Economics of Education Review 14 1999 31–50
Cyclical economic conditions and school attendance in Costa Rica
Edward Funkhouser
Department of Economics, University of California, Santa Barbara, CA 93106, USA Received 14 August 1995; accepted 22 August 1997
Abstract
In this paper, I examine the importance of declining economic conditions in school attendance decisions for house- holds with teenagers aged 12–17 in Costa Rica. I estimate a reduced-form model for school attendance that includes
variables to measure household labor market integration, other household factors, teenager characteristics, regional effects and year effects. The year effects — to measure cyclical factors — show a large drop in school attendance not
otherwise explained by the household variables included between the years of macroeconomic recession 1981–1983. These changes coincide with fiscal austerity imposed towards the end of 1982. In addition, household labor market
characteristics and teenager characteristics are significant determinants of teenager school attendance.
1998 Elsevier
Science Ltd. All rights reserved.
1. Introduction
For many developing countries that have successfully implemented general primary education, the most
important aspects of educational policy are now related to increases in coverage and quality at the secondary
level. In most countries these policies have included changes in the compulsory schooling age, increased
availability of secondary schools, increases in teacher qualifications, and the provision of better resources to
existing schools. But whereas differences in school attendance at the primary level is closely related to avail-
ability of educational opportunities for students of pri- mary age, for teenagers of working age, availability of
educational opportunities at the secondary level is only one component of the schooling decision. In addition,
changes in labor market conditions affect the allocation of time between school attendance, work, and leisure
time.
Recent research on school attendance in developing countries has focused on the role of household character-
istics, including
income and
poverty, in
school enrollments Behrman and Wolfe, 1984; Schultz, 1988;
0272-775798 - see front matter
1998 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 2 - 7 7 5 7 9 7 0 0 0 5 3 - 8
Rosenzweig, 1990; Behrman and Deolalikar, 1991; Paes de Barros and Silva Pinto de Mendonca, 1991; Behrman
et al., 1992; Deolalikar, 1992, the quality–quantity trade-off in children Knodel and Wongsith, 1991, and
differences by gender and birth order in school enrollments Deolalikar, 1992; Parish and Willis, 1992.
There has been little research on the role of cyclical economic factors in increasing the number of secondary
students who do not continue their schooling. Given the importance of the world recession of the early 1980s on
household incomes in the developing world, these issues may be quite important. The finding of importance of
household economic variables in the school enrollment decision would have implications for educational policy
during periods of economic recession. Governments con- cerned about changes in attendance should consider
countercyclical policies to stimulate secondary school attendance in order to stabilize attendance levels during
cyclical swings.
One reason for lack of research in this area may be that aggregate data from the United States and Latin
America does not show a strong relationship between cyclical factors and school attendance. Evidence from
32 E. Funkhouser Economics of Education Review 14 1999 31–50
the developed countries on the effect of labor market conditions on school attendance has focused on intergen-
erational transmission of education through parent wealth and the effects of demographic cohort size on
future labor market outcomes and educational attain- ment. A common finding in the former studies is a posi-
tive relationship between both household wealth and par- ent education and educational attainment of children
Fuller et al., 1982; Margo, 1987; Haveman et al., 1991.
In the latter group, several studies have shown that wages of otherwise similar workers tend to be inversely
related to the size of birth cohort Welch, 1979; Berger, 1985; Bloom et al., 1987. Studies of the effect of these
wage patterns on school attendance have been less unani- mous, though most studies have tended to show that both
the quantity of schooling and the amount of time required to achieve a given level of schooling increase
among persons in larger birth cohorts Wachter and Wascher, 1984; Falaris and Peters, 1992.
But since the school attendance decision depends on the labor market outcomes of both the student and other
family members, the sign of changes in school attend- ance from a general decline in wages, for example, is
not known a priori. Though the own-wage substitution effect from lower market wages leads to an increase in
the demand for schooling, the income effect from a decline in household income could be associated with a
decline in schooling demand. Therefore, aggregate evi- dence that does not reveal a strong relationship between
cyclical wage changes and school attendance may be masking two opposing patterns. Other macroeconomic
factors, such as changes in government spending, also affect school attendance during economic recession.
Aggregate evidence from the United States over the early 1980s does not suggest that enrollment rates were
responsive to cyclical wage changes. The aggregate enrollment rate of persons aged 16 to 24 in the United
States stayed constant at 30 percent over the early 1980s while real wages fell by 5.1 percent.
1
Studies on the effect of own-wage changes on school enrollments have
found mixed evidence for the case of the United States. Card 1992, for example, does not find much effect of
changes in the minimum wage on school enrollments while Neumark and Wascher 1995 conclude that there
is an effect of the minimum wage on the skill level of teens sought by employers and — because the lower
skilled teens are less likely to enroll in school — school enrollments.
Similarly, cross-sectional evidence from the countries of Latin America, shown in Fig. 1, do not reveal a strong
relationship between changes in real gross national pro- duct per capita and percentage point changes in school
enrollments. In Fig. 1A and B, the percentage change
1
See U.S. Bureau of Labor Statistics 1989, Table 63.
in gross national product between 1975 and 1980 is plot- ted against changes in gross school enrollment rates at
the primary and secondary levels.
2
During this period, gross national product per capita was growing in all
countries except Nicaragua and Peru and school enrollment rates were increasing in most countries. Dur-
ing this period of expansion there is not a statistically significant relationship between GNP per capita and
enrollment rates at either the primary or secondary lev- els.
In Fig. 1C and D, the relationship between GDP per capita and enrollments for the period 1980–1983 are
shown. During these years, the mean for both economic growth and enrollment rates are lower than in the earlier
period. As in the earlier period, for both primary and secondary enrollment rates, the relationship between
growth in GNP per capita and change in enrollment rates is statistically insignificant.
In Fig. 1E and F, the relationship for the period 1983–1988 — the last year in which there is data for a
large sample of countries — is shown. In contrast to the earlier period, most countries experienced declines in
real gross national product per capita and many also suf- fered declines in gross enrollment rates. But even during
this period of contraction as well there is not a strong relationship between the magnitudes of declines in GDP
per capita and school enrollments. There is a positive relationship between changes in GNP per capita and
enrollment rates at the primary level and a negative relationship at the secondary level, with only the primary
relationship being statistically significant. To the extent that there is a pattern in these data, it appears that
changes in primary school enrollments are positively related to economic growth within a country and changes
in secondary school enrollments are negatively related to economic growth.
3
This latter finding would indicate that the intertemporal substitution in response to own-wage
changes dominates the household income response. In Costa Rica, though, secondary school enrollments
and percentage change in GNP per capita both fell during the period 1980–1983. In Fig. 1D, it can be seen that,
of the reported countries, Costa Rica CR experienced the largest drop in secondary school enrollments during
the period 1980–1983. Costa Rica also experienced the largest percentage drop in GNP per capita during these
2
The gross school enrollment rate is school enrollment of all students at a given level as a percentage of the number of chil-
dren in the age groups that would normally be at that level. The data on gross school enrollment rates and real gross domestic
product are found in Wilkie 1990.
3
When controls for the initial level of GNP are included, the pattern of lack of correlation remains similar with one
exception — the sign of the relationship between change in secondary enrollments and GNP per capita for the period 1983–
1988 becomes positive.
33 E. Funkhouser Economics of Education Review 14 1999 31–50
Fig. 1. Change in primary enrollment and secondary enrollment, respectively, A,B 1975–1980, C,D 1980–1983, E,F 1983–
1988, controlling for initial GNP per capita.
years. Following 1982, Costa Rica was one of the first Latin American countries to successfully implement sta-
bilization policies to control inflation, the government deficit, and the external account. Since these policies
accentuate effects on both the supply and demand for secondary level education, the decline in school
enrollments associated with the decline in economic activity in the early 1980s make Costa Rica a particularly
interesting case for the study of the effects of cyclical factors on school enrollments.
In this paper, I examine the labor market determinants of school attendance using household data in a
developing country that experienced both a rapid increase in the supply of education in the 1970s and a
substantial worsening in the labor market in the early 1980s. I utilize annual data for 1980–1985 in Costa Rica
to estimate a reduced-form school attendance decision model that allows estimation of both own-wage and
cross-wage effects. Costa Rica has high levels of literacy and education
that resulted from social investments beginning in the last century. By the early 1980s, literacy was over 90
percent and primary education was general in both rural and urban areas. Expansion at the secondary and tertiary
levels has been more recent.
4
But Costa Rica was also among the countries most affected by the debt crisis and
recession following the 1979 oil price shock. The period from 1980 to 1982 included a 9.4 percent drop in real
GDP, an increase in inflation to 108.9 percent annual
4
See for example, Gonzalez Gonzalez 1984.
34 E. Funkhouser Economics of Education Review 14 1999 31–50
Fig. 1. Continued.
rate in September 1982, and significant adjustment in the external sector.
Basic patterns in wages and enrollments during the 1980s are shown in Fig. 2. The pattern in real wages of
workers affiliated with the social insurance system and the mean wages reported in the annual household survey
are shown in Fig. 2A. The long-run increasing trend in real wages was broken by a large drop in real wages
between 1980 and 1983 during the period of economic crisis. The real minimum wage, an alternative measure of
the opportunity wage of teenagers, also fell substantially during this period Fig. 2B.
In the next two graphs of Fig. 2, I report changes in school enrollments using school enrollment data from
the educational system Ministerio de Educacion Pub- lica, 1969, 1991 and from tabulations for attendance
from the National Household Survey conducted in July of each year from 1976 to 1992. During the period of
the drop in real wages, school enrollments in grades 7– 12 fell in absolute numbers Fig. 2C and the proportion
of children in the annual household survey aged 12–17 reporting school attendance also fell Fig. 2D. These
aggregate data do show a correlation between wages and school enrollments. I now turn to explaining these pat-
terns.
2. Model