Intergenerational Transmission of Education in Brazil: Has Inequality

Work in progress – comments welcome
8/1/2012
Intergenerational Transmission of Education in Brazil:
Has Inequality Decreased?
Letícia Marteleto
Population Studies Center and Dept. of Sociology
University of Michigan
[email protected]
April 2001
Abstract: This paper focuses on educational attainment for cohorts of children born pre- and
post- Demographic Transition in Brazil. Using a nationally representative data set, this paper
examines the cohort effects of socio-economic and demographic changes of the last three
decades on inequality in intergenerational transmission of education in Brazil on national and
regional levels. Results from the PNAD suggest that the curve of the relationship between
mothers’ education and children’s schooling is flattening. This is true for Brazil as a whole, as
well as for the Southeast and Northeast regions, which are marked by large socio-economic
and demographic disparities. Despite mothers’ low levels of education, children in younger
cohorts have an educational advantage compared to their older cohort counterparts. The
strength of the inequality on the intergenerational transmission of education has decreased
within and across cohorts, but still remains strong, particularly across regions.
*Paper submitted to be considered for presentation at the Annual Meetings of Anpocs. An earlier
versions of this paper was presented at the 2001 Annual Meetings of the Population Association of
America, Washington D.C. During the different stages of this research, support was provided by
CAPES Foundation, the Population Council, Spencer and Mellon Foundations. I would like to thank
David Lam, Rachel Lucas, Pamela Smock and Yu Xie for helpful comments. Please send
correspondence to [email protected].
1
Work in progress – comments welcome
8/1/2012
Introduction
The purpose of this paper is to examine the trend of the effects of social origin on
children’s education across cohorts separated by two decades of profound demographic and
socio-economic change in Brazil. In the context of declining fertility that resulted in long-term
modifications in the population age structure, I attempt to assess whether Brazil has become a
more meritocratic society with respect to its educational distribution. The adjustment of
individuals and social and educational institutions to the changing composition of dependent
groups resulted from the demographic transition has potentially altered Brazil’s unequal
educational distribution. The question of whether the country’s educational system has shown
improvements related to demographic change have potential lessons for other developing
countries that have experienced fertility decline. Nonetheless, Brazil presents a unique setting
for this study because of its poor educational performance. The country has shown consistent
high levels of educational inequality and low levels of schooling.
In order to assess whether the demographic change has affected Brazil’s distribution of
education across generations, I first focus on how such low levels of educational attainment
are generated within cohorts. Studies in several countries have shown that individuals’
educational attainment is highly correlated with their parents’ socio-economic background
(Mare 1979, 1981; Featherman and Hauser 1975; in Brazil, Barros and Lam 1996; Lam and
Levison 1991). Children whose parents have low levels of education tend to get less education
themselves, reinforcing existing inequality across generations. In addition to intergenerational
transmission of education, this paper also focuses on another central aspects of educational
inequality in Brazil: regional inequalities.
2
Work in progress – comments welcome
8/1/2012
As in several other countries, education is the focus of attention because of its potential
to reduce inequality in the present and in the next generation. Brazil’s unequal distribution of
education, both in terms of quantity and quality, is believed as contributing to inequality in
labor market earnings, and is also viewed as playing an important role in the intergenerational
transmission of inequality (Lam and Levison 1991; Barros and Lam 1996; Levison 1991;
Duryea 1997; Lam 1999). In such a context of high income and educational inequality, the
benefits education generates will be unequally distributed throughout the population.
In this paper I assess the role that mother’s education has played on children’s
educational attainment and school enrollment for cohorts born in 1963 and 1983, that is, in
Brazil’s pre- and post- Demographic Transition. The structure of this paper
Theoretical Framework
A process of educational expansion has recently been documented throughout the
world (Shavit and Blossfeld 1993). From one cohort to another, larger proportions of children
from all social strata have been enabled to attend school (Meyer et al. 1977). Still, children’s
educational prospects are found to strongly interact with parental education in several
countries (Featherman and Hauser 1975; Shavit and Blossfeld 1993). Despite increasing
access to education, and therefore decreasing inequality in educational distribution, empirical
studies on educational stratification in different countries have found a persistent stability of
the effects of parents’ socio-economic background on children’s educational attainment
(Blossfeld and Shavit 1993; Featherman and Hauser 1975; Mare 1980).
The economics and sociology literatures provide theoretical frameworks for the strong
relationship between parents and children’s educational attainment. The theoretical foundation
of the economics literature on educational outcomes is a utility function in which the costs and
3
Work in progress – comments welcome
8/1/2012
benefits of education are compared. Families and students compare the costs and benefits of
attending school against the costs of nonattendance. Direct costs are tuition, fees, books,
supplies, transportation and lodging expenses. The opportunity cost of not working is also
considered within the human capital theory framework. When an individual is in school,
his/her earnings are usually less than if s/he was not in school since s/he can not work as much
or as regularly as if s/he was not in school. The benefits of education are higher wages and
quality of life outcomes such as lower infant mortality rates, longer life expectancy, etc. Most
investments in human capital, such as formal education, raise observed earnings at older ages,
the expected return to education. In the short run, the difference between what could have
been and what is earned is an indirect cost of schooling (Becker 1964). Manski and Wise
(1983) state that the decision to enroll in school and achieve a certain educational level is the
result of student and family’s perceptions of the costs and benefits of education. Quantity and
quality of schooling therefore depend in large measure on the advantages or disadvantages
parents confer on their children. In such a framework, the utility function of parents with
distinct levels of education is different.
The sociology literature adds the roles of norms and expectations to this basic
framework. The cultural capital theory, first advanced by Bourdieu and Passeron (1964;
1977), contends that the key component of this strong association of parents’ education and
children’s education is that children from families with low levels of parental education are
likely to lack those abilities - norms and expectations - normally transmitted by the family and
valued by schools. Those abilities are broadly defined as cultural resources, societal values
such as good manners and social skills, and language skills. This idea reinforces the pattern of
4
Work in progress – comments welcome
8/1/2012
school selection, which favors children from families that already possess dominant cultural
advantages.
Empirical studies in Brazil confirm that the educational prospects of children are
closely related to their social origin, specifically parents’ education (Barros and Lam 1996;
Lam and Levison 1991). Schooling plays a key role in maintaining social inequality from one
generation to the other in Brazil. While past work has examined the extent of inequalities in
intergenerational distribution of education in one point in time, no study has examined how
patterns on these trends have operated across cohorts. In particular, no study has addressed
whether cohorts of children separated by a period of enormous socio-economic and
demographic change have fared differently in terms of their education.
Economic growth and demographic changes shape the macro conditions in which
educational inequalities take place. These changes may have affected parental and children
decision-making about children’s educational attainment and school enrollment. In this paper,
I examine whether the trend towards increasing children’s educational attainment and school
enrollment have translated into a more equal educational distribution in order to address the
question of whether Brazil has become a more meritocratic society. The spread of children’s
educational attainment and school enrollment altered over the last two decades, as documented
in Paper 3, but did it eliminate a strong pattern of educational inequality, namely the strong
association between mother’s education and children’s schooling outcomes?
There are theoretical reasons to believe that Brazil has become a more meritocratic
society in its educational distribution. Proponents of classical modernization theory of
educational attainment state that more schooling is a result of societies becoming industrial
and modern, which translates into decreasing levels of educational inequality (Parsons 1970;
5
Work in progress – comments welcome
8/1/2012
Treiman 1970). According to this perspective, the educational system expands in response to
the functional requirements of an industrial society in which education plays an increasing
role. Because of contextual particularities of the Brazilian development process, the trend on
educational attainment may not conform to classical modernization theory of educational
attainment. The development trend in Brazil has been toward more reliance on the resources
of individual households and less on public social services and social programs. This type of
development is likely to further improvements on children’s education, but it is also likely to
result in much less improvement in the schooling attainment of children at the bottom of the
socio-economic distribution, which may contribute to perpetuating educational inequality.
The major shifts in the Brazilian demographic patterns that have occurred over the last
decades may also have affected the country’s educational stratification. Brazil underwent the
Demographic Transition in the 1970s. Decreasing rates of population growth and declines in
relative cohort sizes resulting from secular declines in birth rates are components of such a
transition that have potentially affected the level of schooling inequality in the population.
Social mobility occurs within a demographic context and the evolution and transformation of
educational hierarchies is the joint outcome of demographic processes, intergenerational
transmission, and opportunity structure (Mare 1995). Basic demographic trends, such as
changing population age structure, may have unanticipated consequences for the process of
educational stratification. At the elementary and secondary levels, rates of school enrollment
would increase with lower rates of population growth, so that demand for education at these
levels would closely parallel future trends in the population aged 6-17 (Serow and Espenshade
1978). An additional consequence of slowing population growth is that the quality of
6
Work in progress – comments welcome
8/1/2012
education can rise – quality as measured by constant resources invested per student (Serow and
Espenshade 1978).
The resulting smaller proportions of primary school age groups relative to the total
population may create favorable demographic conditions, a “window of opportunity”, for
benefiting countries with poor basic educational systems and high levels of educational
inequality as Brazil (Carvalho and Wong 1995). Conceptualizing birth cohort as an important
determinant of well-being (Easterlin 1980), this paper addresses the question of whether the
advantageous demographic conditions for children born after the Demographic Transition, as
opposed to prior, alleviated educational inequality for recent cohorts. Comprehensive analysis
under this framework to examine changes in educational structure will inform policy makers to
better serve their communities as they face unprecedented smaller demands of students in
primary school, a “window of opportunity” for alleviating educational inequality.
In contradiction to Modern Theories of Educational distribution, the maximally
maintained inequality hypothesis developed by Hout, Raftery, and Bell (1993) states that the
effects of social origin on education transitions remain the same from one cohort to another
despite macro changes, except if forced to change by increasing school enrollment. Unless
school enrollment becomes universal, the effect of social origin on schooling remains strong.
Another feature of the country’s unequal educational structure is regional inequalities,
especially between the Northeast and Southeast. These regions are enormously different in
terms of socio-economic characteristics and demographic indicators as the densely populated
Southeast exhibits demographic indicators that resemble the ones of developed countries,
while birth and infant mortality rates are generally higher in the Northeast, as discussed in
7
Work in progress – comments welcome
8/1/2012
Paper 2. These large differences on socio-economic indicators have been stable across time,
and are likely to reflect a very unequal regional allocation of the limited Brazilian investments
in education (Barros and Lam 1996).
Educational attainment is a characteristic that individuals accumulate and carry
throughout their lives, while school enrollment is an activity and a component of eventual
educational attainment (Mare 1995). Increased rates of school enrollment reflect increasing
rates of educational attainment continuation. School enrollment is the current rate at which
schooling is produced in the population. Because of the differences on current status and
accumulation between enrollment and attainment, and also because of the high disparities
between these two measures in Brazil (Lam and Marteleto 2000), this paper will encompass
school enrollment and educational attainment separately.
Research Questions
In this paper, I will address the questions:
1. Has the effect of social background – measured by mother’s education -- on
educational attainment and school enrollment diminished for the smaller cohorts of the post
fertility decline? In other words, has Brazil become a more meritocratic society since going
through the Demographic Transition? I will then assess these regional inequalities on Brazil’s
educational distribution.
2. Are the effects of mother’s education on child’s schooling different in the Southeast
and Northeast regions of Brazil? Do these effects follow Brazil’s pattern? Have these effects
changed for cohorts within such different regions?
8
Work in progress – comments welcome
8/1/2012
Methods
In order to address these questions I estimate models of educational attainment and
school enrollment for children in the cohorts of 1963 and 1983 that focus on the impact of
mother’s education on children’s schooling in Brazil. I address the strength of
intergenerational transmission of education across cohorts and regional inequalities by
estimating separate models by cohort and region in addition to models for the country as a
whole.
I will first model years of school attainment by estimating equation (1) using ordinary
least square regressions:
(1)
S i  a  bM i  cDi  ei
where Si equals the years of schooling for 14 year-old i; M i is a set of dummy variables
indicating mother’s educational attainment; D i is a vector of demographic and residence
characteristics, and e i is a normally distributed error term.
The sets of regressions will consist of separate models for the cohorts of 1963 and 1983
for the whole country, Northeast and Southeast. The usual controls used in models of
children’s educational outcomes such as sex, rural versus urban location, family income, and
region of residence are also included in the models along with mother’s education, the central
piece of the analysis. Next, I will discuss the data used for this study and describe the
analytical sample.
Description of Data and Analytical Sample
In this paper, I will use data from the 1977 and 1997 Pesquisa Nacional por Amostra de
Domicílios/PNAD (National Research of Household Sample), annual household surveys
conducted by the Instituto Brasileiro de Geografia e Estatística (IBGE), the Brazilian
9
Work in progress – comments welcome
8/1/2012
statistical bureau. The PNAD is comparable with the U.S. Current Population Survey (CPS),
and is implemented in September of each year.
The PNAD is a nationally representative survey of extremely good quality. For 1977, the
PNAD contains 498,679 individuals in 100,039 households, compared to 365,870 individuals
in 89,939 households in 1997. The large sample permits sufficient sub-sample sizes for
analysis of specific groups, such as 14 year-olds. For 1977, there are 12,834 14 year-olds in
the sample, compared to 7,861 14-year-olds in 1997. The educational experiences of children
at different ages are sufficiently diverse that it is sensible that they are analyzed separately
(Mare 1993). I have chosen 14 year-olds because this is the eldest age until school enrollment
is legally required in Brazil. This is an ideal age for this analysis also because children who
have been successful in school should be making a transition from primary to secondary
education at 14 years-old. Because the substantive aim is to compare the experience of
different cohorts born before and after crucial socio-economic and demographic changes,
throughout the analysis I will refer to cohorts of children and not to the years of the survey.
Data from 14 year-olds in the 1963 cohort come from the 1977 PNAD, while data from 14
year-olds in the 1983 cohort come from the 1997 PNAD.
The PNAD is appropriate for this study because it contains standard demographic and
socio-economic variables such as sex, age, income, and schooling for all members of the
household. The PNAD provides a complete distribution of single years of schooling.
Mothers’ years of education will be grouped at Grade 12 and over (at least one university year
or more) in the analysis. Another feature of the PNAD that makes it suitable for this study is
that the repeated cross-sections allow for the construction of true cohort histories for schooling
and other social, economic and demographic variables. Data from 1977 and 1997 are
10
Work in progress – comments welcome
8/1/2012
comparable, with the exception of a few discrepancies. Information on race and ethnicity was
not collected in the 1977 PNAD making it impossible to compare ethnic distributions by the
selected cohorts. Also note that the PNAD does not cover the rural part of the Northern region
in 1977 nor in 1997. This probably overestimates the educational and socio-economic
statistics of the Northern region.
Because I want to look at the effects of mothers’ education on children’s education I have
to limit the analytical sample to children who are daughters or sons of the head of the family,
i.e., the ones whose mothers can be identified. This creates a selection bias. Children who are
not daughters or sons of the head of the family may be significantly different from the full
sample of 14 year-olds.
Table 1 shows the socio-economic characteristics used in the analytical models for the full
sample and for children of the head of the family in the 1963 and 1983 cohorts. This provides
evidence on the extent of selection bias. About 9 in 10 children in both cohorts live with at
least one of their parents. Children of the head of the family are not significantly different
from the full sample of children on their distribution across rural/urban location, family
income, region or gender. Comparisons of these groups on school enrollment and years of
education by selected characteristics are provided in Tables x and y of the appendix. There are
no notable differences on enrollment rates nor on educational attainment among the full and
restricted samples. Because the majority of 14 years-olds live with at least one of their
parents, and also because children of the head of the family are not different from all children
there does not appear to be a severe selection bias.
Further descriptions of the analytical sample that are provided in table 1 are
noteworthy. Columns 2 and 4 provide comparisons of the distribution of 1963 and 1983
11
Work in progress – comments welcome
8/1/2012
cohorts across socio-economic characteristics. The general life conditions of these cohorts are
somewhat different. Almost two thirds of 14 year-olds born in 1963 live in urban areas (67%)
compared to nearly four fifths of 14 year-olds born in 1983 (78%). Brazil’s increased
urbanization across the 1960s and 1970s suggests changes in the overall value of children and
therefore on their educational outcomes. With regards to regional distribution, about 3 in 4
children in both cohorts were living in the Southeast and Northeast together. The majority of
Brazilian children live in these regions, reinforcing the importance of studying Brazil’s
Northeast and Southeast separately.
The distribution of children by mother’s education has changed dramatically across
cohorts. The mothers of about 4 in 10 children born in 1963 have no education. Among
children born in 1983, 2 in 10 have mothers without any formal training, half of what was
found for the 1963 cohort. Nearly seven times more children have mothers who attended at
least one year of university in the younger than in the older cohort.
Table 2 provides means and standard deviations for the analytical sample by cohort. It
is worth mentioning that the mean years of mother’s education has nearly doubled over the
twenty years of profound demographic and socio-economic change that separate this study’s
cohorts. Mother’s education went from 2.6 in the older cohort to 5.0 in the younger cohort.
The proportion of 14 year-olds enrolled in school is higher for the younger than for the
older cohorts. Similarly, educational attainment of young people in Brazil has increased
dramatically in the last twenty years. The average education of 14 year-olds grew from 3.4 for
the cohort born in 1963 to 4.7 for the younger cohort. The significant improvement of
mothers’ education may have contributed to the non-trivial increase of 1.3 years of education.
It may also be that the magnitude of these improvements in educational outcomes is different
12
Work in progress – comments welcome
8/1/2012
for specific groups. In order to elucidate that, in the next section I will provide results on
educational outcomes by selected socio-economic variables.
Results
Table 3 provides overall enrollment rate and schooling attainment, as well as the
distribution of these educational measures by socio-economic characteristics separately by
cohort. I will first discuss the distribution of enrollment rates and schooling attainment of 14
year-olds across their socio-economic characteristics. Next, I will compare these distributions
by cohort.
Table 3 demonstrates Brazil’s regional disparities. Columns 1 and 2 show that the
Northeast presents the lowest schooling averages for both older and younger cohorts. In
contrast, the Southeast presents the highest schooling levels for both cohorts. Brazil’s regional
disparities on schooling attainment are decreasing, but remain large. Columns 1 and 2 also
show that, among the older cohort, children in the Northeast had on average 2 fewer years of
schooling than their counterparts in the Southeast. Among children in the younger cohort,
those in the Northeast had on average 1.92 fewer years of schooling than those in the
Southeast. In the older cohort, Northeast children had 52% less education than their Southeast
counterparts, while in the younger cohort Northeast children had on average 35% less
education than those in the Southeast. Even though the Northeast shows the lowest level of
educational attainment in the country, the rates of school enrollment for both cohorts are not
substantially lower for the Northeast as compared with the Southeast. This indicates that
school enrollment is not necessarily translated into educational attainment in Brazil,
reinforcing the role of grade retention and school drop-out (Lam and Marteleto 2000).
13
Work in progress – comments welcome
8/1/2012
Most interestingly are gender differences in levels of school enrollment and attainment
across and within cohorts. Table 3 shows that, among the older cohort, 78% of boys are
enrolled in school compared to 72% of girls. The trend of higher school enrollment for boys is
reversed in the younger cohort: 90% of girls are enrolled in school compared to 81% of boys.
It is interesting that, in the older cohort, even though boys are enrolled in school at higher rates
than girls, girls have on average more years of schooling than boys. The trend of girls’ higher
levels of schooling remains in the younger cohort. The recent pattern of girls’ higher levels of
both educational attainment and school enrollment in detriment of boys is remarkably different
from findings in other developing countries (Hannum 1997; Mensch and Lloyd 1998) and are
in concordance with findings in Brazil (Dureya and Arends-Kuenning 1999).
There are also striking differences on children’s school enrollment and educational
attainment by their mother’s educational levels within and across cohorts. The first two
columns of Table 3 show that children have on average more years of schooling as their
mothers have higher levels of education. The last two columns provide further evidence of the
strong association of mother’s education and children’s educational outcomes. Among both
cohorts, the proportion of children enrolled in school increases as mother’s education rises.
The positive association between mother’s education and children’s educational attainment
and school enrollment was expected and is robust across cohorts. These results suggest that
there still exists an effect of mother’s education in determining children’s schooling. But has
the penalty of having a mother with little or no education versus high levels of education
diminished across cohorts? In this paper, I am particularly interested in whether the
substantial improvement in mother’s educational attainment over the last two decades explains
the large cohort differentials on children’s educational attainment. In the next section, I
14
Work in progress – comments welcome
8/1/2012
attempt to address these questions in order to assess whether Brazil has become a more
meritocratic society with regards to educational distribution.
Brazil: Cohort Results
Table 4 provides the coefficients and standard errors of OLS regressions of children’s
educational attainment on mother’s education and selected socio-economic characteristics for
cohorts of 14 year-olds born in 1963 and 1983, for the whole country1. Nested models were
estimated but are not shown.
Results in Table 4 show that children’s education is highly determined by the
educational attainment of children’s parents in Brazil. Even though children from both cohorts
were penalized for having mothers with low levels of education, such a penalty has decreased
for the younger cohort when compared with the older cohort. For example, Table 4 reveals
that a 14 year-old born in 1963 had on average .486 more years of schooling if his/her mother
had 1 year of education instead of no education. In the 1983 cohort the increase in schooling
from children whose mothers had zero to one years of education was only .23 years of
schooling. This shows that, on average, children in the younger cohort suffer a smaller penalty
for having a mother with low levels of education than did children in the older cohort.
The significance of the differences in the coefficients of mother’s education between
the cohorts was statistically tested using an F-test. The data for the two cohorts was pooled
and interactions between mother’s education and cohort were added to the model (not shown).
1
I have specified mother’s education in the models of children’s educational attainment in several ways:
Linear, dummy variables grouped according to the four major educational categories – no education, primary,
secondary, university or more – and dummy variables for each additional year of education until 15, with zero as
omitted category. The most flexible model, fifteen dummy variables, achieved the smallest chi-square and it is
the model that is shown.
15
Work in progress – comments welcome
8/1/2012
The hypothesis of no difference between the coefficients was rejected at the .001 significance
level. A strong component of educational inequality, the strength of the intergenerational
transmission of education, has decreased in Brazil.
Figure 1 shows the mean schooling of 14 year-olds across cohorts by years of mothers’
education. This Figure demonstrates that the difference in schooling between children whose
mothers completed secondary education (11 years of education) versus no education is about 4
years for the 1963 cohort. This difference decreased to 3 years of schooling for the 1983
cohort. This finding suggests that the gap in the intergenerational transmission of education
has declined in Brazil.
Translating the results to predicted probabilities also provides an intuitive illustration
of the trends in educational attainment and school enrollment across cohorts and levels of
mother’s education. Figure 2 shows predicted probabilities of years of schooling for separate
cohorts by mother’s educational attainment, controlling for selected socio-economic
characteristics. This Figure indicates that, for example, a 14 year-old born in 1963 whose
mother had no education had on average 2.1 years of schooling, while a 14 year-old in the
same conditions born in 1983 had on average 3.0 years of schooling. The penalty on
schooling of having a mother with low levels of education is smaller for children in the
younger cohort than for the older cohort.
As education expands -- higher proportions of children attending school and attaining
higher levels of schooling -- the curve of the relationship of mother’s education and child’s
schooling is flattening. This reinforces the finding that children in younger cohorts are not
being penalized as much as they were in the past on their own educational attainment because
16
Work in progress – comments welcome
8/1/2012
of their mothers educational attainment. The strength of the inequality on the intergenerational
transmission of education has decreased within and across cohorts.
It is also worth noting that children living in rural areas presented significantly lower
levels of educational attainment than children living in urban areas in the older cohort.
However, this penalty was significantly reduced for the younger cohort: While in the older
cohort children in rural areas had 1.15 fewer years of schooling than children in urban areas, in
the younger cohort, this penalty was reduced to less .62 years of schooling. The significant
improvements Brazil has been through in the past decades certainly contributed to the decrease
in the educational penalty for children in rural areas. Information on number and quality of
schools in rural and urban areas would complement analyses to the level of educational supply.
These findings show that Brazil as a whole is becoming a more meritocratic society
with respect to educational distribution, although the strength of inequality from
intergenerational transmission of education still persists for young cohorts of postdemographic transition. But have the levels of educational inequality decreased evenly across
cohorts in the extremely different Northeast and Southeast or is regional inequality another
component of Brazil’s unequal society? In order to address this question, I estimate models
similar to the ones showed for the Northeast and Southeast regions separately. The principal
reason for a regional analysis is the large differences in socio-economic and demographic
figures in the Northeast and Southeast, and the fact that the pace and onset of fertility decline
was significantly different in these regions (Wood and Carvalho 1988). Because fertility rates
differ widely among regions in Brazil these localities may show different educational
experiences as Brazil approached lower fertility levels.
17
Work in progress – comments welcome
8/1/2012
Regional Differences on Mother’s Education and Child’s Schooling
In order to ease the regional comparisons of the models across cohorts, Table 5 shows
models of schooling attainment by region and cohort separately.
Tables 7 and 8 provide results of decompositions of children’s schooling by cohort and
region. The regional decompositions of school attainment show that if children in the
Northeast had the mothers of the Southeast – with greater levels of educational attainment –
they would have higher levels of schooling in both cohorts, but not exactly the same.
Nonetheless, mother’s education accounts for a high proportion of the regional difference in
both cohorts. This finding provides evidence that children in the Northeast are behind on their
educational attainment mainly because of their mothers’ low levels of schooling, and also
because of structural factors such as the conditions of the region and schools, even though
these processes are intrinsically related. This finding confirms, at the educational level, the
high levels of regional inequalities of the country. The difference between the actual and
estimated schooling for Northeast children is 1.71 in the older cohort and 1.40 in the younger
cohort. These numbers show that the difference persists across cohorts. These results also
show that Brazil presents a complicated situation in terms of inequality as intergenerational
transmission of education and regional inequalities interact.
Conclusions
To summarize, three important findings emerge: First, educational attainment has
increased for more recent cohorts of Brazilian children. This is true for the Southeast, the
Northeast, and Brazil as a whole. As education expands -- higher proportions of children
attending school and attaining higher levels of schooling -- the curve of the relationship of
18
Work in progress – comments welcome
8/1/2012
mother’s education and child’s schooling is flattening. Children in younger cohorts are not
being penalized as much as they were in the past on their own educational attainment because
of their mothers low levels of educational attainment. The strength of the inequality on the
intergenerational transmission of education has decreased within and across cohorts.
Second, cohort decompositions show that the increase in schooling stems from the
changing distribution of mother’s educational attainment across children’s cohorts, as well as
from modifications in the effect of mother’s education on children’s education itself. The
effect of mother’s education on children’s schooling has decreased for the younger cohort.
This shows that educational distribution has become more equal in Brazil. The strength of the
unequal intergenerational transmission of education has declined. However, the association
still remains strong.
Third, results from the regional analyses demonstrate that the effect of mother’s
education on children’s schooling is stronger in the Northeast than in the Southeast,
confirming the high regional inequalities of the country at the educational level. The regional
decompositions show that children in the Northeast are behind in terms of educational
attainment to a large extent because of their mothers’ lower levels of education. As Brazil
underwent socio-economic and demographic changes in the last twenty years, the country still
presents persistent high levels of inequality on intergenerational transmission of education, as
well as extreme regional inequalities.
19
Work in progress – comments welcome
8/1/2012
Reference List (incomplete)
Barros, Ricardo and David Lam. "Income and Education Inequality and Children's Schooling
Attainment." Opportunity Foregone: Education, Growth, and Inequality in Brazil.
Nancy Birdsall and Richard Sabot. Washington, D.C.: Inter-American Development
Bank, 1996.
Birdsall, Nancy and Richard Sabot. 1996. Opportunity Foregone: Education in Brazil.
Washington, D.C.: Inter-American Development Bank.
Deaton, Angus. The Analysis of Household Surveys: A Microeconometric Approach to
Development Policy. Baltimore: Johns Hopkins University Press, 1997.
Duryea, Suzanne and Miguel Skézely. 2000. A Micro-Macro Approach. Unpublished
Manuscript.
Filmer, Deon and Lant Prichett. "The Effect of Household Wealth on Educational Attainment:
Evidence From 35 Countries." Population and Development Review 25, no. 1 (1999):
85-120.
Featherman, David and Robert Hauser. 1975. Changes in the socioeconomic stratification of
the races, 1962-1973. Madison: University of Wisconsin Press.
Harbison, Ralph and Eric Hanushek. Educational Performance of the Poor: Lessons From
Rural Northeast Brazil. Oxford: Oxford University Press, 1992.
Hernandez, Donald J. America's Children: Resources From Family, Government, and the
Economy. New York: Russell Sage Foundation, 1993.
IBGE. Indicadores Socias: Uma Analise Da Decada De 80. Rio de Janeiro: 1995.
Lam, David and Deborah Levison. Declining inequality in schooling in Brazil and its effects
on inequality in earnings. Journal of Development Economics 37:199-225, 1991.
Lam, David and Suzanne Duryea. "Effects of Schooling on Fertility, Labor Supply, and
Investment in Children With Evidence From Brazil." Journal of Human Resources,
1999.
Lam, David. 1999. “Generating Extreme Inequality: Schooling, Earnings, and
Intergenerational Transmission of Human Capital in South Africa and Brazil.”
Population Studies Center, University of Michigan, Research Report 99-439.
Lam, David and Letícia Marteleto. “Grade repetition, School Enrollment, and Economic
Shocks in Brazil.” Paper presented at the 2000 PAA Meeting, Los Angeles, 23-25
March. 2000.
Levison, Deborah. 1991. "Children's Labor Force Participation and Schooling in Brazil." The
University of Michigan.
20
Work in progress – comments welcome
8/1/2012
Mare, Robert. 1997. “Differential fertility, intergenerational educational mobility, and racial
inequality.” Social Science Research 26:263-91.
_____ . 1996. “Demography and the Evolution of Educational Inequality.” In James Baron,
David Grusky, and Donald Treiman (Eds.). Social Differentiation and Social
Inequality: Essays in Honor of John C. Pock. Boulder, CO: Westview Press
Martine, George. 1996. "Brazil's Fertility Decline, 1965-95: A Fresh Look at Key Factors."
Population and Development Review 22, no. 1 pp. 47-75.
Mendes, Marcia Martins and Dias Vera Regina. 1995. "Implicacoes Da Dinâmica
Demográfica Sobre o Sistema Educacional.” Rio de Janeiro, RJ: IBGE.
Plank, David N. 1996. The Means of Our Salvation: Public Education in Brazil, 1930-1995.
Oxford: WestviewPress.
Wood, Charles and Jose Alberto Magno de Carvalho. 1988. The Demography of Inequality in
Brazil. Cambridge: Cambridge University Press.
21
Work in progress – comments welcome
8/1/2012
Table 1. Socio-Economic Characteristics of 14 Year-Olds [%]
Cohorts of 1963 and 1983, Brazil
Cohort of 1963
Cohort of 1983
All
Children of
All
Children of
Children
the Head
Children
the Head
Rural/Urban Location
Urban
63.59
62.71
77.92
77.76
Rural
36.41
37.29
22.08
22.24
Region
Southeast = 0
42.13
42.79
40.09
40.96
North = 1
2.24
2.01
5.76
5.51
Northeast = 2
32.40
31.40
33.14
32.13
South = 3
19.65
20.38
14.32
14.61
Central = 4
3.58
3.42
6.69
6.78
Gender
Male
49.37
50.77
49.56
50.60
Female
50.63
49.23
50.44
49.40
Mother’s Education
No Education (0)
N/A
36.94
N/A
19.43
Attended First Primary (1-4)
N/A
47.08
N/A
38.72
Attended Second Primary (5-8)
N/A
11.33
N/A
22.81
Attended High School (9-11)
N/A
3.26
N/A
11.72
Attended University or more (12+)
N/A
1.37
N/A
7.30
Family Income (Quintiles)
First Quintile
16.24
15.83
20.06
19.53
Second Quintile
19.16
19.27
21.90
21.52
Third Quintile
22.01
22.49
19.24
19.22
Fourth Quintile
23.83
24.23
20.37
20.64
Fifth Quintile
18.76
18.18
18.43
19.09
[N]
12,834
11,269
7,861
7,131
Source: PNADs 1977, 1997.
Table 2. Summary Statistics for Outcome and Explanatory Variables
Cohorts of 14 Year-olds born in 1963 and 1983, Brazil
Cohort of 1963
Cohort of 1983
Mean
Std. Dev.
Mean
Std. Dev.
Enrollment
Years of Schooling
Rural/Urban Location
Region
Gender
Mother’s Education
Family Income
[N]
3.42
.37
1.40
.49
2.61
5949.23
7,162
2.32
.48
1.31
.50
2.94
15053.31
4.74
.22
1.41
.49
4.97
887.41
6,672
2.12
.41
1.33
.50
4.21
1300.65
Source: PNADs 1977, 1997.
22
Work in progress – comments welcome
8/1/2012
Table 3. School Enrollment and Schooling by Socio-Economic Characteristics
Cohorts of 14 Year-olds born in 1963 and 1983, Brazil
Mean Years of
Enrollment Rates
Schooling
[%]
Cohort of
Cohort of
Cohort of
Cohort of
1963
1983
1963
1983
Enrolled in School
75.00
88.68
Educational Attainment
3.40
4.72
Rural/Urban Location
Urban
16
5.09
83.59
90.98
Rural
2.14
3.44
75.75
80.66
Region
Southeast
14
5.41
75.96
90.68
North
3.42
4.03
94.97
89.69
Northeast
1.98
3.49
83.59
86.12
South
4.01
5.64
58.91
88.45
Central
3.81
4.96
87.80
88.47
Gender
Male
3.19
4.39
77.67
86.96
Female
3.62
5.05
72.38
90.45
Mother’s Education
No Education (0)
2.13
2.91
66.19
76.35
Attended First Primary (1-4)
3.80
4.55
75.29
87.01
Attended Second Primary (5-8)
5.00
5.31
89.91
94.88
Attended High School (9-11)
5.81
6.13
97.69
98.23
Attended University or more (12+)
6.37
6.58
100.00
99.40
Family Income (Quintiles)
First Quintile
1.86
3.24
68.96
81.83
Second Quintile
2.47
3.98
68.05
83.32
Third Quintile
3.28
4.74
70.57
87.59
Fourth Quintile
4.02
5.52
75.65
93.42
Fifth Quintile
5.09
6.20
88.83
97.52
[N]
7,162
6,672
7,162
6,672
Source: PNADs 1977, 1997
23
Work in progress – comments welcome
8/1/2012
Table 4. OLS Regression Results
School Attainment - Cohorts of 14 Year-olds born in 1963 and 1983, Brazil
Cohort of 1963
Cohort of 1983
Coef.
Std.Error
Coef.
Std.Error
Mother’s Education (no education omitted)
One
.486***
.094
.232***
.114
Two
.857***
.071
.587***
.084
Three
.999***
.067
.618***
.078
Four
1.329***
.063
1.087***
.064
Five
1.602***
.095
.915 ***
.083
Six
2.868***
.173
1.136***
.111
Seven
1.928***
.199
1.342***
.117
Eight
1.928***
.128
1.412***
.085
Nine
1.292***
.369
1.630 ***
.197
Ten
2.327***
.301
1.774***
.164
Eleven
2.318***
.141
1.749***
.080
Twelve or +
2.434***
.183
1.807***
.187
Rural=1
-.993***
.049
-.449***
.051
Female=1
.402***
.041
.611 ***
.039
Log Household Income
.453***
.026
.520***
Region (Southeast omitted)
South
.263***
.057
.328***
.089
North
-1.158***
.148
-.989***
.049
Center-West
-.620***
.115
-.243 ***
.060
Northeast
-1.115***
.054
-.974***
.081
Constant
-.533
.218
.764
.058
R2
.446
.421
[N]
7,162
6,672
Source: PNADs 1977, 1997.
Notes: ***Significant at 1%; **Significant at 5%; *Significant at 10%.
24
Work in progress – comments welcome
8/1/2012
Table 5. OLS Regression Results
School Attainment - Cohorts of 14 Year-olds born in 1963 and 1983, Northeast & Southeast, Brazil
Northeast
Southeast
Cohort of 1963
Cohort of 1983
Cohort of 1963
Cohort of 1983
Std.
Std.
Std.
Std.
Coef.
Error
Coef.
Error
Coef.
Error
Coef.
Error
Mother’s Education
(omitted=0)
One
0.450***
0.133 0.422**
0.175
0.617***
0.188 -0.118
0.252
Two
1.027***
0.115 0.651***
0.140
0.843***
0.122 0.616***
0.156
Three
1.102***
0.116 0.922***
0.138
0.862***
0.113 0.237***
0.138
Four
1.336***
0.118 1.281***
0.126
1.306***
0.104 0.875***
0.109
Five
2.103***
0.166 1.201***
0.165
1.067***
0.182 0.551***
0.149
Six
2.181***
0.470 1.207***
0.251
1.777***
0.245 1.008***
0.176
Seven
2.569***
0.384 1.668***
0.270
1.757***
0.312 1.005***
0.183
Eight
1.944**
0.322 1.872***
0.203
1.801***
0.185 1.055***
0.137
Nine
1.411**
0.676 2.342***
0.450
0.998**
0.602 1.510***
0.308
Ten
2.890***
0.493 2.661***
0.423
1.643***
0.479 1.502***
0.246
Eleven
2.606***
0.301 2.448***
0.162
1.946***
0.225 1.206***
0.141
Twelve or +
3.735***
0.480 2.936***
0.250
1.910***
0.263 1.356***
0.170
Rural=1
-1.307***
0.075 -0.642***
0.084 -0.691***
0.091 -0.382***
0.108
Female=1
0.506***
0.067 0.806***
0.076
0.383***
0.070 0.541***
0.066
Log Family Income 0.333***
0.042 0.463***
0.046
0.594***
0.046 0.529***
0.041
Constant
-0.686
0.326 -0.122
0.267 -1.688
0.382 0.989***
0.278
R2
0.408
0.374
0.290
0.249
[N]
2,398
2,216
2,666
2,095
Source: PNADs 1977, 1977.
Notes: ***Significant at 1%; **Significant at 5%; *Significant at 10%.
25
Work in progress – comments welcome
8/1/2012
Table 6
Decomposition of Increasing Schooling of 14 Year-olds in cohorts of 1963 and 1983
Cohort of
Cohort of
Change
Predicted Change
1963
1983
(% of actual)
Mean Schooling of 14 Year-olds
3.40
4.44
1.04
Coefficients of 1977 and distribution of
4.24
0.84
1997
Change attributable to change in
81%
distribution.
Coefficients of 1997 and distribution of
4.10
0.34
1977
Change attributable to change in the
67%
coefficient
Source: PNADs 1977, 1997.
Table 7
Decomposition of Increasing Schooling of 14 Year-olds in cohort of 1963, Northeast and
Southeast Brazil
Northeast
Southeast
Change
Predicted Change
(% of actual)
Mean Schooling of 14 Year-olds
1.97
4.17
2.20
Coefficients of Southeast and
3.68
1.71
distribution of Northeast
Change attributable to change in
78%
distribution.
Coefficients of Northeast and
2.71
1.46
distribution of Southeast
Change attributable to change in the
34%
coefficient
Source: PNAD 1977
Table 8
Decomposition of Increasing Schooling of 14 Year-olds in cohort of 1983, Northeast and
Southeast Brazil
Northeast
Southeast
Change
Predicted Change
(% of actual)
Mean Schooling of 14 Year-olds
3.48
5.41
1.93
Coefficients of Southeast and
4.88
1.40
distribution of Northeast
Change attributable to change in
73%
distribution.
Coefficients of Northeast and
4.14
1.27
distribution of Southeast
Change attributable to change in the
34%
coefficient
Source: PNAD 1997
26
Work in progress – comments welcome
8/1/2012
Figure 1. Estimated Schooling for 14 Year-olds of 1963 and 1983 Cohorts,
Brazil
7
6
5
4
3
2
1
0
0
1
2
3
4
5
6
7
8
Mother's Years of Schooling
Cohort 1963
9
10
11
12
Cohort 1983
Figure 2. Predicted Schooling for 14 Year-olds of 1963 and 1983 Cohorts,
Brazil
7
6
5
4
3
2
1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
Mother's Years of Schooling
Discrete Coding, Cohort 1963
Discrete Coding, Cohort 1983
Linear Effect, Cohort 1963
Linear Effect, Cohort 1983
27
Work in progress – comments welcome
8/1/2012
Figure 3. Predicted Years of Schooling for Cohort of 1963 and 1983,
Northeast and Southeast
7
6
5
Ne, 1963
Se, 1963
Ne, 1983
Se, 1983
4
3
2
1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
Mother's Years of Schooling
28