38 E. Funkhouser Economics of Education Review 14 1999 31–50
arbitrary fashion, within the household, time allocation decisions of adults are made prior to the time allocation
decision of children. With a sequential decision process in which adults make labor market decisions before teen-
agers make their time allocation decisions, teenagers condition their work and school decision on the actual
labor market outcomes of adults. In this case, pooling data across years and assuming that each teenager makes
the time allocation decision independently of other teen- agers in the household and the price of consumption
goods can be normalized to one, a first-order approxi- mation of Eq. 3 can be written:
T
sj
5 a 1 b
s
w ˆ
j
5
O
n i 5 j
b
wi
w
i
1 b
y
Y 1 b
r
r
j
4 1
O
k j
V
j
b
vj
1 X
f
b
x
1 f
c
1 f
t
1 e
j
in which the predicted wage, w
j
, is the opportunity cost of attending the next year of school for teenager j, f
t
is vector of year effects, and e
j
is an error that includes unobservable individual effects, d
j
, and any random effects. T
sj
is the indicator function for school attendance of child j:
E 5 1 if T
sj
5 E 5 0 if
T
sj
The estimation procedure involves four steps. First, mean wages within each canton and urban cell, con-
trolling for other factors, are calculated for each year from the sample of all workers to proxy for opportunities
in the local labor market. The second step is the calcu- lation of predicted wages for each teenager using Eq.
1d estimated separately by year. The predicted wages for individual j at time t estimated from the sample of
working teenagers is then applied to all teenagers aged 12–17 as a measure of P
s
. This equation is identified by the labor market opportunities variable. Third, the return
to the next year of education is calculated from the esti- mation of a wage equation for all workers in the same
year as the school attendance decision is made is used as a proxy for the future income differential resulting
from the investment in an additional year of education. The fourth step is the estimation of the reduced form
attendance equation, Eq. 4.
3. Data
In this study, I utilize the Encuesta Nacional de Hogares, Empleo, and Desempleo conducted by the
Direccion General de Estadistica y Censos DIGESTYC in Costa Rica. The survey was conducted with an inde-
pendent sample in July of each year from 1980 to 1983 and in 1985
7
and includes demographic and labor market information for household members aged 12 and above.
8
Each person is asked whether they currently attend school and the last grade attended. There are 22 798 chil-
dren aged 12–17 in the 5 years of the survey from 1980 to 1983 and 1985. 56.4 percent of all children report that
they attend school. Of those who attend school, 2.66 per- centage points report that they also work. And of the
43.6 percent who do not attend school, 19.1 percentage points report their main activity as working.
Information on other household members — including employment status of the head of the household, the
number of other working household members, the num- ber of non-working household members, number of chil-
dren under 12, income of head,
9
income of next highest earner,
10
and other income — was obtained by recon- structing households. These data were then merged with
the individual level data for teenagers. The opportunity wage of each teenager is predicted
from a wage equation using the sample of working teena- gers. The identifying variables in this equation are the
mean wage and unemployment rate of all workers in the local labor market calculated separately for the urban
and rural areas in each canton in which the teenager resides. The regressions from which these predicted
values are calculated are shown in Appendix A. With the exception of the equation for 1981, the pattern in
teenage wages is a positive relationship between teenage wages and each of mean wages in the canton calculated
from all workers, experience, and education. Girls earn substantially less than males in all years. And there is
no clear pattern in relative wages of teenagers in urban areas once education and mean wages have been con-
trolled for.
The return to education for each additional year of education was estimated utilizing the complete sample
for each year and including a dummy variable for each
7
The survey has been conducted from 1976 to 1992. In addition, there are also March and November surveys which
do not include information on education. The survey was not conducted in July 1984 in order to conduct a Census at that
time. There are two reasons that the period 1980–1985 is included in this study. First, these are the years before and dur-
ing the economic recession and any effect on school enrollments should be observed during these years. Second, the
survey format is different in surrounding years.
8
Before 1980, earnings information is available only for wage and salary workers. Starting in 1980, earnings questions
were asked of all workers.
9
Nominal values were deflated by the July value of the con- sumer price index for the San Jose metropolitan area Banco
Central de Costa Rica, 1986.
10
Highest non-head earner is used because relationship to household head is not identified in the 1980–1986 data.
39 E. Funkhouser Economics of Education Review 14 1999 31–50
year of education. The returns estimated using this pro- cedure are shown in Appendix B. These values were
matched by calendar year to the next year of education for each teenager in the survey.
11
3.1. School attendance patterns of teenagers aged 12– 17
School attendance patterns of teenagers aged 12–17 using adjusted sample weights
12
are shown by year in Fig. 3 and by age in Fig. 4. In each figure, Panel A
includes the percentage of teenagers in school, Panel B includes the percentage of teenagers only working, and
Panel C includes the percentage of teenagers neither in school nor working. The three lines in the figures report
these percentages for all teenagers, teenagers in rural areas, and teenagers in urban areas. As can be seen in
all of the graphs, the proportion of teenagers aged 12– 17 attending school is much lower in rural areas than in
urban areas. And the proportion that reports neither school attendance nor work is much higher in rural areas
than in urban areas.
The overall pattern by year shows a slight drop in the percentage attending school Fig. 3A, a slight increase
in the percentage only working Fig. 3B, and a more gradual increase in the percentage that neither are in
school nor work Fig. 3C over the period of economic recession 1981–1983. Nearly all of the change, though,
occurred in rural areas. This initial evidence indicates that the effect of cyclical conditions on school attendance
differed between rural and urban areas.
The patterns by age also reveal differences between rural and urban areas in the activities of teenagers. There
is a large drop in the proportion attending school between the ages of 12 and 15 in rural areas from 75.9
percent to 30.6 percent compared to a more gradual decline over these age groups in urban areas from 91.6
percent for those 12 years old to 69.7 percent for those 15 years old.
In Fig. 4, school and work patterns are shown by age of the child. 81.5 percent of children age 12 are reported
to be attending school only. By age 17, only 35.8 percent are only attending school. The drop-off in attendance
rates is quite large immediately following the completion of primary education between the ages of 12 and 15.
Approximately half of the decline corresponds to an increase in those who work and half corresponds to those
who neither work nor attend school. Above 15 years of
11
An alternative approach would be for children to have expectations about the return to each year of education at the
time they are expected to be in the labor market.
12
In addition, weights were calculated that reflect the popu- lation reported in the 1973 and 1984 Censuses in place of the
reported weights in the data files.
age, all declines in attendance correspond to increases in those who only work. There is little change in the pro-
portion who attend school and report working as age increases.
Fig. 3 showed a slight drop in the proportion of teena- gers aged 12–17 that were attending school at the time
of the survey. The permanence of the decline in school attendance rates during the years of economic recession
can be examined by following age cohorts over time. With the assumption that secondary schooling 11 years
is completed by age 23, final educational years of school- ing indicates whether attendance declines reflect either
lower final attainment or a longer period required to attain a given target level of education.
In Fig. 5, the proportion of each age cohort defined by age in 1980 along the horizontal axis with at least
11 years of education is shown at age 18 bottom line, age 19, ages 20–22, and ages 23–25 top line. In this
figure, I include grade completion information from all of the household surveys from 1976 to 1992. For
example, 25 percent of the cohort that was born in 1966 and was 14 years old in 1980 had finished at least 11
years of school four years later in 1984. Reading upwards, 27 percent had 11 years of education in 1985
at 19 years old and approximately 30 percent had 11 years by the ages 20–25. The difference between any
two lines indicates the proportion of a birth cohort that attained 11 years of education between the indicated
ages. Because cohorts that were born earlier were older in 1980, the general increase in educational attainment
of sequential cohorts is seen from the increase in the proportion of each cohort that has a secondary education
at each age level as the figure moves from right to left.
The age groups who were obtaining their secondary education during the years of economic crisis 1980–
1982 were those aged 12–17 in 1980. Two patterns are of interest. First, at all ages, there is a leveling off in the
proportion that had attained 11 years of schooling for birth cohorts that were younger than 18 in 1980, with
the final rate of each cohort between 29 and 33 percent. For the youngest cohorts during the recession and those
that became teenagers after the recession age less than 12 in 1980, the secondary completion rates at age 20–
22 are lower than earlier cohorts. This corresponds to the decreased enrollments seen in Fig. 2C and D. Second,
secondary education completion occurred earlier for birth cohorts that were under 14 in 1980. These cohorts
were entering secondary school age during or after the economic crisis.
3.2. Characteristics by type of activity Basic characteristics by type of activity are shown in
Table 1. The first four columns include information for teenagers from rural areas; the second four columns
include information for teenagers from urban areas.
40 E. Funkhouser Economics of Education Review 14 1999 31–50
Fig. 3. Percentage of teenagers A in school, B only working, C neither in school nor working.
41 E. Funkhouser Economics of Education Review 14 1999 31–50
Fig. 4. Percentage of teenagers a in school, b only working, c neither in school nor working.
42 E. Funkhouser Economics of Education Review 14 1999 31–50
Fig. 5. Percentage of birth cohort with 111 years school.
Table 1 Characteristics of teenagers aged 12–17 by school and work status
Rural Urban
Only school Work and
Only work Neither
Only school Work and
Only work Neither
school school
1 2
3 4
5 6
7 8
Age 13.92
15.23 15.42
14.77 14.38
15.65 15.71
15.25 Years educ.
6.40 6.63
5.65 5.55
7.27 7.19
6.26 6.02
Metropolitan 0.520
0.564 0.542
0.506 0.642
0.644 0.636
0.566 area
Female 0.529
0.307 0.228
0.719 0.519
0.349 0.302
0.595 Total persons
7.15 7.55
7.64 7.52
6.30 6.41
6.76 6.75
Females , 0.85
0.84 0.91
0.89 0.60
0.69 0.78
0.73 12
Males , 12 0.87
0.96 0.97
0.97 0.64
0.81 0.77
0.82 Head works
0.809 0.798
0.787 0.763
0.812 0.775
0.781 0.76
Head female 0.139
0.181 0.145
0.132 0.211
0.324 0.286
0.239 Head educ.
4.25 3.67
2.86 3.06
7.00 5.09
4.27 4.62
Head primary 4036 3316
3153 3410
6145 4315
4092 4284
labor income N
5573 326
3153 4135
6898 284
994 1378
Source: Tabulations by author.
There are several patterns that are common to both rural and urban areas. First, those who work tend to be older
and to have less education than those who were attending school at the time of the survey. Second, though approxi-
mately half of those enrolled in school are female, females are much less likely to be working than are
males; 72 percent of the teenagers that neither attend school nor work in rural areas and 60 percent in urban
areas are female. Third, the education of the head of the household is positively correlated with school attendance
by teenagers, though this may reflect household compo- sition rather than parent’s education. And lastly, employ-
ment rates and mean income of the head of households in which the teenager attends school are higher than
those in which teenagers work.
The main difference between urban and rural areas is in the levels of variables, rather than the pattern across
types of activity. Urban children obtain more education, live in smaller households, are more likely to live in a
household that is headed by a female, live in households
43 E. Funkhouser Economics of Education Review 14 1999 31–50
in which the head has higher mean education, and live in households with higher income.
4. Estimation results