utilization rates of public services. The methodology is known as Benefit Incidence
Analysis3 and it measures how well public services are targeted to certain groups of the
population, for example the poor, regions of interest and girls.
Detailed public spending information is available from a comprehensive public
expenditure review on education published by the Development Information Group of the
Development Bank of South Africa and the World Bank (DBSA. 1Q93)  Individual
patterns of access and utilization of public services are obtained frolm thb So-uth Xfrica
Living Standards Survey, coordinated by the University of Cape Town s S.uthern kfnca
Labour and Development Research Unit (SALDRU, 1993 ) and funded hk the
Governments of Denmark, the Netherlands and Norway, with technical assistance
provided by the World Bank. DBSA and SALDRU data are used to calculate per caplwa
education subsidies for different socioeconomic groups, races, regions and by gender
Section I explains the disaggregation by demographic groups used throughout this
paper.  Section II looks at education indicators. Public education spending and the
allocation of public resources within the education sector are explored in Section III.
Section IV analyzes the distribution patterns of public education spending across
socioeconomic groups and gender. The last Section discusses the most relevant policy
implications coming from this study.
Buckland, Peter and John Fielden (1994). Public Expenditure on Education in South Africa, 1987/88 to 1991/92,
6
Ibid., p. 8
3 See Van de Walle and Nead (1994), Meerman (1979), and Selowsky (1979).



I



IL SOCIOECONOMIC AND DEMOGRAPHIC GROUPS
The geographical concentration of the poor in some areas of South Africa is
crucial in targeting public education resources. The poverty assessment determined that
"... nearly 70 percent of the poor live in the former Homelands .. ."4 This disparity in
income levels between former Homelands and non-former Homelands justified the study
of the population living in the former Homelands as a separate demographic group.
In the former Homelands, Africans constitute almost a 100 percent of the
population with few inhabitants from other racial groups (see Appendix Table A. 1).
Afiicans in former Homelands (H-Africans) are two-thirds of the African population (see
Appendix Table A.2), and account for 63 percent of the households in the poorest quintile
and 55 percent of those in the second poorest quintile (see Figure 1).s
Figure I
Racial Composition within Quintiles
>    . _   _   _   n   i    CD ~~~~Africnl
o0                             Homeland
a60
U Indian
t ~~~~~~~~~EColoured
g20                             Afr _ l .  N w_ -
0~~2:t~~iII!.   ~~Homeland
Source: SALDRU (1993)
Regional income disparities in South Africa clearly suggest that the analysis of the
distribution of education resources needs to pay special attention to three new provinces:
the Eastern Cape, the Northern Transvaal and Kwazulu, where more than three-quarters
of H-Africans live (see Table 1).
4 RDP (1995), p. 12
5 The poverty assessment determined that about 40 percent of South African households are poor.
Households are ranked by per adult equivalent expenditures, the poorest 20 percent constitutes the bottom
quintile, and so on for the next quintile, up to the top quintile which contains the richest 20 percent of all
households. Since households in bottom quintiles are larger than households in higher quintiles, then the
total population found in poverty is larger than the number of poor households.
2



Table 1
Percentage share offormer homeland population
by new province
New province            Share
N.Cape                         0.00
O.F.S.                         2.34
North West                    11.07
E.Transvaal                   10.13
W.Cape                         0.00
N.Transvaal                   22.88
E.Cape                        23.42
Gauteng                        0.00
Kwazulu                       30.15
Total                        100.00
Source: SALDRU (1993)
The analysis by demographic groups is feasible because data on both public
education spending and enrollments is available. DBSA data allows us to calculate per
student subsidies for all races separately and for H-Africans and Africans outside former
Homelands (NH-Africans). The SALDRU household survey contains information on both
the household's residence by new province and its racial composition.
3



IIL SCHOOL ENROLLMENT
Over-aged children attending primary school are common among the poor
Net and gross enrollment rates for primary school are similar across income
quintiles and races (see Tables 2 and 3).6 Overall, the net enrollment rate is about 90
percent and the gross rate is 106 percent. Thus, nine out of every ten children between
the ages of six and twelve years are currently enrolled in primary school while the spread
between the net and gross enrollment rates indicates that over-aged children are attending
primary school.7
Over-aged children attending primary school are common among the poor. The
spread between the net and gross enrollment rates is four times larger for the poorest than
for the richest income quintile. The poorest quintile has a primary net enrollment rate of
85 percent compared with a 112 percent gross enrollment rates while the rates for the
richest quintile are 90 percent and 97 percent, respectively. The main disadvantage of
attending primary school at an older age is that these children are often needed to help to
support the household. This becomes a disincentive for them to complete that level of
education and to proceed into the next level.
Disparities in enrollments are widefor secondary but widestfor tertiary education
The overall net enrollment rate in secondary education is 60 percent but only 11
percent in tertiary education while the gross enrollment rates are 97 percent and 14
percent, respectively.8 Disparities in enrollment rates across income quintiles and races
are wide at the secondary level but widest at tertiary level (see Tables 2 and 3) The
poorest quintile has about half the net enrollment rate of the nchest quintile at the
secondary level, and only one-tenth at the tertiary level.
6 The net enrollment rate is defined as the total number of school age children currently enrolled in each
educational level as a percentage of the total school age population at each level. The gross enrollment
rate is defined as the total number of enrollments in each educational level, of anv age, as a percentage of
the total school age population at each level.
'7 The 87 percent net enrollment rate for the 6 to 12 year old age group is lower than the 92 percent
current attendance rate reported in Table 4. This is because there are sample observations of children who
are 12 years of age and report having completed primary school.
8 Enrollment rates at the secondary level are calculated for 13 to 17 year olds and for 18 to 22 year olds at
the tertiary level.
4



The spread between the net and gross enrollment rate in secondary education for
the poorest quintile is twice as large as for the richest quintile, indicating that over-aged
children are more common among the poor.9 Late completion of primary school among
the poor accentuates the proportion of poor over-aged children in higher levels of
education and decreases their prospects of attending tertiary education. The lack of higher
education among the poor is a strong factor in perpetuating lower income levels.
H-Africans and NH-Africans have the lowest net enrollment rates in secondary
education. However, disparities across races are widest in tertiary education. Afiicans
and Coloureds have enrollment rates at the tertiary level which are less than 10 percent
compared with Indians and Whites whose rates are four times as high.
Girls are as likely as boys to attend primary school and more likely than boys to attend
secondary school
Gender differences are not substantial in primary school (see Appendix Tables A.3
and A.4). However, males at the primary level have slightly higher gross enrollment rates
than females, reflecting a higher proportion of over-aged males than over-aged females in
primary school. Disparities between girls and boys are wider at the secondary level and
girls have higher enrollments than boys. Girls have a 63 percent overall net enrollment
rate in secondary school compared with 56 percent for boys. This pattem favoring girls
occurs for every income quintile. Tertiary education is the only educational level in which
girls have lower enrollment rates than boys, but this occurs only among the rich.
9 The largest spread between the net and gross enrollment rates occurs at the third quintile. This may be
related to a lower drop-out rate and a higher probability of staying in school even if repetition occurs.
5



Table 2
Net enrollment rates, by quintile, race and level of schooling, 1993
Races
Hh quintile    Africanl      Africanl   Coloured   Indian       W7iite       All
Non-Homeland   Homeland
Primary education
Poorest         84            85          88        **         **          85
II            86            87         94         **         **          87
m             88            88         87         **         **          88
IV            89            88         91         93         90          89
Richest         89                                  91         90          90
Total          86            86          90        92         90          87
Secondary education
Poorest         37            49          49                               46
II            60            56         59         **         **          57
III           63            67          74        **         **          67
IV            73            73          79        89         80          78
Richest         77            **                    93         84          83
Total          54             56         70        89         82          60
Tertiary education
Poorest          5             4          4         **         **           4
II            5             5           2         **         **          5
III            7             10         3         **         **           8
IV            24            23          19                   **          20
Richest                        **         **        41         38          38
Total           10            7          9         36         32          11
** Sample <= 30 observations
Source: SALDRU (1993)
6



Table 3
Gross enrollment rates, by quindle, race and level of schooling, 1993
Races
Hh quintile  Afiican/Non-  Afiicanl  Coloured   Indian       White        All
Homeland Homeland
Primary education
Poorest        117         110        101        **         **          112
II          107          107       113         **         **         108
mI           103         105        97         **         **          103
IV           97          96         98         98         98          97
Richest        108                               91         98          97
Total         108         108        101        96         98          106
Secondary education
Poorest        69           85        63                    **          81
nI          104          t00        68         **         **          98
m            11         112        96         **         **          110
IV           109         107        103        99         101         108
Richest        105                     **        103        98          101
Total          96          98        88         99         98          97
_______  Tertiary education
Poorest         5           5          4         **         **           5
II           6            8         4          **         **          7
III           9           13         5         **         **          11
IV           27          30         21         **                     24
Richest        **           **        **         52         48          48
Total          I I         10         12        44         40           1 4
** Sample <= 30 observations
Source: SALDRU (1993)
7



The mnost important reason for being out-of-school is that school expenses are too high
School expenses are too high for more than one in every five primary and
secondary school-age children who are out-of-school (see Table 4).1O Thus, the full cost
of education to households is a strong disincentive to attend school in South Africa."I In
addition, a substantial number of secondary school-age children who are out-of-school
responded that they could not cope with school work, 20 percent, or, that became
pregnant'2, 16 percent.
Table 4
Reasons for being out-of-school by age group
(as % of non-enrolled in age group)
Age <rups
Reason for being out-of-school   6 to 12    13 to 17    18 to 22      All
year olds    year oIds   Year olds
School expenses are too high         21          23          15          1 7
Needed to get work                   -           10          19          15
Illness/disabled                     6           9           4           5
Became pregnant                      -           16          15          -
No school locally                    4           4           1           2
Could not cope with school work      8           20          12          I.,
School boycott                        -          2           l
Social unrest/violence                           3           I           l
Completed education                              3 3,
Reqtured for work at home/farm        1           4
Other                                59           7
Total non-enrolled                   100         100         1,         I
Non-enrolled as percentage of
population group                      8          10
Source: SALDRU Survey (1993)
Among 18 to 22 years olds who are out-of-school high expenses are an important
reason, but the most important reason other than completed school is that they needed to
10 For a large proportion of primary school-age children, almost 60 percent, the reason for being out-of-
school needs to be further explored. These children are all grouped into the "other" category as their
response for being out-of-school.
1 1 Household education expenditures are disaggregated in Section V.
12 Explored by gender in Table 7.
8



work, 20 percent. In addition, the non-enrolled population as a share of the age group is
considerably higher for this eldest age group than for other age groups, more than 50
percent are not enrolled in any type of school. These response pattems highlight the need
to promote early enrollment and decrease repetition rates, as the likelihood of being out-
of-school increases drastically for the population older than 18 years of age.
Primary school-age Africans find school expenses too high and no school locally
Among primary school-age Africans who are out-of school, the two main reasons
for non-enrollment are that school expenses are too high and that there is no school locally
(see Table 5). These two reasons account for more than 60 percent of H-African and 30
percent of NH-African school-age children who are not currently enrolled More than one
in every five primary school-age Africans find school expenses too high Almost four in
every ten primary school-age H-Africans are out of school because there is no school
locally.
Table 5
Reasons for being out-of-schoolfor 6 to 12 year olds by race
(as % of non-enrolled)
{ L__________________________ i______ _ IRaces_
Reason for being         African/    Africanl  Coloured  Indian 1Rhite      All
out-of-school       Non- Homeland Homeland
School expenses too high          17           26        4                         21
Needed to get work                              -        -        .
Illness/disabled                   6           8                          a         6
Became pregnant                                 I                 -                 -
No school locally                 10           38        6                 1        4
Could not cope with school work    6           9              -           -         8
School boycott              |                   1                                   0
Social unrest/violence                           1                                 0
Completed education                                      -                          -
Required for work at home/farm     1            1                                   1-  -
Other                             58           53        94               96        59
Total non-enrolled                100          100      100       -       100      100
Non-enrolled as percentage of
population group                   8            9        6      0          5        8
Source: SALDRU (1993)
9



More than one in every ten secondary school-age Africans and Coloureds are out-of-
school
A much larger proportion of secondary school-age Africans and Coloureds are
out-of-school compared with other races. More than one in every ten secondary school-
age Afiicans and Coloureds are out-of-school, in contrast with only four percent of
Indians and two percent of Whites (see Table 6). The reasons for being out-of-school are
more differentiated across all races for this age group compared with the primary school-
age group. Africans find school expenses too high. In addition, one in every five H-
Africans report that they could not cope with school work, reason that could be related to
the fact that one in every ten report illness or disability. Almost one in every five NH-
Africans needed to work. Almost one in every three Coloureds report that they either
needed to work or could not cope with school work.
Table 6
Reasonsfor being out-of-schoolfor 13 to 17year olds by race
(as % of non-enrolled)
Races
Reason for being out-of-school    Africanl    African!  Coloured  Indian   White    All
Non-Homeland Homeland
School expenses too high            20          28         3        -        -       23
Needed to get work                  18           4        27       33        -        10
Illness/disabled                    4           11         9        -        -        9
Became pregnant                     16           18        9                 -        16
No school locally                   8            2         3        -       24        4
Could not cope with school work     16          21        27        17       -        20
School boycott                      3            1         -        -                 2
Social unrestlviolence               1           3         3        17       -        3
Completed education                  2           2         -        33      76        3
Required for work at home/farm      2            5         9        -        -        4
Other                               11           5         9                 -        7
Total non-enolled                   100         100       100      100      100      100
Non-enrolled as percentage of
population group                    13           11        10       4        2        10
Source: SALDRU (1993)
10



One in every three secondary school-age girls who is out-of-school is because she
became pregnant
One in every three 13 to 17 year old girls who are not enrolled in school is because
she became pregnant (see Table 7). A joint analysis by gender and race shows that
Africans have higher teenage pregnancy rates than other racial groups. Among out-of-
school males in the 13 to 17 year old group, one in every three could not cope with school
work. These findings are similar to other countries, mainly that a smaller proportion of
females than males reports being out-of-school because they cannot cope with school
work. Gender dfferentials in the reasons for being out-of-school among primary school-
age children are not substantial, the most significant reason is that school expenses are too
high.
Table 7
Reasonsfor being out-of-schoolfor 6 to 12 and 13 to 17year olds by gender
(as % of non-enrolled)
6 to 12 year aIds          13 to 17 vear olds
Reason for being out-of-school   Female   Male   All    Female    Male       All
School expenses too high          20       21      21       23        24       23
Needed to get work                 -        -       -        8        12        10
Illness/disabled                   5       8        6        9         8        9
Became pregnant                    I        -       0        30        -        16
No school locally                  4       5        4        4         4        4
Could not cope with school work    7       8        8        11       31       20
School boycott                     -        I       0        1         2        2
Social unrest/violence                      I       0        2         4        3
Completed education                                 -        2         5        3
Required for work at hometfarm     I        I       1        4         4        4
Other                             63       57       59       8         6        7
Total non-enrolled                100      100     100      100       100      100
Non-enrolled as percentage of
pulation group                     7       9        8        11       10        10
Source: SALDRU (1993)
11



IV. PUBLIC SPENDING ON EDUCATION
Increases in public education spending outpaced population growth in the early 1990s
Public spending on education increased steadily in the early 1990s (see Table 8).
The increase in real terms for public education spending is more than twice as large as the
population growth rate, 5.2 percent compared with 2.2 percent, respectively.'3  In
addition, the allocation of state expenditures to public education increased by more than
public education spending as a share of the Gross Domestic Product (GDP).
Table 8
Public Financing of Education
Public Spending   Public Spending
Fiscal Year   on Education as % of  on Education as %
____________  Total State Expenditures  GDP
1987188          22.9               5.8
1988/89          21.9               5.5
1989/90          22.0               5.7
1990/91          22.3               6.3
1991/92          23.9               6.5
1992/93          23.8               7.3
1 1993/94            24.4               7.3
Source: Buckland and Fielden (1994)
Public primary education spending increased but public technical education spending
decreased
Primary and secondary education get more than three-quarters of public education
resources (see Figure 2). Technical colleges, teacher's education and technikons combined
obtain less than seven percent. Universities alone get a considerable proportion of public
education spending compared with other tertiary-level programs, their share is around ten
percent.
The share of public resources dedicated to primary education increased while the
shares of technical education and universities decreased.'4 The share of public spending
allocated to primary education increased from 38 percent in 1987/88 to about 43 percent
in 1991/92. Secondary education spending remained constant at about 30 percent during
13 Buckland, Peter and John Fielden (1994). Op. cit., p. 8
14 Ibid.
12



this period. Technical education spending experienced a larger decline compared with
University spending. Technical education spending decreased from 2 to 1.5 percent, while
University spending decreased from 13 to 10 percent between 1987/88 and 1991/92.
Figure 2
Share ofpublic education spending by educational levels, 1991/92
M)k       0.4N                     Pror 43
AduklVoc i 
rhrn*y 10X
7Techjk., Z4%
Sectnda,y 31%
Source: Buckland and Fielden (1994)
Disparities are substantial in public education spending per student
Disparities in public spending per student by department are substantial. in
particular at the primary and secondary levels (see Table 9)  The.c dipa-t1e1C are an
indication of quality differentials in education by region since the p'pu.;Anf. Ir the former
Homelands receives the lowest education spending per student and thrcc r e- proAinces.
the Eastem Cape, the Northern Transvaal and Kwazulu, concentrate more thar three-
quarters of the former Homeland population.
Per student spending for the House of Assembly (histoncall! catenng W'hite
pupils) is R 3,099 in primary and R 4,675 in secondary. For the same educational levels,
per student spending for the House of Delegates (historically catering Indians) is R 2,565
and R 3,353 and for the House of Representatives (historically catering Coloureds) is R
2,308 and R 2,735, respectively. Public resources per student allocated to the African
Departments of Education are considerably lower than those of the Houses. Per student
spending for the Department of Education and Training (historically catering NH-
Africans) is R 1,012 at the primary level and R 1,014 at the secondary level. The SGT and
TBVC Departments of Education (historically catering H-Africans) have the lowest per
student spending for primary and secondary, R 660 and R 790, respectively.
13



Table 9
Annualpublic recurrent spending on education, 1993
by level, department, race and per student
Department and race         Primary     Secondary      Tertiary *
RECURRENT EXPENDITURES (DBSA - R'000)
DET                                 1,679,275      972,051       801,762
SGT & TBVC                         2,954,317      1,972,808      369,238
House of Assembly                   1,773,402     1,819,737    1.305,795
House of Representatives            1,304,329      796,467       244,761
House of Delegates                   370,593       402.475       176.183
Total                              8,081,916      5,963,538    2.897,739
ENROLLMENTS BY RACE (DBSA)
Africans/Non-Homeland (DET)         1,696,522      646,542       144.716
Africans/Homeland (SGT & TBVC)    4,175,150       1,559.577       67.107
Whites (HoA)                         537,712       369,889       234.114
Coloureds (HoR)                      630,200       230,442        35.411
Indians (HoD)                        149,285        95,894        30,847
Total                              7,188,869      2,902.344      512.195
ENROLLMEN7S BY RACE (SALDRU)
Africans/Non-Homeland               1,659,454      958.539       117.638
Africans/Homeland                  4,473,855      2,497.889      195.166
Whites                               572,171       389.236       143.882
Coloureds                            565,019       291.243        36.186
Indians                              144,476        120.037       21.244
Total                               7,414,975     4.256.944      514.116
SPENDING PER STUDEArT (DBSA_
Africans/Non-Homeland (DET)              990          1.503        5,540
Africans/Homeland (SGT & TBVC)           708          1.265        5.502
Whites                                 3,298         4,920         5.578
Coloureds                              2,070          3,456        6.912
Indians                                2,482         4.197         5,712
Average                                1,124          2.055        5.657
SPENDING PER STUDEAT (SALDR U)
Africans/Non-Homeland                   1,012         1,014        6.816
Africans/Homeland                        660           790         1.892
Whites                                 3,099          4,675        9,075
Coloureds                              2,308          2.735        6,764
Indians                                2,565          3,353        8.293
Average                                 1,090         1,401        5,636
* Public spending and enrollments at the tertiary level are calculated by race
Source: DBSA (1993), SALDRU (1993)
14






V. THIE DISTREBUTION OF PUBLIC EDUCATION SPENDING
This study uses the Benefit Incidence Analysis methodology to analyze the
distribution of public spending across different socioeconomic and demographic groups.
The most important advantage of this technique is to measure how well public services are
targeted to certain groups in the population, for example the poor, particular regions of
interest and girls and women.
Figure 3
Annual spending per student by educational level and race, 1993
6000
4000                                       Secondary
Teray
2000
Source: DBSA (1993) and SALDRU (1993)
The actual allocation of public spending to the different socioeconomic and
demographic groups is determined by a combination of supply and demand factors. First,
government spending allocations to and within the sector on each type of public service.
Second, household behavior regarding utilization rates of public services by
socioeconomic groups.  Thus, the incidence analysis integrates two sources of
information: unitary subsidies by type of service (i.e., the per student annual subsidy for
the education sector); and information on individual utilization rates of public services
15



(i.e., total enrollments by educational level) disaggregated by socioeconomic group, region
and gender. In this way, public spending on education is distributed across socioeconomic
groups subject to their enrollment rates and the annual public subsidy per student by
educational level. The incidence of public spending is the result of: (i) a public policy
decision, the allocation of public expenditures to and within each sector, and (ii) a private
decision, the behavior of households.
Targeting of public education resources across socioeconomic and demographic
groups in South Affica is analyzed in this section using information on the pattern of
government recurrent spending per student obtained from DBSA (1993) and the pattern
of enrollment rates obtained from SALDRU (1993). The combination of the variability in
public education spending and the variability in enrollment rates determines existing
disparities. The incidence analysis uses per student spending calculated on the basis of
SALDRU (1993) enrollments for primary and secondary education, and per student
spending calculated on the basis of DBSA (1993) enrollments for tertiary education (see
Figure 3).15
Public education spending in South Africa is not pro-poor
Public education spending in South Africa is not pro-poor because the share of
public resources going to the poor and the ultra-poor is substantially smaller than their
share of the population.16 If public education resources were equally distributed across
the population, then every household quintile would receive a percentage share equal to its
share of the population.
The poor in South Africa receive about 40 percent of education resources but they
are 53 percent of the population (see Figure 4). The ultra-poor in South Afnca receive
around 20 percent of education resources but they are 29 percent of the population In
15 Per student spending in Table 9 by educational levels and race uses public recurrent educatuon
expenditures from DBSA (1993) and school enrollments from two different sources DBSA (1993) and
SALDRU (1993). Recurrent expenditures and school enrollments reported by DBSA (1993) are for the
1991/92 school year. School enrollments obtained from SALDRU (1993) are an update for the 1993/94
school year. Household surveys like SALDRU are considered a reliable method for estimating primarN
and secondary enrollments since students are usually living with their parents  However, household
surveys could underestimate tertiary-level enrollments because it is common for students at this level to be
living in single-headed households or boarding facilities.
16 Quintiles are created by ranking households, on the basis of per adult equivalent expenditures, from
poorest to richest and then aggregating them into five groups with 20 percent of households in each
quintile. The poverty profile defined as 'poor" the poorest 40 percent of households and as 'ltra-poor"
the poorest 20 percent of households (RDP, 1995). The poor are 53 percent of the population living in the
poorest 40 percent of households and the ultra-poor are 29 percent of the population living in the poorest
20 percent of households. A larger concentration of the population among poorer household quintiles
occurs because poorer households tend to have larger families than richer households (see Appendix
Figure A. 1).
16



contrast, the richest household quintile receives almost twice its share of the population,
23.4 percent of public education resources for 12.5 percent of the population.
This pattern of distribution of public education resources across household
quintiles is conditional on both supply and demand effects: (a) public education spending
per student is low for poor and ultra-poor students;17 and, (b) the poor and ultra-poor
have very low enrollments at the secondary and tertiary levels."'
Figure 4
Al public education spending by household quintile and race
30.0                                            10-                        -
II4              l      230 4sogeLmnd
20.1
200                            * sp-d2ng                                           whae
P.J &dU I: ndww
10.0                      ~4                                                         ooe
Source: DBSA (1993) and SALDRU (1993)
The poor and the ultra-poor receive the highest allocanon o!  u., cjuataon
resources from primary schools.  Although a smaller share of publ,  rirn    education
spending than their share of the population goes to the poor. thn  rr,\cr  almost 50
percent of all primary education resources (see Figure 5). The spendini share  kithin
quintiles by race indicates that African students are the large majonts of pnmary
enrollments among the poor and the ultra-poor but receive less than their share of the
population.
17 This is because the distribution of the population by race within household quintiles shows that a large
majority of the poor and ultra-poor are Africans, mainly inhabitants of the former Homelands (see
Appendix Figure A.2). H-Africans get the sma2lest per student public education spending at every level of
education and NH-Africans get the second smnallest per student public education spending at the primary
and secondary educational levels (see Table 9).
1s See Tables 6 and 7.
17



Figure 5
Primarylevdl public education spending by household quintile and race
30.0 -                                                                      A1._Xn/
20.0                                                                      UNlikeg
UIndian
Ppdadon    40                              celowad
.20                             Homdland
0.0                                          0
Source: DBSA (1993) and SALDRU (1993)
Less than 40 percent of public secondary spending is allocated to the poor while
they are 53 percent of the population (see Figure 6). The ultra-poor receive less than two-
thirds the public secondary spending as compared with their share of the population. By
contrast, the richest quintile receives 25 percent of public spending at the secondary level
for 12.5 percent of the population. The spending share within quintiles by race indicates
that African students are the large majority of secondary enrollments among the poor and
the ultra-poor but these groups receive less than their share of the population.
Figure 6
Secondary-level public education spending by household quintile and race
30.0 -n0
ol20.0- rf 16.(°| | X - C 6e- . | 9                                 wl** Wite
20.0  ~ ~      ~     USpeeding        60~It
cl  Population                                *InIdian
40
10.0                                       a                               ]olne
20                           UAfrican/Non-
0.0 ~~~~~~~~~~~~~~~~~~~~~Homeland
o~ ~- o. -                                                  -. l  lLX2  lL*          mwn
Source: DBSA (1993) and SALDRU (1993)
18



Although the poor are more than half the population in South Africa, they receive
less than 25 percent of all public tertiary education spending (see Figure 7). The ultra-
poor receive about one-third the public tertiary education spending as compared with their
share of the population. By contrast, the richest quintile receives 32.2 percent of public
spending at the tertiary level for 12.5 percent of the population. This is mainly due to
significantly low tertiary-level enrollments amnong the poor and the ultra-poor.19 The
spending share within quintiles by race indicates that African students make the large
majority of tertiary enrollments among the poor and the ultra-poor but they receive less
than their share of the population.
Figure 7
Tertiary-level public education spending by household quintile and race
30.0 -           78100
J ~~~~~~~~CAfricaW
a  Pepsi            iI3 an60
U iondad
20.0                        Sp0                                          Hedw4ng
60
Source. DBSA (1993) and SALDRU (1993)
Race-specific subsidies reveal severe inequities in the South African education system
Severe inequities prevail in the distribution of public education resources across
income groups, regions and races in South Africa. The major source of disparities comes
from substantial differences in public spending per student by race.20 The incidence
analysis of public education spending by income groups suggests that race-specific
subsidies have a significant impact on the distribution of public education resources across
population groups in South Africa.
Although one-third of all enrollments in South Africa are from ultra-poor students,
they receive only one-fifth of all public education spending (see Figure 8). By contrast,
19 Ibid
20 See Table 9 and Figure 3.
19



the richest income quintile has less than one-tenth of all enrollments but receives almost
one-fourth of all public resources in education. This result is obtained when using race-
specific subsidies, compiled by DBSA (1993), to do the incidence analysis of public
education spending. Unfortunately, progress in the distribution of education resources in
South Africa will become increasingly difficult to assess because race disaggregated
infonnation is no longer being collected in South Africa.
If average, instead of race-specific, subsidies are used to perform the incidence
analysis of public education spending in 1993, then public resources in education seem to
be about equally distributed with every income quintile receiving almost its share of
enrollments (see Figure 8).
Figure 8
All public education enrollments and spending by household quintile, 1993
3    44
30           ~~~~0 a,rolu.ut
30 -                       * meage subsidies
z             ~~ ~ ~~~26  * race-speoic subsies
25  ~ ~~2624                                          23
25        21                                  20
&~~~~20 ~~~~19 19
20~~~~~~~~~~~I
U                              1-s~~~~~~~~~~~~~~~~~~1
P                                      13             12 rZ   a
9
..10
5
0           -
poorest      H1          IIIIT                  Richett
.Source: DBSA (1993) and SALDRU (1993)
Public education spending per capita is lowestfor Africans
Public education spending in per capita terms is lowest for Africans at every
educational level (see Figure 9). This result is a combination of supply and demand
effects. African departments of education have the smallest allocation of public spending
per student and Africans have the lowest enrollments at every educational level.
H-Africans receive R 153 annually from primary schools, the smallest public
primary education spending in per capita terms. NH-Africans receive R 167 compared
with Coloureds R 424, Indian R 372, and White R 384. The largest public primary
education spending per capita goes to Coloureds because of a demographic effect, they
20



have large families, and because of an enrollment effect, they have high enrollments at the
primary level.
Figure 9
Annual edrucation spending per capita, by level of schooling and race
1000 .rnr
.} t    t 0
1990 ~      0uc:DsA(93 adSLRU(93
From all ubli eduatio  spedingaggrgate forall eve          Secondary
4a0 tTriany
n    39b                         AD Educarion
200
Source: DBSA (1993) and SALDRU (1993)
From all public education spending aggregated for all levels - primary, secondary
and tertiary - in per capita termns annually, H-Africans receive R 31 1, NI--Africans receive
R 328, Coloureds receive R 765, Indians receive R 899 and WThites receive R 95 1 (see
Appendix Table A.5). Differences in public education spending per student are the
predominant effect explaining disparities at the primary and secondary levels, while by
contrast at the tertiary level differences in enrollments are the main source of disparities.
Targeting the Eastern Cape and the Northern Transvaal can improve the poverty focus
of public education spending
Targeting public education spending towards the poor and ultra-poor in South
Africa requires actions on both the demand for and the supply of education. The poverty
focus of public investments in human capital in South Africa can be substantially improved
if education incentives are targeted to increase secondary and tertiary enrollments of H-
African students and simultaneously a larger share of public education resources is
allocated to them. However, not all enrollments of H-African students belong to the poor
21



and ultra-poor. Therefore, it is also necessary to geographically identify poor areas within
the former Homelands in order to improve targeting of public spending in education.
Important gains in allocating public education resources to the poor and ultra-poor
can be achieved by targeting the Eastem Cape and the Northem Transvaal. The poverty
profile already identified these two provinces as having by far the highest poverty rates
and also explained that "..the high poverty rates in the Eastern Cape and the Northem
Transvaal are closely linked to the very high poverty rates in the former Homelands." In
addition, these provinces concentrate more than 50 percent of H-Africans.
The poor are unable to afford the direct costs of education
The direct costs of sending children to school are a critical factor determining the
demand for education. The analysis of the reasons for being out-of-school in South
Africa, included in Section U of this paper, already determined that education expenses are
the most important reason discouraging attendance to school. Most South African
families have to spend a significant amount of their resources to send children to school.
The direct costs of attending school include school fees, uniforms, transport and meals;
books and stationary and miscellaneous items, such as: boarding fees, contributions to
school buildings, extra costs for teachers and extramural activities
Poor families have a total direct cost of more than 40 percent of per capita non-
food household expenses per each child sent to primary school compared vwth only 6
percent for non-poor families (see Table 10). Direct costs increase to half of per capita
non-food household expenses for poor families sending children to -.ecoridA\ %:h"ol in
contrast to 10 percent for the non-poor. Tertiary education ts       I. 1  the pbor
Each poor student enrolled at the tertiary level needs more thar. i     a t& Iw  the
per capita amount spent on non-food items compared with less th.i  fv  . ee  t  rln-
poor.
Public spending per student at every level is larger than the drec! oi  inuned b%
poor families sending their children to school. Public spending per primanr stuJc'n is two
and a half times the amount spent by poor families on per capita non-food items (sc5 Table
10). The state contribution to education is almost three times per secondary student and
about eight times per tertiary student the amount spent by poor households on non-food
items per child.
22



Table 10
Household expenses on education by educational level, 1993 total and per student
Household expenses on education         Total education  Public spending
Educational level       Share of   School   Uniforms   Books &   Misc. '   Total    as a share of  per student as a
and enrollment group     enrollments   fees    transport  stationary                      per capita     share ofper
& meals                                   non-food     capita non-food
Hh expenses    Hh expenses
PRIMARY
All primary enrollments         100%       35%        47%        6%        12%      100%         12%             52%
Poor primary enrollments        60%        17%        73%        5%        5%       100%         41%            246%
Non-poor primary enrollments    40%        40%        39%        6%        14%      100%         6%              23%
SECONDARY
All secondary enrollments       100%       31%       47%         7%        14%      100%         15%             62%
Poor secondary enrollments      52%        14%        71%        5%        10%      100%         50%            297%
Non-poor secondary              48%        36%        40%        8%        16%      100%         10%             38%
enrollments
TERTIARY
All tertiary enrollments        100%       67%        12%        9%        11%      100%         29%             49%
Poor tertiary enrollments       25%        56%        32%        12%       0%       100%        258%            798%
Non-poor tertiary enrollments   75%        69%        10%        9%        12%      100%         29%             44%
* This category aggregates: Boarding fees, contributions to school buildings, extra costs for teachers and extramural activities; due to low frequency of
household expenditures
Source: SALDRU (1993)



Although the contribution of the state to education in South Africa is larger than
the direct costs to families, on average households contribute a significant proportion of
the total education cost for each child enrolled in school. Including household education
expenses and public education spending, households contribute 20 percent of the total
cost per each child in primary school, 20 percent for children in secondary school and
close to 40 percent for those in tertiary education. Families spend R 267 per primary
student annually compared with R 1,149 allocated from public funds, R 484 compared
with R 2,055 at the secondary level, respectively from families and from the state, and, R
3,318 from households in contrast with R 5,657 from the state at the tertiary level.
Reduced education expenses can increase enrollments among the poor: Uniforms,
transport and meals for primary and secondary and exemption offees for tertiary
Uniforms, transport and meals are the most important education expense for poor
families sending children to primary or secondary school, while school fees are the largest
expense for poor and non-poor households with tertiary-level students (see Table 10).
Poor families with children attending primary and secondary schools spend close to 75
percent of their household education expenses on uniforms, transport and meals in
contrast to only a 15 percent spent by non-poor households. At the tertiary level, school
fees are the largest household expense on education for both the poor and the non-poor,
on average 56 and 69 percent of total education expenditures, respectively.
Unaffordable school fees discourage tertiary-level enrollments among the poor.
The analysis of enrollment rates by income quintile, in Section II of this paper, shows that
the poorest three quintiles have significantly lower tertiary-level enrollments compared
with the richest two quintiles. Therefore, an alternative to reduce the direct costs of
tertiary education for the poor can be to increase the number of scholarships exempting
them from the payment of school fees. Cost recovery mechanisms that do not discriminate
against the poor can be simultaneously introduced to free public education funds to
finance the additional expense.
24



VL CROSS-COUNTRY COMPARISONS OF EDUCATION
INCIDENCE ANALYSIS
The poor in South Afiica receive a smaller share of public education spending at
every educational level than their share of the population but the shares of spending are
larger than their total household expenditure share. This pattern is characterized as weakly
pro-poor because the distributions of public education spending at every educational level
lie in between absolute equality and the Lorenz distribution of total household
expenditures. The distributions of education resources by educational level are shown in
Figure 10 across income quintiles together with the Lorenz distribution of total household
expenditures.2' The diagonal line (or 450 line) is also known as the line of absolute
equality since it goes through those points where the cumulative share of the population
equals the cumulative share of public education spending.
Figure 10
Distribution of public education spending and Lorenz distribution, 1993
400
t 10                   Prww        X
60
i ~ ~ ~ ~ ~ ~ 7 20* e,;/ ;    osR
0     20 2    0     60      so    100
Cumulahve share of popuLaton
Souerce: DBSA (1993) and SALDRU (1993)
21 In this section, quintiles are created by ranking individuals on the basis of per adult equivalent
expenditures, from poorest to richest and then aggregating them into five groups with 20 percent of
individuals in each quintile. This way of creating population quintiles, by ranking individuals instead of
households, is conunonly used in most countries where incidence analysis of public spending is available
and therefore can be used for cross-country comparisons.
25



Public education spending in South Afiica is less inequitable than the Lorenz
distribution or distribution of total household expenditures. The share of all public
education resources going to the poorest quintile is 14 percent compared with only a three
percentage share of total household expenditures. By contrast, the richest quintile
receives a 35 percentage share of all public education spending but concentrates 64
percent of total household expenditures.
Public primary education spending going to the poorest quintile is 19 percent
compared with 28 percent for the richest quintile. At the secondary level, the poorest
quintile receives an 11 percent and the richest quintile a 39 percent of public education
resources.   The distribution of public spending in tertiary education is almost as
inequitable as the distribution of total household expenditures. The poorest quintile
receives only a six percentage share of public tertiary-level education resources compared
with a 47 percent going to the richest quintile.
Table 11
Incidence of public education spending on the poorest and richest quintiles,
selected countries, percentage share
Education spending benefiting:
Country            Year       the poorest 20% of the   the richest 20% of the
_______ _ _  population (%)     population ("o)
AFRICA
Cote d'Ivoire             1993               14                      35
Ghana                     1992                16                     21
Kenva                    1992/3               17                     21
Madagascar                1993                8                      41
Malawi                   1994/5               16                     25
South Africa             1993                 14                     35
Tanzania                  1993                13                     23
LATIN AMERICA
Chile                     1986                25                     17
Colombia                  1992                23                     14
Mexico                    1992                14                     27
Uruguay                   1989                33                      15
Sources: Davon (1995b); Bernier, Chao, and Demery (1994); Castro-Leal (1995); Demery and Verghis (1994); World Bank (1995).
Note: If public education spending were equally distributed across population quintiles, then the poorest and richest income quintiles
(every number in the table above) would receive a 20 percentage share of spending.
26



South Africa has one of the worst distributions ofpublic education spending
Public education spending in South Africa is weakly pro-poor in contrast to
countries like Chile, Colombia and Uruguay, which are strongly pro-poor in their
distribution of public education resources (see Table 11). If public spending were equally
distributed across the population, then every quintile would receive an exactly equal share
of total public resources in the sector. The cross-country comparison is particularly
significant with Chile because the Lorenz distribution, or distribution of household
expenditures, is as inequitable as in South Africa. However, Chile is embarked in an
education reform to progressively distribute public funds in education.
In contrast with other African countries, South Africa has about the most
inequitable distribution of public education spending with the exception of Madagascar.
South Africa also has significant disparities across races for public education spending
which are wide at the secondary level and widest at the tertiary level.
27






VIL POIJCY IMPLICATIONS
Reduce inequities in education by geographically targeting public spending to the poor
Inequality in the distribution of education resources widens by educational level
from primary to tertiary across different income groups and also across different races.
Public education spending per capita is lowest for Africans at every educational level.
However, not all African students belong to the poor and the ultra-poor. Therefore,
policies to target public education resources to the poor have to rely on more precise
means of identifying them. The geographical concentration of the poor in some areas of
South Africa is crucial in targeting public education resources. The poverty assessment
already determined that "... nearly 70 percent of the poor live in the former Homelands ..."
Thus, gains in allocating public education resources to the poor and the ultra-poor can be
achieved by targeting the Eastern Cape and the Northern Transvaal. These two provinces
have by far the highest poverty rates and concentrate more than 50 percent of the former
Homeland population.
Reduce education expenses on uniforms, transport and mealsfor poor primary and
secondary students
Targeting public education spending towards the poor and the ultra-poor in South
Africa requires actions on both the supply of and the demand for education. The direct
costs of sending children to school are a critical factor determining the demand for
education. The analysis of the reasons for being out-of-school in South Africa shows that
education expenses are the most important reason discouraging attendance to school.
Most South African families have to spend a significant amount of resources in school
fees, uniforms, transport, meals and books and stationary in order to send their children to
school. Poor families pay more than 40 percent of per capita non-food household
expenditures per each child sent to primary school compared with only 6 percent for non-
poor families.  Education costs increase to half of per capita non-food household
expenditures for poor families sending children to secondary school, in contrast to 10
percent for the non-poor.
Uniforms, transport and meals are the most important education expense for poor
families sending children to primary and secondary school. Poor families with children
attending primary and secondary schools spend close to 75 percent of their household
education expenses on uniforms, transport and meals in contrast to only a 15 percent spent
by non-poor households. This analysis suggests that reduced costs of uniforms, transport
and meals for poor primary and secondary school-age children can increase enrollments
among the poor.
28



Higher education scholarship program and credit market development
Public resources dedicated to tertiary education largely favor non-poor students.
About one-third of public tertiary-level spending benefits the richest income quintile;
which contains only nine percent of the population 18 -to 22 years old. At present, public
tertiary education in South Africa is prohibitive for the poor. Each poor student enrolled
at the tertiary level needs more than two-and-a-half times the per capita amount spent on
non-food items in order to pay for school expenses compared with less than one-third for
non-poor students. At the tertiary level, school fees are the largest household expense on
education for both the poor and the non-poor, on average 56 and 69 percent of total
education expenses, respectively. This analysis suggests that exempting poor students
from the payment of school fees at the tertiary level can increase enrollments among the
poor.
A strategy to finance the additional expense of providing scholarships in tertiary
education for the poor is to design cost recovery mechanisms that do not discriminate
against them. An alternative for freeing public education funds and simultaneously
increasing cost recovery is the development of tertiary education credit markets. This type
of programs have been designed for other countries and best results are obtained when
student loans are channeled through existing commercial credit institutions to ensure loan
recovery, increase credibility and maintain cost-effectiveness.
29



REFERENCES
Buckland, Peter and John Fielden. 1994. Public Erpenditure on Education in South
Africa, 1987/88 to 1991/2. Center for Education Policy Development,
Johannesburg and The World Bank, Washington, D.C..
Castro-Leal, Florencia, Julia Dayton and Lionel Demery. 1996. The Benefit Incidence
Analysis of Social Spending: An Analysis of Cross-Country Differences. The
World Bank, Washington, D.C. (mimeo).
Development Bank of Southern Africa (DBSA). 1993. Public Expenditure on Education
in South Africa, 1987/8 to 1991/2. Volume 1: Financial Information and Volume
2: Contextual Information. Development Information Group of DBSA and the
World Bank. South Africa.
Meerman, Jacob. 1979. Public Expenditure in Malaysia: Who Benefits and Why? OUP
for the World Bank, Oxford.
Ministry in the Office of the President: Reconstruction and Development Programme
(RDP). 1995. Key Indicators of Poverty in South Africa. South Africa
University of Cape Town's Southern Africa Labour and Development Research Unit
(SALDRU). 1993. South Africa Living Standards Survei' South Afnica
Selowsky, Marcelo. 1979. Who Benefitsfrom Government Expeniditure 2 A Case Study of
Colombia. A World Bank Research Publication, Washington, D.C.
Van de Walle, Dominique and Kimberly Nead. 1994. Public Spending and the Poor.
Theory and Evidence. PRDPE. World Bank, Washington, D.C.
30



I



APPENDIX A
Table A. I
Raciol composition offormer homelands
Race                %
African                  99.95
Coloured                  0.01
Indian                     0.02
White                     0.02
Total                    100.00
Source: SALDRU (1993)
Table A.2
Africans by area of residence
Race            %
H-Africans        65.68 i
NH-Africans       34.32
Total            100.00
Source: SALDRU (1993)
31



TableA.3
Net enolIment rates, by quintile, gender
and level of schooling, 1993
Gender
Hh        Female        Male               All
quintile
Primary education
Poorest        87           84               85
II             87           87               87
m              88           88               88
IV             87           91               89
Richest        91           89               90
Total          87           87               87
Secondary education
Poorest        51           40               46
II             61           52               57
III            68           67               67
IV             81           76               78
Richest        87           79               83
Total          63           56               60
Tertiary education
Poorest         5           3                 4
nI             6            4                 5
III            10           6                 8
IV             19           21               20
Richest        32           44                38
Total          11           10                I1
Sources: SALDRU (1993)
32



Table A.4
Gross enrollment rates, by quintile, gender
and level of schooling, 1993
[              Gender
hih       Female      Male              All
quintile
Pnmary education
Poorest      109         114              112
II         105        111               108
m          101         105              103
IV          95         100               97
Richest       96         99                97
Total       104         108              106
Secondary education
Poorest       81         81                81
n          104         92               98
m          117         102              110
IV         115         100              108
Richest      104         97               101
Total       100         93                97
Tertiary education
Poorest       6           4                5
n          8           6                 7
m           12          8                1 
IV         23          25                24
Richest       40         57                41%
Total        14         13                14
Source: SALDRU (1993)
33



Figure A.)
Population distribution by household quintiles
30.012.                      Ho,,eJeoIz ^r
20.0 919.
10.0                                      1.
0.9
.,   z
Source: SALDRU (1993)
Figure A.2
.Distribution of the population by race within
household quintiles
so                                     Hiomeland
*Inodian
~~~ 40                        E ~~~~~~Coloured
20                                   E  4frcalNon-
I                              ~~~~~~~~~Homeland
Source: SALDRU (1993)
34



Table A. 5
Annual education spending, by level of schooling and race, 1993
(Rand per capita)
Education spending by race
Educational level    H-African    NH-African   Coloured   Indian    White   All
Primary               153           167         424       372       384      213
Secondary             102           97          259       404       394      157
Tertiary               56            65          81        122      174      76
All education         311           328         765       899       951      446
Source: DBSA (1993), SALDRU (1 993)