441 P. Tsakloglou, M. Antoninis Economics of Education Review 18 1999 439–452
Table 2 Cost structure of the Greek public education system, 1987–1988
No. of Current spending
a
Investment Total
Average spending students
spending
b
spending
a
per student Ministry of
Ministry of Current
Total Education
Interior Primary education
Total no. of students 1 023 581
100.00 Public schools
962 232 94.01
72 955 618 4 352 726
14 703 889 92 012 233
80 343 95 624
Private schools 61 349
5.99 Secondary education
Total no. of students 846 658
100.00 Public schools
808 539 95.50
81 640 755 12 340 636 13 319 444 107 300 835
116 236 132 709
Private schools 38 119
4.50 Tertiary education
Total no. of students 183 242
100.00 44 058 355
12 487 222 56 545 577
240 438 308 584
Universities 117 193
63.96 34 160 332
9 465 557 43 625 889
291 488 372 257
Technological 66 049
36.04 9 898 023
3 021 667 12 919 690
149 859 195 608
Education Institutions
a
In thousands of current drachmas.
b
Average spending of the 10 preceding years in thousands of 1988 drachmas. Source: Ministry of Finance 1989.
Current public spending on education is managed by two ministries. The bulk more than 90 of the total,
which relates to the payment of teachers’ salaries, text- books, mess allowances and scholarships is channelled
via the Ministry of Education. A small amount less than 10 of the total covering transport, schools’ mainte-
nance and rent costs, is administered by the Ministry of Interior and the local authorities. Investment expenditure
is managed through the Public Investment Budget and, as could be anticipated, exhibits a fluctuating pattern
Antoninis Tsakloglou, 1997. For this reason, instead of current investment, in our analysis we use a moving
average of the investment level of the 10 years preceding each particular year, thus implicitly assuming that capital
investments in education depreciate constantly within a 10-year period.
Table 2 provides an overview of the Greek education system in 198788 in terms of numbers of students both
in public and
private schools, total
expenditure distinguished
between current
and investment
expenditure stated in current 1988 thousand drachmas and average yearly cost per student attending a public
school in current 1988 drachmas for each of the three
rowing of differentials between tertiary education graduates and the rest of the population; see, also, Lambropoulos and Psachar-
opoulos 1992.
levels of the education system. The analysis of the distri- butional impact of public education spending is based on
the information included in this table.
3. Data and methodology
The data used in the paper are the micro-data of the 198788 Greek Household Budget Survey which was
carried out by the National Statistical Service of Greece. The survey covers all the private non-institutional
households of the country and its sampling fraction is 21000 around 6500 households or 20000 individuals.
It contains detailed information about consumption expenditures actual and imputed, incomes and socio-
economic characteristics of the households and the household members. For the purposes of the paper, the
unobservable welfare level of a household is approxi- mated by its level of consumption expenditure. This
choice was dictated by two factors. Firstly, current con- sumption is usually considered a better indicator of the
long-term welfare level of economic agents than current income and the use of consumption expenditure data in
the framework of the present study can help in side-step- ping problems arising from comparisons over different
phases of the life-cycle. Secondly, according to the National Statistical Service of Greece, the consumption
442 P. Tsakloglou, M. Antoninis Economics of Education Review 18 1999 439–452
expenditure data of the Household Budget Survey are considered as more reliable than the corresponding
income data. The concept of consumption expenditure used here
includes, apart from the value of purchased goods and services, consumption of own production, consumption
of income in kind and imputed rent for owner-occupied accommodation evaluated at market prices
3
. Several adjustments were made to the data. A few households
were removed from the sample on reliability grounds and the sample was re-weighted in order to reflect more accu-
rately the entire population using Population Census weights. Further, all consumption expenditure values
were expressed in constant mid-1988 prices in order to remove the impact of inflation 14.1 from the begin-
ning to the end of the survey. Finally, the self-reported estimates of imputed rent for owner-occupied accommo-
dation were replaced by estimates derived using hedonic regression techniques. Similarly, the value of cars pur-
chased during the period of the survey was subtracted from the concept of consumption expenditure and
replaced by the value of imputed car services, also esti- mated using hedonic regression techniques, for all the
households which owned cars for details, see Tsaklog- lou, 1996.
The unit of analysis in the main part of the paper is the household and the distributions used are distributions
of equivalent consumption expenditure per household. The equivalence scales used are the so called “modified
OECD scales” Hagenaars, de Vos Zaidi, 1994 which assign a weight of one to the household head, a weight
of 0.5 to each of the remaining adults and a weight of 0.3 to each child person aged up to 13 in the household.
Division of the total expenditure of each household with its equivalence scale yields its equivalent consumption
expenditure, which is used as an indicator of the house- hold’s welfare level.
Household Budget Surveys are probably the best sources of information for distributional studies in
Greece but, nevertheless, they are far from ideal for the purposes of the present study. The following information
from the survey can be used for the purposes of this paper. For each member in the sample aged 14 or more
there is information on the highest level of education completed as well as on whether heshe is still a student.
Therefore, it is easy to find out the level of education
3
It can be argued that the concept of “consumption expendi- ture” used here is closer to the unobservable “welfare” than
the concept of “monetary income” that is habitually used in distributional studies. Nevertheless, as in most distributional
studies, due to lack of relevant information this concept does not include the value of goods and services provided by the
state and, less controversially, the imputed value of leisure enjoyed by the economic agents.
attended by them. Although there is no explicit infor- mation on whether those who participate in the education
system attend public or private schools, this information can be extracted implicitly, since the Household Budget
Survey records information on expenditures for private education fees. Since we do not have any information
for pre-school education, this level of the education sys- tem is left out of our analysis. For children aged up to
13 for whom we do not have information about their education status, we assume that if they are 6–11 years
old they attend primary education and if they are 12–13 years old they attend secondary education. This assump-
tion is not unrealistic since, according to the Labour Force Surveys, in the late 1980s only 0.6 of those aged
6–13 did not attend school at all almost all of them non- educables and a further 1.6 did not enrol to secondary
education
after completing
primary education
Kanellopoulos, 1996. Since the estimates in the Household Budget Survey
are expressed in monthly figures, the cost estimates of Table 2 are adjusted accordingly. Hence, we assume that
the monthly transfer from the state to each student in public primary education is 7969 drachmas, to each stud-
ent in public secondary education 11 059 drachmas and to each tertiary education student 25 715 drachmas
4
. Since the Household Budget Survey does not distinguish
between University and Technological Education stu- dents of tertiary education, the average cost per student
is estimated at the tertiary level as a whole, despite sig- nificant differences across institutions.
In line with the general approach of Aaron and McGuire 1970, in the first part of the empirical analysis
of the next section, these transfers are added to the monthly consumption expenditure of the households of
the students, implicitly assuming that the benefits are shared by all household members or, alternatively, that
if the relevant costs were not borne by the state they would have to be borne by the household. This is not an
unrealistic assumption since family links are still strong
4
Only direct transfers are considered. Such indirect subsidies as discounted transport fares are not included in the following
analysis. It is likely that in some remote rural areas as well as in some small islands where class sizes are very small andor
the students are transported to the nearest school at the expense of the state, the cost per student in secondary and, particularly,
primary education is substantially higher than the corresponding costs in urban areas. However, no corresponding cost estimates
are available and, furthermore, it is doubtful whether this higher cost translates into higher quality of the final product education
services. Further, it should be noted that the approach adopted in the paper ignores education externalities static or dynamic
and the shifting of costs and benefits to unintended parties. This is the standard approach in the relevant empirical literature. If
this assumption is relaxed, the results of the paper could change accordingly.
443 P. Tsakloglou, M. Antoninis Economics of Education Review 18 1999 439–452
in Greece and intra-family transfers are a common phenomenon. A problem arises in this case regarding
about 30 of the tertiary education students and a few secondary education students in our sample, who study
in cities other than the cities where their parents live and form their own households or—a few of them—live in
relatives’ households
5
. Due to inadequate information, our
methodology implies
that the
welfare level
equivalent consumption expenditure of these students is determined by the transfers they receive from their
families—in an attempt to equalise the welfare level of all family members—but the benefits of public education
transfers are captured entirely by the students themselves and are not shared with the rest of their families. In the
second part of the empirical analysis it is assumed that these transfers benefit only the students who receive
them. Hence they are added to the equivalent consump- tion expenditure of the students and comparisons are
made not between households but between members of particular age groups
6
. Comparisons of inequality before and after the trans-
fers are made using five indices of inequality; Gini, Atkinson for e 5 0.5 and e 5 2 and the two Theil
indices. These indices satisfy the basic axioms of inequality measurement symmetry, mean-independence,
population-independence and the principle of transfers
7
5
The proportion of tertiary education students who study in places other than where their families live is likely to be sub-
stantially higher, but a considerable proportion of these students were interviewed in the houses of their families during
vacation periods.
6
It has been argued that, at least the students of the post- compulsory levels of education, apart from the benefits they
receive from the state, they incur the cost of foregone labor incomes see, for example, Selden Wasylenko, 1995. Apart
from the fact that this element of private cost is outside the scope of our investigation, it was decided to ignore such effects
for two additional reasons. Firstly, the measure of resources used in our analysis equivalent household consumption
expenditure mitigates substantially such life-cycle effects and, secondly and more importantly, the rate of unemployment
among those aged 14–25 in Greece in 1988 was a staggering 29.0, or 33.9 among those aged 14–19 National Statistical
Service of Greece, 1997, pp. 114–116, thus, rendering any cor- responding calculations almost meaningless.
7
The second of these axioms, mean-independence, is pretty standard in the theory of inequality measurement but may have
serious consequences for our analysis. It states that the level of inequality must remain unchanged if the incomes of all agents
change by the same proportion thus, excluding the possibility that the value of the index may change if the incomes of the
agents are measured in dollars, pounds, drachmas, etc.. How- ever, in the framework of our analysis this axiom implies that
inequality will remain unchanged if public transfers are pro- portional to the consumption expenditures of the households—
an assumption far from uncontroversial. For this reason, in the tables of the next section we also report, apart from estimates
of inequality indices, the absolute level of public transfers.
and are sensitive to different types of transfers Cowell, 1995. The Atkinson index and the second Theil index
are relatively more sensitive to transfers close to the bot- tom of the distribution, the first Theil index more sensi-
tive to transfers close to the top of the distribution and the Gini index to transfers close to the middle of the
distribution. Hence, the combined use of these indices satisfies a wide range of tastes regarding the responsive-
ness of an index to different types of transfers.
4. Empirical results